There are many multiple-model (MM) target-tracking algorithms that are available but there has yet to be a comparison that includes all of them. This work compares seven of the currently most popular MM algorithms in terms of performance, credibility, and computational complexity. The algorithms to be considered are the autonomous multiple-model algorithm, generalized pseudo- Bayesian of first order, generalized pseudo-Bayesian of second order, interacting multiple-model algorithm, B-Best algorithm, Viterbi algorithm, and reweighted interacting multiple-model algorithm. The algorithms were compared using three scenarios consisting of maneuvers that were both in and out of the model set. Based on this comparison, there is no clear-cut best algorithm but the B-best algorithm performs best in terms of tracking errors and the IMM algorithm has the best computational complexity among the algorithms that have acceptable tracking errors.

No description available.

Target tracking refers to the process of estimating the state of a moving object from remote and noisy measurements. In this thesis we consider the Bayesian filtering framework to perform target tracking under nonlinear models, a target moving in continuous time, and measurements that are available in discrete time intervals (known as continuous-discrete). The Bayesian filtering theory establishes the mathematical basis to obtain the posterior probability density function of the state, given the measurement history. This probability density function contains all the information required about the state of the target. It is well documented that there is no exact solution for posterior density under the models mentioned. Hence, the approximation of such density functions have been studied for over four decades. The literature demonstrates that this has led to the development of multiple filters. In target tracking, due to the remote sensing performed, an additional complication emerges. The measurements received are not always from the desired target and could have originated from unknown sources, thus making the tracking more difficult. This problem is known as a measurement origin uncertainty. Additionally to the filters, different methods have been proposed to address the measurement origin uncertainty due to its negative impact, which could cause a false track. Unfortunately, a final solution has yet to be achieved. The first proposal of this thesis is a new approximate Bayesian filter for continuous-discrete systems. The new filter is a higher accuracy version of the cubature Kalman filter. This filter is developed using a fifth-degree spherical radial cubature rule and the Ito-Taylor expansion of order 1.5 for dealing with stochastic differential equations. The second proposal is an improved version of the probabilistic data association method. The proposed method utilizes the maximum likelihood values for selecting the measurements that are used for the data association. In the first experiment, the new filter is tested in a challenging 3-dimensional turn model, demonstrating superiority over other existing filters. In a second and third experiments, the proposed data association method is tested for target tracking in a 2-dimensional scenarios under heavy measurement origin uncertainty conditions. The second and third experiments demonstrate the superiority of the proposed data association method compared to the probabilistic data association. / Thesis / Doctor of Philosophy (PhD)

In this thesis, multiple ground target tracking algorithms are studied. From different aspects of the ground target tracking, three different types of tracking algorithms are proposed according to the specialties of the ground target motion and sensors employed. Firstly, the dependent target tracking for ground targets is studied. State dependency is a common assumption in traditional target tracking algorithms, while this may not be the true in ground target tracking as the motion of targets are constraint to certain path. To enhance the tracking algorithm for ground targets, starting with the dependency assumption, Markov Random Field (MRF) based Probabilistic Data Association (PDA) approach is derived to associate motion dependent targets. The driving behavior model is introduced to describe motion relationship among targets. The Posterior Cramer-Rao Lower Bound (PCRLB) is derived for this new motion model. Experiments and simulations show that the proposed algorithm can reduce the false associations and improve the predictions. Eventually, the proposed approach alleviates issues like the track impurity and coalescence problem and achieves better performance comparing to standard trackers assuming state independence. Ground target tracking using cameras is then studied. To build an efficient multi- target visual tracking algorithm, fast single target visual tracking is an important component. A novel visual tracking algorithm that has high speed and better or comparable performance to state-of-the-art trackers is proposed. The proposed approach solves the tracking task by using a mixed-motion proposal based particle filter with Ridge Regression observation likelihood calculation. This approach largely reduces the exhaustive searching in common state-of-art trackers while maintains efficient representation of the target appearance change. Experiments on 100 public benchmark videos, as well as a high frame rate benchmark, are carried out to compare the performance with the state-of-art published algorithms. The results of the experiment show the proposed tracker achieves good performance while beats other algorithms in speed with a large margin. The proposed visual target tracker is integrated into a new multiple ground tar- get tracking algorithm using a single camera. The multi-target tracker addresses the issues in the target detection, data association and track management aside from the single target tracker. A perspective aware detection algorithm utilizing the re- cent advanced Convolutional Neural Networks (CNN) based detector is proposed to detect multiple ground targets and alleviate the weakness of CNN detectors in detecting small objects. A hierarchical class tree based multi-class data association is presented to solve the multi-class association problem with potential misclassified detections. Track management is also improved utilizing the high efficiency detectors and a Support Vector Machine (SVM) based track deletion is proposed to correctly remove the dead tracks. Benchmarking is presented in experiments and results are analyzed. A case study of applying the proposed algorithm is provided demonstrating the usefulness in real applications. / Thesis / Doctor of Philosophy (PhD)

