To see the other types of publications on this topic, follow the link: Fusion neural network.
Dissertations / Theses on the topic 'Fusion neural network'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'Fusion neural network.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
1
Belghaddar, Yassine. "Data fusion for urban network mapping : application to wastewater networks." Electronic Thesis or Diss., Université de Montpellier (2022-....), 2022. http://www.theses.fr/2022UMONG092.
Full textAbstract:
Nombreuses sont les raisons qui rendent la maitrise des données des réseaux souterrains essentielle~: réduire le coût des réparations et des interventions, lancer des simulations hydrauliques, préserver l'environnement etc. Les données disponibles relatives à ces réseaux et plus spécifiquement ceux d'assainissement sont diverses en terme de types (textes, images, SIG, etc.) et de formats (analogique, numérique). De plus, ces données émanant de sources multiples sont généralement incomplètes, imprécises, incertaines et parfois contradictoires. De ce fait, dans le but d'extraire l'information pertinente à partir de ces données multi-sources/multi-formats, un processus de fusion de données est nécessaire. En effet, les attributs (profondeur, diamètre d'une conduite, etc.) sont toujours associés à une représentation spatiale des objets (conduite, exutoire, etc.). Dans ce cadre, les travaux de cette thèse auront comme premier objectif d'utiliser, adapter et proposer des techniques de fusion et d'intégration de données spatiales pour combiner les données collectées à partir de plusieurs sources. En deuxième lieu, le focus sera mis sur l'objectif de complétion et d'estimation des données manquantes. Le résultat de ce travail pluridisciplinaire sera la mise en place de méthodes de fusion et de complétion de données manquantes permettant la cartographie et la modélisation hydraulique d'un réseau d'assainissement urbain<br>There are many reasons that make data management of underground networks essential: reducing the cost of repairs and expansion, running hydraulic simulations, preserving the environment etc. The available data related to these networks and more specifically wastewater ones are various, and come in different types (texts, images, GIS ,etc.) and formats (analog, digital). In addition, these multisource/multi-format data are usually incomplete, uncertain, imprecise and sometimes contradictory. Consequently, in order to extract relevant information from these inputs, a data fusion process is necessary. In fact, attributes (depth, diameter of a pipe, etc.) are always associated to a spatial representation of the objects (pipes, outfall, etc.). In this context, this work will concentrate first on using, adapting and putting forward data fusion and integration techniques to combine data collected from different sources. The second part of this thesis will be dedicated to impute and estimate missing data of a wastewater network. The result of this multidisciplinary work will be to put forward methods for fusing data and imputing the missing ones, which enables the mapping and the modelling of urban wastewater networks
2
Seneker, Stephen S. "Synesthetic sensor fusion via a cross-wired artificial neural network." [Johnson City, Tenn. : East Tennessee State University], 2002. http://etd-submit.etsu.edu/etd/theses/available/etd-0403102-164937/unrestricted/SenekerS041902.pdf.
Full text
3
Seneker, Stephen Samuel. "Synesthetic Sensor Fusion via a Cross-Wired Artificial Neural Network." Digital Commons @ East Tennessee State University, 2002. https://dc.etsu.edu/etd/673.
Full textAbstract:
The purpose of this interdisciplinary study was to examine the behavior of two artificial neural networks cross-wired based on the synesthesia cross-wiring hypothesis. Motivation for the study was derived from the study of psychology, robotics, and artificial neural networks, with perceivable application in the domain of mobile autonomous robotics where sensor fusion is a current research topic. This model of synesthetic sensor fusion does not exhibit synesthetic responses. However, it was observed that cross-wiring two independent networks does not change the functionality of the individual networks, but allows the inputs to one network to partially determine the outputs of the other network in some cases. Specifically, there are measurable influences of network A on network B, and yet network B retains its ability to respond independently.
4
Koh, Leonard Phin-Liong. "A neural network approach to multisensor data fusion for vessel traffic services." Thesis, Monterey, Calif. : Springfield, Va. : Naval Postgraduate School ; Available from National Technical Information Service, 1995. http://handle.dtic.mil/100.2/ADA294251.
Full text
5
Pasika, Hugh Joseph Christopher. "Neural network sensor fusion : creation of a virtual sensor for cloud-base height estimation /." *McMaster only, 1999.
Find full text
6
Pasika, Hugh Joseph Christopher. "Neural network sensor fusion, creation of a virtual sensor for cloud-base height estimation." Thesis, National Library of Canada = Bibliothèque nationale du Canada, 2000. http://www.collectionscanada.ca/obj/s4/f2/dsk1/tape4/PQDD_0031/NQ66287.pdf.
Full text
7
Correa, Silva Joan Li Guisell, and Sofia Jönsson. "Investigation of Increased Mapping Quality Generated by a Neural Network for Camera-LiDAR Sensor Fusion." Thesis, KTH, Mekatronik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-297463.
Full textAbstract:
This study’s aim was to investigate the mapping part of Simultaneous Localisation And Mapping (SLAM) in indoor environments containing error sources relevant to two types of sensors. The sensors used were an Intel Realsense depth camera and an RPlidar Light Detection AndRanging (LiDAR). Both cameras and LiDARs are frequently used as exteroceptive sensors in SLAM. Cameras typically struggle with strong light in the environment, and LiDARs struggle with reflective surfaces. Therefore, this study investigated the possibility of using a neural network to detect an error in either sensors’ data caused by mentioned error sources. The network identified which sensor produced erroneous data. The sensor fusion algorithm momentarily excluded said sensor’s data, consequently, improving the mapping quality when possible. The quantitative results showed no significant difference in the measured mean squared error and structural similarity between the final maps generated with and without the network, when compared to the ground truth. However, the qualitative analysis showed some advantages with using the network. Many of the camera’s errors were filtered out with the neural network, and led to a more accurate continuous mapping than without the network implemented. The conclusion was that a neural network can to a limited extent recognise the sensors’ data errors, but only the camera data benefited from the proposed solution. The study also produced important findings from the implementation which are presented. Future work recommendations include neural network optimisation, sensor selection, and sensor fusion implementation.<br>Denna studie undersökte kartläggningen i Simultaneous Localisation And Mapping (SLAM) problem, i kontexten av två sensorers felkällor. Sensorerna som användes var en Intel Realsense djupseende kamera samt en LiDAR fran RPlidar. Både kameror och LiDARs är vanliga sensorer i SLAM system, och båda har olika typer av felkällor. Kameror är typiskt känsliga för mycket starkt ljus, medan LiDARs har svårt med reflekterande ytor. Med detta som bakgrund har denna studie undersökt möjligheten att implementera ett neuralt nätverk för att detektera när varje sensor är utsatt för en felkälla (och därmed ger fel data). Nätverkets klassificering används sedan för att i varje tidssteg exkludera den sensors data som det är fel på för att förbättra kartläggningen. De qvantitativa resultaten visade ingen signifikant skillnad mellan kartorna genererade med nätverket och de utan nätverket. Dock visade den kvalitativa analysen att det finns vissa fördelar med att använda det neutrala nätverket. Manga av kamerans fel blev korrigerade när nätverket var implementerat, vilket ledde till mer korrekta kartor under kontinuerlig körning. Slutsatsen blev att ett nätverk kan bli tränat för att identifiera fel i datan, men att kameran drar mest nytta av det. Studien producerade även sekundara resultat som också redovisas. Slutligen rekommenderas optimering av nätverket, val av sensorer, samt uppdaterad algoritm för sensor fusionen som möjliga områden till fortsatt forskning inom området.
8
Rotelli, Matthew D. "Neural networks as a tool for statistical modeling." Diss., This resource online, 1996. http://scholar.lib.vt.edu/theses/available/etd-06062008-151625/.
Full text
9
Ayodeji, Akiwowo. "Developing integrated data fusion algorithms for a portable cargo screening detection system." Thesis, Loughborough University, 2012. https://dspace.lboro.ac.uk/2134/9901.
Full textAbstract:
Towards having a one size fits all solution to cocaine detection at borders; this thesis proposes a systematic cocaine detection methodology that can use raw data output from a fibre optic sensor to produce a set of unique features whose decisions can be combined to lead to reliable output. This multidisciplinary research makes use of real data sourced from cocaine analyte detecting fibre optic sensor developed by one of the collaborators - City University, London. This research advocates a two-step approach: For the first step, the raw sensor data are collected and stored. Level one fusion i.e. analyses, pre-processing and feature extraction is performed at this stage. In step two, using experimentally pre-determined thresholds, each feature decides on detection of cocaine or otherwise with a corresponding posterior probability. High level sensor fusion is then performed on this output locally to combine these decisions and their probabilities at time intervals. Output from every time interval is stored in the database and used as prior data for the next time interval. The final output is a decision on detection of cocaine. The key contributions of this thesis includes investigating the use of data fusion techniques as a solution for overcoming challenges in the real time detection of cocaine using fibre optic sensor technology together with an innovative user interface design. A generalizable sensor fusion architecture is suggested and implemented using the Bayesian and Dempster-Shafer techniques. The results from implemented experiments show great promise with this architecture especially in overcoming sensor limitations. A 5-fold cross validation system using a 12 13 - 1 Neural Network was used in validating the feature selection process. This validation step yielded 89.5% and 10.5% true positive and false alarm rates with 0.8 correlation coefficient. Using the Bayesian Technique, it is possible to achieve 100% detection whilst the Dempster Shafer technique achieves a 95% detection using the same features as inputs to the DF system.
10
Thomas, Kim. "Incident detection on arterials using neural network data fusion of simulated probe vehicle and loop detector data /." [St. Lucia, Qld.], 2005. http://www.library.uq.edu.au/pdfserve.php?image=thesisabs/absthe18433.pdf.
Full text
11
Garratt, Matthew Adam, and m. garratt@adfa edu au. "Biologically Inspired Vision and Control for an Autonomous Flying Vehicle." The Australian National University. Research School of Biological Sciences, 2008. http://thesis.anu.edu.au./public/adt-ANU20090116.154822.
Full textAbstract:
This thesis makes a number of new contributions to control and sensing for unmanned vehicles. I begin by developing a non-linear simulation of a small unmanned helicopter and then proceed to develop new algorithms for control and sensing using the simulation. The work is field-tested in successful flight trials of biologically
inspired vision and neural network control for an unstable rotorcraft. The techniques are more robust and more easily implemented on a small flying vehicle than previously attempted methods.¶
Experiments from biology suggest that the sensing of image motion or optic
flow in insects provides a means of determining the range to obstacles and terrain. This biologically inspired approach is applied to control of height in a helicopter, leading to the Worlds first optic flow based terrain following controller for an unmanned helicopter in forward flight. Another novel optic flow based controller is developed for the control of velocity in hover. Using the measurements of height from other sensors, optic flow is used to provide a measure of the helicopters lateral and longitudinal velocities relative to the ground plane. Feedback of these velocity measurements enables automated hover with a drift of only a few cm per second, which is sufficient to allow a helicopter to land autonomously in gusty conditions
with no absolute measurement of position.¶
New techniques for sensor fusion using Extended Kalman Filtering are developed to estimate attitude and velocity from noisy inertial sensors and optic flow measurements. However, such control and sensor fusion techniques can be computationally
intensive, rendering them difficult or impossible to implement on a small
unmanned vehicle due to limitations on computing resources. Since neural networks can perform these functions with minimal computing hardware, a new technique of control using neural networks is presented. First a hybrid plant model consisting of exactly known dynamics is combined with a black-box representation of the unknown dynamics. Simulated trajectories are then calculated for the plant using an optimal controller. Finally, a neural network is trained to mimic the optimal controller. Flight test results of control of the heave dynamics of a helicopter confirm
the neural network controllers ability to operate in high disturbance conditions and suggest that the neural network outperforms a PD controller. Sensor fusion and control of the lateral and longitudinal dynamics of the helicopter are also shown to
be easily achieved using computationally modest neural networks.
