To see the other types of publications on this topic, follow the link: 3D point clouds.
Dissertations / Theses on the topic '3D point clouds'
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 '3D point clouds.'
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
Srivastava, Siddharth. "Features for 3D point clouds." Thesis, IIT Delhi, 2019. http://eprint.iitd.ac.in:80//handle/2074/8061.
Full text
2
Filho, Carlos André Braile Przewodowski. "Feature extraction from 3D point clouds." Universidade de São Paulo, 2018. http://www.teses.usp.br/teses/disponiveis/55/55134/tde-30072018-111718/.
Full textAbstract:
Computer vision is a research field in which images are the main object of study. One of its category of problems is shape description. Object classification is one important example of applications using shape descriptors. Usually, these processes were performed on 2D images. With the large-scale development of new technologies and the affordable price of equipment that generates 3D images, computer vision has adapted to this new scenario, expanding the classic 2D methods to 3D. However, it is important to highlight that 2D methods are mostly dependent on the variation of illumination and color, while 3D sensors provide depth, structure/3D shape and topological information beyond color. Thus, different methods of shape descriptors and robust attributes extraction were studied, from which new attribute extraction methods have been proposed and described based on 3D data. The results obtained from well known public datasets have demonstrated their efficiency and that they compete with other state-of-the-art methods in this area: the RPHSD (a method proposed in this dissertation), achieved 85:4% of accuracy on the University of Washington RGB-D dataset, being the second best accuracy on this dataset; the COMSD (another proposed method) has achieved 82:3% of accuracy, standing at the seventh position in the rank; and the CNSD (another proposed method) at the ninth position. Also, the RPHSD and COMSD methods have relatively small processing complexity, so they achieve high accuracy with low computing time.Visão computacional é uma área de pesquisa em que as imagens são o principal objeto de estudo. Um dos problemas abordados é o da descrição de formatos (em inglês, shapes). Classificação de objetos é um importante exemplo de aplicação que usa descritores de shapes. Classicamente, esses processos eram realizados em imagens 2D. Com o desenvolvimento em larga escala de novas tecnologias e o barateamento dos equipamentos que geram imagens 3D, a visão computacional se adaptou para este novo cenário, expandindo os métodos 2D clássicos para 3D. Entretanto, estes métodos são, majoritariamente, dependentes da variação de iluminação e de cor, enquanto os sensores 3D fornecem informações de profundidade, shape 3D e topologia, além da cor. Assim, foram estudados diferentes métodos de classificação de objetos e extração de atributos robustos, onde a partir destes são propostos e descritos novos métodos de extração de atributos a partir de dados 3D. Os resultados obtidos utilizando bases de dados 3D públicas conhecidas demonstraram a eficiência dos métodos propóstos e que os mesmos competem com outros métodos no estado-da-arte: o RPHSD (um dos métodos propostos) atingiu 85:4% de acurácia, sendo a segunda maior acurácia neste banco de dados; o COMSD (outro método proposto) atingiu 82:3% de acurácia, se posicionando na sétima posição do ranking; e o CNSD (outro método proposto) em nono lugar. Além disso, os métodos RPHSD têm uma complexidade de processamento relativamente baixa. Assim, eles atingem uma alta acurácia com um pequeno tempo de processamento.
3
Truong, Quoc Hung. "Knowledge-based 3D point clouds processing." Phd thesis, Université de Bourgogne, 2013. http://tel.archives-ouvertes.fr/tel-00977434.
Full textAbstract:
The modeling of real-world scenes through capturing 3D digital data has proven to be both useful andapplicable in a variety of industrial and surveying applications. Entire scenes are generally capturedby laser scanners and represented by large unorganized point clouds possibly along with additionalphotogrammetric data. A typical challenge in processing such point clouds and data lies in detectingand classifying objects that are present in the scene. In addition to the presence of noise, occlusionsand missing data, such tasks are often hindered by the irregularity of the capturing conditions bothwithin the same dataset and from one data set to another. Given the complexity of the underlyingproblems, recent processing approaches attempt to exploit semantic knowledge for identifying andclassifying objects. In the present thesis, we propose a novel approach that makes use of intelligentknowledge management strategies for processing of 3D point clouds as well as identifying andclassifying objects in digitized scenes. Our approach extends the use of semantic knowledge to allstages of the processing, including the guidance of the individual data-driven processing algorithms.The complete solution consists in a multi-stage iterative concept based on three factors: the modeledknowledge, the package of algorithms, and a classification engine. The goal of the present work isto select and guide algorithms following an adaptive and intelligent strategy for detecting objects inpoint clouds. Experiments with two case studies demonstrate the applicability of our approach. Thestudies were carried out on scans of the waiting area of an airport and along the tracks of a railway.In both cases the goal was to detect and identify objects within a defined area. Results show that ourapproach succeeded in identifying the objects of interest while using various data types
4
Stålberg, Martin. "Reconstruction of trees from 3D point clouds." Thesis, Uppsala universitet, Avdelningen för systemteknik, 2017. http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-316833.
Full textAbstract:
The geometrical structure of a tree can consist of thousands, even millions, of branches, twigs and leaves in complex arrangements. The structure contains a lot of useful information and can be used for example to assess a tree's health or calculate parameters such as total wood volume or branch size distribution. Because of the complexity, capturing the structure of an entire tree used to be nearly impossible, but the increased availability and quality of particularly digital cameras and Light Detection and Ranging (LIDAR) instruments is making it increasingly possible. A set of digital images of a tree, or a point cloud of a tree from a LIDAR scan, contains a lot of data, but the information about the tree structure has to be extracted from this data through analysis. This work presents a method of reconstructing 3D models of trees from point clouds. The model is constructed from cylindrical segments which are added one by one. Bayesian inference is used to determine how to optimize the parameters of model segment candidates and whether or not to accept them as part of the model. A Hough transform for finding cylinders in point clouds is presented, and used as a heuristic to guide the proposals of model segment candidates. Previous related works have mainly focused on high density point clouds of sparse trees, whereas the objective of this work was to analyze low resolution point clouds of dense almond trees. The method is evaluated on artificial and real datasets and works rather well on high quality data, but performs poorly on low resolution data with gaps and occlusions.
5
Salman, Nader. "From 3D point clouds to feature preserving meshes." Nice, 2010. http://www.theses.fr/2010NICE4086.
Full textAbstract:
La majorité des algorithmes de reconstruction de surface sont optimisés pour s’appliquer à des données de haute qualité. Les résultats obtenus peuvent alors être utilisables si les données proviennent de solutions d’acquisition bon marché. Notre première contribution est un algorithme de reconstruction de surfaces à partir de données de stéréo vision. Il combine les informations liées aux points 3D avec les images calibrées afin de combler l’imprécision des données. L’algorithme construit une soupe de triangles 3D à l’aide des images calibrées et à l’issue d’une phase de prétraitement du nuage de points. Pour épouser au mieux la surface de la scène, on contraint cette soupe de triangle 3D à respecter des critères de visibilité et de photo-consistance. On calcule ensuite un maillage à partir de la soupe de triangles à l’aide d’une technique de reconstruction qui combine les triangulations de Delaunay contraintes et le raffinement de Delaunay. Notre seconde contribution est un algorithme qui construit, à partir d’un nuage de points 3D échantillonnés sur une surface, un maillage de surface qui représente fidèlement les arrêtes vives. Cet algorithme génère un bon compromis entre précision et complexité du maillage. Dans un premier temps, on extrait une approximation des arrêtes vives de la surface sous-jacente à partir du nuage de points. Dans un deuxième temps, on utilise une variante du raffinement de Delaunay pour générer un maillage qui combine les arrêtes vives extraites avec une surface implicite obtenue à partir du nuage de points. Notre méthode se révèle flexible, robuste au bruit ; cette méthode peut prendre en compte la résolution du maillage ciblé et un champ de taille défini par l’utilisateur. Nos deux contributions génèrent des résultats efficaces sur une variété de scènes et de modèles. Notre méthode améliore l’état de l’art en termes de précisionMost of the current surface reconstruction algorithms target high quality data and can produce some intractable results when used with point clouds acquired through profitable 3D acquisitions methods. Our first contribution is a surface reconstruction, algorithm from stereo vision data copes with the data’s fuzziness using information from both the acquired D point cloud and the calibrated images. After pre-processing the point cloud, the algorithm builds, using the calibrated images, 3D triangular soup consistent with the surface of the scene through a combination of visibility and photo-consistency constraints. A mesh is then computed from the triangle soup using a combination of restricted Delaunay triangulation and Delaunay refinement methods. Our second contribution is an algorithm that builds, given a 3D point cloud sampled on a surface, an approximating surface mesh with an accurate representation of surface sharp edges, providing an enhanced trade-off between accuracy and mesh complexity. We first extract from the point cloud an approximation of the sharp edges of the underlying surface. Then a feature preserving variant of a Delaunay refinement process generates a mesh combining a faithful representation of the extracted sharp edges with an implicit surface obtained from the point cloud. The method is shown to be flexible, robust to noise and tuneable to adapt to the scale of the targeted mesh and to a user defined sizing field. We demonstrate the effectiveness of both contributions on a variety of scenes and models acquired with different hardware and show results that compare favourably, in terms of accuracy, with the current state of the art
6
Robert, Damien. "Efficient learning on large-scale 3D point clouds." Electronic Thesis or Diss., Université Gustave Eiffel, 2024. http://www.theses.fr/2024UEFL2003.
