Dissertations / Theses on the topic 'Bayesian target tracking'

Visual tracking, a building block for many applications, has challenges such as occlusions,illumination changes, background clutter and variable motion dynamics that may degrade the tracking performance and are likely to cause failures. In this thesis, we propose Track-Evaluate-Correct framework (self-correlation) for existing trackers in order to achieve a robust tracking. For a tracker in the framework, we embed an evaluation block to check the status of tracking quality and a correction block to avoid upcoming failures or to recover from failures. We present a generic representation and formulation of the self-correcting tracking for Bayesian trackers using a Dynamic Bayesian Network (DBN). The self-correcting tracking is done similarly to a selfaware system where parameters are tuned in the model or different models are fused or selected in a piece-wise way in order to deal with tracking challenges and failures. In the DBN model representation, the parameter tuning, fusion and model selection are done based on evaluation and correction variables that correspond to the evaluation and correction, respectively. The inferences of variables in the DBN model are used to explain the operation of self-correcting tracking. The specific contributions under the generic self-correcting framework are correlation-based selfcorrecting tracking for an extended object with model points and tracker-level fusion as described below. For improving the probabilistic tracking of extended object with a set of model points, we use Track-Evaluate-Correct framework in order to achieve self-correcting tracking. The framework combines the tracker with an on-line performance measure and a correction technique. We correlate model point trajectories to improve on-line the accuracy of a failed or an uncertain tracker. A model point tracker gets assistance from neighbouring trackers whenever degradation in its performance is detected using the on-line performance measure. The correction of the model point state is based on the correlation information from the states of other trackers. Partial Least Square regression is used to model the correlation of point tracker states from short windowed trajectories adaptively. Experimental results on data obtained from optical motion capture systems show the improvement in tracking performance of the proposed framework compared to the baseline tracker and other state-of-the-art trackers. The proposed framework allows appropriate re-initialisation of local trackers to recover from failures that are caused by clutter and missed detections in the motion capture data. Finally, we propose a tracker-level fusion framework to obtain self-correcting tracking. The fusion framework combines trackers addressing different tracking challenges to improve the overall performance. As a novelty of the proposed framework, we include an online performance measure to identify the track quality level of each tracker to guide the fusion. The trackers in the framework assist each other based on appropriate mixing of the prior states. Moreover, the track quality level is used to update the target appearance model. We demonstrate the framework with two Bayesian trackers on video sequences with various challenges and show its robustness compared to the independent use of the trackers used in the framework, and also compared to other state-of-the-art trackers. The appropriate online performance measure based appearance model update and prior mixing on trackers allows the proposed framework to deal with tracking challenges.

Target tracking is a broad subject area extensively studied in many engineering disciplines. In this thesis, target tracking implies the temporal estimation of target features such as the target's direction-of-arrival (DOA), the target's boundary pixels in a sequence of images, and/or the target's position in space. For multiple target tracking, we have introduced a new motion model that incorporates an acceleration component along the heading direction of the target. We have also shown that the target motion parameters can be considered part of a more general feature set for target tracking, e.g., target frequencies, which may be unrelated to the target motion, can be used to improve the tracking performance. We have introduced an acoustic multiple-target tracker using a flexible observation model based on an image tracking approach by assuming that the DOA observations might be spurious and that some of the DOAs might be missing in the observation set. We have also addressed the acoustic calibration problem from sources of opportunity such as beacons or a moving source. We have derived and compared several calibration methods for the case where the node can hear a moving source whose position can be reported back to the node. The particle filter, as a recursive algorithm, requires an initialization phase prior to tracking a state vector. The Metropolis-Hastings (MH) algorithm has been used for sampling from intractable multivariate target distributions and is well suited for the initialization problem. Since the particle filter only needs samples around the mode, we have modified the MH algorithm to generate samples distributed around the modes of the target posterior. By simulations, we show that this mode hungry algorithm converges an order of magnitude faster than the original MH scheme. Finally, we have developed a general framework for the joint state-space tracking problem. A proposal strategy for joint state-space tracking using the particle filters is defined by carefully placing the random support of the joint filter in the region where the final posterior is likely to lie. Computer simulations demonstrate improved performance and robustness of the joint state-space when using the new particle proposal strategy.

This thesis presents a generalized Bayesian framework for a mobile robot to localize itself and a target, while building a map of the environment. The proposed technique builds upon the Bayesian Simultaneous Robot Localization and Mapping (SLAM) method, to allow the robot to localize itself and the environment using map features or landmarks in close proximity. The target feature is distinguished from the rest of features since the robot has to navigate to its location and thus needs to be observed from a long distance. The contribution of the proposed approach is on enabling the robot to track a target object or region, using a multi-stage technique. In the first stage, the target state is corrected sequentially to the robot correction in the Recursive Bayesian Estimation. In the second stage, with the target being closer, the target state is corrected simultaneously with the robot and the landmarks. The process allows the robot's state uncertainty to be propagated into the estimated target's state, bridging the gap between tracking only methods where the target is estimated assuming known observer state and SLAM methods where only landmarks are considered. When the robot is located far, the sequential stage is efficient in tracking the target position while maintaining an accurate robot state using close only features. Also, target belief is always maintained in comparison to temporary tracking methods such as image-tracking. When the robot is closer to the target and most of its field of view is covered by the target, it is shown that simultaneous correction needs to be used in order to minimize robot, target and map entropies in the absence of other landmarks.<br>M.S.<br>This thesis presents a generalized framework with the goal of allowing a robot to localize itself and a static target, while building a map of the environment. This map is used as in the Simultaneous Localization and Mapping (SLAM) framework to enhance robot accuracy and with close features. Target, here, is distinguished from the rest of features since the robot has to navigate to its location and thus needs to be continuously observed from a long distance. The contribution of the proposed approach is on enabling the robot to track a target object or region, using a multi-stage technique. In the first stage, the robot and close landmarks are estimated simultaneously and they are both corrected. Using the robot's uncertainty in its estimate, the target state is then estimated sequentially, considering known robot state. That decouples the target estimation from the rest of the process. In the second stage, with the target being closer, target, robot and landmarks are estimated simultaneously. When the robot is located far, the sequential stage is efficient in tracking the target position while maintaining an accurate robot state using close only features. When the robot is closer to the target and most of its field of view is covered by the target, it is shown that simultaneous correction needs to be used in order to minimize robot, target and map uncertainties in the absence of other landmarks.