12
PACIFICI, FABIO. "Novel neural network-based algorithms for urban classification and change detection from satellite imagery." Doctoral thesis, Università degli Studi di Roma "Tor Vergata", 2010. http://hdl.handle.net/2108/1239.
Full textAbstract:
L`attività umana sta cambiando radicalmente l`ecosistema ambientale, unito anche alla rapida espansione demografica dei sistemi urbani. Benche` queste aree rappresentano solo una minima frazione della Terra, il loro impatto sulla richiesta di energia, cibo, acqua e materiali primi, e` enorme. Per cui, una informazione accurata e tempestiva risulta essere essenziale per gli enti di protezione civile in caso, ad esempio, di catastrofi ambientali.
Negli ultimi anni il forte sviluppo di sistemi satellitari, sia dal punto di vista della risoluzione spaziale che di quella radiometrica e temporale, ha permesso una sempre piu` accurato monitoraggio della Terra, sia con sistemi ottici che con quelli RADAR.
Ad ogni modo, una piu` alta risoluzione (sia spaziale, che spettrale o temporale) presenta tanti vantaggi e miglioramenti quanti svantaggi e limitazioni.
In questa tesi sono discussi in dettaglio i diversi aspetti e tecniche per la classificazione e monitoraggio dei cambiamenti di aree urbane, utilizzando sia sistemi ottici che RADAR. Particolare enfasi e` data alla teoria ed all`uso di reti neurali.<br>Human activity dominates the Earth's ecosystems with structural modifications. The rapid population growth over recent decades and the concentration of this population in and around urban areas have significantly impacted the environment. Although urban areas represent a small fraction of the land surface, they affect large areas due to the magnitude of the associated energy, food, water, and raw material demands. Reliable information in populated areas is essential for urban planning and strategic decision making, such as civil protection departments in cases of emergency.
Remote sensing is increasingly being used as a timely and cost-effective source of information in a wide number of applications, from environment monitoring to location-aware systems. However, mapping human settlements represents one of the most challenging areas for the remote sensing community due to its high spatial and spectral diversity. From the physical composition point of view, several different materials can be used for the same man-made element (for example, building roofs can be made of clay tiles, metal, asphalt, concrete, plastic, grass or stones). On the other hand, the same material can be used for different purposes (for example, concrete can be found in paved roads or building roofs). Moreover, urban areas are often made up of materials present in the surrounding region, making them indistinguishable from the natural or agricultural areas (examples can be unpaved roads and bare soil, clay tiles and bare soil, or parks and vegetated open spaces) [1].
During the last two decades, significant progress has been made in developing and launching satellites with instruments, in both the optical/infrared and microwave regions of the spectra, well suited for Earth observation with an increasingly finer spatial, spectral and temporal resolution. Fine spatial sensors with metric or sub-metric resolution allow the detection of small-scale objects, such as elements of residential housing, commercial buildings, transportation systems and utilities. Multi-spectral and hyper-spectral remote sensing systems provide additional discriminative features for classes that are spectrally similar, due to their higher spectral resolution. The temporal component, integrated with the spectral and spatial dimensions, provides essential information, for example on vegetation dynamics. Moreover, the delineation of temporal homogeneous patches reduces the effect of local spatial heterogeneity that often masks larger spatial patterns.
Nevertheless, higher resolution (spatial, spectral or temporal) imagery comes with limits and challenges that equal the advantages and improvements, and this is valid for both optical and synthetic aperture radar data [2].
This thesis addresses the different aspects of mapping and change detection of human settlements, discussing the main issues related to the use of optical and synthetic aperture radar data. Novel approaches and techniques are proposed and critically discussed to cope with the challenges of urban areas, including data fusion, image information mining, and active learning. The chapters are subdivided into three main parts. Part I addresses the theoretical aspects of neural networks, including their different architectures, design, and training. The proposed neural networks-based algorithms, their applications to classification and change detection problems, and the experimental results are described in Part II and Part III.
13
Bodén, Johan. "A Comparative Study of Reinforcement-based and Semi-classical Learning in Sensor Fusion." Thesis, Karlstads universitet, Institutionen för matematik och datavetenskap (from 2013), 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:kau:diva-84784.
Full textAbstract:
Reinforcement learning has proven itself very useful in certain areas, such as games. However, the approach has been seen as quite limited. Reinforcement-based learning has for instance not been commonly used for classification tasks as it is receiving feedback on how well it did for an action performed on a specific input. This slows the performance convergence rate as compared to other classification approaches which has the input and the corresponding output to train on. Nevertheless, this thesis aims to investigate whether reinforcement-based learning could successfully be employed on a classification task. Moreover, as sensor fusion is an expanding field which can for instance assist autonomous vehicles in understanding its surroundings, it is also interesting to see how sensor fusion, i.e., fusion between lidar and RGB images, could increase the performance in a classification task. In this thesis, a reinforcement-based learning approach is compared to a semi-classical approach. As an example of a reinforcement learning model, a deep Q-learning network was chosen, and a support vector machine classifier built on top of a deep neural network, was chosen as an example of a semi-classical model. In this work, these frameworks are compared with and without sensor fusion to see whether fusion improves their performance. Experiments show that the evaluated reinforcement-based learning approach underperforms in terms of metrics but mainly due to its slow learning process, in comparison to the semi-classical approach. However, on the other hand using reinforcement-based learning to carry out a classification task could still in some cases be advantageous, as it still performs fairly well in terms of the metrics presented in this work, e.g. F1-score, or for instance imbalanced datasets. As for the impact of sensor fusion, a notable improvement can be seen, e.g. when training the deep Q-learning model for 50 episodes, the F1-score increased with 0.1329; especially, when taking into account that the most of the lidar data used in the fusion is lost since this work projects the 3D lidar data onto the same 2D plane as the RGB images.
14
Algashaam, Faisal Mansour A. "Multispectral techniques for biometrics with focus on periocular region." Thesis, Queensland University of Technology, 2018. https://eprints.qut.edu.au/122985/1/Faisal%20Mansour%20A_Algashaam_Thesis.pdf.
Full textAbstract:
During this study the researcher worked on human identification using the eye region called periocular and iris. The research has proposed a number of novel techniques to improve the recognition accuracy of the identification system. This work is potentially applicable to security and surveillance systems such as border control and attendance management.
15
Respati, Sara Wibawaning. "Network-scale arterial traffic state prediction: Fusing multisensor traffic data." Thesis, Queensland University of Technology, 2020. https://eprints.qut.edu.au/202990/1/Sara%20Wibawaning_Respati_Thesis.pdf.
Full textAbstract:
Road traffic congestion is an increasing societal problem. Road agencies and users seeks accurate and reliable travel speed information. This thesis developed a network-scale traffic state prediction based on Convolutional Neural Network (CNN). The method can predict the speed over the network accurately by preserving road connectivity and incorporating historical datasets. When dealing with an extensive network, the thesis also developed a clustering method to reduce the complexity of the prediction. By accurately predict the traffic state over a network, traffic operators can manage the network more effectively and travellers can make informed decision on their journeys.
16
Bergenroth, Hannah. "Use of Thermal Imagery for Robust Moving Object Detection." Thesis, Linköpings universitet, Medie- och Informationsteknik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-177888.
Full textAbstract:
This work proposes a system that utilizes both infrared and visual imagery to create a more robust object detection and classification system. The system consists of two main parts: a moving object detector and a target classifier. The first stage detects moving objects in visible and infrared spectrum using background subtraction based on Gaussian Mixture Models. Low-level fusion is performed to combine the foreground regions in the respective domain. For the second stage, a Convolutional Neural Network (CNN), pre-trained on the ImageNet dataset is used to classify the detected targets into one of the pre-defined classes; human and vehicle. The performance of the proposed object detector is evaluated using multiple video streams recorded in different areas and under various weather conditions, which form a broad basis for testing the suggested method. The accuracy of the classifier is evaluated from experimentally generated images from the moving object detection stage supplemented with publicly available CIFAR-10 and CIFAR-100 datasets. The low-level fusion method shows to be more effective than using either domain separately in terms of detection results.<br><p>Examensarbetet är utfört vid Institutionen för teknik och naturvetenskap (ITN) vid Tekniska fakulteten, Linköpings universitet</p>
17
Mirzaei, Golrokh. "Data Fusion of Infrared, Radar, and Acoustics Based Monitoring System." University of Toledo / OhioLINK, 2014. http://rave.ohiolink.edu/etdc/view?acc_num=toledo1396564236.
Full text
18
Stone, David L. "The Application of Index Based, Region Segmentation, and Deep Learning Approaches to Sensor Fusion for Vegetation Detection." VCU Scholars Compass, 2019. https://scholarscompass.vcu.edu/etd/5708.
Full textAbstract:
This thesis investigates the application of index based, region segmentation, and deep learning methods to the sensor fusion of omnidirectional (O-D) Infrared (IR) sensors, Kinnect sensors, and O-D vision sensors to increase the level of intelligent perception for unmanned robotic platforms. The goals of this work is first to provide a more robust calibration approach and improve the calibration of low resolution and noisy IR O-D cameras. Then our goal was to explore the best approach to sensor fusion for vegetation detection. We looked at index based, region segmentation, and deep learning methods and compared them with a goal of significant reduction in false positives while maintaining reasonable vegetation detection.
The results are as follows:
Direct Spherical Calibration of the IR camera provided a more consistent and robust calibration board capture and resulted in the best overall calibration results with sub-pixel accuracy
The best approach for sensor fusion for vegetation detection was the deep learning approach, the three methods are detailed in the following chapters with the results summarized here.
Modified Normalized Difference Vegetation Index approach achieved 86.74% recognition and 32.5% false positive, with peaks to 80%
Thermal Region Fusion (TRF) achieved a lower recognition rate at 75.16% but reduced false positives to 11.75% (a 64% reduction)
Our Deep Learning Fusion Network (DeepFuseNet) results demonstrated that deep learning approach showed the best results with a significant (92%) reduction in false positives when compared to our modified normalized difference vegetation index approach. The recognition was 95.6% with 2% false positive.
Current approaches are primarily focused on O-D color vision for localization, mapping, and tracking and do not adequately address the application of these sensors to vegetation detection. We will demonstrate the contradiction between current approaches and our deep sensor fusion (DeepFuseNet) for vegetation detection. The combination of O-D IR and O-D color vision coupled with deep learning for the extraction of vegetation material type, has great potential for robot perception. This thesis will look at two architectures: 1) the application of Autoencoders Feature Extractors feeding a deep Convolution Neural Network (CNN) fusion network (DeepFuseNet), and 2) Bottleneck CNN feature extractors feeding a deep CNN fusion network (DeepFuseNet) for the fusion of O-D IR and O-D visual sensors. We show that the vegetation recognition rate and the number of false detects inherent in the classical indices based spectral decomposition are greatly improved using our DeepFuseNet architecture.
We first investigate the calibration of omnidirectional infrared (IR) camera for intelligent perception applications. The low resolution omnidirectional (O-D) IR image edge boundaries are not as sharp as with color vision cameras, and as a result, the standard calibration methods were harder to use and less accurate with the low definition of the omnidirectional IR camera. In order to more fully address omnidirectional IR camera calibration, we propose a new calibration grid center coordinates control point discovery methodology and a Direct Spherical Calibration (DSC) approach for a more robust and accurate method of calibration. DSC addresses the limitations of the existing methods by using the spherical coordinates of the centroid of the calibration board to directly triangulate the location of the camera center and iteratively solve for the camera parameters. We compare DSC to three Baseline visual calibration methodologies and augment them with additional output of the spherical results for comparison. We also look at the optimum number of calibration boards using an evolutionary algorithm and Pareto optimization to find the best method and combination of accuracy, methodology and number of calibration boards. The benefits of DSC are more efficient calibration board geometry selection, and better accuracy than the three Baseline visual calibration methodologies.
