Dissertations / Theses on the topic 'Reconnaissance de gestes en temps réel'
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 'Reconnaissance de gestes en temps réel.'
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.
Barnachon, Mathieu. "Reconnaissance d'actions en temps réel à partir d'exemples." Phd thesis, Université Claude Bernard - Lyon I, 2013. http://tel.archives-ouvertes.fr/tel-00820113.
Full textCoupeté, Eva. "Reconnaissance de gestes et actions pour la collaboration homme-robot sur chaîne de montage." Thesis, Paris Sciences et Lettres (ComUE), 2016. http://www.theses.fr/2016PSLEM062/document.
Full textCollaborative robots are becoming more and more present in our everyday life. In particular, within the industrial environment, they emerge as one of the preferred solution to make assembly line in factories more flexible, cost-effective and to reduce the hardship of the operators’ work. However, to enable a smooth and efficient collaboration, robots should be able to understand their environment and in particular the actions of the humans around them.With this aim in mind, we decided to study technical gestures recognition. Specifically, we want the robot to be able to synchronize, adapt its speed and understand if something unexpected arises.We considered two use-cases, one dealing with copresence, the other with collaboration. They are both inspired by existing task on automotive assembly lines.First, for the co-presence use case, we evaluated the feasibility of technical gestures recognition using inertial sensors. We obtained a very good result (96% of correct recognition with one operator) which encouraged us to follow this idea.On the collaborative use-case, we decided to focus on non-intrusive sensors to minimize the disturbance for the operators and we chose to use a depth-camera. We filmed the operators with a top view to prevent most of the potential occultations.We introduce an algorithm that tracks the operator’s hands by calculating the geodesic distances between the points of the upper body and the top of the head.We also design and evaluate an approach based on discrete Hidden Markov Models (HMM) taking the hand positions as an input to recognize technical gestures. We propose a method to adapt our system to new operators and we embedded inertial sensors on tools to refine our results. We obtain the very good result of 90% of correct recognition in real time for 13 operators.Finally, we formalize and detail a complete methodology to realize technical gestures recognition on assembly lines
Cassel, Ryan. "Analyse du mouvement humain par un système de vision : une approche globale pour l'analyse et la reconnaissance en temps réel de mouvements acrobatiques." Paris 11, 2005. http://www.theses.fr/2005PA112282.
Full textAcrobatics is an area of athletics that is exacting in terms of gesture analysis and recognition. It comprises body rotations along two separate axes that result in fast and complex movements. Acrobatics intervene in multiple disciplines such as gymnastics, trampoline, ski, and diving. Devices for capturing movements based on markers and multiple cameras for movement analysis are problematic to implement in the context of training and are not exploitable in competition. A single camera may be used but the movement's complexity makes it very difficult to use traditional machine vision techniques without markers to carry out the analysis. Our approach offers a monocular system of analysis and recognition of acrobatic movements in real time, based on global measurements. Information relating to the acrobat's movements-without identifying specific body parts-constitutes our global measurements. Thus, we have developed movement models based on acrobatics characteristics and on global measurements extracted from image sequences. Moreover, we present a system capable of analyzing acrobatic movements with a view toward improvements of athletic performance, or for identifying the performance level of an acrobat. Analysis and recognition are based on measures of the movements identified by extracting and tracking the acrobat
Granger, Nicolas. "Deep-learning for high dimensional sequential observations : application to continuous gesture recognition." Thesis, Université Paris-Saclay (ComUE), 2019. http://www.theses.fr/2019SACLL002/document.
Full textThis thesis aims to improve the intuitiveness of human-computer interfaces. In particular, machines should try to replicate human's ability to process streams of information continuously. However, the sub-domain of Machine Learning dedicated to recognition on time series remains barred by numerous challenges. Our studies use gesture recognition as an exemplar application, gestures intermix static body poses and movements in a complex manner using widely different modalities. The first part of our work compares two state-of-the-art temporal models for continuous sequence recognition, namely Hybrid Neural Network--Hidden Markov Models (NN-HMM) and Bidirectional Recurrent Neural Networks (BDRNN) with gated units. To do so, we reimplemented the two within a shared test-bed which is more amenable to a fair comparative work. We propose adjustments to Neural Network training losses and the Hybrid NN-HMM expressions to accommodate for highly imbalanced data classes. Although recent publications tend to prefer BDRNNs, we demonstrate that Hybrid NN-HMM remain competitive. However, the latter rely significantly on their input layers to model short-term patterns. Finally, we show that input representations learned via both approaches are largely inter-compatible. The second part of our work studies one-shot learning, which has received relatively little attention so far, in particular for sequential inputs such as gestures. We propose a model built around a Bidirectional Recurrent Neural Network. Its effectiveness is demonstrated on the recognition of isolated gestures from a sign language lexicon. We propose several improvements over this baseline by drawing inspiration from related works and evaluate their performances, exhibiting different advantages and disadvantages for each
Poupet, Victor. "Automates cellulaires : temps réel et voisinages." Lyon, École normale supérieure (sciences), 2006. http://www.theses.fr/2006ENSL0390.
