An algorithm for tracking single visual targets is developed in this study. Feature detection is the necessary and appropriate image processing technique for this algorithm. The main point of this approach is to use the data supplied by the feature detection as the observation from a group of targets having similar motion dynamics. Therefore a single visual target is regarded as a group of multiple targets. Accurate data association and state estimation under clutter are desired for this application similar to other multi-target tracking applications. The group tracking approach is used with the well-known probabilistic data association technique to cope with data association and estimation problems. The applicability of this method particularly for visual tracking and for other cases is also discussed.

Behavioral research involves the study of the behaviors of one or more agents (often animals) in order to better understand the agents' thoughts and actions. Identifying subject movements and behaviors based upon those movements is a critical, time-consuming step in behavioral research. To successfully perform behavior analysis, three goals must be met. First, the agents of interest are observed, and their movements recorded. Second, each individual must be uniquely identified. Finally, behaviors must be identified and recognized. I explore a system that can uniquely identify and track agents, then use these tracks to automatically build behavioral models and recognize similar behaviors in the future. I address the tracking and identification problems using a combination of laser range finders, active RFID sensors, and probabilistic models for real-time tracking. The laser range component adds environmental flexibility over vision based systems, while the RFID tags help disambiguate individual agents. The probabilistic models are important to target identification during the complex interactions with other agents of similar appearance. In addition to tracking, I present work on automatic methods for generating behavioral models based on supervised learning techniques using the agents' tracked data. These models can be used to classify new tracked data and identify the behavior exhibited by the agent, which can then be used to help automate behavior analysis.

<p>Due to their rapid technological advancement, mobile sensors such as unmanned aerial vehicles (UAVs) are seeing growing application in the area of multi-target tracking and identity management (MTIM). For efficient and sustainable performance of a MTIM system with mobile sensors, proper algorithms are needed to both effectively estimate the states/identities of targets from sensing data and optimally guide the mobile sensors based on the target estimates. One major challenge in MTIM is that a target may be temporarily lost due to line-of-sight breaks or corrupted sensing data in cluttered environments. It is desired that these targets are kept tracking and identification, especially when they reappear after the temporary loss of detection. Another challenging task in MTIM is to correctly track and identify targets during track coalescence, where multiple targets get close to each other and could be hardly distinguishable. In addition, while the number of targets in the sensors’ surveillance region is usually unknown and time-varying in practice, many existing MTIM algorithms assume their number of targets to be known and constant, thus those algorithms could not be directly applied to real scenarios.</p> <p>In this research, a set of solutions is developed to address three particular issues in MTIM that involves the above challenges: 1) using a single mobile sensor with a limited sensing range to track multiple targets, where the targets may occasionally lose detection; 2) using a network of mobile sensors to actively seek and identify targets to improve the accuracy of multi-target identity management; and 3) tracking and managing the identities of an unknown and time-varying number of targets in clutter.</p>

This dissertation concerns with the formulation of an improved multi-target multi-Bernoulli (MeMBer) filter and the use of the joint multi-target (JoM) estimator in an effective and efficient manner for a specific implementation of MeMBer filters. After reviewing random finite set (RFS) formalism for multi-target tracking problems and the related Bayes estimators the major contributions of this dissertation are explained in detail. The second chapter of this dissertation is dedicated to the analysis of the relationship between the multi-Bernoulli RFS distribution and the MeMBer corrector. This analysis leads to the formulation of an unbiased MeMBer filter without making any limiting assumption. Hence, as opposed to the popular cardinality balanced multi-target multi-Bernoulli (CBMeMBer) filter, the proposed MeMBer filter can be employed under the cases when sensor detection probability is moderate to low. Furthermore, a statistical refinement process is introduced to improve the stability of the estimated cardinality of targets obtained from the proposed MeMBer filter. The results from simulations demonstrate the effectiveness of the improved MeMBer filter. In Chapters III and IV, the Bayesian optimal estimators proposed for the RFS based multi-target tracking filters are examined in detail. First, an optimal solution to the unknown constant in the definition of the JoM estimator is determined by solving a multi-objective optimization problem. Thus, the JoM estimator can be implemented for tracking of a Bernoulli target using the optimal joint target detection and tracking (JoTT) filter. The results from simulations confirm assertions about its performance obtained by theoretical analysis in the literature. Finally, in the third chapter of this dissertation, the proposed JoM estimator is reformulated for RFS multi-Bernoulli distributions. Hence, an effective and efficient implementation of the JoM estimator is proposed for the Gaussian mixture implementations of the MeMBer filters. Simulation results demonstrate the robustness of the proposed JoM estimator under low-observable conditions. / Thesis / Doctor of Philosophy (PhD)

