A Novel Multiple Hypothesis Tracking Algorithm Integrated with Detection Processing

Abstract

In current multi-targets tracking system, the multiple hypotheses tracking (MHT) algorithm is considered as the preferred algorithm for data association. However, this method is straightforward, and ignores the couple relationship between detector and tracker, which may result system performance loss. In this paper, we propose a novel multiple hypothesis tracking algorithm integrated with detection processing (MHT-IDP) for multi-target tracking. It explores the idea of integrating the detector with the tracker, in which the tracker guides the detector where to search a target, and the detector provides what is discovered. Moreover, we find that the Bayesian detection threshold is lower near the predicted measurement. Simulation results demonstrate that the MHT-IDP algorithm can improve the clutter suppression effect, enhance the detection performance of the target, and achieve better performance in tracking accuracy compared with the MHT algorithm.