Full textAbstract:
Au cours de la dernière décennie, l'apprentissage profond a été le moteur des progrès dans l'analyse automatisée de structures de données complexes aussi diverses que le texte, l'image, l'audio et la vidéo. En particulier, les modèles de type transformer et l'apprentissage auto-supervisé ont récemment déclenché une course généralisée visant à apprendre des représentations textuelles et visuelles expressives en entrainant le modèle au plus grand nombre de paramètres, sur le plus gros jeu de données possible, à l'aide des plus grandes ressources de calcul. Cette thèse emprunte un chemin différent en proposant des méthodes d'apprentissage profond économes en ressources, pour l'analyse de nuages de points 3D à grande échelle. L'efficacité des approches présentées se décline sous différentes formes : entrainement rapide, peu de paramètres, faible coût de calcul, économe en mémoire et exploitation de données disponibles de manière réaliste. Ce faisant, nous nous efforçons de concevoir des solutions pouvant être utilisées par les chercheurs et les praticiens avec des exigences matérielles minimales.Nous introduisons d'abord un modèle de segmentation sémantique 3D qui combine l'efficacité des méthodes basées superpoints avec l'expressivité des transformers. Nous construisons une représentation hiérarchique des données qui réduit considérablement la taille du problème d'analyse de nuage de points 3D, facilitant le traitement de scènes de grande échelle.Notre réseau se révèle égaler, voire surpasser, les approches de pointe sur une gamme de capteurs et d'environnements d'acquisition, tout en réduisant le nombre de paramètres et le temps d'entrainement de un à deux ordres de grandeur. Nous étendons ensuite ce cadre à la segmentation panoptique de nuages de points à grande échelle.Les méthodes existantes de segmentation d'instance et panoptique doivent résoudre un problème de correspondance complexe entre les instances prédites et réelles pour calculer leur fonction de coût. Au lieu de cela, nous formulons cette tâche comme un problème de clustering de graphe, qu'un petit réseau est entrainé pour résoudre à partir d'objectifs locaux uniquement, sans nécessiter le calcul d'instances durant l'entraînement. Notre modèle peut traiter des scènes de dix millions de points à la fois sur un seul GPU en quelques secondes, ouvrant la voie à la segmentation panoptique 3D à des échelles sans précédent. Enfin, nous proposons d'exploiter la complémentarité des modalités image et nuage de points pour améliorer l'analyse de scènes 3D. Nous nous plaçons dans un cadre d'acquisition réaliste, où plusieurs images arbitrairement positionnées observent la même scène, avec de potentielles occultations. Contrairement aux approches existantes de fusion 2D-3D, nous apprenons à sélectionner des informations à partir de différentes vues du même objet en fonction de leurs conditions d'observation respectives : distance caméra-objet, taux d'occultation, distorsion optique, etc. Notre implémentation efficace atteint l'état de l'art tant pour des scènes d'intérieur que d'extérieur, avec des exigences minimales : nuages de points bruts, images positionnées de manière arbitraire et les poses de leurs caméras. Dans l'ensemble, cette thèse soutient le principe que, dans des régimes où les données sont rares, exploiter la structure du problème permet de développer des architectures à la fois efficaces et performantesFor the past decade, deep learning has been driving progress in the automated understanding of complex data structures as diverse as text, image, audio, and video. In particular, transformer-based models and self-supervised learning have recently ignited a global competition to learn expressive textual and visual representations by training the largest possible model on Internet-scale datasets, with the help of massive computational resources. This thesis takes a different path, by proposing resource-efficient deep learning methods for the analysis of large-scale 3D point clouds.The efficiency of the introduced approaches comes in various flavors: fast training, few parameters, small compute or memory footprint, and leveraging realistically-available data.In doing so, we strive to devise solutions that can be used by researchers and practitioners with minimal hardware requirements.We first introduce a 3D semantic segmentation model which combines the efficiency of superpoint-based methods with the expressivity of transformers. We build a hierarchical data representation which drastically reduces the size of the 3D point cloud parsing problem, facilitating the processing of large point clouds en masse. Our self-attentive network proves to match or even surpass state-of-the-art approaches on a range of sensors and acquisition environments, while boasting orders of magnitude fewer parameters, faster training, and swift inference.We then build upon this framework to tackle panoptic segmentation of large-scale point clouds. Existing instance and panoptic segmentation methods need to solve a complex matching problem between predicted and ground truth instances for computing their supervision loss.Instead, we frame this task as a scalable graph clustering problem, which a small network is trained to address from local objectives only, without computing the actual object instances at train time. Our lightweight model can process ten-million-point scenes at once on a single GPU in a few seconds, opening the door to 3D panoptic segmentation at unprecedented scales. Finally, we propose to exploit the complementarity of image and point cloud modalities to enhance 3D scene understanding.We place ourselves in a realistic acquisition setting where multiple arbitrarily-located images observe the same scene, with potential occlusions.Unlike previous 2D-3D fusion approaches, we learn to select information from various views of the same object based on their respective observation conditions: camera-to-object distance, occlusion rate, optical distortion, etc. Our efficient implementation achieves state-of-the-art results both in indoor and outdoor settings, with minimal requirements: raw point clouds, arbitrarily-positioned images, and their cameras poses. Overall, this thesis upholds the principle that in data-scarce regimes,exploiting the structure of the problem unlocks both efficient and performant architectures
7
Al, Hakim Ezeddin. "3D YOLO: End-to-End 3D Object Detection Using Point Clouds." Thesis, KTH, Skolan för elektroteknik och datavetenskap (EECS), 2018. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-234242.
Full textAbstract:
For safe and reliable driving, it is essential that an autonomous vehicle can accurately perceive the surrounding environment. Modern sensor technologies used for perception, such as LiDAR and RADAR, deliver a large set of 3D measurement points known as a point cloud. There is a huge need to interpret the point cloud data to detect other road users, such as vehicles and pedestrians. Many research studies have proposed image-based models for 2D object detection. This thesis takes it a step further and aims to develop a LiDAR-based 3D object detection model that operates in real-time, with emphasis on autonomous driving scenarios. We propose 3D YOLO, an extension of YOLO (You Only Look Once), which is one of the fastest state-of-the-art 2D object detectors for images. The proposed model takes point cloud data as input and outputs 3D bounding boxes with class scores in real-time. Most of the existing 3D object detectors use hand-crafted features, while our model follows the end-to-end learning fashion, which removes manual feature engineering. 3D YOLO pipeline consists of two networks: (a) Feature Learning Network, an artificial neural network that transforms the input point cloud to a new feature space; (b) 3DNet, a novel convolutional neural network architecture based on YOLO that learns the shape description of the objects. Our experiments on the KITTI dataset shows that the 3D YOLO has high accuracy and outperforms the state-of-the-art LiDAR-based models in efficiency. This makes it a suitable candidate for deployment in autonomous vehicles.För att autonoma fordon ska ha en god uppfattning av sin omgivning används moderna sensorer som LiDAR och RADAR. Dessa genererar en stor mängd 3-dimensionella datapunkter som kallas point clouds. Inom utvecklingen av autonoma fordon finns det ett stort behov av att tolka LiDAR-data samt klassificera medtrafikanter. Ett stort antal studier har gjorts om 2D-objektdetektering som analyserar bilder för att upptäcka fordon, men vi är intresserade av 3D-objektdetektering med hjälp av endast LiDAR data. Därför introducerar vi modellen 3D YOLO, som bygger på YOLO (You Only Look Once), som är en av de snabbaste state-of-the-art modellerna inom 2D-objektdetektering för bilder. 3D YOLO tar in ett point cloud och producerar 3D lådor som markerar de olika objekten samt anger objektets kategori. Vi har tränat och evaluerat modellen med den publika träningsdatan KITTI. Våra resultat visar att 3D YOLO är snabbare än dagens state-of-the-art LiDAR-baserade modeller med en hög träffsäkerhet. Detta gör den till en god kandidat för kunna användas av autonoma fordon.
8
Biasutti, Pierre. "2D Image Processing Applied to 3D LiDAR Point Clouds." Thesis, Bordeaux, 2019. http://www.theses.fr/2019BORD0161/document.
Full textAbstract:
L'intérêt toujours grandissant pour les données cartographiques fiables, notamment en milieu urbain, a motivé le développement de systèmes de cartographie mobiles terrestres. Ces systèmes sont conçus pour l'acquisition de données de très haute précision, telles que des nuages de points LiDAR 3D et des images optiques. La multitude de données, ainsi que leur diversité, rendent complexe le traitement des données issues de ce type de systèmes. Cette thèse se place dans le contexte du traitement de l'image appliqué au nuages de points LiDAR 3D issus de ce type de système.Premièrement, nous nous intéressons à des images issues de la projection de nuages de points LiDAR dans des grilles de pixels 2D régulières. Ces projections créent généralement des images éparses, dans lesquelles l'information de certains pixels n'est pas connue. Nous proposons alors différentes méthodes pour des applications telles que la génération d'orthoimages haute résolution, l'imagerie RGB-D et l'estimation de la visibilité des points d'un nuage.De plus, nous proposons d'exploiter la topologie d'acquisition des capteurs LiDAR pour produire des images de faible résolution: les range-images. Ces images offrent une représentation efficace et canonique du nuage de points, tout en étant directement accessibles à partir du nuage de points. Nous montrons comment ces images peuvent être utilisées pour simplifier, voire améliorer, des méthodes pour le recalage multi-modal, la segmentation, la désoccultation et la détection 3DThe ever growing demand for reliable mapping data, especially in urban environments, has motivated the development of "close-range" Mobile Mapping Systems (MMS). These systems acquire high precision data, and in particular 3D LiDAR point clouds and optical images. The large amount of data, along with their diversity, make MMS data processing a very complex task. This thesis lies in the context of 2D image processing applied to 3D LiDAR point clouds acquired with MMS.First, we focus on the projection of the LiDAR point clouds onto 2D pixel grids to create images. Such projections are often sparse because some pixels do not carry any information. We use these projections for different applications such as high resolution orthoimage generation, RGB-D imaging and visibility estimation in point clouds.Moreover, we exploit the topology of LiDAR sensors in order to create low resolution images, named range-images. These images offer an efficient and canonical representation of the point cloud, while being directly accessible from the point cloud. We show how range-images can be used to simplify, and sometimes outperform, methods for multi-modal registration, segmentation, desocclusion and 3D detection
9
IRFAN, MUHAMMAD ABEER. "Joint geometry and color denoising for 3D point clouds." Doctoral thesis, Politecnico di Torino, 2021. http://hdl.handle.net/11583/2912976.
Full text
10
Fucili, Mattia. "3D object detection from point clouds with dense pose voters." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2019. http://amslaurea.unibo.it/17616/.
Full textAbstract:
Il riconoscimento di oggetti è sempre stato un compito sfidante per la Computer Vision. Trova applicazione in molti campi, principalmente nell’industria, come ad esempio per permettere ad un robot di trovare gli oggetti da afferrare. Negli ultimi decenni tali compiti hanno trovato nuovi modi di essere raggiunti grazie alla riscoperta delle Reti Neurali, in particolare le Reti Neurali Convoluzionali. Questo tipo di reti ha raggiunto ottimi risultati in molte applicazioni per il riconoscimento e la classificazione degli oggetti. La tendenza, ora, `e quella di utilizzare tali reti anche nell’industria automobilistica per cercare di rendere reale il sogno delle automobili che guidano da sole. Ci sono molti lavori importanti sul riconoscimento delle auto dalle immagini. In questa tesi presentiamo la nostra architettura di Rete Neurale Convoluzionale per il riconoscimento di automobili e la loro posizione nello spazio, utilizzando solo input lidar. Salvando le informazioni riguardanti le bounding box attorno all’auto a livello del punto ci assicura una buona previsione anche in situazioni in cui le automobili sono occluse. I test vengono eseguiti sul dataset più utilizzato per il riconoscimento di automobili e pedoni nelle applicazioni di guida autonoma.