Full textIn this thesis we have worked on the impact of the choice of a neighborhood on the algorithmic abilities of cellular automata. We have specifically studied the lower complexity classes such as the real time (that corresponds to the shortest time necessary for a cellular automaton to read all the letters of the input word) and the real time plus a constant. It is indeed known that neighborhoods are equivalent in linear time and it is therefore necessary to consider shorter times. We have obtained neighborhood equivalence results with respect to the real time (neighborhood classes such that cellular automata working on any of those neighborhoods can recognize the same languages in real time) and linear or constant speed-up theorems for many classes of neighborhoods
Grandjean, Anaël. "Reconnaissance de langage en temps réel sur automates cellulaires 2D." Thesis, Montpellier, 2016. http://www.theses.fr/2016MONTT331/document.
Full textCellular automata were introduced in the 50s by J. von Neumann and S. Ulamas an efficient way of modeling massively parallel computation. Many variations of the model can be considered such as varying the dimension of the computation space or the communication capabilities of the computing cells. In a cellular automaton each cell can communicate only with a finite number of other cells called its neighbors. My work focuses on the impact of the choice of the neighbors on the algorithmic properties of the model. My first goal was to generalize some classical properties of computation models to the widest possible class of neighborhoods, in particular I prove a linear speedup theorem for any two dimensional neighborhood. I then study the difference between the complexity classes defined by different neighborhoods, show the existence of neighborhoods defining incomparable classes, and some sets of neighborhoods defining identical classes. Finally, I also discuss the impact of the dimension of the automata on their computational power
Malasné, Nicolas. "Localisation et reconnaissance de visages en temps réel : algorithmes et architectures." Dijon, 2002. http://www.theses.fr/2002DIJOS045.
Full textBorello, Alex. "Reconnaissance de langages en temps réel par des automates cellulaires avec contraintes." Thesis, Aix-Marseille 1, 2011. http://www.theses.fr/2011AIX10127.
Full textThis document deals with cellular automata as a model of computation used to recognise languages. In such a domain, it is always difficult to provide negative results, that is, typically, to prove that a given language is not recognised in some function of time by some class of automata. The document focuses in particular on the low-complexity classes such as real time, about which a lot of questions remain open since several decades.In a first part, several techniques to weaken further still these classes of languages are investigated, thereby bringing examples of negative results. A second part is dedicated to the comparison of cellular automata with another model language recognition, namely multi-head finite automata. This leads to speed-up theorem when finite automata are oblivious, which makes them a priori weaker than in the general case but leaves them a nontrivial power
Lahaye, Jean-Claude. "Etude et réalisation d'un système de vision temps réel par reconnaissance d'éléments rectilignes." Grenoble INPG, 1986. http://www.theses.fr/1986INPG0129.
Full textMénier, Clément. "Système de vision temps-réel pour les intéractions." Grenoble INPG, 2007. http://www.theses.fr/2007INPG0041.
Full textThis thesis focuses on the the real time acquisition of 3D information on a scene from multiple camera in the context of interactive applications. A complete vision system from image acquisition to motion and shape modeling is presented. The distribution of tasks on a PC cluster, and more precisely the parallelization of different shape modeling algorithms, enables a real time execution with a low latency. Several applications are developped and validate the practical implementation of this system. An original approach of motion modeling is lso presented. It allows for limbs tracking and identification white not requiring prior information on the shape of the user
Dupont, Marc. "Reconnaissance gestuelle par gant de données pour le contrôle temps réel d’un robot mobile." Thesis, Lorient, 2017. http://www.theses.fr/2017LORIS437/document.
Full textAlthough gesture recognition has been studied for several decades, much research stays in the realm of indoors laboratory experiments. In this thesis, we address the problem of designing a truly usable, real- world gesture recognition system, focusing mainly on the real-time control of an outdoors robot for use by military soldiers. The main contribution of this thesis is the development of a real-time gesture recognition pipeline, which can be taught in a few minutes with: very sparse input ("small data"); freely user-invented gestures; resilience to user mistakes during training; and low computation requirements. This is achieved thanks to two key innovations: first, a stream-enabled, DTW-inspired technique to compute distances between time series; and second, an efficient stream history analysis procedure to automatically determine model hyperparameters without user intervention. Additionally, a custom, hardened data glove was built and used to demonstrate successful gesture recognition and real-time robot control. We finally show this work's flexibility by furthermore using it beyond robot control to drive other kinds of controllable systems
Zouari-Ben, Letaïfa Leila. "Vers le temps réel en transcription automatique de la parole grand vocabulaire." Paris,ENST, 2007. http://www.theses.fr/2007ENST0024.