In the context of vegetation detection, the fusion of omnidirectional (O-D) Infrared (IR) and color vision sensors may increase the level of vegetation perception for unmanned robotic platforms. A literature search found no significant research in our area of interest. The fusion of O-D IR and O-D color vision sensors for the extraction of feature material type has not been adequately addressed. We will look at augmenting indices based spectral decomposition with IR region based spectral decomposition to address the number of false detects inherent in indices based spectral decomposition alone. Our work shows that the fusion of the Normalized Difference Vegetation Index (NDVI) from the O-D color camera fused with the IR thresholded signature region associated with the vegetation region, minimizes the number of false detects seen with NDVI alone. The contribution of this work is the demonstration of two new techniques, Thresholded Region Fusion (TRF) technique for the fusion of O-D IR and O-D Color. We also look at the Kinect vision sensor fused with the O-D IR camera. Our experimental validation demonstrates a 64% reduction in false detects in our method compared to classical indices based detection.
We finally compare our DeepFuseNet results with our previous work with Normalized Difference Vegetation index (NDVI) and IR region based spectral fusion. This current work shows that the fusion of the O-D IR and O-D visual streams utilizing our DeepFuseNet deep learning approach out performs the previous NVDI fused with far infrared region segmentation. Our experimental validation demonstrates an 92% reduction in false detects in our method compared to classical indices based detection. This work contributes a new technique for the fusion of O-D vision and O-D IR sensors using two deep CNN feature extractors feeding into a fully connected CNN Network (DeepFuseNet).
19
Bränn, Jesper. "Smartphone sensors are sufficient to measure smoothness of car driving." Thesis, KTH, Skolan för datavetenskap och kommunikation (CSC), 2017. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-208895.
Full textAbstract:
This study aims to look at whether or not it is sufficient to only use smartphone sensors to judge if someone who is driving a car is driving aggressively or smoothly. To determine this, data were first collected from the accelerometer, gyroscope, magnetometer and GPS sensors in the smartphone as well as values based on these sensors from the iOS operating system. After this the data, together with synthesized data based on the collected data, were used to train an artificial neural network.The results indicate that it is possible to give a binary judgment on aggressive or smooth driving with a 97% accuracy, with little model overfitting. The conclusion of this study is that it is sufficient to only use smartphone sensors to make a judgment on the drive.<br>Den här studien ämnar till att bedöma huruvida smartphonesensorer är tillräckliga för att avgöra om någon kör en bil aggressivt eller mjukt. För att kunna avgöra detta så samlades först data in från accelerometer, gyroskop, magnetometer och GPS-sensorerna i en smartphone, tillsammans med värden baserade på dessa data från iOS-operativ-systemet. Efter den datan var insamlad tränades ett artificiellt neuronnät med datan.Resultaten indikerar att det är möjligt att ge ett binärt utlåtande om aggressiv kontra mjuk körning med 97% säkerhet, och med liten överanpassning. Detta innebär att det är tillräckligt att enbart använda smartphonesensorer för att avgörande om körningen var mjuk eller aggressiv.
20
Decoux, Benoît. "Un modèle connexionniste de vision 3-D : imagettes rétiniennes, convergence stéréoscopique, et apprentissage auto-supervisé de la fusion." Rouen, 1995. http://www.theses.fr/1995ROUES056.
Full textAbstract:
Les études destinées à apporter l'apprentissage non-supervisé à la vision stéréoscopique artificielle s'inscrivent dans la recherche en auto-organisation des systèmes, et constituent une avancée dans la modélisation de la vision stéréo naturelle. Le principal objectif de cette thèse est de participer à cette recherche. Après quelques données sur la vision naturelle, des propriétés importantes des réseaux neuronaux sont présentées. L'accent est mis ensuite sur les propriétés d'auto-organisation de ces derniers, ainsi que sur leurs capacités sensorimotrices. Un passage en revue non-exhaustif des modèles connexionnistes de vision stéréo existant, est alors effectué. Enfin, un modèle connexionniste de vision stéréo est proposé. Ce modèle comporte deux processus complémentaires : 1) la convergence stéréo met en correspondance des régions, par minimisation d'une disparité globale. Elle simule un processus de convergence visio-motrice; 2) la fusion stéréo recherche alors la correspondance entre des éléments caractéristiques. La fusion est obtenue après une phase d'apprentissage auto-supervisé. Le type de l'apprentissage est ainsi dénommé parce que la règle utilisée est une règle d'apprentissage supervisé, mais dans laquelle l'information de supervision est extraite automatiquement des entrées visuelles par le modèle. Les scènes visuelles sont perçues au moyen d'un ensemble d'imagettes rétiniennes : il s'agit de petites images de différents champs visuels et résolutions.
21
Bekmen, Onur. "A Software Environment For Behavior-based Mobile Robot Control." Master's thesis, METU, 2007. http://etd.lib.metu.edu.tr/upload/12607965/index.pdf.
Full textAbstract:
Robotic science can be defined as a modern multi-disciplinary branch of science, which hosts many technological elements with a huge theoretic base. From electrical and electronics engineering point of view, construction of intelligent agents that produce and/or collects information by interacting the surrounding environment and that can achieve some goal via learning, is investigated in robotic science. In this scope, behavior-based robotic control has emerged in recent years, which can be defined as a hierarchically higher control mechanism over classical control theory and applications.
In this thesis, software which is capable of producing behavior-based control over mobile robots is constructed. Research encapsulates an investigation on behavior-based robotic concept by comparison of different approaches. Although there are numerous commercial and freeware software products for robotics, the number of open source, detail documented software on behavior-based control concept together with easy usage is limited. Aimed to fulfill a necessity in this field, an open source software environment is implemented in which different algorithms
and applications can be developed. In order to evaluate the effectiveness and the capabilities of the implemented software, a fully detailed simulation is conducted.
This simulation covers multi-behavior coordination concept for a differential drive mobile robot navigating in a collision free path through a target point which is detected by sensors, in an unstructured environment, that robot has no priori information about, in which static and moving obstacles exists. Coordination is accomplished by artificial neural network with back-propagation training algorithm. Behaviors are constructed using fuzzy control concept. Mobile robot has no information about sizes, number of static and/or dynamic obstacles. All the information is gathered by its simulated sensors (proximity, range, vision sensors). Yielded results are given in detail.
22
Wang, Fengzhen. "Neural networks for data fusion." Thesis, National Library of Canada = Bibliothèque nationale du Canada, 1997. http://www.collectionscanada.ca/obj/s4/f2/dsk2/ftp02/NQ30179.pdf.
Full text
23
LOCCI, AMEDEO. "Sviluppo di una piattaforma inerziale terrestre assistita da reti neurali artificiali." Doctoral thesis, Università degli Studi di Roma "Tor Vergata", 2008. http://hdl.handle.net/2108/684.
Full textAbstract:
Molte delle tecnologie impiegate nei moderni sistemi di ausilio alla navigazione terrestre, risalgono a circa un quarto di secolo. Generalmente, i principali sistemi in uso negli apparati per la navigazione terrestre, risultano essere i Sistemi di Posizionamento Globale (GPS) ed i Sistemi di Navigazione Inerziale (INS). Tali apparati tecnologici, necessitano di componenti fondamentali di qualità più o meno spinta che ne determina costi di mercato progressivamente crescenti, in grado di soddisfare le più diverse applicazioni di navigazione, passando da quella veicolare terrestre, sino a quella aerospaziale, ma includendo anche applicazione di handling o gestione di flotte veicolari. Negli ultimi anni, i sistemi di navigazione hanno visto applicazioni su larga scala, come sistemi di ausilio alla navigazione terrestre, in particolare con l’impiego del sistema GPS. Per ovviare agli inconvenienti che esso presenta, ultimamente si sono sviluppati sistemi integrati low-cost per la navigazione terrestre con applicazioni su larga scala, tra i quali si distingue, per rilevanti aspetti tecnici ed economici, il sistema integrato INS/GPS. Per ovviare alla bassa qualità dei sensori impiegati, si è dovuto ricorrere all’impiego di modelli matematici in grado di correggere le risposte dei sensori. Da qui, negli anni, uno dei maggiori strumenti di utilizzo si è dimostrato essere il filtro di Kalman nel data fusion INS/GPS; nonostante esso presenti diverse limitazioni.
La necessità della correzione dell’INS è legata sia all’ottenimento di una risposta più affidabile nei brevi percorsi rispetto al GPS, sia alla generazione di un data base ove vengano raccolte tutta una serie di informazioni tra le caratteristiche cinematiche, geometriche ed ambientali necessarie, qual’ora venisse ad essere assente o di scarsa qualità il segnale GPS. A tale scopo, si cerca di rendere intelligente il sistema di navigazione: Il filtro di Kalman ha fatto scuola negli anni passati ed ha indirizzato la ricerca verso l’impiego di sistemi intelligenti, quali ad esempio la logica fuzzy, gl algoritmi genetici o le reti neurali. Quest’ultime sono in grado di rendere i sistemi integrati INS/GPS dei sistemi intelligenti, cioè capaci di prendere delle decisioni autonome nella correzione dei dati, a fronte di capacità di apprendimento.
Ciò che qui si è perseguito è stato l’impiego delle Reti Neurali Artificiali, al fine di sopperire alla sola risposta del sistema INS qualora si verificasse l’outage del GPS. In tali situazioni, a causa delle inevitabili derive causate dagli effetti del random walk dell’INS, si necessita di eseguire una correzione dei dati inerziali, demandata alle Reti Neurali. Si è resa necessaria un’investigazione dei modelli più attinenti con tali tipologia di dati, nonché un tuning delle rete stesse. Quindi è stato sviluppato sia un sistema di memorizzazione dei dati sensibili, basato sull’aggiornamento della memoria della rete, ( i pesi ), sia un sistema di correzione dei dati inerziali. La valutazione è stata eseguita su test case evidenziando, rispetto agli impieghi di correzione con sistemi classici un netto miglioramento di performance.
Tali applicazioni, come anche evidenziato in lavori scientifici, possono considerarsi come un metodo per gli sviluppi futuri delle nuove piattaforme integrate per la navigazione terrestre. Inoltre esse sono in grado di fornire informazioni di assetto, necessarie per la guida dei sistemi autonomi. ed i loro bassi costi di mercato, ne consentono l’impiego su vastissima scala, conseguendo vantaggi in ambito di sicurezza stradale e di ricostruzione di eventi accidentali.<br>Many of the technologies used in the modern systems of aid to terrestrial navigation go back to approximately a quarter of century. Generally, the main systems in use in the equipments for terrestrial Navigation are the Global Positioning Systems (GPS) and the Inertial Navigation Systems (INS). Such technological apparatus, need fundamental components of high or quite almost quality, which cause progressively growing market costs and can satisfy the most various range of applications in the navigation field, from the vehicular terrestrial one to the aerospace application, but also including handling applications or the management of vehicular fleets. In the last few years, the navigation systems have seen applications on a large scale, as the terrestrial navigation aid systems, in particular with the use of the GPS system. In order to avoid the disadvantages that it can cause, integrated low-cost systems for terrestrial navigation have been lately developed with applications on a large scale, among which, for important technical and economic aspects, the integrated INS/GPS system. The need to avoid the low quality of the used sensors led to the application of mathematical models able to correct the sensors’ answers. From here, one of the most used instrument throughout the years has been the Kalman filter, as an optimal linear Gaussian estimator in the data fusion INS/GPS. However, as multi-sensor integration methodology, it has various limits.
The INS correction is needed either to get a more reliable answer in the short route compared the GPS-based navigation system, and to generate a data base in which a series of information are collected, among the necessary cinematic, geometric and environmental characteristics, in case the GPS signal is absent or of low quality. For that purpose, the attempt is to make the navigation system intelligent: the Kalman filter has been shown to be in the last years the reference model which has addressed the research to the use of intelligent models, such as for example the fuzzy logic, the genetic algorithms or the neural networks. The latter can make the integrated INS/GPS systems intelligent, or capable to take independent decisions in the data correction, after a learning process.