11
Wang, Yutao. "Outlier formation and removal in 3D laser scanned point clouds." Thesis, University of British Columbia, 2014. http://hdl.handle.net/2429/51265.
Full textAbstract:
3D scanners have become widely used in many industrial applications in reverse engineering, quality inspection, entertainment industry, etc. Despite the popularity of 3D scanners, the raw scanned data, referred to as point cloud, is often contaminated by outliers not belonging to the scanned surface. Moreover, when the scanned surface is highly reflective, outliers become much more extensive due to specular reflections. Such outliers cause considerable issues to point cloud applications and thus need to be removed through an outlier detection process. Considering the commonness of reflective surfaces in mechanical parts, it is critical to investigate the outlier formation mechanism and develop methods to effectively remove outliers. However, research on how outliers are formed in scanning reflective surfaces is very limited. Meanwhile, existing outlier removal methods show limited effectiveness in detecting extensive outliers.
This thesis investigates the outlier formation mechanism in scanning reflective surfaces using laser scanners, and develops outlier removal algorithms to effectively and efficiently detect outliers in the scanned point clouds. The overall objective is to remove outliers in a raw data to obtain a clean point cloud in order to ensure the performance of point cloud applications. In particular, two outlier formation models, mixed reflections and multi-path reflections, are proposed and verified through experiments. The effects of scanning orientation on outlier formation are also experimentally investigated. A guidance of proper scan path planning is provided in order to reduce the occurrence of outliers. Regarding outlier removal, a rotating scan approach is proposed to efficiently remove view-dependent outliers. A flexible and effective algorithm is also presented to detect the challenging non-isolated outliers as well as other outliers.Applied Science, Faculty of
Mechanical Engineering, Department of
Graduate
12
Yanes, Luis. "Haptic Interaction with 3D oriented point clouds on the GPU." Thesis, University of East Anglia, 2015. https://ueaeprints.uea.ac.uk/58556/.
Full textAbstract:
Real-time point-based rendering and interaction with virtual objects is gaining popularity and importance as different haptic devices and technologies increasingly provide the basis for realistic interaction. Haptic Interaction is being used for a wide range of applications such as medical training, remote robot operators, tactile displays and video games. Virtual object visualization and interaction using haptic devices is the main focus; this process involves several steps such as: Data Acquisition, Graphic Rendering, Haptic Interaction and Data Modification. This work presents a framework for Haptic Interaction using the GPU as a hardware accelerator, and includes an approach for enabling the modification of data during interaction. The results demonstrate the limits and capabilities of these techniques in the context of volume rendering for haptic applications. Also, the use of dynamic parallelism as a technique to scale the number of threads needed from the accelerator according to the interaction requirements is studied allowing the editing of data sets of up to one million points at interactive haptic frame rates.
13
Paffenholz, Jens-André [Verfasser]. "Direct geo-referencing of 3D point clouds with 3D positioning sensors / Jens-André Paffenholz." Hannover : Technische Informationsbibliothek und Universitätsbibliothek Hannover (TIB), 2013. http://d-nb.info/1036695646/34.
Full text
14
Paffenholz, Jens-André [Verfasser]. "Direct geo-referencing of 3D point clouds with 3D positioning sensors / Jens-André Paffenholz." Hannover : Gottfried Wilhelm Leibniz Universität Hannover, 2012. http://d-nb.info/1183907168/34.
Full text
15
Rashidi, Abbas. "Improved monocular videogrammetry for generating 3D dense point clouds of built infrastructure." Diss., Georgia Institute of Technology, 2014. http://hdl.handle.net/1853/52257.
Full textAbstract:
Videogrammetry is an affordable and easy-to-use technology for spatial 3D scene recovery. When applied to the civil engineering domain, a number of issues have to be taken into account.
First, videotaping large scale civil infrastructure scenes usually results in large video files filled with blurry, noisy, or simply redundant frames. This is often due to higher frame rate over camera speed ratio than necessary, camera and lens imperfections, and uncontrolled motions of the camera that results in motion blur. Only a small percentage of the collected video frames are required to achieve robust results. However, choosing the right frames is a tough challenge.
Second, the generated point cloud using a monocular videogrammetric pipeline is up to scale, i.e. the user has to know at least one dimension of an object in the scene to scale up the entire scene. This issue significantly narrows applications of generated point clouds in civil engineering domain since measurement is an essential part of every as-built documentation technology.
Finally, due to various reasons including the lack of sufficient coverage during videotaping of the scene or existence of texture-less areas which are common in most indoor/outdoor civil engineering scenes, quality of the generated point clouds are sometimes poor. This deficiency appears in the form of outliers or existence of holes or gaps on surfaces of point clouds. Several researchers have focused on this particular problem; however, the major issue with all of the currently existing algorithms is that they basically treat holes and gaps as part of a smooth surface. This approach is not robust enough at the intersections of different surfaces or corners while there are sharp edges. A robust algorithm for filling holes/gaps should be able to maintain sharp edges/corners since they usually contain useful information specifically for applications in the civil and infrastructure engineering domain.
To tackle these issues, this research presents and validates an improved videogrammetric pipeline for as built documentation of indoor/outdoor applications in civil engineering areas. The research consists of three main components:
1. Optimized selection of key frames for processing. It is necessary to choose a number of informative key frames to get the best results from the videogrammetric pipeline. This step is particularly important for outdoor environments as it is impossible to process a large number of frames existing in a large video clip.
2. Automated calculation of absolute scale of the scene. In this research, a novel approach for the process of obtaining absolute scale of points cloud by using 2D and 3D patterns is proposed and validated.
3. Point cloud data cleaning and filling holes on the surfaces of generated point clouds. The proposed algorithm to achieve this goal is able to fill holes/gaps on surfaces of point cloud data while maintaining sharp edges.
In order to narrow the scope of the research, the main focus will be on two specific applications:
1. As built documentation of bridges and building as outdoor case studies.
2. As built documentation of offices and rooms as indoor case studies.
Other potential applications of monocular videogrammetry in the civil engineering domain are out of scope of this research. Two important metrics, i.e. accuracy, completeness and processing time, are utilized for evaluation of the proposed algorithms.
16
Engelmann, Francis [Verfasser], Bastian [Akademischer Betreuer] Leibe, and Siyu [Akademischer Betreuer] Tang. "3D scene understanding on point clouds / Francis Engelmann ; Bastian Leibe, Siyu Tang." Aachen : Universitätsbibliothek der RWTH Aachen, 2021. http://nbn-resolving.de/urn:nbn:de:101:1-2021100303242549426277.
Full text
17
Abuzaina, Anas. "On evidence gathering in 3D point clouds of static and moving objects." Thesis, University of Southampton, 2015. https://eprints.soton.ac.uk/381290/.
Full textAbstract:
The recent and considerable progress in 3D sensing technologies mandates the development of efficient algorithms to process the sensed data. Many of these algorithms are based on computing and matching of 3D feature descriptors in order to estimate point correspondences between 3D datasets. The dependency on 3D feature description and computation can be a significant limitation to many 3D perception tasks; the fact that there are a variety of criteria used to describe 3D features, such as surface normals and curvature, makes feature-based approaches sensitive to noise and occlusion. In many cases, such as smooth surfaces, computation of feature descriptors can be non-informative. Moreover, the process of computing and matching features requires more computational overhead than using points directly. On the other hand, there has not been much focus on employing evidence gathering frameworks to obtain solutions for 3D perception problems. Evidence gathering approaches, which use data directly, have proved to provide robust performance against noise and occlusion. More importantly, evidence gathering approaches do not require initialisation or training, and avoid the need to solve the correspondence problem. The capability to detect, extract and reconstruct 3D bjects without relying on feature matching and estimating correspondences between 3D datasets has not been thoroughly investigated, and is certainly desirable and has many practical applications. In this thesis we present theoretical formulations and practical solutions to 3D perceptual tasks, that are based on evidence gathering. We propose a new 3D reconstruction algorithm for rotating objects that is based on motion-compensated temporal accumulation. We also propose two fast and robust Hough Transform based algorithms for 3D static parametric object detection and 3D moving parametric object extraction. Furthermore, we introduce two algorithms for 3D motion parameter estimation that are based on Reuleaux's and Chasles' kinematic theorems. The proposed algorithms estimate 3D motion parameters directly from the data by exploiting the geometry of rigid transformation. Moreover, they provide an alternative to the both local and global feature description and matching pipelines commonly used by numerous 3D object recognition and registration algorithms. Our objective is to provide new means for understanding static and dynamic scenes, captured by new 3D sensing technologies as we believe that these technologies will be dominant in the perception field as they are going under rapid development. We provide alternative solutions to commonly used feature based approaches by using new evidence gathering based methods for the processing of 3D range data.
18
Wiklander, Marcus. "Classification of tree species from 3D point clouds using convolutional neural networks." Thesis, Umeå universitet, Institutionen för fysik, 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:umu:diva-174662.
Full textAbstract:
In forest management, knowledge about a forest's distribution of tree species is key. Being able to automate tree species classification for large forest areas is of great interest, since it is tedious and costly labour doing it manually. In this project, the aim was to investigate the efficiency of classifying individual tree species (pine, spruce and deciduous forest) from 3D point clouds acquired by airborne laser scanning (ALS), using convolutional neural networks. Raw data consisted of 3D point clouds and photographic images of forests in northern Sweden, collected from a helicopter flying at low altitudes. The point cloud of each individual tree was connected to its representation in the photos, which allowed for manual labeling of training data to be used for training of convolutional neural networks. The training data consisted of labels and 2D projections created from the point clouds, represented as images. Two different convolutional neural networks were trained and tested; an adaptation of the LeNet architecture and the ResNet architecture. Both networks reached an accuracy close to 98 %, the LeNet adaptation having a slightly lower loss score for both validation and test data compared to that of ResNet. Confusion matrices for both networks showed similar F1 scores for all tree species, between 97 % and 98 %. The accuracies computed for both networks were found higher than those achieved in similar studies using ALS data to classify individual tree species. However, the results in this project were never tested against a true population sample to confirm the accuracy. To conclude, the use of convolutional neural networks is indeed an efficient method for classification of tree species, but further studies on unbiased data is needed to validate these results.
19
Azhari, Faris. "Automated crack detection and characterisation from 3D point clouds of unstructured surfaces." Thesis, Queensland University of Technology, 2022. https://eprints.qut.edu.au/234510/1/Faris_Azhari_Thesis.pdf.
Full textAbstract:
This thesis proposes a novel automated crack detection and characterisation method on unstructured surfaces using 3D point cloud. Crack detection on unstructured surfaces poses a challenge compared to flat surfaces such as pavements and concrete, which typically utilise image-based sensors. The detection method utilises a point cloud-based deep learning method to perform point-wise classification. The detected points are then automatically characterised to estimate the detected cracks’ properties such as width profile, orientation, and length. The proposed method enables the deployment of autonomous systems to conduct reliable surveys in environments risky to humans.