Full textLarge vocabulary speech recognition systems based on hidden Markov models (HMM) make use of many tens of thousands of Gaussian distributions to achieve improved recognition. Hence, the computation of the state likelihood is time consuming. As the performance and the speed of such recognition systems are closely related to the number of HMM Gaussians, reducing the number of Gaussians without decreasing the system performance is of major interest. Assuming that only a few Gaussians dominate the state likelihood, Gaussian selection techniques have been developed to detect them. These techniques are based on classification and can be divided into two categories: state and model based methods. In order to improve the state-based Gaussian selection we propose an original clustering and a multi-level Gaussian selection. The clustering algorithm use a new Gaussian similarity distance. In model based methods the classification is applied to the Gaussian distributions of all the models. The contextuel information is lost du to merging distributions of the different contexts. So we introduce a contextual Gaussian selection. In recent years, as an alternative to the Gaussian selection, sub-vector quantization was successfully used to reduce the acoustic models complexity. Unfortunally, these techniques make use of the classification by merging different contexts. Hence we investigate a contextual sub-vector quantization. The proposed algorithms are evaluated within a framework of large vocabulary continuous speech recognition. Results outperform some existing methods
Petreto, Andrea. "Débruitage vidéo temps réel pour systèmes embarqués." Electronic Thesis or Diss., Sorbonne université, 2020. http://www.theses.fr/2020SORUS060.
Full textIn many applications, noisy video can be a major problem. There are denoising methods with highly effective denoising capabilities but at the cost of a very high computational complexity. Other faster methods are limited in their applications since they does not handle high levels of noise correctly. For many applications, it is however very important to preserve a good image quality in every situation with sometimes strong embedding constraints. In this work, the goal is to propose an embedded solution for live video denoising. The method needs to remain efficient with even under high level of noise. We limit our work to embedded CPU under 30W of power consumption. This work led to a new video denoising algorithm called RTE-VD: Real-Time Embedded Video Denoising. RTE-VD is composed of 3 steps: stabilization, movement compensation by dense optical flow estimation and spatio-temporal filtering. On an embedded CPU (Jetson AGX), RTE-VD runs at 30 frame per seconds on qHD videos (960x580 pixels). In order to achieve such performance, many compromises and optimizations had to be done. We compare RTE-VD to other state-of-the-art methods in both terms of denoising capabilities and processing time. We show that RTE-VD brings a new relevant tradeoff between quality and speed
Brazey, Denis. "Reconnaissance de formes et suivi de mouvements en 4D temps-réel : Restauration de cartes de profondeur." Thesis, Rouen, INSA, 2014. http://www.theses.fr/2014ISAM0019.
Full textIn this dissertation, we are interested in several issues related to 3D data processing. The first one concerns people detection and tracking in depth map sequences. We propose an improvement of an existing method based on a segmentation stage followed by a tracking module. The second issue is head detection and modelling in 3D point clouds. In order to do this, we adopt a probabilistic approach based on a new spherical mixture model. The last considered application deals with the restoration of deteriorated depth maps. To solve this problem, we propose to use a surface approximation method based on interpolation Dm-splines with scale transforms to approximate and restore the image. Presented results illustrate the efficiency of the developed algorithms
Pétillot, Y. "Vers une implantation de corrélateurs optiques temps réel : applications a divers problèmes de reconnaissance des formes." Brest, 1996. http://www.theses.fr/1996BRES2015.
Full textTancrez, Pierre-Marie. "Un système de vision linéaire pour l'inspection et la caractérisation en temps réel de produits plats en défilement continu." Lille 1, 1989. http://www.theses.fr/1989LIL10119.
Full textBordieu, Christophe. "Utilisation des réseaux de neurones artificiels pour la détection et la reconnaissance des gaz en temps réel." Bordeaux 1, 1998. http://www.theses.fr/1998BOR10562.
Full textRamaux, Nicolas. "Supervision de systèmes dynamiques par reconnaissance de scénarios." Compiègne, 1998. http://www.theses.fr/1998COMP1126.
Full textBennegadi, Karim. "Étude et réalisation d'opérateurs câblés pour le traitement d'images en temps réel." Compiègne, 1990. http://www.theses.fr/1990COMPD263.
