In this paper we present an online approach for joint detection and tracking for multiple targets with multiple sensors using sequential Monte Carlo (SMC) methods. There are two main contributions in the paper. The first contribution is the extension of the deterministic detection method proposed in our previous publications to a full SMC context in which the track initiation and termination are executed using Bayesian Monte Carlo methods. In effect the dimensions of the particles are variable, and the number of targets can be obtained by the MAP estimation of the dimensions of these particles. The second contribution is the tracking of maneuvering targets without using multiple-model approaches. This can be achieved by recursively estimating the heading directions of the targets, followed by the sampling of the target state along these directions. In effect the use of multiple models to model target maneuvers may not be necessary. Furthermore there is no limitation on which the number of targets that can be simultaneously handled by proposed algorithm. With the employment of multiple sensors, a central-level tracking strategy is adopted, where the observations from all active sensors are fused together for detection and tracking and a set of global tracks is maintained. To further save in the increased computational load arising as a result of the multisensor scenario, only those observations from different sensors that are close to each other according to a distance metric are used for data association. To cope with the data association between the observations from all active sensors and the targets at a given time, we adopt an efficient 2-D data assignment algorithm. Computer simulations demonstrate that the proposed approach is robust in performing joint detection and tracking for multiple maneuvering targets even though the environment is hostile with high clutter rate and low target detection probability

In this paper we propose methods for tracking multiple maneuvering objects using variable rate particle filters with multiple sensors. Unlike more standard approaches the proposed method assumes that the states change at different and unknown rates compared with the observation process, and hence is able to model parsimoniously the maneuvering behaviour of an object. Furthermore, a Poisson model is used to model both target and clutter measurements, avoiding the data association difficulties associated with traditional tracking approaches. Computer simulations demonstrate the potential of the proposed method for tracking highly maneuverable targets in a hostile environment with high clutter density.

In this paper, we present a new sequential Monte Carlo (SMC) algorithm for online joint multitarget tracking (MTT) and detection in the presence of spurious objects, e.g., clutter. The proposed method provides an efficient solution to deal with two major challenges in MIT problems: 1) time-varying number of targets, and 2) measurement-to-target association. By detecting regions of interest within the surveillance region and monitoring their appearance and disappearance, we are able to estimate the number of targets, even when the environment is hostile with low target detection probability and high clutter density. Adopting an efficient 2-D data assignment algorithm that computes all feasible assignments subject to certain constraints, we are able to efficiently and effectively marginalize the association hypotheses from the likelihood junction. Subsequently, we utilize SMC methods, also known as particle filters, to recursively and jointly estimate the multitarget states. Computer simulations and performance evaluation demonstrate the robustness of the proposed method for multitarget detection and tracking within a hostile environment in terms of high clutter density and low target detection probability.

In this paper, we propose an extension of the soft-gating approach for measurement-to-target assignment for multitarget tracking. Given the latest observation and a set of multitarget particles, the proposed method combines efficient m-best 2D data assignment and sampling methods to compute a feasible measurement-to-target assignment with an associated probability for each particle. The particles containing the multitarget states and the association vectors can then be used to recursively estimate the posterior distribution of the targets using sequential Monte Carlo methods. Computer simulations demonstrate the robustness and effectiveness of the proposed method for data association and multitarget tracking.