20
Serra, Sabina. "Deep Learning for Semantic Segmentation of 3D Point Clouds from an Airborne LiDAR." Thesis, Linköpings universitet, Datorseende, 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-168367.
Full textAbstract:
Light Detection and Ranging (LiDAR) sensors have many different application areas, from revealing archaeological structures to aiding navigation of vehicles. However, it is challenging to interpret and fully use the vast amount of unstructured data that LiDARs collect. Automatic classification of LiDAR data would ease the utilization, whether it is for examining structures or aiding vehicles. In recent years, there have been many advances in deep learning for semantic segmentation of automotive LiDAR data, but there is less research on aerial LiDAR data. This thesis investigates the current state-of-the-art deep learning architectures, and how well they perform on LiDAR data acquired by an Unmanned Aerial Vehicle (UAV). It also investigates different training techniques for class imbalanced and limited datasets, which are common challenges for semantic segmentation networks. Lastly, this thesis investigates if pre-training can improve the performance of the models. The LiDAR scans were first projected to range images and then a fully convolutional semantic segmentation network was used. Three different training techniques were evaluated: weighted sampling, data augmentation, and grouping of classes. No improvement was observed by the weighted sampling, neither did grouping of classes have a substantial effect on the performance. Pre-training on the large public dataset SemanticKITTI resulted in a small performance improvement, but the data augmentation seemed to have the largest positive impact. The mIoU of the best model, which was trained with data augmentation, was 63.7% and it performed very well on the classes Ground, Vegetation, and Vehicle. The other classes in the UAV dataset, Person and Structure, had very little data and were challenging for most models to classify correctly. In general, the models trained on UAV data performed similarly as the state-of-the-art models trained on automotive data.
21
Paudel, Danda Pani. "Local and global methods for registering 2D image sets and 3D point clouds." Thesis, Dijon, 2015. http://www.theses.fr/2015DIJOS077/document.
Full textAbstract:
Pas de résuméIn this thesis, we study the problem of registering 2D image sets and 3D point clouds under threedifferent acquisition set-ups. The first set-up assumes that the image sets are captured using 2Dcameras that are fully calibrated and coupled, or rigidly attached, with a 3D sensor. In this context,the point cloud from the 3D sensor is registered directly to the asynchronously acquired 2D images.In the second set-up, the 2D cameras are internally calibrated but uncoupled from the 3D sensor,allowing them to move independently with respect to each other. The registration for this set-up isperformed using a Structure-from-Motion reconstruction emanating from images and planar patchesrepresenting the point cloud. The proposed registration method is globally optimal and robust tooutliers. It is based on the theory Sum-of-Squares polynomials and a Branch-and-Bound algorithm.The third set-up consists of uncoupled and uncalibrated 2D cameras. The image sets from thesecameras are registered to the point cloud in a globally optimal manner using a Branch-and-Prunealgorithm. Our method is based on a Linear Matrix Inequality framework that establishes directrelationships between 2D image measurements and 3D scene voxels
22
Graehling, Quinn R. "Feature Extraction Based Iterative Closest Point Registration for Large Scale Aerial LiDAR Point Clouds." University of Dayton / OhioLINK, 2020. http://rave.ohiolink.edu/etdc/view?acc_num=dayton1607380713807017.
Full text
23
Digne, Julie. "Inverse geometry : from the raw point cloud to the 3d surface : theory and algorithms." Phd thesis, École normale supérieure de Cachan - ENS Cachan, 2010. http://tel.archives-ouvertes.fr/tel-00610432.
Full textAbstract:
Many laser devices acquire directly 3D objects and reconstruct their surface. Nevertheless, the final reconstructed surface is usually smoothed out as a result of the scanner internal de-noising process and the offsets between different scans. This thesis, working on results from high precision scans, adopts the somewhat extreme conservative position, not to loose or alter any raw sample throughout the whole processing pipeline, and to attempt to visualize them. Indeed, it is the only way to discover all surface imperfections (holes, offsets). Furthermore, since high precision data can capture the slightest surface variation, any smoothing and any sub-sampling can incur in the loss of textural detail.The thesis attempts to prove that one can triangulate the raw point cloud with almost no sample loss. It solves the exact visualization problem on large data sets of up to 35 million points made of 300 different scan sweeps and more. Two major problems are addressed. The first one is the orientation of the complete raw point set, an the building of a high precision mesh. The second one is the correction of the tiny scan misalignments which can cause strong high frequency aliasing and hamper completely a direct visualization.The second development of the thesis is a general low-high frequency decomposition algorithm for any point cloud. Thus classic image analysis tools, the level set tree and the MSER representations, are extended to meshes, yielding an intrinsic mesh segmentation method.The underlying mathematical development focuses on an analysis of a half dozen discrete differential operators acting on raw point clouds which have been proposed in the literature. By considering the asymptotic behavior of these operators on a smooth surface, a classification by their underlying curvature operators is obtained.This analysis leads to the development of a discrete operator consistent with the mean curvature motion (the intrinsic heat equation) defining a remarkably simple and robust numerical scale space. By this scale space all of the above mentioned problems (point set orientation, raw point set triangulation, scan merging, segmentation), usually addressed by separated techniques, are solved in a unified framework.
24
Ochmann, Sebastian Klaus [Verfasser]. "Automatic Reconstruction of Parametric, Volumetric Building Models from 3D Point Clouds / Sebastian Klaus Ochmann." Bonn : Universitäts- und Landesbibliothek Bonn, 2019. http://d-nb.info/1188731599/34.
Full text
25
Patiño, Mejía Isabel Cristina [Verfasser], and Andreas [Akademischer Betreuer] Zell. "Estimating Head Measurements from 3D Point Clouds / Isabel Cristina Patiño Mejía ; Betreuer: Andreas Zell." Tübingen : Universitätsbibliothek Tübingen, 2019. http://d-nb.info/1201644429/34.
Full text
26
Dehbi, Youness [Verfasser]. "Statistical relational learning of semantic models and grammar rules for 3D building reconstruction from 3D point clouds / Youness Dehbi." Bonn : Universitäts- und Landesbibliothek Bonn, 2016. http://d-nb.info/112228585X/34.
Full text
27
Arvidsson, Simon, and Marcus Gullstrand. "Predicting forest strata from point clouds using geometric deep learning." Thesis, Jönköping University, JTH, Avdelningen för datavetenskap, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:hj:diva-54155.
Full textAbstract:
Introduction: Number of strata (NoS) is an informative descriptor of forest structure and is therefore useful in forest management. Collection of NoS as well as other forest properties is performed by fieldworkers and could benefit from automation. Objectives: This study investigates automated prediction of NoS from airborne laser scanned point clouds over Swedish forest plots.Methods: A previously suggested approach of using vertical gap probability is compared through experimentation against the geometric neural network PointNet++ configured for ordinal prediction. For both approaches, the mean accuracy is measured for three datasets: coniferous forest, deciduous forest, and a combination of all forests. Results: PointNet++ displayed a better point performance for two out of three datasets, attaining a top mean accuracy of 46.2%. However only the coniferous subset displayed a statistically significant superiority for PointNet++. Conclusion: This study demonstrates the potential of geometric neural networks for data mining of forest properties. The results show that impediments in the data may need to be addressed for further improvements.
28
Avdiu, Blerta. "Matching Feature Points in 3D World." Thesis, Tekniska Högskolan, Högskolan i Jönköping, JTH, Data- och elektroteknik, 2012. http://urn.kb.se/resolve?urn=urn:nbn:se:hj:diva-23049.
Full textAbstract:
This thesis work deals with the most actual topic in Computer Vision field which is scene understanding and this using matching of 3D feature point images. The objective is to make use of Saab’s latest breakthrough in extraction of 3D feature points, to identify the best alignment of at least two 3D feature point images. The thesis gives a theoretical overview of the latest algorithms used for feature detection, description and matching. The work continues with a brief description of the simultaneous localization and mapping (SLAM) technique, ending with a case study on evaluation of the newly developed software solution for SLAM, called slam6d. Slam6d is a tool that registers point clouds into a common coordinate system. It does an automatic high-accurate registration of the laser scans. In the case study the use of slam6d is extended in registering 3D feature point images extracted from a stereo camera and the results of registration are analyzed. In the case study we start with registration of one single 3D feature point image captured from stationary image sensor continuing with registration of multiple images following a trail. Finally the conclusion from the case study results is that slam6d can register non-laser scan extracted feature point images with high-accuracy in case of single image but it introduces some overlapping results in the case of multiple images following a trail.
29
Kersten, Thomas. "Untersuchungen zur Qualität und Genauigkeit von 3D-Punktwolken für die 3D-Objektmodellierung auf der Grundlage von terrestrischem Laserscanning und bildbasierten Verfahren." Doctoral thesis, Saechsische Landesbibliothek- Staats- und Universitaetsbibliothek Dresden, 2018. http://nbn-resolving.de/urn:nbn:de:bsz:14-qucosa-231616.