Multi-target tracking is the process that allows the extraction of object motion patterns of interest from a scene. Motion patterns are often described through metadata representing object locations and shape information. In the first part of this thesis we discuss the state-of-the-art methods aimed at accomplishing this task on monocular views and also analyse the methods for evaluating their performance. The second part of the thesis describes our research contribution to these topics. We begin presenting a method for multi-target tracking based on track-before-detect (MTTBD) formulated as a particle filter. The novelty involves the inclusion of the target identity (ID) into the particle state, which enables the algorithm to deal with an unknown and unlimited number of targets. We propose a probabilistic model of particle birth and death based on Markov Random Fields. This model allows us to overcome the problem of the mixing of IDs of close targets. We then propose three evaluation measures that take into account target-size variations, combine accuracy and cardinality errors, quantify long-term tracking accuracy at different accuracy levels, and evaluate ID changes relative to the duration of the track in which they occur. This set of measures does not require pre-setting of parameters and allows one to holistically evaluate tracking performance in an application-independent manner. Lastly, we present a framework for multi-target localisation applied on scenes with a high density of compact objects. Candidate target locations are initially generated by extracting object features from intensity maps using an iterative method based on a gradient-climbing technique and an isocontour slicing approach. A graph-based data association method for multi-target tracking is then applied to link valid candidate target locations over time and to discard those which are spurious. This method can deal with point targets having indistinguishable appearance and unpredictable motion. MT-TBD is evaluated and compared with state-of-the-art methods on real-world surveillance.

Network resources (such as bandwidth on a link) are not unlimited, and must be shared by all networked applications in some manner of fairness. This calls for the development and implementation of effective strategies that enable optimal utilization of these scarce network resources among the various applications that share the network. Although several rate controllers have been proposed in the literature to address the issue of optimal rate allocation, they do not appear to capture other factors that are of critical concern. For example, consider a battlefield data fusion application where a fusion center desires to allocate more bandwidth to incoming flows that are perceived to be more accurate and important. For these applications, network users should consider transmission rates of other users in the process of rate allocation. Hence, a rate controller should consider application specific rate coordination directives given by the underlying application. The work reported herein addresses this issue of how a rate controller may establish and maintain the desired application specific rate coordination directives. We identify three major challenges in meeting this objective. First, the application specific performance measures must be formulated as rate coordination directives. Second, it is necessary to incorporate these rate coordination directives into a rate controller. Of course, the resulting rate controller must co-exist with ordinary rate controllers, such as TCP Reno, in a shared network. Finally, a mechanism for identifying those flows that require the rate allocation directives must be put in place. The first challenge is addressed by means of a utility function which allows the performance of the underlying application to be maximized. The second challenge is addressed by utilizing the Network Utility Maximization (NUM) framework. The standard utility function (i.e. utility function of the standard rate controller) is augmented by inserting the application specific utility function as an additive term. Then the rate allocation problem is formulated as a constrained optimization problem, where the objective is to maximize the aggregate utility of the network. The gradient projection algorithm is used to solve the optimization problem. The resulting solution is formulated and implemented as a window update function. To address the final challenge we resort to a machine learning algorithm. We demonstrate how data features estimated utilizing only a fraction of the flow can be used as evidential input to a series of Bayesian Networks (BNs). We account for the uncertainty introduced by partial flow data through the Dempster-Shafer (DS) evidential reasoning framework.

In modern systems for air surveillance, it is important to have a high quality situationassessment. SAAB has a system for air surveillance, and in this thesis possibleimprovements of the tracking performance of this system are explored. The focushas been on improving the tracking of highly maneuverable targets observed withlow sampling rate. To evaluate improvements of the tracking performance, a componentthat is similar to the one used in SAAB’s present tracker was implementedin an Interacting Multiple Model (IMM) structure. The use of an Auxiliary ParticleFilter for improving the tracking performance is explored, and a way to fita particle filter into SAAB’s existing IMM framework is proposed. The differentfilters were implemented in Matlab, and evaluation was done by the meansof Monte Carlo simulations. The results from Monte Carlo simulations show significantimprovement when tracking in two dimensions. However, the results inthree dimensions do not display any substantial overall improvement when usingthe particle filter compared to using SAAB’s present filter. It is therefore notworthwhile to switch the filter used in SAAB’s present tracker for a particle filter,at least not under the high SNR circumstances presented in this thesis. However,further studies within this area are recommended before any final decisions aremade.