This thesis deals with the problem of online visual tracking of multiple humans in an enclosed environment. The focus is to develop techniques to deal with the challenges of varying number of targets, inter-target occlusions and interactions when every target gives rise to multiple measurements (pixels) in every video frame. This thesis contains three different contributions to the research in multi-target tracking. Firstly, a multiple target tracking algorithm is proposed which focuses on mitigating the inter-target occlusion problem during complex interactions. This is achieved with the help of a particle filter, multiple video cues and a new interaction model. A Markov chain Monte Carlo particle filter (MCMC-PF) is used along with a new interaction model which helps in modeling interactions of multiple targets. This helps to overcome tracking failures due to occlusions. A new weighted Markov chain Monte Carlo (WMCMC) sampling technique is also proposed which assists in achieving a reduced tracking error. Although effective, to accommodate multiple measurements (pixels) produced by every target, this technique aggregates measurements into features which results in information loss. In the second contribution, a novel variational Bayesian clustering-based multi-target tracking framework is proposed which can associate multiple measurements to every target without aggregating them into features. It copes with complex inter-target occlusions by maintaining the identity of targets during their close physical interactions and handles efficiently a time-varying number of targets. The proposed multi-target tracking framework consists of background subtraction, clustering, data association and particle filtering. A variational Bayesian clustering technique groups the extracted foreground measurements while an improved feature based joint probabilistic data association filter (JPDAF) is developed to associate clusters of measurements to every target. The data association information is used within the particle filter to track multiple targets. The clustering results are further utilised to estimate the number of targets. The proposed technique improves the tracking accuracy. However, the proposed features based JPDAF technique results in an exponential growth of computational complexity of the overall framework with increase in number of targets. In the final work, a novel data association technique for multi-target tracking is proposed which more efficiently assigns multiple measurements to every target, with a reduced computational complexity. A belief propagation (BP) based cluster to target association method is proposed which exploits the inter-cluster dependency information. Both location and features of clusters are used to re-identify the targets when they emerge from occlusions. The proposed techniques are evaluated on benchmark data sets and their performance is compared with state-of-the-art techniques by using, quantitative and global performance measures.

In this thesis a number of improvements have been established for specific methods which utilize sequential Monte Carlo (SMC), aka. Particle filtering (PF) techniques. The first problem is the Bayesian multi-target tracking (MTT) problem for which we propose the use of non-parametric Bayesian models that are based on time varying extension of Dirichlet process (DP) models. The second problem studied in this thesis is an important application area for the proposed DP based MTT method<br>the tracking of vocal tract resonance frequencies of the speech signals. Lastly, we investigate SMC based parameter estimation problem of nonlinear non-Gaussian state space models in which we provide a performance improvement for the path density based methods by utilizing regularization techniques.

The multi-target tracking problem essentially involves the recursive joint estimation of the state of unknown and time-varying number of targets present in a tracking scene, given a series of observations. This problem becomes more challenging because the sequence of observations is noisy and can become corrupted due to miss-detections and false alarms/clutter. Additionally, the detected observations are indistinguishable from clutter. Furthermore, whether the target(s) of interest are point or extended (in terms of spatial extent) poses even more technical challenges. An approach known as random finite sets provides an elegant and rigorous framework for the handling of the multi-target tracking problem. With a random finite sets formulation, both the multi-target states and multi-target observations are modelled as finite set valued random variables, that is, random variables which are random in both the number of elements and the values of the elements themselves. Furthermore, compared to other approaches, the random finite sets approach possesses a desirable characteristic of being free of explicit data association prior to tracking. In addition, a framework is available for dealing with random finite sets and is known as finite sets statistics. In this thesis, advanced signal processing techniques are employed to provide enhancements to and develop new random finite sets based multi-target tracking algorithms for the tracking of both point and extended targets with the aim to improve tracking performance in cluttered environments. To this end, firstly, a new and efficient Kalman-gain aided sequential Monte Carlo probability hypothesis density (KG-SMC-PHD) filter and a cardinalised particle probability hypothesis density (KG-SMC-CPHD) filter are proposed. These filters employ the Kalman- gain approach during weight update to correct predicted particle states by minimising the mean square error between the estimated measurement and the actual measurement received at a given time in order to arrive at a more accurate posterior. This technique identifies and selects those particles belonging to a particular target from a given PHD for state correction during weight computation. The proposed SMC-CPHD filter provides a better estimate of the number of targets. Besides the improved tracking accuracy, fewer particles are required in the proposed approach. Simulation results confirm the improved tracking performance when evaluated with different measures. Secondly, the KG-SMC-(C)PHD filters are particle filter (PF) based and as with PFs, they require a process known as resampling to avoid the problem of degeneracy. This thesis proposes a new resampling scheme to address a problem with the systematic resampling method which causes a high tendency of resampling very low weight particles especially when a large number of resampled particles are required; which in turn affect state estimation. Thirdly, the KG-SMC-(C)PHD filters proposed in this thesis perform filtering and not tracking , that is, they provide only point estimates of target states but do not provide connected estimates of target trajectories from one time step to the next. A new post processing step using game theory as a solution to this filtering - tracking problem is proposed. This approach was named the GTDA method. This method was employed in the KG-SMC-(C)PHD filter as a post processing technique and was evaluated using both simulated and real data obtained using the NI-USRP software defined radio platform in a passive bi-static radar system. Lastly, a new technique for the joint tracking and labelling of multiple extended targets is proposed. To achieve multiple extended target tracking using this technique, models for the target measurement rate, kinematic component and target extension are defined and jointly propagated in time under the generalised labelled multi-Bernoulli (GLMB) filter framework. The GLMB filter is a random finite sets-based filter. In particular, a Poisson mixture variational Bayesian (PMVB) model is developed to simultaneously estimate the measurement rate of multiple extended targets and extended target extension was modelled using B-splines. The proposed method was evaluated with various performance metrics in order to demonstrate its effectiveness in tracking multiple extended targets.