Full textIn order to facilitate the realisation of the hardware processing units in image analysis and pattern recognition, an autonomous environment was designed. To group these hardware units in a single system, to ease their extension and to normalize them for modularity purpose, a hardware image bus was designed, creating the link between different image sources as video digitizing cards or computer buses, and the set of hardware processing units. A full software package that consists in two libraries, complete the hardware part. The first library presents an enhanced user interface that can overlay the image, and the second is a software toolbox used to ease new routines integration, user interface extension and provides a methodology to translate algorithms from software form to hardware form. The hardware bus based part, provides a high data transfer rate using synchronous mode, and the software part is written in C language for modularity and portability purposes
Minault, Sophie. "Sibylle : un système expert pour l'interprétation en temps réel de signaux monodimensionnels : application au signal vocal." Paris 11, 1987. http://www.theses.fr/1987PA112099.
Full textDousson, Christophe. "Suivi d'évolutions et reconnaissance de chroniques." Toulouse 3, 1994. http://www.theses.fr/1994TOU30264.
Full textFilion, Julien. "Reconnaissance de plan probabiliste par exploration partielle des hypothèses." Mémoire, Université de Sherbrooke, 2015. http://hdl.handle.net/11143/8005.
Full textBouabdallaoui, Benyounes. "Reconnaissance électrochimique directe et en temps réel d’oligonucléotides, à base de polypyrrole fonctionnalisé : application aux puces à ADN." Paris, CNAM, 2006. http://www.theses.fr/2006CNAM1026.
Full textDNA biosensors have raised an increasing interest for their very wide potential applications. The miniaturization achieved for these biosensors has led to the terminology of DNA chips. The basic operating process of DNA biosensors involves an array of single stranded DNA probe grafed on a physical transducer, whose properties are modified upon hybridization of the DNA probe with a complementary single stranded DNA target. Up to recently, indirect transducing techniques have been mainly proposed. In this work, a direct real-time transducing method is presented based on the electrochemical response of a conjugated polymer. A polypyrrole film, bearing ODN probes, undergoes a modification of its electrochemical signature upon hybridization of the grafted ODN probes with their complementary ODN targets. This modification is monitored in real-time by performing cyclic voltammetry, and the higt affinity of base pairing ensures the very higt selectivity of the recognition process. With the aim of improving the sensibility of these polypyrrole-basedDNA chips, we report in this work the analysis of the mode of operation of these biodevices, and particularly of the parameters which control their detection threshold. A detection threshold in the femtomolar range has been experimentally obtained, which confirms the interesting potential of using electroactive conjugated polymers both as matrix for grafted ODN probes and also as physical transducers of the biological recognition process
Bisson, Francis. "La reconnaissance de plan des adversaires." Mémoire, Université de Sherbrooke, 2012. http://hdl.handle.net/11143/5764.
Full textFarabet, Clément. "Analyse sémantique des images en temps-réel avec des réseaux convolutifs." Phd thesis, Université Paris-Est, 2013. http://tel.archives-ouvertes.fr/tel-00965622.
Full textCloup, Pascal. "Etude et réalisation d'un système d'analyse gestuelle en trois dimensions par traitement d'images en temps réel." Valenciennes, 1989. https://ged.uphf.fr/nuxeo/site/esupversions/e2ada66a-822b-40d9-864e-973d4f37cc07.
Full textVercruysse, Michel. "Mesure en temps réel de l'orthogonalité des fils de trame d'un tissu en défilement et son traitement automatique." Lille 1, 1991. http://www.theses.fr/1991LIL10034.
Full textBoulay, Bernard. "Human posture recognition for behaviour understanding." Nice, 2007. http://www.theses.fr/2007NICE4000.