We consider the topic of target tracking in this thesis. Target tracking is one of the applications in wireless sensor networks (WSNs). Clustering approach prolongs sensor¡¦s lifetime and provides better data aggregation for WSNs. Most previous researches assumed that cluster regions are disjointed, while others assigned overlapping cluster regions, and utilized them in some applications, including inter-cluster routing and time synchronization. However, in overlapping clustering, processing of redundant sensing data may impair system performance. We present a regular distributed overlapping WSN in this thesis. The network is based on two kinds of sensors: (1) high-capability sensors, which are assigned as cluster heads (CHs), responsible for data processing and inter-cluster communication, (2) normal sensors, which are in a larger number when comparing with the high-capability sensors, the function of normal sensors are to provide data to the CHs. We define several operating modes of CHs and sensors. WSN works more efficient under the settings. Since a target may be located in the overlapping region, redundant data processing problem exists. To solve the problem, we utilize Cholesky decomposition to decorrelate the measurement noise covariance matrices. The correlation will be eliminated during the process. In addition, we modify extended information filter (EIF) and adapt to the decorrelated data. The CHs track the target, fuse the information from other CHs, and implement distributed positioning. The simulations are based on ultra-wideband (UWB) environment, we have verified that the proposed scheme works more efficient under the setting of different modes. The performance with decorrelated measurement is better than that with correlated ones.

In decentralized data fusion system, if the probability model of the noise is Gaussian and the innovation informations from the sensors are uncorrlated,the information filtering technique can be the best method to fuse the information from different sensors. However, in the realistic environments, information filter cannot provide the best solution of state estimation and data integration when the noises are non-Gaussian and correlated. Since particle filter are capable of dealing with non-linear and non-Gaussian problems, it is an intuitive approach to replace the information filter by particle filter with some suitable data fusion techniques.In this thesis, we investigate a decentralized data fusion system with improved particle filters for target tracking. In order to achieve better tracking performance, the Iterated Extended Kalman Filter framework is used to incorporate the newest observations into the proposal distribution of the particle filter. In our proposed architecture, each sensor consists of one particle filter, which is used in generating the local statistics of the system state. Gaussian mixture model (GMM) is adopted to approximate the posterior distribution of the weighted particles in the filters, thereby more compact representations of the distribution for transmmision can be obtained. To achieve information sharing and integration, the GMM-Covariance Intersection algorithm is used in formulating the decentralized fusion solutions. Simulation resluts of target tracking cases in a sensor system with two sensor nodes are given to show the effectiveness and superiorty of the proposed architecture.

In solving target tracking problems, the Kalman filter (KF) is one of the most widely used estimators. Whether the state of target movement adapts to the changes in the observations depends on the model assumptions. The interacting multiple model (IMM) algorithm uses interaction of a bank of parallel Kalman filters to solve the hypothetical model of tracking maneuvering target. Based on the function of soft switching, the IMM algorithm, with parallel Kalman filters of different dimensions, can perform well by adjusting the model weights. Nonetheless, the uncertainty in measured data and the types of sensing systems used for target tracking may still hinder the signal processing in the IMM. In order to improve the performance of target tracking and signal estimation, the concept of data fusion can be adapted in the IMM-based structures. Multiple IMM based estimators can be used in the structure of multi-sensor data fusion. In this thesis, we propose a dual-IMM estimator structure, in which data fusion of the two IMM estimators is achieved by updating associated model probabilities. Suppose that two sensors for measuring the moving target is affected by the different degrees of noise, the measured data can be processed first through two separate IMM estimators. Then, the IMM-based estimators exchange with each other the estimates, model probabilities and model transition probabilities. The dual-IMM estimator will integrate the shared data based on the proposed dual-IMM algorithm. The dual-IMM estimator can be used to avoid degraded performance of single IMM with insufficient data or undesirable environmental effects. The simulation results show that a single IMM estimator with smaller measurement noise level can be used to compensate the other IMM, which is affected by larger measurement noise. Improved overall performance from the dual-IMM estimator is obtained. Generally speaking, the two IMM estimators in the proposed structure achieve better performance when same level of measurement noise is assumed. The proposed dual-IMM estimator structure can be easily extended to multiple-IMM structure for estimation and data fusion.