Understanding the content of a video sequence is not a particularly difficult problem for humans. We can easily identify objects, such as people, and track their position and pose within the 3D world. A computer system that could understand the world through videos would be extremely beneficial in applications such as surveillance, robotics, biology. Despite significant advances in areas like tracking and, more recently, 3D static scene understanding, such a vision system does not yet exist. In this work, I present progress on this problem, restricted to videos of objects that move in smoothly and which are relatively easily detected, such as people. Our goal is to identify all the moving objects in the scene and track their physical state (e.g., their 3D position or pose) in the world throughout the video. We develop a Bayesian generative model of a temporal scene, where we separately model data association, the 3D scene and imaging system, and the likelihood function. Under this model, the video data is the result of capturing the scene with the imaging system, and noisily detecting video features. This formulation is very general, and can be used to model a wide variety of scenarios, including videos of people walking, and time-lapse images of pollen tubes growing in vitro. Importantly, we model the scene in world coordinates and units, as opposed to pixels, allowing us to reason about the world in a natural way, e.g., explaining occlusion and perspective distortion. We use Gaussian processes to model motion, and propose that it is a general and effective way to characterize smooth, but otherwise arbitrary, trajectories. We perform inference using MCMC sampling, where we fit our model of the temporal scene to data extracted from the videos. We address the problem of variable dimensionality by estimating data association and integrating out all scene variables. Our experiments show our approach is competitive, producing results which are comparable to state-of-the-art methods.

Target tracking is an important part of video surveillance from a UAV. Tracking a target in an urban environment can be difficult because of the number of occlusions present in the environment. If multiple UAVs are used to track a target and the target behavior is learned autonomously by the UAV then the task may become easier. This thesis explores the hypothesis that an existing cooperative control algorithm can be enhanced by a language modeling algorithm to improve over time the target tracking performance of one or more ground targets in a dense urban environment. Observations of target behavior are reported to the Sequence Memoizer which uses the observations to create a belief model of future target positions. This belief model is combined with a kinematic belief model and then used in a cooperative auction algorithm for UAV path planning. The results for tracking a single target using the combined belief model outperform other belief models and improve over the duration of the mission. Results from tracking multiple targets indicate that algorithmic enhancements may be needed to find equivalent success. Future target tracking algorithms should involve machine learning to enhance tracking performance.

This thesis contributes new algorithms and implementations for particle filter-based target tracking. From an algorithmic perspective, modifications that improve a batch-based acoustic direction-of-arrival (DOA), multi-target, particle filter tracker are presented. The main improvements are reduced execution time and increased robustness to target maneuvers. The key feature of the batch-based tracker is an image template-matching approach that handles data association and clutter in measurements. The particle filter tracker is compared to an extended Kalman filter~(EKF) and a Laplacian filter and is shown to perform better for maneuvering targets. Using an approach similar to the acoustic tracker, a radar range-only tracker is also developed. This includes developing the state update and observation models, and proving observability for a batch of range measurements. From an implementation perspective, this thesis provides new low-power and real-time implementations for particle filters. First, to achieve a very low-power implementation, two mixed-mode implementation strategies that use analog and digital components are developed. The mixed-mode implementations use analog, multiple-input translinear element (MITE) networks to realize nonlinear functions. The power dissipated in the mixed-mode implementation of a particle filter-based, bearings-only tracker is compared to a digital implementation that uses the CORDIC algorithm to realize the nonlinear functions. The mixed-mode method that uses predominantly analog components is shown to provide a factor of twenty improvement in power savings compared to a digital implementation. Next, real-time implementation strategies for the batch-based acoustic DOA tracker are developed. The characteristics of the digital implementation of the tracker are quantified using digital signal processor (DSP) and field-programmable gate array (FPGA) implementations. The FPGA implementation uses a soft-core or hard-core processor to implement the Newton search in the particle proposal stage. A MITE implementation of the nonlinear DOA update function in the tracker is also presented.

Maritime target tracking from an airborne radar faces many issues due to the features of theenvironment, the targets to be tracked and the movement of the radar platform. Therefore, aunique tracking algorithm is not always able to reach the best possible performance for everyencountered situation. It needs to self-adapt to the environment and to the targets which areobserved in order to always be as ecient as possible. Adaptability is thus a key issue of radartracking.Several implementations of the mathematical Bayesian estimation theory, commonly called lters,have been used in the literature in order to estimate as precisely as possible targets trajectory.Depending on the situations and the assumptions that are considered, some of themare expected to perform better. This thesis suggests to look deeper into the tracking techniquesthat can be found in the literature and compare them in order to dene more precisely the advantagesof each of them over the others. This should enable to wisely choose the method thatis most likely to provide the best performance for a given situation. In particular, the nonlinearconversion between the Cartesian coordinates with which the state vector is dened and thespherical coordinates used for the measurements is investigated. A measure of nonlinearity isintroduced, studied and used to compare the extended Kalman lter and the particle lter.The size of the detected maritime targets is a special feature that makes it possible to draw amaneuverability-based classication which enables to adapt the tracking technique to be used.Joint tracking and classication (JTC) has already been described in the literature with a specicmeasurement model. This thesis makes this model more realistic using a random distribution ofthe reection point on the target's shape. The tracking method is modied to take into accountthis new measurement model and some simulations are run.This modied JTC algorithm proves to be more ecient than the JTC structure presented inthe literature. Eventually, this thesis shows that nonlinearity is a paramount issue that needsto be considered to implement an ecient self-adapatable radar tracking algorithm, this beingespecially true for extended targets.<br>Maritim malfoljning fran en luftburen radar star infor manga problem pa grund av miljons karaktar, de mal som ska sparas och radarplattformens rorelse. Darfor kan en unik sparningsalgoritminte na basta mojliga prestanda for varje situation som uppstar. Den maste anpassa sig sjalvtill miljon och till de mal som overvakas for att bli sa eektiv som mojligt. Anpassningsformagaar alltsa en viktig fraga inom radarsparning.Flera implementeringar av den matematiska Bayesianska berakningsteorin, vanligtvis kalladelter, har anvants i litteraturen for att forutsaga malbanor sa exakt som mojligt. Beroendepa situationer och antaganden som beaktas forvantas vissa av dem bli battre. Denna avhandlingforeslar att noggrant undersoka sparningsteknikerna som kan hittas i litteraturen ochjamfora dem for att mer precist deniera fordelarna av var och en framfor de andra. Det skulleunderlatta ett klokt val av metoden som mest sannolikt ger basta prestanda for varje given situation.Sarskilt undersoks den icke-linjara omvandlingen mellan kartesiska koordinatsystemet,som denierar tillstandsvektorn, och sfariska koordinater som anvands for matningarna. Ettmatt pa icke-linjaritet presenteras, studeras och anvands for att jamfora ett utokat Kalmanltermed partikelltret.Storleken pa de detekterade maritima malen ar en speciell egenskap som gor det mojligt attgora en klassicering baserad pa manovrerbarhet som hjalper till att anpassa sparningsteknikensom ska anvandas. Simultan foljning och klassiering, "joint tracking and classication" (JTC)pa engelska, har redan beskrivits i litteraturen med en specik matmodell. Denna avhandlinggor modellen mer realistisk med hjalp av en slumpmassig fordelning av reektionspunkten pamalets form. Sparningsmetoden ar modierad for att beakta denna nya matmodell och nagrasimuleringar utfors.Denna modierade JTC-struktur visar sig mer eektiv an JTC-strukturen som presenteras ilitteraturen. Slutligen visar denna avhandling att icke-linjaritet ar en viktig fraga som mastebeaktas for att erhalla en eektiv radarsparningsalgoritm som kan anpassa sig sjalv. Dettagaller sarskilt for utstrackta mal.