L'opération neOCampus, initiée en 2013 par l'Université Paul Sabatier, a pour objectif de créer un campus connecté, innovant, intelligent et durable en exploitant les compétences de 11 laboratoires et de plusieurs partenaires industriels. Pluridisciplinaires, ces compétences sont croisées dans le but d'améliorer le confort au quotidien des usagers du campus (étudiants, corps enseignant, personnel administratif) et de diminuer son empreinte écologique. L'intelligence que nous souhaitons apporter au Campus du futur exige de fournir à ses bâtiments une perception de son activité interne. En effet, l'optimisation des ressources énergétiques nécessite une caractérisation des activités des usagers afin que le bâtiment puisse s'y adapter automatiquement. L'activité humaine étant sujet à plusieurs niveaux d'interprétation nos travaux se focalisent sur l'extraction des déplacements des personnes présentes, sa composante la plus élémentaire. La caractérisation de l'activité des usagers, en termes de déplacements, exploite des données extraites de caméras et de microphones disséminés dans une pièce, ces derniers formant ainsi un réseau épars de capteurs hétérogènes. Nous cherchons alors à extraire de ces données une signature audiovisuelle et une localisation grossière des personnes transitant dans ce réseau de capteurs. Tout en préservant la vie privée de l'individu, la signature doit être discriminante, afin de distinguer les personnes entre elles, et compacte, afin d'optimiser les temps de traitement et permettre au bâtiment de s'auto-adapter. Eu égard à ces contraintes, les caractéristiques que nous modélisons sont le timbre de la voix du locuteur, et son apparence vestimentaire en termes de distribution colorimétrique. Les contributions scientifiques de ces travaux s'inscrivent ainsi au croisement des communautés parole et vision, en introduisant des méthodes de fusion de signatures sonores et visuelles d'individus. Pour réaliser cette fusion, des nouveaux indices de localisation de source sonore ainsi qu'une adaptation audiovisuelle d'une méthode de suivi multi-cibles ont été introduits, représentant les contributions principales de ces travaux. Le mémoire est structuré en 4 chapitres. Le premier présente un état de l'art sur les problèmes de ré-identification visuelle de personnes et de reconnaissance de locuteurs. Les modalités sonores et visuelles ne présentant aucune corrélation, deux signatures, une vidéo et une audio sont générées séparément, à l'aide de méthodes préexistantes de la littérature. Le détail de la génération de ces signatures est l'objet du chapitre 2. La fusion de ces signatures est alors traitée comme un problème de mise en correspondance d'observations audio et vidéo, dont les détections correspondantes sont cohérentes et compatibles spatialement, et pour lesquelles deux nouvelles stratégies d'association sont introduites au chapitre 3. La cohérence spatio-temporelle des observations sonores et visuelles est ensuite traitée dans le chapitre 4, dans un contexte de suivi multi-cibles. / The neOCampus operation, started in 2013 by Paul Sabatier University in Toulouse, aims to create a connected, innovative, intelligent and sustainable campus, by exploiting the skills of 11 laboratories and several industrial partners. These multidisciplinary skills are combined in order to improve users (students, teachers, administrative staff) daily comfort and to reduce the ecological footprint of the campus. The intelligence we want to bring to the campus of the future requires to provide to its buildings a perception of its intern activity. Indeed, optimizing the energy resources needs a characterization of the user's activities so that the building can automatically adapt itself to it. Human activity being open to multiple levels of interpretation, our work is focused on extracting people trajectories, its more elementary component. Characterizing users activities, in terms of movement, uses data extracted from cameras and microphones distributed in a room, forming a sparse network of heterogeneous sensors. From these data, we then seek to extract audiovisual signatures and rough localizations of the people transiting through this network of sensors. While protecting person privacy, signatures must be discriminative, to distinguish a person from another one, and compact, to optimize computational costs and enables the building to adapt itself. Having regard to these constraints, the characteristics we model are the speaker's timbre, and his appearance, in terms of colorimetric distribution. The scientific contributions of this thesis are thus at the intersection of the fields of speech processing and computer vision, by introducing new methods of fusing audio and visual signatures of individuals. To achieve this fusion, new sound source location indices as well as an audiovisual adaptation of a multi-target tracking method were introduced, representing the main contributions of this work. The thesis is structured in 4 chapters, and the first one presents the state of the art on visual reidentification of persons and speaker recognition. Acoustic and visual modalities are not correlated, so two signatures are separately computed, one for video and one for audio, using existing methods in the literature. After a first chapter dedicated to the state of the art in re-identification and speaker recognition methods, the details of the computation of the signatures is explored in chapter 2. The fusion of the signatures is then dealt as a problem of matching between audio and video observations, whose corresponding detections are spatially coherent and compatible. Two novel association strategies are introduced in chapter 3. Spatio-temporal coherence of the bimodal observations is then discussed in chapter 4, in a context of multi-target tracking.

