Multi-object Bayesian filters with amplitude information in clutter background

In many radar or sonar tracking applications, the amplitude information (AI) is known to improve data association and target state estimation in most of multi-object filters. However, when considering targets in noisy backgrounds, existing multi-object filters rely on a number of assumptions, relating to the uniformity of the spatial distribution of the clutter and amplitude distribution of the clutter being Rayleigh. These assumptions are seldom held under realistic conditions, and as such, the underlying multi-object filters deliver a sub-optimal tracking performance. In this paper, we incorporate the AI as part of the multi-object filtering process to render very novel filters that can handle multi-object tracking in much more difficult and realistic conditions. In particular, we propose an inverse Gamma Gaussian Model for the target and clutter state, consisting of kinematic state and return power. We then develop the inverse Gamma Gaussian Mixture (IGGM) implementation of the RFS filters with AI. Simulations show that proposed filters, in particular when combined with clutter estimation and its RFS approximation, are more robust in handling a number of realistic cases when compared against existing filters.