Many domains in the real world are richly structured, containing a diverse set of agents characterized by different set of features and related to each other in a variety of ways. Moreover, uncertainty both on the objects observations and on their relations can be present. This is the case of many problems as, for example, multi-target tracking, activity recognition, automatic surveillance and traffic monitoring. The common ground of these types of problems is the necessity of recognizing and understanding the scene, the activities that are going on, who are the actors, their role and estimate their positions. When the environment is particularly complex, including several distinct entities whose behaviors might be correlated, automated reasoning becomes particularly challenging. Even in cases where humans can easily recognize activities, current computer programs fail because they lack of commonsense reasoning, and because the current limitation of automated reasoning systems. As a result surveillance supervision is so far mostly delegated to humans. The explicit representation of the interconnected behaviors of agents can provide better models for capturing key elements of the activities in the scene. In this Thesis we propose the use of relations to model particular correlations between agents features, aimed at improving the inference task. We propose the use of relational Dynamic Bayesian Networks, an extension of Dynamic Bayesian Networks with First Order Logic, to represent the dependencies between an agent’s attributes, the scene’s elements and the evolution of state variables over time. In this way, we can combine the advantages of First Order Logic (that can compactly represent structured environments), with those of probabilistic models (that provide a mathematically sound framework for inference in face of uncertainty). In particular, we investigate the use of Relational Dynamic Bayesian Networks to represent the dependencies between the agents’ behaviors in the context of multi-agents tracking and activity recognition. We propose a new formulation of the transition model that accommodates for relations and present a filtering algorithm that extends the Particle Filter algorithm in order to directly track relations between the agents. The explicit recognition of the relationships between interacting objects can improve the understanding of their dynamic domain. The inference algorithm we develop in this Thesis is able to take into account relations between interacting objects and we demonstrate with experiments that the performance of our relational approach outperforms those of standard non-relational methods. While the goal of emulating human-level inference on scene understanding is out of reach for the current state of the art, we believe that this work represents an important step towards better algorithms and models to provide inference in complex multi-agent systems. Another advantage of our probabilistic model is its ability to make inference online, so that the appropriate cause of action can be taken when necessary (e.g., raise an alarm). This is an important requirement for the adoption of automatic surveillance systems in the real world, and avoid the common problems associated with human surveillance.

Thesis (Ph.D)--Electrical and Computer Engineering, Georgia Institute of Technology, 2009.<br>Committee Chair: Rusell M. Mersereau; Committee Member: Biing Hwang (Fred) Juang; Committee Member: Christopher E. Heil; Committee Member: Georgia Vachtsevanos; Committee Member: James H. McClellan. Part of the SMARTech Electronic Thesis and Dissertation Collection.

[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.

<p>Integrated tracking and detection, based on unthresholded measurements, also referred to as track before detect (TBD) is a hard nonlinear and non-Gaussian dynamical estimation and detection problem. However, it is a technique that enables the user to track and detect targets that would be extremely hard to track and detect, if possible at all with ''classical'' methods. TBD enables us to be better able to detect and track weak, stealthy or dim targets in noise and clutter and particles filter have shown to be very useful in the implementation of TBD algorithms. </p><p>This Master's thesis has investigated the use of particle filters on radar measurements, in a TBD approach.</p><p>The work has been divided into two major problems, a time efficient implementation and new functional features, as estimating the radar cross section (RCS) and the extension of the target. The later is of great importance when the resolution of the radar is such, that specific features of the target can be distinguished. Results will be illustrated by means of realistic examples.</p>