[Truncated abstract] The multi-object filtering problem is a logical and fundamental generalization of the ubiquitous single-object vector filtering problem. Multi-object filtering essentially concerns the joint detection and estimation of the unknown and time-varying number of objects present, and the dynamic state of each of these objects, given a sequence of observation sets. This problem is intrinsically challenging because, given an observation set, there is no knowledge of which object generated which measurement, if any, and the detected measurements are indistinguishable from false alarms. Multi-object filtering poses significant technical challenges, and is indeed an established area of research, with many applications in both military and commercial realms. The new and emerging approach to multi-object filtering is based on the formal theory of random finite sets, and is a natural, elegant and rigorous framework for the theory of multiobject filtering, originally proposed by Mahler. In contrast to traditional approaches, the random finite set framework is completely free of explicit data associations. The random finite set framework is adopted in this dissertation as the basis for a principled and comprehensive study of multi-object filtering. The premise of this framework is that the collection of object states and measurements at any time are treated namely as random finite sets. A random finite set is simply a finite-set-valued random variable, i.e. a random variable which is random in both the number of elements and the values of the elements themselves. Consequently, formulating the multiobject filtering problem using random finite set models precisely encapsulates the essence of the multi-object filtering problem, and enables the development of principled solutions therein. '...' The performance of the proposed algorithm is demonstrated in simulated scenarios, and shown at least in simulation to dramatically outperform traditional single-object filtering in clutter approaches. The second key contribution is a mathematically principled derivation and practical implementation of a novel algorithm for multi-object Bayesian filtering, based on moment approximations to the posterior density of the random finite set state. The performance of the proposed algorithm is also demonstrated in practical scenarios, and shown to considerably outperform traditional multi-object filtering approaches. The third key contribution is a mathematically principled derivation and practical implementation of a novel algorithm for multi-object Bayesian filtering, based on functional approximations to the posterior density of the random finite set state. The performance of the proposed algorithm is compared with the previous, and shown to appreciably outperform the previous in certain classes of situations. The final key contribution is the definition of a consistent and efficiently computable metric for multi-object performance evaluation. It is shown that the finite set theoretic state space formulation permits a mathematically rigorous and physically intuitive construct for measuring the estimation error of a multi-object filter, in the form of a metric. This metric is used to evaluate and compare the multi-object filtering algorithms developed in this dissertation.