The aim which has been reached is the use of the Artificial Neural Networks (ANN), in order to compensate for the answer of the INS system, in case the GPS outage occurred. In such situations, due to the unavoidable drifts caused by the INS random walk effects, a correction of the inertial data is needed, through the Neural Networks. A research of the models more related to this kind of data has been necessary, as well as a tuning of the networks. Therefore it has been developed either a storage system for the sensitive data, based on the updating of the network memory, (the weights), and a correction system of the inertial data. The evaluation has been carried out on many test cases and it has shown a definite improvement in performance compared to the use of the correction with conventional systems.
Such applications, as also emphasized in scientific works, can be considered as a method for the future developments of the new integrated platforms for terrestrial navigation. Moreover, they can supply attitude configuration needed for the control of the autonomous systems and their low market costs allow a large scale applications, and also advantages in the road safety field and the reconstruction of accidental events.
24
He, Juan Xia. "Assessing and Improving Methods for the Effective Use of Landsat Imagery for Classification and Change Detection in Remote Canadian Regions." Thesis, Université d'Ottawa / University of Ottawa, 2016. http://hdl.handle.net/10393/34221.
Full textAbstract:
Canadian remote areas are characterized by a minimal human footprint, restricted accessibility, ubiquitous lichen/snow cover (e.g. Arctic) or continuous forest with water bodies (e.g. Sub-Arctic). Effective mapping of earth surface cover and land cover changes using free medium-resolution Landsat images in remote environments is a challenge due to the presence of spectrally mixed pixels, restricted field sampling and ground truthing, and the often relatively homogenous cover in some areas. This thesis investigates how remote sensing methods can be applied to improve the capability of Landsat images for mapping earth surface features and land cover changes in Canadian remote areas. The investigation is conducted from the following four perspectives: 1) determining the continuity of Landsat-8 images for mapping surficial materials, 2) selecting classification algorithms that best address challenges involving mixed pixels, 3) applying advanced image fusion algorithms to improve Landsat spatial resolution while maintaining spectral fidelity and reducing the effects of mixed pixels on image classification and change detection, and, 4) examining different change detection techniques, including post-classification comparisons and threshold-based methods employing PCA(Principal Components Analysis)-fused multi-temporal Landsat images to detect changes in Canadian remote areas. Three typical landscapes in Canadian remote areas are chosen in this research. The first is located in the Canadian Arctic and is characterized by ubiquitous lichen and snow cover. The second is located in the Canadian sub-Arctic and is characterized by well-defined land features such as highlands, ponds, and wetlands. The last is located in a forested highlands region with minimal built-environment features. The thesis research demonstrates that the newly available Landsat-8 images can be a major data source for mapping Canadian geological information in Arctic areas when Landsat-7 is decommissioned. In addition, advanced classification techniques such as a Support-Vector-Machine (SVM) can generate satisfactory classification results in the context of mixed training data and minimal field sampling and truthing. This thesis research provides a systematic investigation on how geostatistical image fusion can be used to improve the performance of Landsat images in identifying surface features. Finally, SVM-based post-classified multi-temporal, and threshold-based PCA-fused bi-temporal Landsat images are shown to be effective in detecting different aspects of vegetation change in a remote forested region in Ontario. This research provides a comprehensive methodology to employ free Landsat images for image classification and change detection in Canadian remote regions.
25
Grahn, Fredrik, and Kristian Nilsson. "Object Detection in Domain Specific Stereo-Analysed Satellite Images." Thesis, Linköpings universitet, Datorseende, 2019. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-159917.
Full textAbstract:
Given satellite images with accompanying pixel classifications and elevation data, we propose different solutions to object detection. The first method uses hierarchical clustering for segmentation and then employs different methods of classification. One of these classification methods used domain knowledge to classify objects while the other used Support Vector Machines. Additionally, a combination of three Support Vector Machines were used in a hierarchical structure which out-performed the regular Support Vector Machine method in most of the evaluation metrics. The second approach is more conventional with different types of Convolutional Neural Networks. A segmentation network was used as well as a few detection networks and different fusions between these. The Convolutional Neural Network approach proved to be the better of the two in terms of precision and recall but the clustering approach was not far behind. This work was done using a relatively small amount of data which potentially could have impacted the results of the Machine Learning models in a negative way.
26
Marek, Jan. "Rekonstrukce chybějících části obličeje pomocí neuronové sítě." Master's thesis, Vysoké učení technické v Brně. Fakulta informačních technologií, 2020. http://www.nusl.cz/ntk/nusl-433506.
Full textAbstract:
Cílem této práce je vytvořit neuronovou síť která bude schopna rekonstruovat obličeje z fotografií na kterých je část obličeje překrytá maskou. Jsou prezentovány koncepty využívané při vývoji konvolučních neuronových sítí a generativních kompetitivních sítí. Dále jsou popsány koncepty používané v neuronových sítích specificky pro rekonstrukci fotografií obličejů. Je představen model generativní kompetitivní sítě využívající kombinaci hrazených konvolučních vrstev a víceškálových bloků schopný realisticky doplnit oblasti obličeje zakryté maskou.
27
Ronchi, Emanuele. "Neural Networks Applications and Electronics Development for Nuclear Fusion Neutron Diagnostics." Doctoral thesis, Uppsala universitet, Institutionen för fysik och astronomi, 2009. http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-108583.
Full textAbstract:
This thesis describes the development of electronic modules for fusion neutron spectroscopy as well as several implementations of artificial neural networks (NN) for neutron diagnostics for the Joint European Torus (JET) experimental reactor in England. The electronics projects include the development of two fast light pulser modules based on Light Emitting Diodes (LEDs) for the calibration and stability monitoring of two neutron spectrometers (MPRu and TOFOR) at JET. The particular electronic implementation of the pulsers allowed for operation of the LEDs in the nanosecond time scale, which is typically not well accessible with simpler circuits. Another electronic project consisted of the the development and implementation at JET of 32 high frequency analog signal amplifiers for MPRu. The circuit board layout adopted and the choice of components permitted to achieve bandwidth above 0.5 GHz and low distortion for a wide range of input signals. The successful and continued use of all electronic modules since 2005 until the present day is an indication of their good performance and reliability. The NN applications include pulse shape discrimination (PSD), deconvolution of experimental data and tomographic reconstruction of neutron emissivity profiles for JET. The first study showed that NN can perform neutron/gamma PSD in liquid scintillators significantly better than other conventional techniques, especially for low deposited energy in the detector. The second study demonstrated that NN can be used for statistically efficient deconvolution of neutron energy spectra, with and without parametric neutron spectroscopic models, especially in the region of low counts in the data. The work on tomography provided a simple but effective parametric model for describing neutron emissivity at JET. This was then successfully implemented with NN for fast and automatic tomographic reconstruction of the JET camera data. The fast execution time of NN, i.e. usually in the microsecond time scale, makes the NN applications presented here suitable for real-time data analysis and typically orders of magnitudes faster than other commonly used codes. The results and numerical methods described in this thesis can be applied to other diagnostic instruments and are of relevance for future fusion reactors such as ITER, currently under construction in Cadarache, France.
28
Nguyen, Tien Dung. "Multimodal emotion recognition using deep learning techniques." Thesis, Queensland University of Technology, 2020. https://eprints.qut.edu.au/180753/1/Tien%20Dung_Nguyen_Thesis.pdf.
Full textAbstract:
This thesis investigates the use of deep learning techniques to address the problem of machine understanding of human affective behaviour and improve the accuracy of both unimodal and multimodal human emotion recognition. The objective was to explore how best to configure deep learning networks to capture individually and jointly, the key features contributing to human emotions from three modalities (speech, face, and bodily movements) to accurately classify the expressed human emotion. The outcome of the research should be useful for several applications including the design of social robots.
29
Rajbhandari, Samyam. "Locality Optimizations for Regular and Irregular Applications." The Ohio State University, 2016. http://rave.ohiolink.edu/etdc/view?acc_num=osu1469033289.
Full text
30
Howard, Shaun Michael. "Deep Learning for Sensor Fusion." Case Western Reserve University School of Graduate Studies / OhioLINK, 2017. http://rave.ohiolink.edu/etdc/view?acc_num=case1495751146601099.
Full text
31
Li, Bing Nan. "Wavelet neural networks : the fusion of HC and SC for computerized physiological signal interpretation." Thesis, University of Macau, 2009. http://umaclib3.umac.mo/record=b2145135.
Full text
32
Wadi, Issam M. "The assessment of blanking process characteristics using acoustic emission, sensory fusion and neural networks." Thesis, University of Strathclyde, 1999. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.366921.
Full text
33
Fay, Rebecca. "Feature selection and information fusion in hierarchical neural networks for iterative 3D-object recognition." [S.l. : s.n.], 2007. http://nbn-resolving.de/urn:nbn:de:bsz:289-vts-60447.
Full text
34
Manne, Mihira. "MACHINE VISION FOR AUTOMATICVISUAL INSPECTION OF WOODENRAILWAY SLEEPERS USING UNSUPERVISED NEURAL NETWORKS." Thesis, Högskolan Dalarna, Datateknik, 2009. http://urn.kb.se/resolve?urn=urn:nbn:se:du-3977.
Full textAbstract:
The motivation for this thesis work is the need for improving reliability of equipment and quality of service to railway passengers as well as a requirement for cost-effective and efficient condition maintenance management for rail transportation. This thesis work develops a fusion of various machine vision analysis methods to achieve high performance in automation of wooden rail track inspection.The condition monitoring in rail transport is done manually by a human operator where people rely on inference systems and assumptions to develop conclusions. The use of conditional monitoring allows maintenance to be scheduled, or other actions to be taken to avoid the consequences of failure, before the failure occurs. Manual or automated condition monitoring of materials in fields of public transportation like railway, aerial navigation, traffic safety, etc, where safety is of prior importance needs non-destructive testing (NDT).In general, wooden railway sleeper inspection is done manually by a human operator, by moving along the rail sleeper and gathering information by visual and sound analysis for examining the presence of cracks. Human inspectors working on lines visually inspect wooden rails to judge the quality of rail sleeper. In this project work the machine vision system is developed based on the manual visual analysis system, which uses digital cameras and image processing software to perform similar manual inspections. As the manual inspection requires much effort and is expected to be error prone sometimes and also appears difficult to discriminate even for a human operator by the frequent changes in inspected material. The machine vision system developed classifies the condition of material by examining individual pixels of images, processing them and attempting to develop conclusions with the assistance of knowledge bases and features.A pattern recognition approach is developed based on the methodological knowledge from manual procedure. The pattern recognition approach for this thesis work was developed and achieved by a non destructive testing method to identify the flaws in manually done condition monitoring of sleepers.In this method, a test vehicle is designed to capture sleeper images similar to visual inspection by human operator and the raw data for pattern recognition approach is provided from the captured images of the wooden sleepers. The data from the NDT method were further processed and appropriate features were extracted.The collection of data by the NDT method is to achieve high accuracy in reliable classification results. A key idea is to use the non supervised classifier based on the features extracted from the method to discriminate the condition of wooden sleepers in to either good or bad. Self organising map is used as classifier for the wooden sleeper classification.In order to achieve greater integration, the data collected by the machine vision system was made to interface with one another by a strategy called fusion. Data fusion was looked in at two different levels namely sensor-level fusion, feature- level fusion. As the goal was to reduce the accuracy of the human error on the rail sleeper classification as good or bad the results obtained by the feature-level fusion compared to that of the results of actual classification were satisfactory.
35
Oliveira, e. Cruz Rafael Menelau. "Methods for dynamic selection and fusion of ensemble of classifiers." Universidade Federal de Pernambuco, 2011. https://repositorio.ufpe.br/handle/123456789/2436.