La détection et le pistage de cibles de surface, maritimes ou terrestres, constituent l’un des champs d’application de la surveillance par radar aéroporté. Dans ce contexte spécifique, il s’agit d’estimer les trajectoires d’un ou de plusieurs objets mobiles au cours du temps à partir de mesures radar bruitées. Cependant, plusieurs contraintes s’additionnent au problème d’estimation des trajectoires :1. le nombre d’objets présents dans la région d’intérêt est inconnu et peut évoluer au cours du temps,2. les mesures fournies par le radar ne correspondent pas toutes à des objets mobiles car certaines sont dues à l’environnement ; il s’agit de fausses alarmes,3. une mesure n’est pas toujours disponible pour chaque objet à chaque instant ; il s’agit de non-détections,4. les cibles de surface peuvent être très diverses en termes de capacité de manoeuvre.Pour tenir compte des trois premières exigences, les modèles d’ensembles finis aléatoires peuvent être envisagés pour procéder aux estimations simultanées du nombre d’objets et de leur trajectoire dans un formalisme bayésien. Pour répondre à la quatrième contrainte, une classification des objets à pister peut s’avérer utile. Aussi, dans le cadre de cette thèse, nous nous intéressons à deux traitements adaptatifs qui intègrent ces deux principes.Tout d’abord, nous proposons une approche conjointe de pistage et de classification dédiée au cas d’un objet évoluant en présence de fausses alarmes. Notre contribution réside dans le développement d’un algorithme incorporant un filtre fondé sur un ensemble fini aléatoire de Bernoulli. L’algorithme résultant combine robustesse aux fausses alarmes et capacité à classer l’objet. Cette classification peut être renforcée grâce à l’estimation d’un paramètre discriminant comme la longueur, qui est déduite d’une mesure d’étalement distance.Le second traitement adaptatif présenté dans cette thèse est une technique de pistage de groupes de cibles dont les mouvements sont coordonnés. Chaque groupe est caractérisé par un paramètre commun définissant la coordination des mouvements de ses cibles. Cependant, ces dernières conservent une capacité de manoeuvre propre par rapport à la dynamique de groupe. S’appuyant sur le formalisme des ensembles finis aléatoires, la solution proposée modélise hiérarchiquement la configuration multi-groupes multi-cibles. Au niveau supérieur, la situation globale est représentée par un ensemble fini aléatoire dont les éléments correspondent aux groupes de cibles. Ils sont constitués du paramètredu groupe et d’un ensemble fini aléatoire multi-cibles. Ce dernier contient les vecteurs d’état des cibles du groupe dont le nombre peut évoluer au cours du temps. L’algorithme d’estimation développé est lui-aussi organisé de manière hiérarchique. Un filtre multi-Bernoulli labélisé (LMB) permet d’estimer le nombre de groupes, puis pour chacun d’entre eux, leur probabilité d’existence ainsi que leur paramètre commun. Pour ce faire, le filtre LMB interagit avec un banc de filtres multi-cibles qui opèrent conditionnellement à une hypothèse de groupe. Chaque filtre multi-cibles estime le nombre et les vecteurs d’état des objets du groupe. Cette approche permet de fournir à l’opérationnel des informations sur la situation tactique<br>Detecting and tracking maritime or ground targets is one of the application fields for surveillance by airborne radar systems. In this specific context, the goal is to estimate the trajectories of one or more moving objects over time by using noisy radar measurements. However, several constraints have to be considered in addition to the problem of estimating trajectories:1. the number of objects inside the region of interest is unknown and may change over time,2. the measurements provided by the radar can arise from the environment and do not necessarily correspond to a mobile object; the phenomenon is called false detection,3. a measurement is not always available for each object; the phenomenon is called non-detection,4. the maneuverability depends on the surface targets.Concerning the three first points, random finite set models can be considered to simultaneously estimate the number of objects and their trajectories in a Bayesian formalism. To deal with the fourth constraint, a classification of the objects to be tracked can be useful. During this PhD thesis, we developped two adaptive approaches that take into account both principles.First of all, we propose a joint target tracking and classification method dedicated to an object with the presence of false detections. Our contribution is to incorporate a filter based on a Bernoulli random finite set. The resulting algorithm combines robustness to the false detections and the ability to classify the object. This classification can exploit the estimation of a discriminating parameter such as the target length that can be deduced from a target length extent measurement.The second adaptive approach presented in this PhD dissertation aims at tracking target groups whose movements are coordinated. Each group is characterized by a common parameter defining the coordination of the movements of its targets. However, the targets keep their own capabilities of maneuvering relatively to the group dynamics. Based on the random finite sets formalism, the proposed solution represents the multi-target multi-group configuration hierarchically. At the top level, the overall situation is modeled by a random finite set whose elements correspond to the target groups. They consist of the common parameter of the group and a multi-target random finite set. The latter contains the state vectors of the targets of the group whose number may change over time. The estimation algorithm developed is also organized hierarchically. A labeled multi-Bernoulli filter (LMB) makes it possible to estimate the number of groups, and for each of them, to obtain their probability of existence as well as their common parameter. For this purpose, the LMB filter interacts with a bank of multi-target filters working conditionally to a group hypothesis. Each multi-target filter estimates the number and state vectors of the objects in the group. This approach provides operational information on the tactical situation

Cette thèse s’intéresse au problème de l’inférence bayésienne dans les modèles probabilistes dynamiques. Plus précisément nous nous focalisons sur les méthodes de Monte Carlo pour l’intégration. Nous revisitons tout d’abord le mécanisme d’échantillonnage d’importance avec rééchantillonnage, puis son extension au cadre dynamique connue sous le nom de filtrage particulaire, pour enfin conclure nos travaux par une application à la poursuite multi-cibles.En premier lieu nous partons du problème de l’estimation d’un moment suivant une loi de probabilité, connue à une constante près, par une méthode de Monte Carlo. Tout d’abord,nous proposons un nouvel estimateur apparenté à l’estimateur d’échantillonnage d’importance normalisé mais utilisant deux lois de proposition différentes au lieu d’une seule. Ensuite,nous revisitons le mécanisme d’échantillonnage d’importance avec rééchantillonnage dans son ensemble afin de produire des tirages Monte Carlo indépendants, contrairement au mécanisme usuel, et nous construisons ainsi deux nouveaux estimateurs.Dans un second temps nous nous intéressons à l’aspect dynamique lié au problème d’inférence bayésienne séquentielle. Nous adaptons alors dans ce contexte notre nouvelle technique de rééchantillonnage indépendant développée précédemment dans un cadre statique.Ceci produit le mécanisme de filtrage particulaire avec rééchantillonnage indépendant, que nous interprétons comme cas particulier de filtrage particulaire auxiliaire. En raison du coût supplémentaire en tirages requis par cette technique, nous proposons ensuite une procédure de rééchantillonnage semi-indépendant permettant de le contrôler.En dernier lieu, nous considérons une application de poursuite multi-cibles dans un réseau de capteurs utilisant un nouveau modèle bayésien, et analysons empiriquement les résultats donnés dans cette application par notre nouvel algorithme de filtrage particulaire ainsi qu’un algorithme de Monte Carlo par Chaînes de Markov séquentiel<br>This thesis deals with integration calculus in the context of Bayesian inference and Bayesian statistical filtering. More precisely, we focus on Monte Carlo integration methods. We first revisit the importance sampling with resampling mechanism, then its extension to the dynamic setting known as particle filtering, and finally conclude our work with a multi-target tracking application. Firstly, we consider the problem of estimating some moment of a probability density, known up to a constant, via Monte Carlo methodology. We start by proposing a new estimator affiliated with the normalized importance sampling estimator but using two proposition densities rather than a single one. We then revisit the importance sampling with resampling mechanism as a whole in order to produce Monte Carlo samples that are independent, contrary to the classical mechanism, which enables us to develop two new estimators. Secondly, we consider the dynamic aspect in the framework of sequential Bayesian inference. We thus adapt to this framework our new independent resampling technique, previously developed in a static setting. This yields the particle filtering with independent resampling mechanism, which we reinterpret as a special case of auxiliary particle filtering. Because of the increased cost required by this technique, we next propose a semi independent resampling procedure which enables to control this additional cost. Lastly, we consider an application of multi-target tracking within a sensor network using a new Bayesian model, and empirically analyze the results from our new particle filtering algorithm as well as a sequential Markov Chain Monte Carlo algorithm

The master's thesis deals with target tracking. For this a decentralized sensor network using distributed particle filter with likelihood consensus is used. This consensus is based on a sparse representation of local likelihood function in a suitable chosen dictionary. In this thesis two dictionaries are compared: the widely used Fourier dictionary and our proposed B-splines. At the same time, thanks to the sparsity of distributed data, it is possible to implement compressed sensing method. The results are compared in terms of tracking error and communication costs. The thesis also contains scripts and functions in MATLAB.