Full textDurant cette thèse nous avons proposé une approche temps réel, générique et fonctionnelle pour reconnaître la posture des personnes filmées par une caméra statique. Notre approche est conçue pour être complètement automatique et indépendante du point de vue de la caméra. La reconnaissance de posture à partir de séquence vidéo est un problème difficile. Ce problème s'inscrit dans le champ de recherche plus général de l'interprétation de séquence vidéo. L'approche proposée prend en entrée des informations provenant d'algorithmes de vision telles que la silhouette de la personne observée (une image binaire où une couleur représente la personne et l'autre le fond) ou sa position dans la scène. La première contribution est la modélisation d'un avatar 3D de posture. Un avatar 3D de posture est composé d'un modèle 3D humain (définissant les relations entre les différentes parties du corps), d'un ensemble de paramètre (définissant les positions des différentes parties du corps) et d'un ensemble de primitive (définissant l'aspect visuel des parties du corps). La seconde contribution est la proposition d'une approche hybride combinant l'utilisation de modèles 3D et de techniques 2D. Les avatars 3D de postures sont utilisés dans le processus de reconnaissance pour avoir une certaine indépendance du point de vue de la caméra. Les techniques 2D représentent les silhouettes des personnes détectées pour garder un temps réel de calcul. Cette thèse montre comment les avatars 3D peuvent être utilisés pour obtenir une approche générique et fonctionnelle pour reconnaître les postures. Cette approche est composée de deux parties : la détection de postures qui reconnaît la posture de la personne détectée en utilisant seulement l'information calculée sur l'image considérée, et le filtrage temporel de posture qui reconnaît la posture en utilisant l'information provenant des images précédentes. Une troisième contribution a été faite en comparant différentes représentations 2D des silhouettes au niveau du temps de calcul nécessaire et de leur dépendance à la qualité de la silhouette. Quatre représentations ont été retenues : une représentation combinant différentes valeurs géométriques, les moment de Hu, la skeletonisation et les projections horizontale et verticale. Une quatrième contribution est la caractérisation des cas ambigus. Des ambiguïtés au niveau de la reconnaissance peuvent se produire en utilisant seulement une caméra statique. Une posture ambiguë est définie par plusieurs postures vii qui ont des silhouettes visuellement similaires. Des données de synthèse sont générées pour évaluer l'approche proposée pour différents points de vue. Ainsi, les postures ambiguës sont identifiées en considérant la posture et son orientation. L'approche est aussi évaluée pour des données réelles en proposant un modèle de vérité terrain pour la reconnaissance de posture. Une cinquième contribution a été proposée en appliquant le résultat de notre approche à la reconnaissance d'action. Une méthode utilisant des machines à états finis a ainsi été proposée pour reconnaître des actions faisant intervenir une seule personne. Chaque état de la machine est composé d'une ou plusieurs postures. Cette méthode est appliquée avec succès pour détecter les chutes et la marche. Bien que notre approche donne de très bon taux de reconnaissance, il subsiste quelques limitations. La principale limitation de l'approche est qu'elle est limitée en nombre de postures d'intérêt pour des raisons de temps de calcul et de discrimination entre les postures considérées. La seconde limitation est le temps nécessaire à la génération des silhouettes des avatars 3D de posture. En utilisant l'information sur le déplacement de la personne dans la scène, l'algorithme de reconnaissance de posture traite entre 5 et 6 images par seconde. Des améliorations peuvent être faites pour résoudre ces limitations. En particulier, nous pourrions adapter automatiquement l'ensemble des postures d'intérêt au cas considéré, en utilisant par exemple la posture reconnue précédemment pour restreindre les postures 3D dont nous voulons extraire les silhouettes
Ghorayeb, Hicham. "Conception et mise en oeuvre d'algorithmes de vision temps réel pour la vidéo surveillance intelligente." Paris, ENMP, 2007. http://www.theses.fr/2007ENMP1463.
Full textIn this dissertation, we present our research work held at the Center of Robotics (CAOR) of the Ecole des Mines de Paris which tackles the problem of intelligent video analysis. The primary objective of our research is to prototype a generic framework for intelligent video analysis. We optimized this framework and configured it to cope with specific application requirements. We consider a people tracker application extracted from the PUVAME project. This application aims to improve people security in urban zones near to bus stations. Then, we have improved the generic framework for video analysis mainly for background subtraction and visual object detection. We have developed a library for machine learning specialized in boosting for visual object detection called LibAdaBoost. To the best of our knowledge LibAdaBoost is the first library in its kind. We make LibAdaBoost available for the machine learning community under the LGPL license. Finally we wanted to adapt the visual object detection algorithm based on boosting so that it could run on the graphics hardware. To the best of our knowledge we were the first to implement visual object detection with sliding technique on the graphics hardware. The results were promising and the prototype performed three to nine times better than the CPU. The framework was successfully implemented and integrated to the RTMaps environment. It was evaluated at the final session of the project PUVAME and demonstrated its fiability over various test scenarios elaborated specifically for the PUVAME project
Ziani, Ahmed. "Interprétation en temps réel de séquence vidéo par exploitation des modèles graphiques probabilistes." Littoral, 2010. http://www.theses.fr/2010DUNK0271.