Full textAbstract:
Made available in DSpace on 2014-06-12T15:58:13Z (GMT). No. of bitstreams: 2
arquivo3310_1.pdf: 8155353 bytes, checksum: 2f4dcd5adb2b0b1a23c40bf343b36b34 (MD5)
license.txt: 1748 bytes, checksum: 8a4605be74aa9ea9d79846c1fba20a33 (MD5)
Previous issue date: 2011<br>Faculdade de Amparo à Ciência e Tecnologia do Estado de Pernambuco<br>Ensemble of Classifiers (EoC) é uma nova alternative para alcançar altas taxas de reconhecimento
em sistemas de reconhecimento de padrões. O uso de ensemble é motivado pelo fato
de que classificadores diferentes conseguem reconhecer padrões diferentes, portanto, eles são
complementares. Neste trabalho, as metodologias de EoC são exploradas com o intuito de
melhorar a taxa de reconhecimento em diferentes problemas. Primeiramente o problema do
reconhecimento de caracteres é abordado. Este trabalho propõe uma nova metodologia que
utiliza múltiplas técnicas de extração de características, cada uma utilizando uma abordagem
diferente (bordas, gradiente, projeções). Cada técnica é vista como um sub-problema possuindo
seu próprio classificador. As saídas deste classificador são utilizadas como entrada para
um novo classificador que é treinado para fazer a combinação (fusão) dos resultados. Experimentos
realizados demonstram que a proposta apresentou o melhor resultado na literatura pra
problemas tanto de reconhecimento de dígitos como para o reconhecimento de letras.
A segunda parte da dissertação trata da seleção dinâmica de classificadores (DCS). Esta
estratégia é motivada pelo fato que nem todo classificador pertencente ao ensemble é um especialista
para todo padrão de teste. A seleção dinâmica tenta selecionar apenas os classificadores
que possuem melhor desempenho em uma dada região próxima ao padrão de entrada para classificar
o padrão de entrada. É feito um estudo sobre o comportamento das técnicas de DCS
demonstrando que elas são limitadas pela qualidade da região em volta do padrão de entrada.
Baseada nesta análise, duas técnicas para seleção dinâmica de classificadores são propostas.
A primeira utiliza filtros para redução de ruídos próximos do padrão de testes. A segunda é
uma nova proposta que visa extrair diferentes tipos de informação, a partir do comportamento
dos classificadores, e utiliza estas informações para decidir se um classificador deve ser selecionado
ou não. Experimentos conduzidos em diversos problemas de reconhecimento de
padrões demonstram que as técnicas propostas apresentam um aumento de performance significante
36
Sundelius, Carl. "Deep Fusion of Imaging Modalities for Semantic Segmentation of Satellite Imagery." Thesis, Linköpings universitet, Datorseende, 2018. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-145193.
Full textAbstract:
In this report I summarize my master’s thesis work, in which I have investigated different approaches for fusing imaging modalities for semantic segmentation with deep convolutional networks. State-of-the-art methods for semantic segmentation of RGB-images use pre-trained models, which are fine-tuned to learn task-specific deep features. However, the use of pre-trained model weights constrains the model input to images with three channels (e.g. RGB-images). In some applications, e.g. classification of satellite imagery, there are other imaging modalities that can complement the information from the RGB modality and, thus, improve the performance of the classification. In this thesis, semantic segmentation methods designed for RGB images are extended to handle multiple imaging modalities, without compromising on the benefits, that pre-training on RGB datasets offers. In the experiments of this thesis, RGB images from satellites have been fused with normalised difference vegetation index (NDVI) and a digital surface model (DSM). The evaluation shows that the modality fusion can significantly improve the performance of semantic segmentation networks in comparison with a corresponding network with only RGB input. However, the different investigated approaches to fuse the modalities proved to achieve similar performance. The conclusion of the experiments is, that the fusion of imaging modalities is necessary, but the method of fusion has shown to be of less importance.
37
Papazoglou, Dimitri Pierre. "Additively Manufactured Lattices for Orthopedic Implants and Process Monitoring of Laser-Powder Bed Fusion Using Neural Networks." University of Dayton / OhioLINK, 2019. http://rave.ohiolink.edu/etdc/view?acc_num=dayton1556894605220274.
Full text
38
MAGGIOLO, LUCA. "Deep Learning and Advanced Statistical Methods for Domain Adaptation and Classification of Remote Sensing Images". Doctoral thesis, Università degli studi di Genova, 2022. http://hdl.handle.net/11567/1070050.
Full textAbstract:
In the recent years, remote sensing has faced a huge evolution. The constantly growing availability of remote sensing data has opened up new opportunities and laid the foundations for many new challenges. The continuous space missions and new constellations of satellites allow in fact more and more frequent acquisitions, at increasingly higher spatial resolutions, and at an almost total coverage of the globe. The availability of such an huge amount data has highlighted the need for automatic techniques capable of processing the data and exploiting all the available information. Meanwhile, the almost unlimited potential of machine learning has changed the world we live in. Artificial neural Networks have break trough everyday life, with applications that include computer vision, speech processing, autonomous driving but which are also the basis of commonly used tools such as online search engines. However, the vast majority of such models are of the supervised type and therefore their applicability rely on the availability of an enormous quantity of labeled data available to train the models themselves. Unfortunately, this is not the case with remote sensing, in which the enormous amounts of data are opposed to the almost total absence of ground truth. The purpose of this thesis is to find the way to exploit the most recent deep learning techniques, defining a common thread between two worlds, those of remote sensing and deep learning, which is often missing.
In particular, this thesis proposes three novel contributions which face current issues in remote sensing. The first one is related to multisensor image registration and combines generative adversarial networks and non-linear optimization of crosscorrelation-like functionals to deal with the complexity of the setting. The proposed method was proved able to outperform state of the art approaches.
The second novel contribution faces one of the main issues in deep learning for remote sensing: the scarcity of ground truth data for semantic segmentation. The proposed solution combines convolutional neural networks and probabilistic graphical models, two very active areas in machine learning for remote sensing, and approximate a fully connected conditional random field. The proposed method is capable of filling part of the gap which separate a densely trained model from a weakly trained one. Then, the third approach is aimed at the classification of high resolution satellite images for climate change purposes. It consist of a specific formulation of an energy minimization which allows to fuse multisensor information and the application a markov random field in a fast and efficient way for global scale applications.
The results obtained in this thesis shows how deep learning methods based on artificial neural networks can be combined with statistical analysis to overcome their limitations, going beyond the classic benchmark environments and addressing practical, real and large-scale application cases.
39
Lundberg, Gustav. "Automatic map generation from nation-wide data sources using deep learning." Thesis, Linköpings universitet, Statistik och maskininlärning, 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-170759.
Full textAbstract:
The last decade has seen great advances within the field of artificial intelligence. One of the most noteworthy areas is that of deep learning, which is nowadays used in everything from self driving cars to automated cancer screening. During the same time, the amount of spatial data encompassing not only two but three dimensions has also grown and whole cities and countries are being scanned. Combining these two technological advances enables the creation of detailed maps with a multitude of applications, civilian as well as military.This thesis aims at combining two data sources covering most of Sweden; laser data from LiDAR scans and surface model from aerial images, with deep learning to create maps of the terrain. The target is to learn a simplified version of orienteering maps as these are created with high precision by experienced map makers, and are a representation of how easy or hard it would be to traverse a given area on foot. The performance on different types of terrain are measured and it is found that open land and larger bodies of water is identified at a high rate, while trails are hard to recognize.It is further researched how the different densities found in the source data affect the performance of the models, and found that some terrain types, trails for instance, benefit from higher density data, Other features of the terrain, like roads and buildings are predicted with higher accuracy by lower density data.Finally, the certainty of the predictions is discussed and visualised by measuring the average entropy of predictions in an area. These visualisations highlight that although the predictions are far from perfect, the models are more certain about their predictions when they are correct than when they are not.
40
TOOSI, AMIRHOSEIN. "Feature Fusion for Fingerprint Liveness Detection." Doctoral thesis, Politecnico di Torino, 2018. http://hdl.handle.net/11583/2711594.
Full textAbstract:
For decades, fingerprints have been the most widely used biometric trait in identity
recognition systems, thanks to their natural uniqueness, even in rare cases such as
identical twins. Recently, we witnessed a growth in the use of fingerprint-based
recognition systems in a large variety of devices and applications. This, as a consequence,
increased the benefits for offenders capable of attacking these systems. One
of the main issues with the current fingerprint authentication systems is that, even
though they are quite accurate in terms of identity verification, they can be easily
spoofed by presenting to the input sensor an artificial replica of the fingertip skin’s
ridge-valley patterns.
Due to the criticality of this threat, it is crucial to develop countermeasure
methods capable of facing and preventing these kind of attacks. The most effective
counter–spoofing methods are those trying to distinguish between a "live" and a
"fake" fingerprint before it is actually submitted to the recognition system. According
to the technology used, these methods are mainly divided into hardware and software-based
systems. Hardware-based methods rely on extra sensors to gain more pieces
of information regarding the vitality of the fingerprint owner. On the contrary,
software-based methods merely rely on analyzing the fingerprint images acquired
by the scanner. Software-based methods can then be further divided into dynamic,
aimed at analyzing sequences of images to capture those vital signs typical of a real
fingerprint, and static, which process a single fingerprint impression. Among these
different approaches, static software-based methods come with three main benefits.
First, they are cheaper, since they do not require the deployment of any additional
sensor to perform liveness detection. Second, they are faster since the information
they require is extracted from the same input image acquired for the identification
task. Third, they are potentially capable of tackling novel forms of attack through an
update of the software. The interest in this type of counter–spoofing methods is at the basis of this
dissertation, which addresses the fingerprint liveness detection under a peculiar
perspective, which stems from the following consideration. Generally speaking, this
problem has been tackled in the literature with many different approaches. Most of
them are based on first identifying the most suitable image features for the problem
in analysis and, then, into developing some classification system based on them. In
particular, most of the published methods rely on a single type of feature to perform
this task. Each of this individual features can be more or less discriminative and often
highlights some peculiar characteristics of the data in analysis, often complementary
with that of other feature. Thus, one possible idea to improve the classification
accuracy is to find effective ways to combine them, in order to mutually exploit their
individual strengths and soften, at the same time, their weakness. However, such a
"multi-view" approach has been relatively overlooked in the literature.
Based on the latter observation, the first part of this work attempts to investigate
proper feature fusion methods capable of improving the generalization and robustness
of fingerprint liveness detection systems and enhance their classification strength.
Then, in the second part, it approaches the feature fusion method in a different way,
that is by first dividing the fingerprint image into smaller parts, then extracting an
evidence about the liveness of each of these patches and, finally, combining all these
pieces of information in order to take the final classification decision.
The different approaches have been thoroughly analyzed and assessed by comparing
their results (on a large number of datasets and using the same experimental
protocol) with that of other works in the literature. The experimental results discussed
in this dissertation show that the proposed approaches are capable of obtaining
state–of–the–art results, thus demonstrating their effectiveness.
41
Sala, Davi Alberto. "Controle de posição com múltiplos sensores em um robô colaborativo utilizando liquid state machines." reponame:Biblioteca Digital de Teses e Dissertações da UFRGS, 2017. http://hdl.handle.net/10183/163752.