Dans le contexte de la surveillance spatiale, nous nous intéressons à un amas de débris évoluant en orbite autour de la Terre et observé par un capteur radar.Il est alors constaté que l'ensemble des débris se disperse selon une forme bananoïdale due à leur mouvement contraint par les lois de Kepler.Cette répartition est représentative d'échantillons gaussiens concentréssur le groupe de Lie SE(3) et peut être complètement caractérisée par unematrice de covariance inconnue.Nous proposons dans cette thèse une reformulation originale sur groupe de Liedu modèle d'observation de l'amas. Ce dernier est alors modélisé comme une cibleétendue caractérisée par sa forme et et son centroïde. De cette manière, nous reconsidéronsl'estimation de ces derniers comme un problème d'inférence sur variété.La géométrie de l’amas est ainsi intrinsèquement prise en compte. Deux algorithmes sur groupes de Liesont alors proposés afin d'estimer respectivement de manière statique et dynamique les paramètres de l'amas.Dans une première partie du manuscrit, l'enjeu de la surveillance spatiale est souligné et les principales méthodes de pistage de débris sont rappelées.Dans une seconde partie, les fondements des groupes de Lie sontprésentés. La troisième partie est axée sur les contributions de la thèse etpropose un modèle et deux algorithmes d'estimation de la forme et du centroïde d’un amas qui sont ensuite testés sur différents scénarios de simulation.La dernière partie est consacrée à une contribution théorique danslaquelle est mise en place une borne d'erreur d'estimation bayésienne sur groupe de Lie<br>In the context of space surveillance, we are interested in a cluster of debris evolving in orbit around the Earth and observed by a radar sensor.It is then observed that the debris spreads out taking a bananoid shape due to their movement constrained by Kepler's laws.This distribution is representative of concentrated Gaussian samples on the Lie group SE (3) and can be completely characterized by anunknown covariance matrix.We propose in this thesis an original reformulation of the cluster observation model on Lie groups. The latter is then modeled as an extended targetcharacterized by its shape and its centroid. In this way, we reconsiderits estimation as a manifold inference problem.The geometry of the cluster is thus intrinsically taken into account. Two algorithms on Lie groups are then proposed in order to estimate respectively statically and dynamically the parameters of the cluster.In the first part of the manuscript, the issue of space surveillance is underlined and the main methods for tracking debris are recalled.In a second part, the foundations of Lie groups arepresented. The third part focuses on the contributions of the thesis andproposes a model and two algorithms for estimating the shape and centroid of a cluster which are then tested on different simulation scenarios.The last part is devoted to a theoretical contribution inwhich is proposed a bound for Bayesian estimation error on Lie groups

Cette thèse s’intéresse au problème de l’inférence bayésienne dans les modèles probabilistes dynamiques. Plus précisément nous nous focalisons sur les méthodes de Monte Carlo pour l’intégration. Nous revisitons tout d’abord le mécanisme d’échantillonnage d’importance avec rééchantillonnage, puis son extension au cadre dynamique connue sous le nom de filtrage particulaire, pour enfin conclure nos travaux par une application à la poursuite multi-cibles.En premier lieu nous partons du problème de l’estimation d’un moment suivant une loi de probabilité, connue à une constante près, par une méthode de Monte Carlo. Tout d’abord,nous proposons un nouvel estimateur apparenté à l’estimateur d’échantillonnage d’importance normalisé mais utilisant deux lois de proposition différentes au lieu d’une seule. Ensuite,nous revisitons le mécanisme d’échantillonnage d’importance avec rééchantillonnage dans son ensemble afin de produire des tirages Monte Carlo indépendants, contrairement au mécanisme usuel, et nous construisons ainsi deux nouveaux estimateurs.Dans un second temps nous nous intéressons à l’aspect dynamique lié au problème d’inférence bayésienne séquentielle. Nous adaptons alors dans ce contexte notre nouvelle technique de rééchantillonnage indépendant développée précédemment dans un cadre statique.Ceci produit le mécanisme de filtrage particulaire avec rééchantillonnage indépendant, que nous interprétons comme cas particulier de filtrage particulaire auxiliaire. En raison du coût supplémentaire en tirages requis par cette technique, nous proposons ensuite une procédure de rééchantillonnage semi-indépendant permettant de le contrôler.En dernier lieu, nous considérons une application de poursuite multi-cibles dans un réseau de capteurs utilisant un nouveau modèle bayésien, et analysons empiriquement les résultats donnés dans cette application par notre nouvel algorithme de filtrage particulaire ainsi qu’un algorithme de Monte Carlo par Chaînes de Markov séquentiel<br>This thesis deals with integration calculus in the context of Bayesian inference and Bayesian statistical filtering. More precisely, we focus on Monte Carlo integration methods. We first revisit the importance sampling with resampling mechanism, then its extension to the dynamic setting known as particle filtering, and finally conclude our work with a multi-target tracking application. Firstly, we consider the problem of estimating some moment of a probability density, known up to a constant, via Monte Carlo methodology. We start by proposing a new estimator affiliated with the normalized importance sampling estimator but using two proposition densities rather than a single one. We then revisit the importance sampling with resampling mechanism as a whole in order to produce Monte Carlo samples that are independent, contrary to the classical mechanism, which enables us to develop two new estimators. Secondly, we consider the dynamic aspect in the framework of sequential Bayesian inference. We thus adapt to this framework our new independent resampling technique, previously developed in a static setting. This yields the particle filtering with independent resampling mechanism, which we reinterpret as a special case of auxiliary particle filtering. Because of the increased cost required by this technique, we next propose a semi independent resampling procedure which enables to control this additional cost. Lastly, we consider an application of multi-target tracking within a sensor network using a new Bayesian model, and empirically analyze the results from our new particle filtering algorithm as well as a sequential Markov Chain Monte Carlo algorithm

Táto práca sa zaoberá odstraňovaním šumu, ktorý vzniká z tzv. multilateračných meraní leteckých cieľov. Na tento účel bude využitá najmä teória Bayesovských odhadov. Odvodí sa aposteriórna hustota skutočnej (presnej) polohy lietadla. Spolu s polohou (alebo aj rýchlosťou) lietadla bude odhadovaná tiež geometria trajektórie lietadla, ktorú lietadlo v aktuálnom čase sleduje a tzv. procesný šum, ktorý charakterizuje ako moc sa skutočná trajektória môže od tejto líšiť. Odhad spomínaného procesného šumu je najdôležitejšou časťou tejto práce. Je odvodený prístup maximálnej vierohodnosti a Bayesovský prístup a ďalšie rôzne vylepšenia a úpravy týchto prístupov. Tie zlepšujú odhad pri napr. zmene manévru cieľa alebo riešia problém počiatočnej nepresnosti odhadu maximálnej vierohodnosti. Na záver je ukázaná možnosť kombinácie prístupov, t.j. odhad spolu aj geometrie aj procesného šumu.

