Maintaining multimodality through mixture tracking

Abstract

In recent years particle filters have become a tremendously popular tool to perform tracking for non-linear and/or non-Gaussian models. This is due to their simplicity, generality and success over a wide range of challenging applications. Particle filters, and Monte Carlo methods in general, are however poor at consistently maintaining the multi-modality of the target distributions that may arise due to ambiguity or the presence of multiple objects. To address this shortcoming this paper proposes to model the target distribution as a non-parametric mixture model, and presents the general tracking recursion in this case. It is shown how a Monte Carlo implementation of the general recursion leads to a mixture of particle filters that interact only in the computation of the mixture weights, thus leading to an efficient numerical algorithm, where all the results pertaining to standard particle filters apply. The ability of the new method to maintain posterior multi-modality is illustrated on a synthetic example and a real world tracking problem involving the tracking of football players in a video sequence.

A Bayesian Framework for Multi-cue 3D Object Tracking

“…dealing with discrete/continuous state spaces [9,11], multiple target tracking [12,15,21], or multiple sources of information [10,17]. The latter has involved techniques such as importance sampling [10] or democratic integration [17,19], and have been used to combine visual cues such as edge and texture in a particle filter framework.…”

Evaluation of Approaches for Tracking Virus Particles in Fluorescence Microscopy Images

“…When tracking multiple objects, multiple modes arise in the posterior distribution. The multimodality can be modeled via a non-parametric M -component mixture model, which can be computed using a mixture of particle filters [7]. Alternatively, one may track multiple objects by instantiating one independent particle filter per object (e.g., [8]).…”

A Multi-Scan Mixture Particle Filter for Joint Detection and Tracking of Multiple Targets

“…As an example, Bayesian [12,13] and artificial neural networks [14,15] techniques have been successfully used, considering a single-scan strategy approach. As pointed out in [16], the posterior distribution of multiple target state is a multi-mode distribution and each mode corresponds to either a target or clutter. A mixture particle filter method is developed in [16], where each mode is modeled with an individual particle filter that forms part of the mixture.…”

“…As pointed out in [16], the posterior distribution of multiple target state is a multi-mode distribution and each mode corresponds to either a target or clutter. A mixture particle filter method is developed in [16], where each mode is modeled with an individual particle filter that forms part of the mixture. The mixture particle filter avoids the dimension problem by exploring the particle filter's ability to track multiple targets in a single-target state space.…”

“…Different from [16], a general global posterior distribution is adopted in this work, which consists of existing components and new components. The existing components could be further divided into target components and undetermined components (including potential targets and clutter).…”