Full textAbstract:
A ideia de usar redes neurais biologicamente inspiradas na computação tem sido amplamente utilizada nas últimas décadas. O fato essencial neste paradigma é que um neurônio pode integrar e processar informações, e esta informação pode ser revelada por sua atividade de pulsos. Ao descrever a dinâmica de um único neurônio usando um modelo matemático, uma rede pode ser implementada utilizando um conjunto desses neurônios, onde a atividade pulsante de cada neurônio irá conter contribuições, ou informações, da atividade pulsante da rede em que está inserido. Neste trabalho é apresentado um controlador de posição no eixo Z utilizando fusão de sensores baseado no paradigma de Redes Neurais Recorrentes. O sistema proposto utiliza uma Máquina de Estado Líquido (LSM) para controlar o robô colaborativo BAXTER. O framework foi projetado para trabalhar em paralelo com as LSMs que executam trajetórias em formas fechadas de duas dimensões, com o objetivo de manter uma caneta de feltro em contato com a superfície de desenho, dados de sensores de força e distância são alimentados ao controlador. O sistema foi treinado utilizando dados de um controlador Proporcional Integral Derivativo (PID), fundindo dados de ambos sensores. Resultados mostram que a LSM foi capaz de aprender o comportamento do controlador PID em diferentes situações.<br>The idea of employing biologically inspired neural networks to perform computation has been widely used over the last decades. The essential fact in this paradigm is that a neuron can integrate and process information, and this information can be revealed by its spiking activity. By describing the dynamics of a single neuron using a mathematical model, a network in which the spiking activity of every single neuron will get contributions, or information, from the spiking activity of the embedded network. A positioning controller based on Spiking Neural Networks for sensor fusion suitable to run on a neuromorphic computer is presented in this work. The proposed framework uses the paradigm of reservoir computing to control the collaborative robot BAXTER. The system was designed to work in parallel with Liquid State Machines that performs trajectories in 2D closed shapes. In order to keep a felt pen touching a drawing surface, data from sensors of force and distance are fed to the controller. The system was trained using data from a Proportional Integral Derivative controller, merging the data from both sensors. The results show that the LSM can learn the behavior of a PID controller on di erent situations.
42
Borges, Gabriel de Morais. "Estudo e aplicação de diferentes métodos para redução de falsos alarmes no monitoramento de frequência cardíaca." reponame:Biblioteca Digital de Teses e Dissertações da UFRGS, 2015. http://hdl.handle.net/10183/141939.
Full textAbstract:
O monitoramento automático de pacientes é um recurso essencial em hospitais para o bom gerenciamento de cuidados médicos. Enquanto que alarmes devido a condições fisiológicas anormais são importantes para o rápido tratamento, estes também podem ser uma fonte de ruídos desnecessários devido a falsos alarmes causados por interferência eletromagnética ou movimentação de sensores. Uma fonte significativa de falsos alarmes é relacionada com a frequência cardíaca, o qual é disparado quando o ritmo cardíaco do paciente está muito rápido ou muito lento. Neste trabalho, a fusão de diferentes sensores fisiológicos é explorada para fazer uma estimativa robusta de frequência cardíaca. Um conjunto de algoritmos utilizando índice de variabilidade cardíaca, inferência bayesiana, redes neurais, lógica fuzzy e votador majoritário são propostos para fundir a informação do eletrocardiograma, pressão sanguínea e fotopletismograma. Três informações básicas são extraídas de cada sensor: variabilidade cardíaca, a diferença de frequência cardíaca entre os sensores e a análise espectral. Estas informações são usadas como entradas para os algoritmos. Quarenta gravações selecionadas do banco de dados MIMIC são usadas para validar o sistema. Finalmente, a frequência cardíaca calculada é comparada com as anotações do banco de dados. Resultados mostram que a fusão utilizando redes neurais apresenta a melhor redução de falsos alarmes de 89.33%, enquanto que a técnica bayesiana apresenta uma redução de 83.76%. A lógica fuzzy mostrou uma redução de 77.96%, o votador majoritário 61.25% e o índice de variabilidade cardíaca de 65.43%. Portanto, os algoritmos propostos mostraram bom desempenho e podem ser muito úteis em monitores de sinais vitais modernos.<br>Automatic patient monitoring is an essential resource in hospitals for good health care management. While alarms due to abnormal physiological conditions are important to deliver fast treatment, it can be also a source of unnecessary noise due to false alarms caused by electromagnetic interference or motion artifacts. One significant source of false alarms are those related to heart rate, which is triggered when the heart rhythm of the patient is too fast or too slow. In this work, the fusion of different physiological sensors is explored in order to create a robust heart rate estimation. A set of algorithms using heart rate variability index, bayesian inference, neural networks, fuzzy logic and majority voting is proposed to fuse information from electrocardiogram, arterial blood pressure and photoplethysmogram. Three basic informations are extracted from each source, namely, heart rate variability, the heart rate difference between sensors and the spectral analysis. These informations are used as inputs to the algorithms. Forty selected recordings from MIMIC database was used to validate the system. Finally, the calculated heart rate is compared with the database annotation. Results show that neural networks fusion presents the best false alarms reduction of 89.33%, while the bayesian technique presents an error reduction of 83.76%. Fuzzy logic showed an error reduction of 77.96%, majority voting 61.25% and the heart rate variability index 65.43%. Therefore, the proposed algorithms showed good performance and can be very useful for modern bedside monitors.
43
Duchene, Pierre. "Caractérisation non destructive des matériaux composites en fatigue : diagnostic de l’état de santé et pronostic de la durée de vie résiduelle par réseaux de neurones." Thesis, Ecole nationale supérieure Mines-Télécom Lille Douai, 2018. http://www.theses.fr/2018MTLD0008.
Full textAbstract:
Ce travail de recherche consiste en la proposition d’une nouvelle approche de caractérisation non destructive de l’endommagement des matériaux composites (carbone/époxy) sollicités en fatigue par des essais d’auto-échauffement (blocs de chargements croissants). Cette approche est basée sur l’utilisation de plusieurs techniques non destructives appliquées in-situ, en temps réel ou différé, dont l’analyse est, soit redondante soit complémentaire. Au total, six techniques ont été utilisées (émission acoustique, thermographie infrarouge, corrélation d’images numériques, acousto-ultrasons, ultrasons C-scan et ondes de Lamb) et leurs résultats post-traités puis fusionnés à l’aide d’algorithmes basés sur les réseaux de neurones. Les résultats obtenus ont permis d’évaluer et de localiser l’endommagement du matériau et d’estimer sa durée de vie résiduelle. Ce faisant, plusieurs avancés scientifiques ont été obtenus en réalisant, par exemple, une localisation 2D des évènements acoustiques à l’aide seulement de deux capteurs avec une précision millimétrique, ou encore le développement d’une nouvelle technique imagée d’acousto-ultrasons permettant un contrôle hors contraintes de l’état d’endommagement du matériau, …et enfin, le pronostic de la durée de vie résiduelle du matériau basé sur une fusion de données par réseaux de neurones<br>This research work consists in a new approach for non-destructive characterisation of damage in composite materials (carbon/epoxy) subjected to fatigue during self-heating tests (increasing load blocks). This approach is based on the use of several non-destructive techniques applied in-situ, in real time or delayed, whose analysis is either redundant or complementary. Six techniques were used (acoustic emission, infrared thermography, digital image correlation, acousto-ultrasound, C-scan ultrasound and lamb waves) and their post-processed results were merged using algorithms based on neural networks. The results obtained made it possible to assess and locate the damage of the material and to estimate its residual life. In doing so, several scientific advances have been obtained by, for example, carrying out a 2D localization of acoustic events using only two sensors with millimetric precision, or the development of a new pictorial acousto-ultrasonic technique allowing an control of the state of material damage at free stress conditions, ... and finally, the prognosis of the residual lifetime of the material based on a data fusion by neural networks
44
Jaureguiberry, Xabier. "Fusion pour la séparation de sources audio." Thesis, Paris, ENST, 2015. http://www.theses.fr/2015ENST0030/document.
Full textAbstract:
La séparation aveugle de sources audio dans le cas sous-déterminé est un problème mathématique complexe dont il est aujourd'hui possible d'obtenir une solution satisfaisante, à condition de sélectionner la méthode la plus adaptée au problème posé et de savoir paramétrer celle-ci soigneusement. Afin d'automatiser cette étape de sélection déterminante, nous proposons dans cette thèse de recourir au principe de fusion. L'idée est simple : il s'agit, pour un problème donné, de sélectionner plusieurs méthodes de résolution plutôt qu'une seule et de les combiner afin d'en améliorer la solution. Pour cela, nous introduisons un cadre général de fusion qui consiste à formuler l'estimée d'une source comme la combinaison de plusieurs estimées de cette même source données par différents algorithmes de séparation, chaque estimée étant pondérée par un coefficient de fusion. Ces coefficients peuvent notamment être appris sur un ensemble d'apprentissage représentatif du problème posé par minimisation d'une fonction de coût liée à l'objectif de séparation. Pour aller plus loin, nous proposons également deux approches permettant d'adapter les coefficients de fusion au signal à séparer. La première formule la fusion dans un cadre bayésien, à la manière du moyennage bayésien de modèles. La deuxième exploite les réseaux de neurones profonds afin de déterminer des coefficients de fusion variant en temps. Toutes ces approches ont été évaluées sur deux corpus distincts : l'un dédié au rehaussement de la parole, l'autre dédié à l'extraction de voix chantée. Quelle que soit l'approche considérée, nos résultats montrent l'intérêt systématique de la fusion par rapport à la simple sélection, la fusion adaptative par réseau de neurones se révélant être la plus performante<br>Underdetermined blind source separation is a complex mathematical problem that can be satisfyingly resolved for some practical applications, providing that the right separation method has been selected and carefully tuned. In order to automate this selection process, we propose in this thesis to resort to the principle of fusion which has been widely used in the related field of classification yet is still marginally exploited in source separation. Fusion consists in combining several methods to solve a given problem instead of selecting a unique one. To do so, we introduce a general fusion framework in which a source estimate is expressed as a linear combination of estimates of this same source given by different separation algorithms, each source estimate being weighted by a fusion coefficient. For a given task, fusion coefficients can then be learned on a representative training dataset by minimizing a cost function related to the separation objective. To go further, we also propose two ways to adapt the fusion coefficients to the mixture to be separated. The first one expresses the fusion of several non-negative matrix factorization (NMF) models in a Bayesian fashion similar to Bayesian model averaging. The second one aims at learning time-varying fusion coefficients thanks to deep neural networks. All proposed methods have been evaluated on two distinct corpora. The first one is dedicated to speech enhancement while the other deals with singing voice extraction. Experimental results show that fusion always outperform simple selection in all considered cases, best results being obtained by adaptive time-varying fusion with neural networks
45
Prokopetc, Kristina. "Precise Mapping for Retinal Photocoagulation in SLIM (Slit-Lamp Image Mosaicing)." Thesis, Université Clermont Auvergne (2017-2020), 2017. http://www.theses.fr/2017CLFAC093/document.