Full textAbstract:
3D-Punktwolken haben die Objektvermessung in den letzten 25 Jahren signifikant verändert. Da Einzelpunktmessungen durch flächenhafte Messungen in Form von Punktwolken bei vielen Anwendungen ersetzt wurden, spricht man auch von einem Paradigmenwechsel in der Vermessung. Ermöglicht wurde diese Änderung in der Messmethodik durch die Innovationen im Instrumentenbau und die rasanten Entwicklungen der Computertechnologie. Luftgestützte und terrestrische Laserscanner sowie handgeführte 3D-Scanner liefern heute direkt dichte Punktwolken, während dichte 3D-Punkt-wolken aus Fotos bildbasierter Aufnahmesysteme indirekt abgeleitet werden, die zur detaillierten 3D-Objektrekonstruktion zunehmend eingesetzt werden. In dieser Arbeit werden Untersuchungen vorgestellt, mit denen das geometrische Genauigkeitsverhalten verschiedener scannender Messsysteme evaluiert und geprüft wurde. Während bei den untersuchten terrestrischen Laserscannern in den Untersuchungen die Genauigkeitsangaben (1 Sigma) der technischen Spezifikationen der Systemhersteller von 3-5 mm für den 3D-Punkt und die Distanzmessung eingehalten wurden, zeigten sich dagegen bei vielen untersuchten 3D-Handscannern signifikante Abweichungen gegenüber den technischen Spezifikationen. Diese festgestellten Abweichungen deuten auf eine gewisse geometrische Instabilität des jeweiligen Messsystems hin, die entweder durch die Bauweise und/oder durch eine ungenaue Systemkalibrierung (besonders hinsichtlich der Maßstäblichkeit) verursacht werden. Daher ist davon auszugehen, dass diese handgeführten 3D-Scanner offensichtlich erst am Anfang ihrer Entwicklungsphase stehen und dass noch genügend Optimierungspotential vorhanden ist. Als flexible und effiziente Alternativen zu den scannenden Messsystemen haben sich seit ca. 10 Jahren die bildbasierten Aufnahmesysteme zunehmend im Markt etabliert. Die in dieser Arbeit vorgestellten Untersuchungen des bildbasierten Aufnahme- und Auswertungsverfahren haben gezeigt, dass diese (mit Farbattributen versehene) 3D-Punktwolken, je nach Bildmaßstab und Oberflächenmaterial des Objektes, durchaus den Genauigkeiten der Laserscanner entsprechen. Gegenüber den Ergebnissen vieler 3D-Handscanner weisen die durch bildbasierte Aufnahmeverfahren generierten Punktwolken qualitativ bessere Resultate auf. Allerdings zeigte der Creaform HandySCAN 700, der auf einem photogrammetrischen Aufnahmeprinzip beruht, als einzige Ausnahme bei der handgeführten 3D-Scannern sehr gute Ergebnisse, die mit Durchschnittswerten besser als 30 Mikrometern sogar in den Bereichen der Referenzsysteme (hier Streifenprojektionssysteme) lagen. Die entwickelten Prüfverfahren und die entsprechenden durchgeführten Untersuchungen haben sich als praxistauglich erwiesen, da man auch unter zur Hilfenahme der VDI/VDE Richtlinie 2634 ver-gleichbare Ergebnisse erzielt, die dem praxisorientierten Anwender Aussagen über die Leistungsfä-higkeit des Messsystems erlauben. Bei den im statischen Modus erfassten Scans kommen noch Fehlereinflüsse durch die Registrierung der Scans hinzu, während bei kinematisch erfassten Scans die Genauigkeiten der verschiedenen (absoluten) Positionierungssensoren auf dem Fehlerhaushalt der Punktwolke addiert werden. Eine sorgfältige Systemkalibrierung der verschiedenen im kinematischen Modus arbeitenden Positionierungs- und Aufnahmesensoren des mobilen Multi-Sensor-Systems ermöglicht eine 3D-Punktgenauigkeit von ca. 3-5 cm, die unter guten Bedingungen mit höherwertigen Sensoren ggf. noch verbessert werden kann. Mit statischen Scans kann eine höhere Genauigkeit von besser als 1 cm für den 3D-Punkt erreicht werden, jedoch sind bei größeren aufzunehmenden Flächen mobile Aufnahmesysteme wesentlich effizienter. Die Anwendung definiert daher das zum Einsatz kommende Messverfahren. 3D-Punktwolken dienen als Grundlage für die Objektrekonstruktion auf verschiedenen Wegen: a) Engineering Modelling als generalisierte CAD-Konstruktion durch geometrische Primitive und b) Mesh Modelling durch Dreiecksvermaschung der Punktwolken zur exakten Oberflächenbeschreibung. Durch die Generalisierung bei der CAD-Konstruktion können sehr schnell Abweichungen vom Sollmaß von bis zu 10 cm (und größer) entstehen, allerdings werden durch die Anpassung auf geometrische Primitive eine signifikante Datenreduktion und eine topologische Strukturierung erreicht. Untersuchungen haben jedoch auch gezeigt, dass die Anzahl der Polygone bei der Dreiecksvermaschung je nach Oberflächenbeschaffenheit des Objektes auf 25% und sogar auf 10% der Originaldatenmenge bei intelligenter Ausdünnung (z.B. krümmungsbasiert) reduziert werden kann, ohne die visuelle und geometrische Qualität des Ergebnisses zu stark zu beeinträchtigen. Je nach Objektgröße können hier Abweichungen von unter einem Millimeter (z.B. bei archäologischen Fundstücken) bis zu 5 cm im Durchschnitt bei größeren Objekten erreicht werden. Heute können Punktwolken eine wichtige Grundlage zur Konstruktion der Umgebung für viele Virtual Reality Anwendungen bilden, bei denen die geometrische Genauigkeit der modellierten Objekte im Einzelfall keine herausragende Rolle spielt3D point clouds have significantly changed the surveying of objects in the last 25 years. Since in many applications, the individual point measurements were replaced through area-based measurements in form of point clouds, a paradigm shift in surveying has been fulfilled. This change in measurement methodology was made possible with the rapid developments in instrument manufacturing and computer technology. Today, airborne and terrestrial laser scanners, as well as hand-held 3D scanners directly generate dense point clouds, while dense point clouds are indirectly derived from photos of image-based recording systems used for detailed 3D object reconstruction in almost any scale. In this work, investigations into the geometric accuracy of some of these scanning systems are pre-sented to document and evaluate their performance. While terrestrial laser scanners mostly met the accuracy specifications in the investigations, 3-5 mm for 3D points and distance measurements as defined in the technical specifications of the system manufacturer, significant differences are shown, however, by many tested hand-held 3D scanners. These observed deviations indicate a certain geometric instability of the measuring system, caused either by the construction/manufacturing and/or insufficient calibration (particularly with regard to the scale). It is apparent that most of the hand-held 3D scanners are at the beginning of the technical development, which still offers potential for optimization. The image-based recording systems have been increasingly accepted by the market as flexible and efficient alternatives to laser scanning systems for about ten years. The research of image-based recording and evaluation methods presented in this work has shown that these coloured 3D point clouds correspond to the accuracy of the laser scanner depending on the image scale and surface material of the object. Compared with the results of most hand-held 3D scanners, point clouds gen-erated by image-based recording techniques exhibit superior quality. However, the Creaform HandySCAN 700, based on a photogrammetric recording principle (stereo photogrammetry), shows as the solitary exception of the hand-held 3D scanners very good results with better than 30 micrometres on average, representing accuracies even in the range of the reference systems (here structured light projection systems). The developed test procedures and the corresponding investigations have been practically proven for both terrestrial and hand-held 3D scanners, since comparable results can be obtained using the VDI/VDE guidelines 2634, which allows statements about the performance of the tested scanning system for practice-oriented users. For object scans comprised of multiple single scan acquired in static mode, errors of the scan registration have to be added, while for scans collected in the kine-matic mode the accuracies of the (absolute) position sensors will be added on the error budget of the point cloud. A careful system calibration of various positioning and recording sensors of the mobile multi-sensor system used in kinematic mode allows a 3D point accuracy of about 3-5 cm, which if necessary can be improved with higher quality sensors under good conditions. With static scans an accuracy of better than 1 cm for 3D points can be achieved surpassing the potential of mobile recording systems, which are economically much more efficient if larger areas have to be scanned. The 3D point clouds are the basis for object reconstruction in two different ways: a) engineering modelling as generalized CAD construction through geometric primitives and b) mesh modelling by triangulation of the point clouds for the exact representation of the surface. Deviations up to 10 cm (and possibly higher) from the nominal value can be created very quickly through the generalization in the CAD construction, but on the other side a significant reduction of data and a topological struc-turing can be achieved by fitting the point cloud into geometric primitives. However, investigations have shown that the number of polygons can be reduced to 25% and even 10% of the original data in the mesh triangulation using intelligent polygon decimation algorithms (e.g. curvature based) depending on the surface characteristic of the object, without having too much impact on the visual and geometric quality of the result. Depending on the object size, deviations of less than one milli-metre (e.g. for archaeological finds) up to 5 cm on average for larger objects can be achieved. In the future point clouds can form an important basis for the construction of the environment for many virtual reality applications, where the visual appearance is more important than the perfect geometric accuracy of the modelled objects
30
Schauer, Marin Rodrigues Johannes [Verfasser], and Andreas [Gutachter] Nüchter. "Detecting Changes and Finding Collisions in 3D Point Clouds / Johannes Schauer Marin Rodrigues ; Gutachter: Andreas Nüchter." Würzburg : Universität Würzburg, 2020. http://d-nb.info/1222910462/34.
Full text
31
SHAH, GHAZANFAR ALI. "Template-based reverse engineering of parametric CAD models from point clouds." Doctoral thesis, Università degli studi di Genova, 2021. http://hdl.handle.net/11567/1048640.
Full textAbstract:
Even if many Reverse Engineering techniques exist to reconstruct real objects in 3D, very few are able to deal directly and efficiently with the reconstruction of editable CAD models of assemblies of mechanical parts that can be used in the stages of Product Development Processes (PDP). In the absence of suitable segmentation tools, these approaches struggle to identify and reconstruct model the different parts that make up the assembly. The thesis aims to develop a new Reverse Engineering technique for the reconstruction of editable CAD models of mechanical parts’ assemblies. The originality lies in the use of a Simulated Annealing-based fitting technique optimization process that leverages a two-level filtering able to capture and manage the boundaries of the parts’ geometries inside the overall point cloud to allow for interface detection and local fitting of a part template to the point cloud. The proposed method uses various types of data (e.g. clouds of points, CAD models possibly stored in database together with the associated best parameter configurations for the fitting process). The approach is modular and integrates a sensitivity analysis to characterize the impact of the variations of the parameters of a CAD model on the evolution of the deviation between the CAD model itself and the point cloud to be fitted. The evaluation of the proposed approach is performed using both real scanned point clouds and as-scanned virtually generated point clouds which incorporate several artifacts that could appear with a real scanner. Results cover several Industry 4.0 related application scenarios, ranging from the global fitting of a single part to the update of a complete Digital Mock-Up embedding assembly constraints. The proposed approach presents good capacities to help maintaining the coherence between a product/system and its digital twin.
32
Manamasa, Krishna Himaja. "Domain adaptation from 3D synthetic images to real images." Thesis, Blekinge Tekniska Högskola, Institutionen för datavetenskap, 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:bth-19303.
Full textAbstract:
Background. Domain adaptation is described as, a model learning from a source data distribution and performing well on the target data. This concept, Domain adaptation is applied to assembly-line production tasks to perform an automatic quality inspection. Objectives. The aim of this master thesis is to apply this concept of 3D domain adaptation from synthetic images to real images. It is an attempt to bridge the gap between different domains (synthetic and real point cloud images), by implementing deep learning models that learn from synthetic 3D point cloud (CAD model images) and perform well on the actual 3D point cloud (3D Camera images). Methods. Through this course of thesis project, various methods for understand- ing the data and analyzing it for bridging the gap between CAD and CAM to make them similar is looked into. Literature review and controlled experiment are research methodologies followed during implementation. In this project, we experiment with four different deep learning models with data generated and compare their performance to know which deep learning model performs best for the data. Results. The results are explained through metrics i.e, accuracy and train time, which were the outcomes of each of the deep learning models after the experiment. These metrics are illustrated in the form of graphs for comparative analysis between the models on which the data is trained and tested on. PointDAN showed better results with higher accuracy compared to the other 3 models. Conclusions. The results attained show that domain adaptation for synthetic images to real images is possible with the data generated. PointDAN deep learning model which focuses on local feature alignment and global feature alignment with single-view point data shows better results with our data.