Full textThe research covers the design and implementation of systems for recognition of scenarios in video image sequences. The upper layers of the recognition system operating primarily graphical probabilistic approaches (Bayesian networks and Hidden Markov models and their extensions) that can effectively handle uncertainties in the interpretation system. A first algorithm for recognition of sequences of events, combining two extensions of HMM (hierarchical and semi-Markov) was proposed. It allows to model complex scenarios based on a hierarchical structure integrating temporal constraints on the duration of each event. Then, we proposed a prediction technique based on the recognition of early tracks and allows quick to dismiss the models may be consistent with the observations. The last part of the work was the development of a global structure and a modular recognition system scenarios. The main advantage of this architecture is to use probabilistic techniques while integrating temporal reasoning capabilities. The logical architecture of the system uses a multi agents. In order to manage real-time constraints of the application, the control strategy of the recognition systems enables a minimum number of agents according to its internal decisions. The agents of the first layer has a role to highlight the basic events and are constructed mainly of Bayesian networks or hidden Markov models. The agents of the second temporal layer are also built from a specific structure type Bayesian network. Their role is to model explicitly the temporal relationships between events highlighted from the first layer. The third level officials involved in the final stage of decision using all of the decisions of intermediate agents. Different approaches to recognition of scenarios were tested on various real images in external and internal environment
Meynard, Jean-Baptiste. "Réalisation et évaluation d'un système de surveillance en temps réel pour les forces armées en opérations." Aix-Marseille 2, 2007. http://www.theses.fr/2007AIX20690.
Full textLeyrit, Laetitia. "Reconnaissance d'objets en vision artificielle : application à la reconnaissance de piétons." Phd thesis, Université Blaise Pascal - Clermont-Ferrand II, 2010. http://tel.archives-ouvertes.fr/tel-00626492.
Full textStoutenburg, Tardieu Cody. "Une comparaison d’algorithmes de reconnaissance de plan à l’aide d’observations in situ." Mémoire, Université de Sherbrooke, 2015. http://hdl.handle.net/11143/7525.
Full textDidelet, Elisabeth. "Les arbres de neurones avec rejet d'ambiguïté : application au diagnostic pour le pilotage en temps réel du réseau téléphonique français." Compiègne, 1992. http://www.theses.fr/1992COMPD520.
Full textGhorayeb, Hicham. "Conception et mise en œuvre d'algorithmes de vision temps-réel pour la vidéo surveillance intelligente." Phd thesis, École Nationale Supérieure des Mines de Paris, 2007. http://pastel.archives-ouvertes.fr/pastel-00003064.
Full textPaumard, José. "Reconnaissance multiéchelle d'objets dans des scènes." Cachan, Ecole normale supérieure, 1996. http://www.theses.fr/1996DENS0025.
Full textSémani-Delmi, Dahbia. "Une méthode supervisée de sélection et de discrimination avec rejet : application au projet Aqu@thèque." La Rochelle, 2004. http://www.theses.fr/2004LAROS113.
Full textOur work is a part of the Aqu@thèque project, of which the principal goal is to elaborate an automatique and real-time fish recognition system. In the first part of this work, we study the classification methods with reject options (ambiguity reject and distance reject). We proposed a new strategy and an original rules of classification for the implementation of the reject options. Our strategy is based on an ambiguity measure which is a combination of fuzzy aggregation operators. The second part of this study deals with the feature selection problem. We focus on the study of the evaluation criteria which measure the goodness of feature subsets. A new evaluation criterion is proposed. It is based on the proposed ambiguity measure and is used with an efficient search algorithm to define an new feature selection method. Finally, we present our automatic and real-time fish recognition system. Our feature selection method and our classification method with reject options are used to define an efficient classifier able to recognize the species of fish with a very good recognition rate (89\% without reject, 92\% with reject option)
Cron, Geneviève. "Diagnostic par reconnaissance des formes floue d'un système dynamique et réparti : Application à la gestion en temps réel du trafic téléphonique français." Compiègne, 1999. http://www.theses.fr/1999COMP1231.
Full textBégard, Julien. "Reconnaissance de formes dans des séquences vidéo : application à la vidéo-surveillance." Paris 6, 2008. http://www.theses.fr/2008PA066538.
Full textThis thesis addresses the problem of detecting people in pictures. The objective is to achieve an efficient system for detecting pedestrians in an automotive context in urban areas. We discussed this issue on two main fronts. The first was to design a discriminant representation of the form of the objects. Due to the wide variety of appearances of pedestrians, it must be rich enough to capture the characteristics necessary to recognize them. We built this description from the information of the gradient calculated in the image and its magnitude: Histograms of Orientation Gradient fortified with the Magnitude. Calculated densely in regions of the image, these descriptors forms vectors that define finely the shape of objects. We compare these descriptors with decision rules formed from machine learning methods. Our system is based on AdaBoost algorithm that we have made changes to get an algorithm using more effectively the information provided. These changes gave two algorithms with comparable performance. We have designed a method to combine these algorithms and get better results. We structured our system in a cascade of sub-detectors that are increasingly complex. The purpose of this structure is to eliminate as soon as possible wrong candidates. Thus, only the right candidates reaching the top of the cascade, consisting of the most complex and time-costly rules decisions. We evaluated and validated this methodology on color and infrared data
Cablé, Baptiste. "Vers la gestion de scénarios pour la reconnaissance et l'anticipation de situations dynamiques." Troyes, 2011. http://www.theses.fr/2011TROY0007.