Full textAbstract:
Cette thèse est issue d’un accord CIFRE entre le groupe de recherche EnCoV de l’Université Clermont Auvergne et la société Quantel Medical (www.quantel-medical.fr). Quantel Medical est une entreprise spécialisée dans le développement innovant des ultrasons et des produits laser en ophtalmologie. Cette thèse présente un travail de recherche visant à l’application du diagnostic assisté par ordinateur et du traitement des maladies de la rétine avec une utilisation du prototype industriel TrackScan développé par Quantel Medical. Plus précisément, elle contribue au problème du mosaicing précis de l’image de la lampe à fente (SLIM) et du recalage automatique et multimodal en utilisant les images SLIM avec l’angiographie par fluorescence (FA) pour aider à la photo coagulation pan-rétienne naviguée. Nous abordons trois problèmes différents.Le premier problème est lié à l’accumulation des erreurs du recalage en SLIM., il dérive de la mosaïque. Une approche commune pour obtenir la mosaïque consiste à calculer des transformations uniquement entre les images temporellement consécutives dans une séquence, puis à les combiner pour obtenir la transformation entre les vues non consécutives temporellement. Les nombreux algorithmes existants suivent cette approche. Malgré le faible coût de calcul et la simplicité de cette méthode, en raison de sa nature de ‘chaînage’, les erreurs d’alignement s’accumulent, ce qui entraîne une dérive des images dans la mosaïque. Nous proposons donc d’utilise les récents progrès réalisés dans les méthodes d’ajustement de faisceau et de présenter un cadre de réduction de la dérive spécialement conçu pour SLIM. Nous présentons aussi une nouvelle procédure de raffinement local.Deuxièmement, nous abordons le problème induit par divers types d’artefacts communs á l’imagerie SLIM. Ceus-sont liés à la lumière utilisée, qui dégrade considérablement la qualité géométrique et photométrique de la mosaïque. Les solutions existantes permettent de faire face aux blouissements forts qui corrompent entièrement le rendu de la rétine dans l’image tout en laissant de côté la correction des reflets spéculaires semi-transparents et reflets des lentilles. Cela introduit des images fantômes et des pertes d’information. En outre, les méthodes génériques ne produisent pas de résultats satisfaisants dans SLIM. Par conséquent, nous proposons une meilleure alternative en concevant une méthode basée sur une technique rapide en utilisant une seule image pour éliminer les éblouissements et la notion de feux spéculaires semi-transparents en utilisant les indications de mouvement pour la correction intelligente de reflet de lentille.Finalement, nous résolvons le problème du recalage multimodal automatique avec SLIM. Il existe une quantité importante de travaux sur le recalage multimodal de diverses modalités d’image rétinienne. Cependant, la majorité des méthodes existantes nécessitent une détection de points clés dans les deux modalités d’image, ce qui est une tâche très difficile. Dans le cas de SLIM et FA ils ne tiennent pas compte du recalage précis dans la zone maculaire - le repère prioritaire. En outre, personne n’a développé une solution entièrement automatique pour SLIM et FA. Dans cette thèse, nous proposons la première méthode capable de recolu ces deux modalités sans une saisie manuelle, en détectant les repères anatomiques uniquement sur une seule image pour assurer un recalage précis dans la zone maculaire. (...)<br>This thesis arises from an agreement Convention Industrielle de Formation par la REcherche (CIFRE) between the Endoscopy and Computer Vision (EnCoV) research group at Université Clermont Auvergne and the company Quantel Medical (www.quantel-medical.fr), which specializes in the development of innovative ultrasound and laser products in ophthalmology. It presents a research work directed at the application of computer-aided diagnosis and treatment of retinal diseases with a use of the TrackScan industrial prototype developed at Quantel Medical. More specifically, it contributes to the problem of precise Slit-Lamp Image Mosaicing (SLIM) and automatic multi-modal registration of SLIM with Fluorescein Angiography (FA) to assist navigated pan-retinal photocoagulation. We address three different problems.The first is a problem of accumulated registration errors in SLIM, namely the mosaicing drift.A common approach to image mosaicking is to compute transformations only between temporally consecutive images in a sequence and then to combine them to obtain the transformation between non-temporally consecutive views. Many existing algorithms follow this approach. Despite the low computational cost and the simplicity of such methods, due to its ‘chaining’ nature, alignment errors tend to accumulate, causing images to drift in the mosaic. We propose to use recent advances in key-frame Bundle Adjustment methods and present a drift reduction framework that is specifically designed for SLIM. We also introduce a new local refinement procedure.Secondly, we tackle the problem of various types of light-related imaging artifacts common in SLIM, which significantly degrade the geometric and photometric quality of the mosaic. Existing solutions manage to deal with strong glares which corrupt the retinal content entirely while leaving aside the correction of semi-transparent specular highlights and lens flare. This introduces ghosting and information loss. Moreover, related generic methods do not produce satisfactory results in SLIM. Therefore, we propose a better alternative by designing a method based on a fast single-image technique to remove glares and the notion of the type of semi-transparent specular highlights and motion cues for intelligent correction of lens flare.Finally, we solve the problem of automatic multi-modal registration of FA and SLIM. There exist a number of related works on multi-modal registration of various retinal image modalities. However, the majority of existing methods require a detection of feature points in both image modalities. This is a very difficult task for SLIM and FA. These methods do not account for the accurate registration in macula area - the priority landmark. Moreover, none has developed a fully automatic solution for SLIM and FA. In this thesis, we propose the first method that is able to register these two modalities without manual input by detecting retinal features only on one image and ensures an accurate registration in the macula area.The description of the extensive experiments that were used to demonstrate the effectiveness of each of the proposed methods is also provided. Our results show that (i) using our new local refinement procedure for drift reduction significantly ameliorates the to drift reduction allowing us to achieve an improvement in precision over the current solution employed in the TrackScan; (ii) the proposed methodology for correction of light-related artifacts exhibits a good efficiency, significantly outperforming related works in SLIM; and (iii) despite our solution for multi-modal registration builds on existing methods, with the various specific modifications made, it is fully automatic, effective and improves the baseline registration method currently used on the TrackScan
46
Jaureguiberry, Xabier. "Fusion pour la séparation de sources audio." Electronic Thesis or Diss., Paris, ENST, 2015. http://www.theses.fr/2015ENST0030.
Full textAbstract:
La séparation aveugle de sources audio dans le cas sous-déterminé est un problème mathématique complexe dont il est aujourd'hui possible d'obtenir une solution satisfaisante, à condition de sélectionner la méthode la plus adaptée au problème posé et de savoir paramétrer celle-ci soigneusement. Afin d'automatiser cette étape de sélection déterminante, nous proposons dans cette thèse de recourir au principe de fusion. L'idée est simple : il s'agit, pour un problème donné, de sélectionner plusieurs méthodes de résolution plutôt qu'une seule et de les combiner afin d'en améliorer la solution. Pour cela, nous introduisons un cadre général de fusion qui consiste à formuler l'estimée d'une source comme la combinaison de plusieurs estimées de cette même source données par différents algorithmes de séparation, chaque estimée étant pondérée par un coefficient de fusion. Ces coefficients peuvent notamment être appris sur un ensemble d'apprentissage représentatif du problème posé par minimisation d'une fonction de coût liée à l'objectif de séparation. Pour aller plus loin, nous proposons également deux approches permettant d'adapter les coefficients de fusion au signal à séparer. La première formule la fusion dans un cadre bayésien, à la manière du moyennage bayésien de modèles. La deuxième exploite les réseaux de neurones profonds afin de déterminer des coefficients de fusion variant en temps. Toutes ces approches ont été évaluées sur deux corpus distincts : l'un dédié au rehaussement de la parole, l'autre dédié à l'extraction de voix chantée. Quelle que soit l'approche considérée, nos résultats montrent l'intérêt systématique de la fusion par rapport à la simple sélection, la fusion adaptative par réseau de neurones se révélant être la plus performante<br>Underdetermined blind source separation is a complex mathematical problem that can be satisfyingly resolved for some practical applications, providing that the right separation method has been selected and carefully tuned. In order to automate this selection process, we propose in this thesis to resort to the principle of fusion which has been widely used in the related field of classification yet is still marginally exploited in source separation. Fusion consists in combining several methods to solve a given problem instead of selecting a unique one. To do so, we introduce a general fusion framework in which a source estimate is expressed as a linear combination of estimates of this same source given by different separation algorithms, each source estimate being weighted by a fusion coefficient. For a given task, fusion coefficients can then be learned on a representative training dataset by minimizing a cost function related to the separation objective. To go further, we also propose two ways to adapt the fusion coefficients to the mixture to be separated. The first one expresses the fusion of several non-negative matrix factorization (NMF) models in a Bayesian fashion similar to Bayesian model averaging. The second one aims at learning time-varying fusion coefficients thanks to deep neural networks. All proposed methods have been evaluated on two distinct corpora. The first one is dedicated to speech enhancement while the other deals with singing voice extraction. Experimental results show that fusion always outperform simple selection in all considered cases, best results being obtained by adaptive time-varying fusion with neural networks
47
Ghosh, Binayak. "Opto-Acoustic Slopping Prediction System in Basic Oxygen Furnace Converters." Thesis, KTH, Skolan för informations- och kommunikationsteknik (ICT), 2017. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-219614.
Full textAbstract:
Today, everyday objects are becoming more and more intelligent and some-times even have self-learning capabilities. These self-learning capacities in particular also act as catalysts for new developments in the steel industry.Technical developments that enhance the sustainability and productivity of steel production are very much in demand in the long-term. The methods of Industry 4.0 can support the steel production process in a way that enables steel to be produced in a more cost-effective and environmentally friendly manner. This thesis describes the development of an opto-acoustic system for the early detection of slag slopping in the BOF (Basic Oxygen Furnace) converter process. The prototype has been installed in Salzgitter Stahlwerks, a German steel plant for initial testing. It consists of an image monitoring camera at the converter mouth, a sound measurement system and an oscillation measurement device installed at the blowing lance. The camera signals are processed by a special image processing software. These signals are used to rate the amount of spilled slag and for a better interpretation of both the sound data and the oscillation data. A certain aspect of the opto-acoustic system for slopping detection is that all signals, i.e. optic, acoustic and vibratory, are affected by process-related parameters which are not always relevant for the slopping event. These uncertainties affect the prediction of the slopping phenomena and ultimately the reliability of the entire slopping system. Machine Learning algorithms have been been applied to predict the Slopping phenomenon based on the data from the sensors as well as the other process parameters.<br>Idag blir vardagliga föremål mer och mer intelligenta och ibland har de självlärande möjligheter. Dessa självlärande förmågor fungerar också som katalysatorer för den nya utvecklingen inom stålindustrin. Teknisk utveckling som stärker hållbarheten och produktiviteten i stålproduktionen är mycket efterfrågad på lång sikt. Metoderna för Industry 4.0 kan stödja stålproduktionsprocessen på ett sätt som gör att stål kan produceras på ett mer kostnadseffektivt och miljövänligt sätt. Denna avhandling beskriver utvecklingen av ett opto-akustiskt system för tidig detektering av slaggsslipning i konverteringsprocessen BOF (Basic Oxygen Furnace). Prototypen har installerats i Salzgitter Stahlwerks, en tysk stålverk för första provning. Den består av en bildövervakningskamera på omvandlarens mun, ett ljudmätningssystem och en oscillationsmätningsenhet som installeras vid blåsans. Kamerans signaler behandlas av en speciell bildbehandlingsprogram. Dessa signaler används för att bestämma mängden spilld slagg och för bättre tolkning av både ljuddata och oscillationsdata. En viss aspekt av det optoakustiska systemet för släckningsdetektering är att alla signaler, dvs optiska, akustiska och vibrerande, påverkas av processrelaterade parametrar som inte alltid är relevanta för slöjningsevenemanget. Dessa osäkerheter påverkar förutsägelsen av slopfenomenerna och i slutändan tillförlitligheten för hela slöjningssystemet. Maskininlärningsalgoritmer har tillämpats för att förutsäga Slopping-fenomenet baserat på data från sensorerna liksom de andra processparametrarna.
48
Bertrand, Sarah. "Analyse d'images pour l'identification multi-organes d'espèces végétales." Thesis, Lyon, 2018. http://www.theses.fr/2018LYSE2127/document.