In conventional tracking algorithms the targets are usually considered as point source objects. However, in realistic scenarios the point source assumption is often not suitable and estimating the states of an object extension characterized by a collectively moving ballistic object group (cluster) becomes a very critical and relevant problem which has applications in the defense area. Recently, a Bayesian approach to extended object tracking using random matrices has been proposed. Within this approach, ellipsoidal object extensions are modeled by random matrices and treated as additional state variables to be estimated. In this work we propose to use a slightly modified version of this new approach that simultaneously estimates the ellipsoidal shape and the kinematics of a group of ballistic targets. Target group that is tracked consists of subsequent projectiles. We use JPDAF framework together with the new approach to emphasize the pros and cons of both approaches. The methods are demonstrated and evaluated in detail by making various simulations.

Ce mémoire décrit mes travaux de thèse menés au sein de l'équipe ComSee (Computers that See) rattachée à l'axe ISPR (Image, Systèmes de Perception et Robotique) de l'Institut Pascal. Celle-ci a été financée par la société Vesalis par le biais d'une convention CIFRE avec l'Institut Pascal, subventionnée par l'ANRT (Association Nationale de la Recherche et de la Technologie). Les travaux de thèse s'inscrivent dans le cadre de l'automatisation de la fouille d'archives vidéo intervenant lors d'enquêtes policières. L'application rattachée à cette thèse concerne la création automatique d'un album photo des individus apparaissant sur une séquence de vidéosurveillance. En s'appuyant sur un détecteur de visages, l'objectif est de regrouper par identité les visages détectés sur l'ensemble d'une séquence vidéo. Comme la reconnaissance faciale en environnement non-contrôlé reste difficilement exploitable, les travaux se sont orientés vers le suivi visuel multi-cibles global basé détections. Ce type de suivi est relativement récent. Il fait intervenir un détecteur d'objets et traite la vidéo dans son ensemble (en opposition au traitement séquentiel couramment utilisé). Cette problématique a été représentée par un modèle probabiliste de type Maximum A Posteriori. La recherche de ce maximum fait intervenir un algorithme de circulation de flot sur un graphe, issu de travaux antérieurs. Ceci permet l'obtention d'une solution optimale au problème (défini par l'a posteriori) du regroupement des détections pour le suivi. L'accent a particulièrement été mis sur la représentation de la similarité entre les détections qui s'intègre dans le terme de vraisemblance du modèle. Plusieurs mesures de similarités s'appuyant sur différents indices (temps, position dans l'image, apparence et mouvement local) ont été testées. Une méthode originale d'estimation de ces similarités entre les visages détectés a été développée pour fusionner les différentes informations et s'adapter à la situation rencontrée. Plusieurs expérimentations ont été menées sur des situations complexes, mais réalistes, de scènes de vidéosurveillance. Même si les qualités des albums construits ne satisfont pas encore à une utilisation pratique, le système de regroupement de détections mis en œuvre au cours de cette thèse donne déjà une première solution. Grâce au point de vue partitionnement de données adopté au cours de cette thèse, le suivi multi-cibles développé permet une extension simple à du suivi autre que celui des visages. / This report describes my thesis work conducted within the ComSee (Computers That See) team related to the ISPR axis (ImageS, Perception Systems and Robotics) of Institut Pascal. It was financed by the Vesalis company via a CIFRE (Research Training in Industry Convention) agreement with Institut Pascal and publicly funded by ANRT (National Association of Research and Technology). The thesis was motivated by issues related to automation of video analysis encountered during police investigations. The theoretical research carried out in this thesis is applied to the automatic creation of a photo album summarizing people appearing in a CCTV sequence. Using a face detector, the aim is to group by identity all the faces detected throughout the whole video sequence. As the use of facial recognition techniques in unconstrained environments remains unreliable, we have focused instead on global multi-target tracking based on detections. This type of tracking is relatively recent. It involves an object detector and global processing of the video (as opposed to sequential processing commonly used). This issue has been represented by a Maximum A Posteriori probabilistic model. To find an optimal solution of Maximum A Posteriori formulation, we use a graph-based network flow approach, built upon third-party research. The study concentrates on the definition of inter-detections similarities related to the likelihood term of the model. Multiple similarity metrics based on different clues (time, position in the image, appearance and local movement) were tested. An original method to estimate these similarities was developed to merge these various clues and adjust to the encountered situation. Several experiments were done on challenging but real-world situations which may be gathered from CCTVs. Although the quality of generated albums do not yet satisfy practical use, the detections clustering system developed in this thesis provides a good initial solution. Thanks to the data clustering point of view adopted in this thesis, the proposed detection-based multi-target tracking allows easy transfer to other tracking domains.