33
Borrmann, Dorit [Verfasser], Andreas [Gutachter] Nüchter, Joachim [Gutachter] Hertzberg, and Claus [Gutachter] Brenner. "Multi-modal 3D mapping - Combining 3D point clouds with thermal and color information / Dorit Borrmann ; Gutachter: Andreas Nüchter, Joachim Hertzberg, Claus Brenner." Würzburg : Universität Würzburg, 2018. http://d-nb.info/1152211501/34.
Full text
34
Borrmann, Dorit Verfasser], Andreas [Gutachter] Nüchter, Joachim [Gutachter] Hertzberg, and Claus [Gutachter] [Brenner. "Multi-modal 3D mapping - Combining 3D point clouds with thermal and color information / Dorit Borrmann ; Gutachter: Andreas Nüchter, Joachim Hertzberg, Claus Brenner." Würzburg : Universität Würzburg, 2018. http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-157085.
Full text
35
Richter, Rico [Verfasser], and Jürgen [Akademischer Betreuer] Döllner. "Concepts and techniques for processing and rendering of massive 3D point clouds / Rico Richter ; Betreuer: Jürgen Döllner." Potsdam : Universität Potsdam, 2018. http://d-nb.info/1217813020/34.
Full text
36
Orts-Escolano, Sergio. "A three-dimensional representation method for noisy point clouds based on growing self-organizing maps accelerated on GPUs." Doctoral thesis, Universidad de Alicante, 2013. http://hdl.handle.net/10045/36484.
Full textAbstract:
The research described in this thesis was motivated by the need of a robust model capable of representing 3D data obtained with 3D sensors, which are inherently noisy. In addition, time constraints have to be considered as these sensors are capable of providing a 3D data stream in real time. This thesis proposed the use of Self-Organizing Maps (SOMs) as a 3D representation model. In particular, we proposed the use of the Growing Neural Gas (GNG) network, which has been successfully used for clustering, pattern recognition and topology representation of multi-dimensional data. Until now, Self-Organizing Maps have been primarily computed offline and their application in 3D data has mainly focused on free noise models, without considering time constraints. It is proposed a hardware implementation leveraging the computing power of modern GPUs, which takes advantage of a new paradigm coined as General-Purpose Computing on Graphics Processing Units (GPGPU). The proposed methods were applied to different problem and applications in the area of computer vision such as the recognition and localization of objects, visual surveillance or 3D reconstruction.
37
Thomas, Hugues. "Apprentissage de nouvelles représentations pour la sémantisation de nuages de points 3D." Thesis, Paris Sciences et Lettres (ComUE), 2019. http://www.theses.fr/2019PSLEM048/document.
Full textAbstract:
Aujourd’hui, de nouvelles technologies permettent l’acquisition de scènes 3D volumineuses et précises sous la forme de nuages de points. Les nouvelles applications ouvertes par ces technologies, comme les véhicules autonomes ou la maintenance d'infrastructure, reposent sur un traitement efficace des nuages de points à grande échelle. Les méthodes d'apprentissage profond par convolution ne peuvent pas être utilisées directement avec des nuages de points. Dans le cas des images, les filtres convolutifs ont permis l’apprentissage de nouvelles représentations, jusqu’alors construites « à la main » dans les méthodes de vision par ordinateur plus anciennes. En suivant le même raisonnement, nous présentons dans cette thèse une étude des représentations construites « à la main » utilisées pour le traitement des nuages de points. Nous proposons ainsi plusieurs contributions, qui serviront de base à la conception d’une nouvelle représentation convolutive pour le traitement des nuages de points. Parmi elles, une nouvelle définition de voisinages sphériques multi-échelles, une comparaison avec les k plus proches voisins multi-échelles, une nouvelle stratégie d'apprentissage actif, la segmentation sémantique des nuages de points à grande échelle, et une étude de l'influence de la densité dans les représentations multi-échelles. En se basant sur ces contributions, nous introduisons la « Kernel Point Convolution » (KPConv), qui utilise des voisinages sphériques et un noyau défini par des points. Ces points jouent le même rôle que les pixels du noyau des convolutions en image. Nos réseaux convolutionnels surpassent les approches de segmentation sémantique de l’état de l’art dans presque toutes les situations. En plus de ces résultats probants, nous avons conçu KPConv avec une grande flexibilité et une version déformable. Pour conclure notre réflexion, nous proposons plusieurs éclairages sur les représentations que notre méthode est capable d'apprendreIn the recent years, new technologies have allowed the acquisition of large and precise 3D scenes as point clouds. They have opened up new applications like self-driving vehicles or infrastructure monitoring that rely on efficient large scale point cloud processing. Convolutional deep learning methods cannot be directly used with point clouds. In the case of images, convolutional filters brought the ability to learn new representations, which were previously hand-crafted in older computer vision methods. Following the same line of thought, we present in this thesis a study of hand-crafted representations previously used for point cloud processing. We propose several contributions, to serve as basis for the design of a new convolutional representation for point cloud processing. They include a new definition of multiscale radius neighborhood, a comparison with multiscale k-nearest neighbors, a new active learning strategy, the semantic segmentation of large scale point clouds, and a study of the influence of density in multiscale representations. Following these contributions, we introduce the Kernel Point Convolution (KPConv), which uses radius neighborhoods and a set of kernel points to play the role of the kernel pixels in image convolution. Our convolutional networks outperform state-of-the-art semantic segmentation approaches in almost any situation. In addition to these strong results, we designed KPConv with a great flexibility and a deformable version. To conclude our argumentation, we propose several insights on the representations that our method is able to learn
38
Leroy, Rémy. "Deep Learning methods for monocular 3D vision systems." Electronic Thesis or Diss., université Paris-Saclay, 2023. http://www.theses.fr/2023UPASG021.
Full textAbstract:
Dans cette thèse, nous étudions l'apport de l'apprentissage profond pour les systèmes de vision 3D monoculaire, de l'acquisition de l'image au traitement. Nous proposons d'abord Pix2Point, une méthode d'estimation de nuage de points 3D à partir d'une seule image en utilisant des informations de contexte, et entraînée avec une fonction de coût de transport optimal. Pix2Point réalise une meilleure couverture des scènes lorsqu'il est entraîné sur des nuages de points lacunaires que les méthodes d'estimation de profondeur monoculaire, entraînées sur des cartes de profondeur lacunaires. Deuxièmement, pour exploiter les indices de profondeur provenant du capteur, nous proposons une méthode de régression de profondeur à partir d'un patch défocalisé. Cette méthode surpasse la classification et la régression directe, sur données simulées et réelles. Enfin, nous abordons la conception d'un système de vision RVB-D, composé d'un capteur dont l'image est traitée par notre réseau de régression de profondeur basée sur la défocalisation et par un réseau de défloutage d'image. Nous proposons un cadre d'optimisation multi-tâches, conjointement aux paramètres des capteurs et des réseaux, et nous l'appliquons à l'optimisation de la mise au point d'une lentille chromatique. Le paysage d'optimisation présente plusieurs optima liés à la tâche de régression en profondeur, tandis que la tâche de défloutage semble moins sensible au paramètre de mise au point. En résumé, cette thèse propose plusieurs contributions exploitant les réseaux de neurones pour l'estimation 3D monoculaire et ouvre la voie d'une conception conjointe de systèmes RVB-DIn this thesis, we explore deep learning methods for monocular 3D vision systems, from image acquisition to processing. We first propose Pix2Point, a method for 3D point cloud prediction from a single image using context information, trained with an optimal transport loss. Pix2Point achieves a better coverage of the scenes when trained on sparse point clouds than monocular depth estimation methods, trained on sparse depth maps. Second, to exploit sensor depth cues, we propose a depth regression method from a defocused patch, which outperforms classification and direct regression, on simulated and real data. Finally, we tackle the design of a RGB-D monocular vision system for which the image is processed jointly by our defocus-based depth regression method and a simple image deblurring network. We propose an end-to-end multi-task optimisation framework of sensor and network parameters, that we apply to the focus optimisation for a chromatic lens. The optimisation landscape presents multiple optima, due to the depth regression task, while the deblurring task appears less sensitive to the focus. This thesis hence contains several contributions exploiting neural networks for monocular 3D estimation and paves the way towards end-to-end design of RGB-D systems
39
Fischer, Andreas, and Andreas Schäfer. "Untersuchungen zum mobilen 3D-Scannen unter Tage bei K+S." Technische Universitaet Bergakademie Freiberg Universitaetsbibliothek "Georgius Agricola", 2016. http://nbn-resolving.de/urn:nbn:de:bsz:105-qucosa-205693.
Full textAbstract:
Im Rahmen einer Diplomarbeit an der TU Bergakademie Freiberg wurden in 2014 die Grundlagen für die Auswertung von 3D-Punktwolken zur automatisierten Nachtragung des Risswerks gelegt. Um die dafür notwendigen 3D-Punktwolken möglichst wirtschaftlich zu erstellen, laufen seit 2015 Untersuchungen und Testmessungen zur Machbarkeit des untertägigen Einsatzes von mobil messenden Laserscannern. Im Folgenden werden verschiedene technische Ansätze sowie die Ergebnisse der Testmessungen und die weiteren geplanten Schritte vorgestelltAs part of a thesis at the Technical University of Freiberg, a basis for the analysis of 3D point clouds was set for refining the mine map automatically. Since 2015 studies and test measurements have been running to create the necessary 3D point clouds as economically as possible, by using an underground mobile scanning system. Below the different technical approaches will be presented as well as the results of the test measurements and the next planned steps
40
Hölscher, Phillip. "Deep Learning for estimation of fingertip location in 3-dimensional point clouds : An investigation of deep learning models for estimating fingertips in a 3D point cloud and its predictive uncertainty." Thesis, Linköpings universitet, Statistik och maskininlärning, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-176675.
