Intertarget Occlusion Handling in Multiextended Target Tracking Based on Labeled Multi-Bernoulli Filter Using Laser Range Finder

Abstract

Multiextended target tracking (METT) has important applications in many fields, including autonomous vehicles, traffic flow monitoring, etc. However, as an acknowledged METT algorithm, the labeled multi-Bernoulli (LMB) filter cannot handle the intertarget occlusion problem, which frequently observed among METT, leading to an estimated trajectory break and even target loss. To tackle this problem, this article proposes an improved LMB filter with intertarget occlusion handling ability for METT. First, an intertarget occlusion probability (IOP) model based on a labeled random finite set using a laser range finder is developed. The IOP model theoretically describes the probability that each tracked target being occluded, considering both partial and full occlusion. Second, a true detection probability (TDP) model is presented to describe the probability that a target will generate corresponding measurement of a sensor in the presence of occlusion. The TDPs of tracked targets provide a way to integrate their estimated occlusion probabilities into the LMB filter to improve its capability of handling occluded tracks. We carried out multiple vehicle tracking simulation and field test using a laser range finder. Both simulation and experiment results illustrate that our improved LMB filter with occlusion handling ability can successfully tackle the intertarget occlusion problem and outperform conventional algorithms.