Ce manuscrit propose une approche méthodologique pour la constitution d’une chaîne complète de vidéosurveillance pour des environnements naturellement encombrés. Nous identifions et levons un certain nombre de verrous méthodologiques et technologiques inhérents : 1) à l’acquisition de séquences vidéo en milieu naturel, 2) au traitement d’images, 3) au suivi multi-cibles, 4) à la découverte et la modélisation de motifs comportementaux récurrents, et 5) à la fusion de données. Le contexte applicatif de nos travaux est la surveillance apicole, et en particulier, l’étude des trajectoires des abeilles en vol devant la ruche. De ce fait, cette thèse se présente également comme une étude de faisabilité et de prototypage dans le cadre des deux projets interdisciplinaires EPERAS et RISQAPI (projets menées en collaboration avec l’INRA Magneraud et le Muséum National d’Histoire Naturelle). Il s’agit pour nous informaticiens et pour les biologistes qui nous ont accompagnés, d’un domaine d’investigation totalement nouveau, pour lequel les connaissances métiers, généralement essentielles à ce genre d’applications, restent encore à définir. Contrairement aux approches existantes de suivi d’insectes, nous proposons de nous attaquer au problème dans l’espace à trois dimensions grâce à l’utilisation d’une caméra stéréovision haute fréquence. Dans ce contexte, nous détaillons notre nouvelle méthode de détection de cibles appelée segmentation HIDS. Concernant le calcul des trajectoires, nous explorons plusieurs approches de suivi de cibles, s’appuyant sur plus ou moins d’a priori, susceptibles de supporter les conditions extrêmes de l’application (e.g. cibles nombreuses, de petite taille, présentant un mouvement chaotique). Une fois les trajectoires collectées, nous les organisons selon une structure de données hiérarchique et mettons en œuvre une approche Bayésienne non-paramétrique pour la découverte de comportements émergents au sein de la colonie d’insectes. L’analyse exploratoire des trajectoires issues de la scène encombrée s’effectue par classification non supervisée, simultanément sur des niveaux sémantiques différents, et où le nombre de clusters pour chaque niveau n’est pas défini a priori mais est estimé à partir des données. Cette approche est dans un premier temps validée à l’aide d’une pseudo-vérité terrain générée par un Système Multi-Agents, puis dans un deuxième temps appliquée sur des données réelles<br>This manuscript provides the basis for a complete chain of videosurveillence for naturally cluttered environments. In the latter, we identify and solve the wide spectrum of methodological and technological barriers inherent to : 1) the acquisition of video sequences in natural conditions, 2) the image processing problems, 3) the multi-target tracking ambiguities, 4) the discovery and the modeling of recurring behavioral patterns, and 5) the data fusion. The application context of our work is the monitoring of honeybees, and in particular the study of the trajectories bees in flight in front of their hive. In fact, this thesis is part a feasibility and prototyping study carried by the two interdisciplinary projects EPERAS and RISQAPI (projects undertaken in collaboration with INRA institute and the French National Museum of Natural History). It is for us, computer scientists, and for biologists who accompanied us, a completely new area of investigation for which the scientific knowledge, usually essential for such applications, are still in their infancy. Unlike existing approaches for monitoring insects, we propose to tackle the problem in the three-dimensional space through the use of a high frequency stereo camera. In this context, we detail our new target detection method which we called HIDS segmentation. Concerning the computation of trajectories, we explored several tracking approaches, relying on more or less a priori, which are able to deal with the extreme conditions of the application (e.g. many targets, small in size, following chaotic movements). Once the trajectories are collected, we organize them according to a given hierarchical data structure and apply a Bayesian nonparametric approach for discovering emergent behaviors within the colony of insects. The exploratory analysis of the trajectories generated by the crowded scene is performed following an unsupervised classification method simultaneously over different levels of semantic, and where the number of clusters for each level is not defined a priori, but rather estimated from the data only. This approach is has been validated thanks to a ground truth generated by a Multi-Agent System. Then we tested it in the context of real data

Research Doctorate - Doctor of Philosophy (PhD)<br>Since its initial conception in the mid 1960's, multiple point-target tracking has been at the heart of many applications; spanning from surveillance and air-traffic control, to medicine and econometrics. In modern times, a generalisation of point-target tracking, referred to as extended target tracking has become increasingly popular, due to its ability to handle the generation of multiple measurement per target in each time-step. Today, multiple extended-target tracking is one of the most crucial components in vehicle autonomy - enabling for autonomous vehicles to avoid imminent collision with moving objects in the scene. The focus of this thesis is on the tracking of multiple extended targets in a robust and computationally efficient manner. One of the key developments of this work is the proposal of a new class of state transition models that afford closed-form predictions for the tracking of extended targets. This model builds upon the immensely popular random matrix model, and employs a non-central inverse Wishart distribution to represent the state transition density of the target extent. Importantly, this action results in a prediction update that experiences less overconfidence in its estimation quality than previous works, and offers an additional tuning parameter to model unforeseen deformations in the target extent. This proposed prediction update is then generalised to obtain an algorithm that can track extended targets undergoing kinematic state dependent rotations - no matter the size of the turn-rate variance. To compliment the above prediction schemes, we additionally derive a new correction update for the factorised random matrix model; which utilises an alternative conditional expectation to produce better estimates of the target extent. To take full advantage of this correction update, a new multiple model approach is derived; which, by additionally considering the above generalised prediction, results in an extended-target tracking filter with superior tracking performance than state-of-the-art alternatives. To combat the problem of combinatorial data association, a new partitioning scheme for multiple extended target tracking is also derived in this work. A key innovation is to employ Gibbs sampling to obtain a subset of high-quality partitions from a Dirichlet process Gaussian mixture model. This partitioning scheme is integrated into the Gamma Gaussian inverse Wishart Poisson multi-Bernoulli mixture filter, and shown to produce better overall tracking performance than previous works. Moreover, it is shown that the proposed partitioning scheme possesses the following highly desirable characteristics: it works equally well for spatially close and distant targets; it does not assert any additional assumptions upon the spatial distribution of each measurement; it is less sensitive to the quality of the predicted hypotheses; and finally, the sampling distribution is guaranteed to converge to the true posterior distribution.