Full textAbstract:
Sensor technology is rapidly developing and, consequently, the generation of point cloud data is constantly increasing. Since the recent release of PointNet, it is possible to process this unordered 3-dimensional data directly in a neural network. The company TLT Screen AB, which develops cutting-edge tracking technology, seeks to optimize the localization of the fingertips of a hand in a point cloud. To do so, the identification of relevant 3D neural network models for modeling hands and detection of fingertips in various hand orientations is essential. The Hand PointNet processes point clouds of hands directly and generate estimations of fixed points (joints), including fingertips, of the hands. Therefore, this model was selected to optimize the localization of fingertips for TLT Screen AB and forms the subject of this research. The model has advantages over conventional convolutional neural networks (CNN). First of all, in contrast to the 2D CNN, the Hand PointNet can use the full 3-dimensional spatial information. Compared to the 3D CNN, moreover, it avoids unnecessarily voluminous data and enables more efficient learning. The model was trained and evaluated on the public dataset MRSA Hand. In contrast to previously published work, the main object of this investigation is the estimation of only 5 joints, for the fingertips. The behavior of the model with a reduction from the usual 21 to 11 and only 5 joints are examined. It is found that the reduction of joints contributed to an increase in the mean error of the estimated joints. Furthermore, the examination of the distribution of the residuals of the estimate for fingertips is found to be less dense. MC dropout to study the prediction uncertainty for the fingertips has shown that the uncertainty increases when the joints are decreased. Finally, the results show that the uncertainty is greatest for the prediction of the thumb tip. Starting from the tip of the thumb, it is observed that the uncertainty of the estimates decreases with each additional fingertip.
41
Stella, Federico. "Learning a Local Reference Frame for Point Clouds using Spherical CNNs." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2020. http://amslaurea.unibo.it/20197/.
Full textAbstract:
Uno dei problemi più importanti della 3D Computer Vision è il cosiddetto surface matching, che consiste nel trovare corrispondenze tra oggetti tridimensionali. Attualmente il problema viene affrontato calcolando delle feature locali e compatte, chiamate descrittori, che devono essere riconosciute e messe in corrispondenza al mutare della posa dell'oggetto nello spazio, e devono quindi essere invarianti rispetto all'orientazione. Il metodo più usato per ottenere questa proprietà consiste nell'utilizzare dei Local Reference Frame (LRF): sistemi di coordinate locali che forniscono un'orientazione canonica alle porzioni di oggetti 3D che vengono usate per calcolare i descrittori. In letteratura esistono diversi modi per calcolare gli LRF, ma fanno tutti uso di algoritmi progettati manualmente. Vi è anche una recente proposta che utilizza reti neurali, tuttavia queste vengono addestrate mediante feature specificamente progettate per lo scopo, il che non permette di sfruttare pienamente i benefici delle moderne strategie di end-to-end learning.
Lo scopo di questo lavoro è utilizzare un approccio data-driven per far imparare a una rete neurale il calcolo di un Local Reference Frame a partire da point cloud grezze, producendo quindi il primo esempio di end-to-end learning applicato alla stima di LRF. Per farlo, sfruttiamo una recente innovazione chiamata Spherical Convolutional Neural Networks, le quali generano e processano segnali nello spazio SO(3) e sono quindi naturalmente adatte a rappresentare e stimare orientazioni e LRF. Confrontiamo le prestazioni ottenute con quelle di metodi esistenti su benchmark standard, ottenendo risultati promettenti.
42
El, Sayed Abdul Rahman. "Traitement des objets 3D et images par les méthodes numériques sur graphes." Thesis, Normandie, 2018. http://www.theses.fr/2018NORMLH19/document.
Full textAbstract:
La détection de peau consiste à détecter les pixels correspondant à une peau humaine dans une image couleur. Les visages constituent une catégorie de stimulus importante par la richesse des informations qu’ils véhiculent car avant de reconnaître n’importe quelle personne il est indispensable de localiser et reconnaître son visage. La plupart des applications liées à la sécurité et à la biométrie reposent sur la détection de régions de peau telles que la détection de visages, le filtrage d'objets 3D pour adultes et la reconnaissance de gestes. En outre, la détection de la saillance des mailles 3D est une phase de prétraitement importante pour de nombreuses applications de vision par ordinateur. La segmentation d'objets 3D basée sur des régions saillantes a été largement utilisée dans de nombreuses applications de vision par ordinateur telles que la correspondance de formes 3D, les alignements d'objets, le lissage de nuages de points 3D, la recherche des images sur le web, l’indexation des images par le contenu, la segmentation de la vidéo et la détection et la reconnaissance de visages. La détection de peau est une tâche très difficile pour différentes raisons liées en général à la variabilité de la forme et la couleur à détecter (teintes différentes d’une personne à une autre, orientation et tailles quelconques, conditions d’éclairage) et surtout pour les images issues du web capturées sous différentes conditions de lumière. Il existe plusieurs approches connues pour la détection de peau : les approches basées sur la géométrie et l’extraction de traits caractéristiques, les approches basées sur le mouvement (la soustraction de l’arrière-plan (SAP), différence entre deux images consécutives, calcul du flot optique) et les approches basées sur la couleur. Dans cette thèse, nous proposons des méthodes d'optimisation numérique pour la détection de régions de couleurs de peaux et de régions saillantes sur des maillages 3D et des nuages de points 3D en utilisant un graphe pondéré. En se basant sur ces méthodes, nous proposons des approches de détection de visage 3D à l'aide de la programmation linéaire et de fouille de données (Data Mining). En outre, nous avons adapté nos méthodes proposées pour résoudre le problème de la simplification des nuages de points 3D et de la correspondance des objets 3D. En plus, nous montrons la robustesse et l’efficacité de nos méthodes proposées à travers de différents résultats expérimentaux réalisés. Enfin, nous montrons la stabilité et la robustesse de nos méthodes par rapport au bruitSkin detection involves detecting pixels corresponding to human skin in a color image. The faces constitute a category of stimulus important by the wealth of information that they convey because before recognizing any person it is essential to locate and recognize his face. Most security and biometrics applications rely on the detection of skin regions such as face detection, 3D adult object filtering, and gesture recognition. In addition, saliency detection of 3D mesh is an important pretreatment phase for many computer vision applications. 3D segmentation based on salient regions has been widely used in many computer vision applications such as 3D shape matching, object alignments, 3D point-point smoothing, searching images on the web, image indexing by content, video segmentation and face detection and recognition. The detection of skin is a very difficult task for various reasons generally related to the variability of the shape and the color to be detected (different hues from one person to another, orientation and different sizes, lighting conditions) and especially for images from the web captured under different light conditions. There are several known approaches to skin detection: approaches based on geometry and feature extraction, motion-based approaches (background subtraction (SAP), difference between two consecutive images, optical flow calculation) and color-based approaches. In this thesis, we propose numerical optimization methods for the detection of skins color and salient regions on 3D meshes and 3D point clouds using a weighted graph. Based on these methods, we provide 3D face detection approaches using Linear Programming and Data Mining. In addition, we adapted our proposed methods to solve the problem of simplifying 3D point clouds and matching 3D objects. In addition, we show the robustness and efficiency of our proposed methods through different experimental results. Finally, we show the stability and robustness of our methods with respect to noise
43
Bobkov, Dmytro [Verfasser], Eckehard [Akademischer Betreuer] Steinbach, Eckehard [Gutachter] Steinbach, and Klaus [Gutachter] Diepold. "Semantic understanding of 3D point clouds of indoor environments / Dmytro Bobkov ; Gutachter: Eckehard Steinbach, Klaus Diepold ; Betreuer: Eckehard Steinbach." München : Universitätsbibliothek der TU München, 2019. http://d-nb.info/1196791856/34.
Full text
44
Huang, Rong [Verfasser], Uwe [Akademischer Betreuer] Stilla, Helmut [Gutachter] Mayer, and Uwe [Gutachter] Stilla. "Change detection of construction sites based on 3D point clouds / Rong Huang ; Gutachter: Helmut Mayer, Uwe Stilla ; Betreuer: Uwe Stilla." München : Universitätsbibliothek der TU München, 2021. http://d-nb.info/1240832850/34.
Full text
45
Jack, Dominic. "Deep learning approaches for 3D inference from monocular vision." Thesis, Queensland University of Technology, 2020. https://eprints.qut.edu.au/204267/1/Dominic_Jack_Thesis.pdf.
Full textAbstract:
This thesis looks at deep learning approaches to 3D computer vision problems, using representations including occupancy grids, deformable meshes, key points, point clouds, and event streams. We focussed on methods targeted towards medium-sized mobile robotics platforms with modest computational power on board. Key results include state-of-the-art accuracies on single-view high resolution voxel reconstruction and event camera classification tasks, point cloud convolution networks capable of performing inference an order of magnitude faster than similar methods, and a 3D human pose lifting model with significantly fewer floating point operations and learnable weights than baseline deep learning methods.
46
Birdal, Tolga [Verfasser], Slobodan [Akademischer Betreuer] Ilic, Darius [Gutachter] Burschka, Yasutaka [Gutachter] Furukawa, and Slobodan [Gutachter] Ilic. "Geometric Methods for 3D Reconstruction from Large Point Clouds / Tolga Birdal ; Gutachter: Darius Burschka, Yasutaka Furukawa, Slobodan Ilic ; Betreuer: Slobodan Ilic." München : Universitätsbibliothek der TU München, 2018. http://d-nb.info/1179914082/34.
Full text
47
Fernandes, maligo Artur otavio. "Unsupervised Gaussian mixture models for the classification of outdoor environments using 3D terrestrial lidar data." Thesis, Toulouse, INSA, 2016. http://www.theses.fr/2016ISAT0053/document.
Full textAbstract:
Le traitement de nuages de points 3D de lidars permet aux robots mobiles autonomes terrestres de construire des modèles sémantiques de l'environnement extérieur dans lequel ils évoluent. Ces modèles sont intéressants car ils représentent des informations qualitatives, et ainsi donnent à un robot la capacité de raisonner à un niveau plus élevé d'abstraction. Le coeur d'un système de modélisation sémantique est la capacité de classifier les observations venant du capteur. Nous proposons un système de classification centré sur l'apprentissage non-supervisé. La prémière couche, la couche intermédiaire, consiste en un modèle de mélange gaussien. Ce modèle est déterminé de manière non-supervisée lors d'une étape de training. Il definit un ensemble de classes intermédiaires qui correspond à une partition fine des classes présentes dans l'environnement. La deuxième couche, la couche finale, consiste en un regroupement des classes intermédiaires dans un ensemble de classes finales qui, elles, sont interprétables dans le contexte de la tâche ciblée. Le regroupement est déterminé par un expert lors de l'étape de training, de manière supervisée, mais guidée par les classes intermédiaires. L'évaluation est basée sur deux jeux de données acquis avec de différents lidars et possédant différentes caractéristiques. L'évaluation est quantitative pour l'un des jeux de données, et qualitative pour l'autre. La concéption du système utilise la procédure standard de l'apprentissage, basée sur les étapes de training, validation et test. L'opération suit la pipeline standard de classification. Le système est simple, et ne requiert aucun pré-traitement ou post-traitementThe processing of 3D lidar point clouds enable terrestrial autonomous mobile robots to build semantic models of the outdoor environments in which they operate. Such models are interesting because they encode qualitative information, and thus provide to a robot the ability to reason at a higher level of abstraction. At the core of a semantic modelling system, lies the capacity to classify the sensor observations. We propose a two-layer classi- fication model which strongly relies on unsupervised learning. The first, intermediary layer consists of a Gaussian mixture model. This model is determined in a training step in an unsupervised manner, and defines a set of intermediary classes which is a fine-partitioned representation of the environment. The second, final layer consists of a grouping of the intermediary classes into final classes that are interpretable in a considered target task. This grouping is determined by an expert during the training step, in a process which is supervised, yet guided by the intermediary classes. The evaluation is done for two datasets acquired with different lidars and possessing different characteristics. It is done quantitatively using one of the datasets, and qualitatively using another. The system is designed following the standard learning procedure, based on a training, a validation and a test steps. The operation follows a standard classification pipeline. The system is simple, with no requirement of pre-processing or post-processing stages
48
Contreras, Samamé Luis Federico. "SLAM collaboratif dans des environnements extérieurs." Thesis, Ecole centrale de Nantes, 2019. http://www.theses.fr/2019ECDN0012/document.