Full textAbstract:
Cette thèse s’inscrit dans le cadre de l’ANR ReVeRIES dont l’objectif est d’utiliser les technologies mobiles pour aider l’homme à mieux connaître son environnement et notamment les végétaux qui l’entourent. Plus précisément, le projet ReVeRIES s’appuie sur une application mobile, nommée Folia développée dans le cadre du projet ANR ReVeS, capable de reconnaître les espèces d’arbres et arbustes à partir de photos de leurs feuilles. Ce prototype se différencie des autres outils car il est capable de simuler le comportement du botaniste. Dans le contexte du projet ReVeRIES, nous nous proposons d’aller beaucoup plus loin en développant de nouveaux aspects : la reconnaissance multimodale d’espèces, l’apprentissage par le jeu et les sciences citoyennes. L’objet de cette thèse porte sur le premier de ces trois aspects, à savoir l’analyse d’images d’organes de végétaux en vue de l’identification.Plus précisément, nous considérons les principaux arbres et arbustes, endémiques ou exotiques, que l’on trouve en France métropolitaine. L’objectif de cette thèse est d’étendre l’algorithme de reconnaissance en prenant en compte d’autres organes que la feuille. Cette multi-modalité est en effet essentielle si nous souhaitons que l’utilisateur apprenne et s’entraîne aux différentes méthodes de reconnaissance, pour lesquelles les botanistes utilisent la variété des organes (i.e. les feuilles, les fleurs, les fruits et les écorces). La méthode utilisée par Folia pour la reconnaissance des feuilles étant dédiée, car simulant le botaniste, ne peut s’appliquer directement aux autres organes. Ainsi, de nouveaux verrous se posent, tant au niveau dutraitement des images qu’au niveau de la fusion de données.Une première partie de la thèse a été consacrée à la mise en place de méthodes de traitement d’images pour l’identification des espèces végétales. C’est l’identification des espèces d’arbres à partir d’images d’écorces qui a été étudiée en premier. Les descripteurs développés prennent en compte la structure de l’écorce en s’inspirant des critères utilisés par les botanistes. Les fruits et les fleurs ont nécessité une étape de segmentation avant leur description. Une nouvelle méthode de segmentation réalisable sur smartphone a été développée pour fonctionner sur la grande variabilité des fleurs et des fruits. Enfin, des descripteurs ont été extraits sur les fruits et les fleurs après l’étape de segmentation. Nous avons décidé de ne pas faire de séparation entre les fleurs et les fruits car nous avons montré qu’un utilisateur novice en botanique ne sait pas toujours faire la différence entre ces deux organes sur des arbres dits «d’ornement» (non fruitiers). Pour les fruits et les fleurs, la prédiction n’est pas seulement faite sur les espèces mais aussi sur les genres et les familles, groupes botaniques traduisant d’une similarité entre ces organes.Une deuxième partie de la thèse traite de la combinaison des descripteurs des différents organes que sont les feuilles, les écorces, les fruits et les fleurs. En plus des méthodes de combinaison basiques, nous proposons de prendre en compte la confusion entre les espèces, ainsi que les prédictions d’appartenance aux taxons botaniques supérieurs à l’espèce.Enfin, un chapitre d’ouverture est consacré au traitement de ces images par des réseaux de neurones à convolutions. En effet, le Deep-Learning est de plus en plus utilisé en traitement d’images, notamment appliqué aux organes végétaux. Nous proposons dans ce contexte de visualiser les filtres de convolution extrayant de l’information, afin de faire le lien entre lesinformations extraites par ces réseaux et les éléments botaniques<br>This thesis is part of the ANR ReVeRIES, which aims to use mobile technologies to help people better understand their environment and in particular the plants that surround them. More precisely, the ReVeRIES project is based on a mobile application called Folia developed as part of the ANR ReVeS project and capable of recognising tree and shrub species based on photos of their leaves. This prototype differs from other tools in that it is able to simulate the behaviour of the botanist. In the context of the ReVeRIES project, we propose to go much further by developing new aspects: multimodal species recognition, learning through play and citizen science. The purpose of this thesis is to focus on the first of these three aspects, namelythe analysis of images of plant organs for identification.More precisely, we consider the main trees and shrubs, endemic or exotic, found in metropolitan France. The objective of this thesis is to extend the recognition algorithm by taking into account other organs in addition to the leaf. This multi-modality is indeed essential if we want the user to learn and practice the different methods of recognition for which botanists use the variety of organs (i.e. leaves, flowers, fruits and bark). The method used by Folia for leaf recognition being dedicated, because simulating the work of a botanist on the leaf, cannot be applied directly to other organs. Thus, new challenges are emerging, both in terms of image processing and data fusion.The first part of the thesis was devoted to the implementation of image processing methods for the identification of plant species. The identification of tree species from bark images was the first to be studied. The descriptors developed take into account the structure of the bark inspired from the criteria used by botanists. Fruits and flowers required a segmentation step before their description. A new segmentation method that can be used on smartphones has been developed to work in spite of the high variability of flowers and fruits. Finally, descriptors were extracted on fruits and flowers after the segmentation step. We decided not to separate flowers and fruits because we showed that a user new to botany does not always know the difference between these two organs on so-called "ornamental" trees (not fruit trees). For fruits and flowers, prediction is not only made on their species but also on their genus and family, botanical groups reflecting a similarity between these organs.The second part of the thesis deals with the combination of descriptors of the different organs: leaves, bark, fruits and flowers. In addition to basic combination methods, we propose to consider the confusion between species, as well as predictions of affiliations in botanical taxa higher than the species.Finally, an opening chapter is devoted to the processing of these images by convolutional neural networks. Indeed, Deep Learning is increasingly used in image processing, particularly for plant organs. In this context, we propose to visualize the learned convolution filters extracting information, in order to make the link between the information extracted by these networks and botanical elements
49
Grayston, TI. "A Simplified Fuzzy-Logic Control System Approach to Obstacle Avoidance combining Stereoscopic Vision and Sonar." Thesis, Honours thesis, University of Tasmania, 2006. https://eprints.utas.edu.au/791/1/Thesis.pdf.
Full textAbstract:
Stereoscopic vision is a technique for calculating the depths of objects in a scene from two images. Ultrasonic ranging is a well-established technique for estimating the distance to objects by bouncing an acoustic pulse off the object and measuring the time-of-flight. As with any types of sensors, these techniques each have their associated strengths and weaknesses. Therefore it is desirable
to be able to use both sensor types simultaneously on a robot such that the benefits of the techniques can each be taken advantage of.
Effective obstacle avoidance is an important challenge in the field of robotics that is integral to achieving the goal of fully autonomous mobile robots. To achieve such behaviour a control system is required for directing the robot on the basis of sensor inputs.
This study presents a simplified fuzzy logic control system that differs from the standard fuzzy logic system in the way that fuzzy sets are generated. The control layers of the system dynamically create fuzzy sets on-the-fly when called upon to do so. The developed control system is used to show that there are benefits to combining stereoscopic vision and sonar for robot obstacle avoidance compared against using these sensors in isolation.
50
Gomez-Cardenas, Carolina. "Outils d'aide à l'optimisation des campagnes d'essais non destructifs sur ouvrages en béton armé." Thesis, Toulouse 3, 2015. http://www.theses.fr/2015TOU30177/document.
Full textAbstract:
Les méthodes de contrôle non destructif (CND) sont essentielles pour estimer les propriétés du béton (mécaniques ou physiques) et leur variabilité spatiale. Elles constituent également un outil pertinent pour réduire le budget d'auscultation d'un ouvrage d'art. La démarche proposée est incluse dans un projet ANR (EvaDéOS) dont l'objectif est d'optimiser le suivi des ouvrages de génie civil en mettant en œuvre une maintenance préventive afin de réduire les coûts. Dans le cas du travail de thèse réalisé, pour caractériser au mieux une propriété particulière du béton (ex : résistance mécanique, porosité, degré de Saturation, etc.), avec des méthodes ND sensibles aux mêmes propriétés, il est impératif de développer des outils objectifs permettant de rationaliser une campagne d'essais sur les ouvrages en béton armé. Dans ce but, premièrement, il est proposé un outil d'échantillonnage spatial optimal pour réduire le nombre de points d'auscultation. L'algorithme le plus couramment employé est le recuit simulé spatial (RSS). Cette procédure est régulièrement utilisée dans des applications géostatistiques, et dans d'autres domaines, mais elle est pour l'instant quasiment inexploitée pour des structures de génie civil. Dans le travail de thèse, une optimisation de la méthode d'optimisation de l'échantillonnage spatial (MOES) originale inspirée du RSS et fondée sur la corrélation spatiale a été développée et testée dans le cas d'essais sur site avec deux fonctions objectifs complémentaires : l'erreur de prédiction moyenne et l'erreur sur l'estimation de la variabilité. Cette méthode est décomposée en trois parties. Tout d'abord, la corrélation spatiale des mesures ND est modélisée par un variogramme. Ensuite, la relation entre le nombre de mesures organisées dans une grille régulière et la fonction objectif est déterminée en utilisant une méthode d'interpolation spatiale appelée krigeage. Enfin, on utilise l'algorithme MOES pour minimiser la fonction objectif en changeant les positions d'un nombre réduit de mesures ND et pour obtenir à la fin une grille irrégulière optimale. Des essais destructifs (ED) sont nécessaires pour corroborer les informations obtenues par les mesures ND. En raison du coût ainsi que des dégâts possibles sur la structure, un plan d'échantillonnage optimal afin de prélever un nombre limité de carottes est important. Pour ce faire, une procédure utilisant la fusion des données fondée sur la théorie des possibilités et développée antérieurement, permet d'estimer les propriétés du béton à partir des ND. Par le biais d'un recalage nécessitant des ED réalisés sur carottes, elle est étalonnée. En sachant qu'il y a une incertitude sur le résultat des ED réalisés sur les carottes, il est proposé de prendre en compte cette incertitude et de la propager au travers du recalage sur les résultats des données fusionnées. En propageant ces incertitudes, on obtient des valeurs fusionnées moyennes par point avec un écart-type. On peut donc proposer une méthodologie de positionnement et de minimisation du nombre des carottes nécessaire pour ausculter une structure par deux méthodes : la première, en utilisant le MOES pour les résultats des propriétés sortis de la fusion dans chaque point de mesure et la seconde par la minimisation de l'écart-type moyen sur la totalité des points fusionnés, obtenu après la propagation des incertitudes des ED. Pour finir, afin de proposer une alternative à la théorie des possibilités, les réseaux de neurones sont également testés comme méthodes alternatives pour leur pertinence et leur simplicité d'utilisation<br>Non-destructive testing methods (NDT) are essential for estimating concrete properties (mechanical or physical) and their spatial variability. They also constitute an useful tool to reduce the budget auscultation of a structure. The proposed approach is included in an ANR project (EvaDéOS) whose objective is to optimize the monitoring of civil engineering structures by implementing preventive maintenance to reduce diagnosis costs. In this thesis, the objective was to characterize at best a peculiar property of concrete (e.g. mechanical strength, porosity, degree of saturation, etc.), with technical ND sensitive to the same properties. For this aim, it is imperative to develop objective tools that allow to rationalize a test campaign on reinforced concrete structures. For this purpose, first, it is proposed an optimal spatial sampling tool to reduce the number of auscultation points. The most commonly used algorithm is the spatial simulated annealing (SSA). This procedure is regularly used in geostatistical applications, and in other areas, but yet almost unexploited for civil engineering structures. In the thesis work, an original optimizing spatial sampling method (OSSM) inspired in the SSA and based on the spatial correlation was developed and tested in the case of on-site auscultation with two complementary fitness functions: mean prediction error and the error on the estimation of the global variability. This method is divided into three parts. First, the spatial correlation of ND measurements is modeled by a variogram. Then, the relationship between the number of measurements organized in a regular grid and the objective function is determined using a spatial interpolation method called kriging. Finally, the OSSM algorithm is used to minimize the objective function by changing the positions of a smaller number of ND measurements and for obtaining at the end an optimal irregular grid. Destructive testing (DT) are needed to corroborate the information obtained by the ND measurements. Because of the cost and possible damage to the structure, an optimal sampling plan to collect a limited number of cores is important. For this aim, a procedure using data fusion based on the theory of possibilities and previously developed is used to estimate the properties of concrete from the ND. Through a readjustment bias requiring DTs performed on carrots, it is calibrated. Knowing that there is uncertainty about the results of DTs performed on carrots, it is proposed to take into account this uncertainty and propagate it through the calibration on the results of the fused data. By propagating this uncertainty, it is obtained mean fused values with a standard deviation. One can thus provide a methodology for positioning and minimizing the number of cores required to auscultate a structure by two methods: first, using the OSSM for the results of fused properties values in each measuring point and the second by the minimization of the average standard deviation over all of the fused points obtained after the propagation of DTs uncertainties. Finally, in order to propose an alternative to the possibility theory, neural networks are also tested as alternative methods for their relevance and usability