This thesis focuses on the problem of automated tracking of tiny cellular and sub-cellular structures, known as particles, in the sequences acquired from total internal reflection fluorescence microscopy (TIRFM) imaging technique. Our primary biological motivation is to develop an automated system for tracking the sub-cellular structures involving exocytosis (an intracellular mechanism) which is helpful for studying the possible causes of the defects in diseases such as diabetes and obesity. However, all methods proposed in this thesis are generalized to be applicable for a wide range of particle tracking applications. A reliable multi-particle tracking method should be capable of tracking numerous similar objects in the presence of high levels of noise, high target density and complex motions and interactions. In this thesis, we choose the Bayesian filtering framework as our main approach to deal with this problem. We focus on the approaches that work based on detections. Therefore, in this thesis, we first propose a method that robustly detects the particles in the noisy TIRFM sequences with inhomogeneous and time-varying background. In order to evaluate our detection and tracking methods on the sequences with known and reliable ground truth, we also present a framework for generating realistic synthetic TIRFM data. To propose a reliable multi-particle tracking method for TIRFM sequences, we suggest a framework by combining two robust Bayesian filters, the interacting multiple model and joint probabilistic data association (IMM-JPDA) filters. The performance of our particle tracking method is compared against those of several popular and state-of-the art particle tracking approaches on both synthetic and real sequences. Although our approach performs well in tracking particles, it can be very computationally demanding for the applications with dense targets with poor detections. To propose a computationally cheap, but reliable, multi-particle tracking method, we investigate the performance of a recent multi-target Bayesian filter based on random finite theory, the probability hypothesis density (PHD) filter, on our application. To this end, we propose a general framework for tracking particles using this filter. Moreover, we assess the performance of our proposed PHD filter on both synthetic and real sequences with high level of noise and particle density. We compare its results from both aspects of accuracy and processing time against our IMM-JPDA filter. Finally, we suggest a framework for tracking particles in a challenging problem where the noise characteristic and the background intensity of sequences change during the acquisition process which make detection profile and clutter rate time-variant. To deal with this, we propose a bootstrap filter using another type of the random finite set based Bayesian filters, the cardinalized PHD (CPHD) filter, composed of an estimator and a tracker. The estimator adaptively estimates the required meta parameters for the tracker such as clutter rate and the detection probability while the tracker estimates the state of the targets. We evaluate the performance of our bootstrap on both synthetic and real sequences under these time-varying conditions. Moreover, its performance is compared against those of our other particle trackers as well as the state-of-the art particle tracking approaches.

abstract: As the demand for spectrum sharing between radar and communications systems is steadily increasing, the coexistence between the two systems is a growing and very challenging problem. Radar tracking in the presence of strong communications interference can result in low probability of detection even when sequential Monte Carlo tracking methods such as the particle filter (PF) are used that better match the target kinematic model. In particular, the tracking performance can fluctuate as the power level of the communications interference can vary dynamically and unpredictably. This work proposes to integrate the interacting multiple model (IMM) selection approach with the PF tracker to allow for dynamic variations in the power spectral density of the communications interference. The model switching allows for a necessary transition between different communications interference power spectral density (CI-PSD) values in order to reduce prediction errors. Simulations demonstrate the high performance of the integrated approach with as many as six dynamic CI-PSD value changes during the target track. For low signal-to-interference-plus-noise ratios, the derivation for estimating the high power levels of the communications interference is provided; the estimated power levels would be dynamically used in the IMM when integrated with a track-before-detect filter that is better matched to low SINR tracking applications.<br>Dissertation/Thesis<br>Masters Thesis Electrical Engineering 2015

碩士<br>國立臺灣大學<br>電機工程學研究所<br>94<br>In visual tracking, multi-target tracking (MTT) systems encounter the problem that unavoidably moving targets may occlude each other and the measurement process of each target becomes dependent. We construct a tracking system with considering joint image likelihood to recognize targets, even though the appearances of the target are identical. Also, the multiple hypotheses of the targets’ depth level are utilized for occlusion handling. In order to enhance system performance, we extend the sampling importance resampling (SIR) particle filter with the separated importance functions for tracking each target and detection. Furthermore, when targets occlude together, the state vector of these targets is transferred into a joint state vector, and the MCMC (Markov Chain Monte Carlo) based particle filter is then proposed for efficient sampling in the high-dimensional joint state during occlusion.　Furthermore, a control strategy for the active camera is proposed in order to move the camera such that the surveillance area will contain the most information. The overall performance is validated in the experiments and shows the robustness with real-time tracking.

University of Technology, Sydney. Faculty of Engineering.<br>Target tracking is a well established field with over fifty years of intense research. While in its core, it deals with estimating targets dynamic states, it is also a critical component of all ” Situation Awareness” and threat assessment systems. These higher layer applications take decisions on important questions like number of targets, positions of them, the instant and position of their initiation, the instant and position of their maneuvers and above all, which of them are threatening and/or friendly. The lower level target tracking algorithms feed the necessary information to these decision taking systems. There are a number of target tracking algorithms to cater for the need of such systems. Most of these available algorithms are based on filtering theory. But it is established that smoothing increases the accuracy of the systems at the expense of a slight lag between the instant of estimation and the instant at which the parameter of interest is being estimated. Hence smoothing is not widely used for practical target tracking applications. However, the situation awareness system is expected to perform better if more precise information is obtained about initiation and termination of the targets along with improved discrimination of true/false targets. This thesis addresses the problem of improved track initiation and maintenance with the smoothing framework to provide better information. It first reviews target tracking and filtering literature. It introduces the concept of random set smoother and derives the IPDA smoother under linear Gaussian assumption. IPDA smoother is also derived by extending the PDA smoother. Finally a theoretical link is established between Random Set smoothing and IPDA smoothing framework. To extend the domain into multiple sensor scenario, the problem of out-of-sequence measurements is also addressed in this thesis under target existence uncertainty. Several realistic scenarios are simulated and the results are verified.