Full textAbstract:
Cette thèse propose des modèles cartographiques à grande échelle d'environnements urbains et ruraux à l'aide de données en 3D acquises par plusieurs robots. La mémoire contribue de deux manières principales au domaine de recherche de la cartographie. La première contribution est la création d'une nouvelle structure, CoMapping, qui permet de générer des cartes 3D de façon collaborative. Cette structure s’applique aux environnements extérieurs en ayant une approche décentralisée. La fonctionnalité de CoMapping comprend les éléments suivants : Tout d’abord, chaque robot réalise la construction d'une carte de son environnement sous forme de nuage de points.Pour cela, le système de cartographie a été mis en place sur des ordinateurs dédiés à chaque voiture, en traitant les mesures de distance à partir d'un LiDAR 3D se déplaçant en six degrés de liberté (6-DOF). Ensuite, les robots partagent leurs cartes locales et fusionnent individuellement les nuages de points afin d'améliorer leur estimation de leur cartographie locale. La deuxième contribution clé est le groupe de métriques qui permettent d'analyser les processus de fusion et de partage de cartes entre les robots. Nous présentons des résultats expérimentaux en vue de valider la structure CoMapping et ses métriques. Tous les tests ont été réalisés dans des environnements extérieurs urbains du campus de l’École Centrale de Nantes ainsi que dans des milieux rurauxThis thesis proposes large-scale mapping model of urban and rural environments using 3D data acquired by several robots. The work contributes in two main ways to the research field of mapping. The first contribution is the creation of a new framework, CoMapping, which allows to generate 3D maps in a cooperative way. This framework applies to outdoor environments with a decentralized approach. The CoMapping's functionality includes the following elements: First of all, each robot builds a map of its environment in point cloud format.To do this, the mapping system was set up on computers dedicated to each vehicle, processing distance measurements from a 3D LiDAR moving in six degrees of freedom (6-DOF). Then, the robots share their local maps and merge the point clouds individually to improve their local map estimation. The second key contribution is the group of metrics that allow to analyze the merging and card sharing processes between robots. We present experimental results to validate the CoMapping framework with their respective metrics. All tests were carried out in urban outdoor environments on the surrounding campus of the École Centrale de Nantes as well as in rural areas
49
Deng, Haowen [Verfasser], Slobodan [Akademischer Betreuer] Ilic, Stefano Luigi [Gutachter] Di, and Slobodan [Gutachter] Ilic. "Learned 3D Local Features for Rigid Pose Estimation on Point Clouds / Haowen Deng ; Gutachter: Luigi Di Stefano, Slobodan Ilic ; Betreuer: Slobodan Ilic." München : Universitätsbibliothek der TU München, 2020. http://d-nb.info/1223616924/34.
Full text
50
Schilling, Anita. "Automatic Retrieval of Skeletal Structures of Trees from Terrestrial Laser Scanner Data." Doctoral thesis, Saechsische Landesbibliothek- Staats- und Universitaetsbibliothek Dresden, 2014. http://nbn-resolving.de/urn:nbn:de:bsz:14-qucosa-155698.
Full textAbstract:
Research on forest ecosystems receives high attention, especially nowadays with regard to sustainable management of renewable resources and the climate change. In particular, accurate information on the 3D structure of a tree is important for forest science and bioclimatology, but also in the scope of commercial applications. Conventional methods to measure geometric plant features are labor- and time-intensive. For detailed analysis, trees have to be cut down, which is often undesirable. Here, Terrestrial Laser Scanning (TLS) provides a particularly attractive tool because of its contactless measurement technique. The object geometry is reproduced as a 3D point cloud. The objective of this thesis is the automatic retrieval of the spatial structure of trees from TLS data. We focus on forest scenes with comparably high stand density and with many occlusions resulting from it. The varying level of detail of TLS data poses a big challenge. We present two fully automatic methods to obtain skeletal structures from scanned trees that have complementary properties. First, we explain a method that retrieves the entire tree skeleton from 3D data of co-registered scans. The branching structure is obtained from a voxel space representation by searching paths from branch tips to the trunk. The trunk is determined in advance from the 3D points. The skeleton of a tree is generated as a 3D line graph. Besides 3D coordinates and range, a scan provides 2D indices from the intensity image for each measurement. This is exploited in the second method that processes individual scans. Furthermore, we introduce a novel concept to manage TLS data that facilitated the researchwork. Initially, the range image is segmented into connected components. We describe a procedure to retrieve the boundary of a component that is capable of tracing inner depth discontinuities. A 2D skeleton is generated from the boundary information and used to decompose the component into sub components. A Principal Curve is computed from the 3D point set that is associated with a sub component. The skeletal structure of a connected component is summarized as a set of polylines. Objective evaluation of the results remains an open problem because the task itself is ill-defined: There exists no clear definition of what the true skeleton should be w.r.t. a given point set. Consequently, we are not able to assess the correctness of the methods quantitatively, but have to rely on visual assessment of results and provide a thorough discussion of the particularities of both methods. We present experiment results of both methods. The first method efficiently retrieves full skeletons of trees, which approximate the branching structure. The level of detail is mainly governed by the voxel space and therefore, smaller branches are reproduced inadequately. The second method retrieves partial skeletons of a tree with high reproduction accuracy. The method is sensitive to noise in the boundary, but the results are very promising. There are plenty of possibilities to enhance the method’s robustness. The combination of the strengths of both presented methods needs to be investigated further and may lead to a robust way to obtain complete tree skeletons from TLS data automaticallyDie Erforschung des ÖkosystemsWald spielt gerade heutzutage im Hinblick auf den nachhaltigen Umgang mit nachwachsenden Rohstoffen und den Klimawandel eine große Rolle. Insbesondere die exakte Beschreibung der dreidimensionalen Struktur eines Baumes ist wichtig für die Forstwissenschaften und Bioklimatologie, aber auch im Rahmen kommerzieller Anwendungen. Die konventionellen Methoden um geometrische Pflanzenmerkmale zu messen sind arbeitsintensiv und zeitaufwändig. Für eine genaue Analyse müssen Bäume gefällt werden, was oft unerwünscht ist. Hierbei bietet sich das Terrestrische Laserscanning (TLS) als besonders attraktives Werkzeug aufgrund seines kontaktlosen Messprinzips an. Die Objektgeometrie wird als 3D-Punktwolke wiedergegeben. Basierend darauf ist das Ziel der Arbeit die automatische Bestimmung der räumlichen Baumstruktur aus TLS-Daten. Der Fokus liegt dabei auf Waldszenen mit vergleichsweise hoher Bestandesdichte und mit zahlreichen daraus resultierenden Verdeckungen. Die Auswertung dieser TLS-Daten, die einen unterschiedlichen Grad an Detailreichtum aufweisen, stellt eine große Herausforderung dar. Zwei vollautomatische Methoden zur Generierung von Skelettstrukturen von gescannten Bäumen, welche komplementäre Eigenschaften besitzen, werden vorgestellt. Bei der ersten Methode wird das Gesamtskelett eines Baumes aus 3D-Daten von registrierten Scans bestimmt. Die Aststruktur wird von einer Voxelraum-Repräsentation abgeleitet indem Pfade von Astspitzen zum Stamm gesucht werden. Der Stamm wird im Voraus aus den 3D-Punkten rekonstruiert. Das Baumskelett wird als 3D-Liniengraph erzeugt. Für jeden gemessenen Punkt stellt ein Scan neben 3D-Koordinaten und Distanzwerten auch 2D-Indizes zur Verfügung, die sich aus dem Intensitätsbild ergeben. Bei der zweiten Methode, die auf Einzelscans arbeitet, wird dies ausgenutzt. Außerdem wird ein neuartiges Konzept zum Management von TLS-Daten beschrieben, welches die Forschungsarbeit erleichtert hat. Zunächst wird das Tiefenbild in Komponenten aufgeteilt. Es wird eine Prozedur zur Bestimmung von Komponentenkonturen vorgestellt, die in der Lage ist innere Tiefendiskontinuitäten zu verfolgen. Von der Konturinformation wird ein 2D-Skelett generiert, welches benutzt wird um die Komponente in Teilkomponenten zu zerlegen. Von der 3D-Punktmenge, die mit einer Teilkomponente assoziiert ist, wird eine Principal Curve berechnet. Die Skelettstruktur einer Komponente im Tiefenbild wird als Menge von Polylinien zusammengefasst. Die objektive Evaluation der Resultate stellt weiterhin ein ungelöstes Problem dar, weil die Aufgabe selbst nicht klar erfassbar ist: Es existiert keine eindeutige Definition davon was das wahre Skelett in Bezug auf eine gegebene Punktmenge sein sollte. Die Korrektheit der Methoden kann daher nicht quantitativ beschrieben werden. Aus diesem Grund, können die Ergebnisse nur visuell beurteiltwerden. Weiterhinwerden die Charakteristiken beider Methoden eingehend diskutiert. Es werden Experimentresultate beider Methoden vorgestellt. Die erste Methode bestimmt effizient das Skelett eines Baumes, welches die Aststruktur approximiert. Der Detaillierungsgrad wird hauptsächlich durch den Voxelraum bestimmt, weshalb kleinere Äste nicht angemessen reproduziert werden. Die zweite Methode rekonstruiert Teilskelette eines Baums mit hoher Detailtreue. Die Methode reagiert sensibel auf Rauschen in der Kontur, dennoch sind die Ergebnisse vielversprechend. Es gibt eine Vielzahl von Möglichkeiten die Robustheit der Methode zu verbessern. Die Kombination der Stärken von beiden präsentierten Methoden sollte weiter untersucht werden und kann zu einem robusteren Ansatz führen um vollständige Baumskelette automatisch aus TLS-Daten zu generieren