Full textOur study deals with the problem of recognition and anticipation of dynamic situations for user assistance. Existing tools like Hidden Markov Models or Petri Nets are already used in this context. However, learning this kind of models is complicated and slow. Thus, the de-signer has to specify every model of situation so that the program can work in real-time. Our solution is a generic algorithm which build itself the representation of the dynamic system. It adapts to the user and the situation in order to make predictions. Dynamic situations are modeled by scenarios. A scenario corresponds to a period during which every event has an influence on other. It is made of an ordered series of states and actions in the form of symbols. The algorithm is a kind of Case-Based Reason-ing method but some modifications are made. Representations and computations are oriented towards simplicity and speed. Moreover, the algorithm is suitable for problems which evolve in time. The approach is applied to two distinct fields. The first application consists in assisting the user of a powered wheelchair. Without knowing initially the environment, the algorithm memorizes the usual paths of the user. This knowledge is used to drive automatically the wheelchair during usual paths. The second ap-plication is dedicated to the assistance of novice players in a multi-player online game. Experience of dynamic situations is learned from all the players and is used to predict the consequences of every battle
Fasquel, Jean-Baptiste. "Une méthode opto-informatique de détection et de reconnaissance d'objets d'intérêt : Application à la détection des lésions cancéreuses du foie et à la vérification en temps-réel des signatures manuscrites." Université Louis Pasteur (Strasbourg) (1971-2008), 2002. http://www.theses.fr/2002STR13234.
Full textDue to recent technological advances, optical processors become faster than specialized digital processors, essentially for linear filterings. The purpose of this thesis is to point out, for two applications, the potential of the coupling of a specialized digital processor with a Vander Lugt optical correlator, by developping an original hybrid " opto-electronic " method for object detection and recognition. The proposed object detection method is based on the digital statistical recombination of a set of optical smoothings, within regions of interest which are previously detected using a fast hybrid technique. It is shown that this hybrid method allows the unsupervised detection of noisy objects of varying sizes. Experimental results validate its potential for the fast detection of liver tumors. The proposed object recognition method, dedicated to the fast verification of handwritten signatures, consists in several statistical classifiers. Each one is based on a set of specific optical filterings allowing to measure the similarity between underlying structures of the signature to be verified and the reference signatures. The different decisions and their fusion are performed with a digital processor. Experimental results validate the proposed hybrid object recognition method
Jaber, Jamal. "Définition et validation d'une architecture électronique rapide de caractérisation et d'étiquetage d'objets dans une image." Nancy 1, 1993. http://www.theses.fr/1993NAN10332.
Full textBigué, Laurent. "Reconnaissance des formes en temps réel par voie optique : étude comparative d'implantations optiques de filtres de corrélation : application au corrélateur optique à transformée de Fourier conjointe." Mulhouse, 1996. http://www.theses.fr/1996MULH0445.
Full textHamdi, Hamza. "Plate-forme multimodale pour la reconnaissance d'émotions via l'analyse de signaux physiologiques : Application à la simulation d'entretiens d'embauche." Phd thesis, Université d'Angers, 2012. http://tel.archives-ouvertes.fr/tel-00997249.
Full textMazeau, Jean-Paul. "De la vidéo virtuelle à la compréhension de l'image en temps réel par l'analyse des composantes chromatiques : des applications dans le domaine de la réalité virtuelle." Paris 8, 1992. http://www.theses.fr/1992PA080698.
Full textThis thesis is divided into 2 parts. The first one deals with the problems of the virtual montage of video sequences recorded on various supports. After the analysis we propose a serial or parallel processing machine capable of accessing all available supports. This first step includes various applications in the fields of experimentation. Teaching, games, etc. The second step deals with the generalization of virtual video beyond the limits of the use of video as a support of visible images. It introduces the concept of (iconicdata) and offers a machine architecture designed to process the image directlyfrom such data, on the basis of the signal. This specific parallel architecture uses the concept of (authorsilicium) which allows for application cnstrained silicium, so as to process, through a set of rules all the relations between the images to be understood and the images that help to understand. This architecture operates on an information frame cycle, i. E. 25 times per second. Processing time depend neither on the number of video patterns to be understood nor on, the complexity of the relations between theses patterns. We consider a set of examples from suitable applications domain such as: robotics, image comprehension, interactive movement, virtual reality
Rahmani, Naïm Mohamed. "Instrumentation pour le traitement numérique du signal électroencéphalographique : Application à la reconnaissance automatique, temps réel, des différents stades de sommeil et de veille chez le rat." Nancy 1, 1990. http://docnum.univ-lorraine.fr/public/SCD_T_1990_0336_RAHMANI.pdf.