Cette thèse s'inscrit dans la problématique de la perception des véhicules autonomes, qui doivent notamment être capables de détecter et de positionner à tout moment les éléments fixes et mobiles de leur environnement. Les besoins sont ensuite multiples, de la détection d'obstacles à la localisation du porteur dans l'espace, et de nombreuses méthodes de la littérature s'y attellent. L'objectif de cette thèse est de reconstituer, à partir de prises de vues de stéréo-vision, une carte en trois dimensions décrivant l'environnement proche ; tout en effectuant une détection, localisation et suivi dans le temps des objets mobiles.La détection et le suivi dans le temps d'un grand nombre de points d'intérêt constitue une première étape. Après avoir effectué une comparaison exhaustive de divers détecteurs de points d'intérêt de la littérature, on propose pour réaliser le suivi de points une implémentation massivement parallélisée de l'algorithme KLT, dans une configuration redondante réalisée pendant cette thèse. Cette implémentation autorise le suivi fiable de milliers de points en temps réel, et se compare favorablement à l'état de l'art.Il s'agit ensuite d'estimer le déplacement du porteur, et de positionner ces points dans l'espace, tâche pour laquelle on propose une évolution robuste d'une procédure bien connue, dite "SVD", suivie d'un filtrage par UKF, qui nous permettent d'estimer très rapidement le mouvement propre du porteur. Les points suivis sont ensuite positionnés dans l'espace, en prenant en compte leur possible mobilité, en estimant continuellement la position la plus probable compte tenu des observations successives.La détection et le suivi des objets mobiles font l'objet d'une dernière partie, dans laquelle on propose une segmentation originale tenant compte des aspects de position et de vitesse. On exploite ainsi une des singularités de notre approche, qui conserve pour chaque point positionné un ensemble cohérent de positions dans le temps. Le filtrage et le suivi des cibles se basent finalement sur un filtre GM-PHD. / This PhD work is to be seen within the context of autonomous vehicle perception, in which the detection and localisation of elements of the surroundings in real time is an obvious requirement. Subsequent perception needs are manyfold, from localisation to obstacle detection, and are the subject of a continued research interest. The goal of this work is to build, in real time and from stereovision acquisition, a 3D map of the surroundings ; while detecting and tracking moving objects.Interest point selection and tracking on picture space are a first step, which we initiate by a thorough comparison of detectors from the literature. As regards tracking, we propose a massively parallel implementation of the standard KLT algorithm, using redundant tracking to provide reliable quality estimation. This allows us to track thousands of points in real-time, which compares favourably to the state of the art.Next step is the ego-motion estimation, along with the positioning of tracked points in 3D space. We first propose an iterative variant of the well known “SVD” process followed by UKF filtering, which allows for a very fast and reliable estimation. Then the position of every followed interest point is filtered on the fly over time, in contrast to most dense approaches from the literature.We finally propose a segmentation of moving objects in the augmented position-speed space, which is made possible by our continuous estimation of feature points position. Target tracking and filtering finally use a GM-PHD approach.

This thesis deals with a concept of system for automatic passenger counting in different  modes of transport. Counting units are placed in top of the door area in the vehicle. Passengers are detected at the disparity map counted from the stereo-camera images. Object tracking is achieved with Global nearest neighbor and Multiple hypothesis tracking algorithm. This system is used for public transportation surveys.