Improvement Joint Detection and Tracking of Small RCS Targets from Image Observations

Abstract

Small targets tracking from image observations is a particularly difficult problem with unknown measurement variances. The Cardinality-Balanced Multi-target Multi-Bernoulli filter (CBMeMBer) is an optimal Bayes approach of multi-object filtering and it has outstanding performance over the opposite filters. Recently, CBMeMBer supported Variational Bayesian (VB) approximations have implemented for multi-target tracking with unknown measurement variances. However, VB-CBMeMBer filter implemented by known probability of detection, which is unsuitable for small targets tracking. Recently, the new approach known by track-before-detect (TBD) which is an efficient algorithm to track small objects. During this work, a completely unique VB-CBMeMBer-TBD filter presented to unravel the fluctuation problems of small target parameters like the probability of detection and therefore the variances of measurement. The simulation results confirm the robustness and effectiveness of the proposed filter.