Full textJeanne, Florian. "Métaphore d’interaction gestuelle en environnement virtuel : application à l’apprentissage de gestes." Thesis, Compiègne, 2017. http://www.theses.fr/2017COMP2398/document.
Full textIn the light of our work, we were interested in precise gesture learning in virtual environment. In these environments, learners can train in safer and longer training sessions than in real environments, using augmented feedback that wouldn’t be available otherwise—such as a 3D trace of canonical movements or the opportunity to see your gesture from a different point of view. However, in the context of gesture guidance, most of the usual visual metaphors do not tackle the dependency they generate. The guidance hypothesis, from motor learning theories, states that the continuous use of feedback during early phases of learning leads to a dependency on the feedback. Learners are no longer able to properly reproduce the requested gesture or task without the help of this feedback. However, in virtual environments the use of augmented feedback remains necessary for training. To deal with this issue, we propose a new dynamic metaphor for 3D interaction in virtual environments, based on the visual modality. Our approach is focus users’ training on their intrinsic feedback, by using a trial-and-error strategy. Our metaphor shows users’ trajectory errors in real time, implicitly spurring them to correct these errors. We assume that when the metaphor is no longer visible, learners are able to properly reproduce the gesture using their past experiences. We conducted a first experiment to evaluate the capacity of our metaphor to properly guide learners during gesture training. Then, a second experiment showed that the metaphor allowed learners to improve their performance and that it further reduced the dependency than usual guiding metaphors
Zemour, Aurélie. "Gestes, espaces et temps funéraires au début du Néolithique (6ème millénaire et 1ère moitié du 5ème millénaire cal-BC) en Italie et en France méridionale : reconnaissance des témoins archéologiques de l'après-mort." Thesis, Nice, 2013. http://www.theses.fr/2013NICE2021.
Full textThis study deeply renews our understanding of the funerary practices of the first farmers living in Italy and in southern France at the dawn of Neolithic. The hypothesis of their uniformity, which has been not truly questioned so far, is now invalidated. On the contrary, the very essence of the funerary ideology, that was spread and perpetuated by these societies during more than one millennium, is the diversity of the funerary practices. Indeed, this variability including variousforms of corpse deposits (primary, secondary, individual, plural) nevertheless displays codesvarying from a culture to another, a group to another, a site to another and from an individual toanother, creating a diverse but coherent system. This study has also highlighted symbolic practices exploiting human remains, and revealed their procedures. Relying on a reasoned visionof the neolithisation process and on the nature of the Impresso-cardial complex, this approach also reaches the role occupied by the funerary system within the multifaceted complementarity between sites as well as the innovation degree of burial gestures, whose Mesolithic origin appears limited. Following an archaeothanatological approach and mobilizing a broad panel of after-death archaeological testimonies on a wide corpus (45 sites, 87 funerary units, 128 individuals) has therefore not only allowed accessing the funerary system of the considered societies, but on the top of that, studying and describing the behaviours they adopted in regard of the corpse and towards Death
Sheng, Yunlong. "Processeur optique de traitement en temps réel d'images vidéo : Application au calcul optique des moments bidimensionnels des images." Besançon, 1986. http://www.theses.fr/1986BESA2032.
Full textGirondel, Vincent. "Contribution à l'analyse et à l'interprétation du mouvement humain : application à la reconnaissance de postures." Phd thesis, Grenoble INPG, 2006. http://tel.archives-ouvertes.fr/tel-00156572.
Full textL'analyse du mouvement humain fait intervenir plusieurs processus de traitement d'images tels que la segmentation d'objets en mouvement, le suivi temporel, la détection de peau, les modèles de corps humain et la reconnaissance d'actions ou de postures. Nous proposons une méthode de suivi temporel en deux étapes permettant de suivre au cours du temps une ou plusieurs personnes même si elles s'occultent entre elles. Cette méthode est basée sur un calcul d'intersection de boîtes englobantes rectangulaires et sur un filtrage partiel de Kalman. Puis nous explicitons une méthode de détection de peau par une approche couleur afin de localiser leurs visages et leurs mains. Toutes ces étapes préliminaires donnent accès à de nombreuses informations bas-niveau. Dans une dernière partie, nous utilisons une partie de ces informations pour reconnaître les postures statiques de personnes parmi les quatre postures suivantes: debout, assis, accroupi et couché. De nombreux résultats illustrent les avantages et les limitations des méthodes proposées, ainsi que leur efficacité et robustesse.