A novel SMC-PHD filter for multi-target tracking without clustering

Abstract

The probability hypothesis density (PHD) filter has been regarded as a practicable multi-target tracking algorithm which can alleviate the computational difficulty of the multi-target Bayes filter due to the multiple high-dimensional space integrals. As one of the major implementations of the PHD filter, sequential Monte Carlo PHD (SMC-PHD) is suitable for highly nonlinear systems. However, in scenarios with missing detections, the false-estimation problem occurs which leads to estimation performance degradation. To reduce the negative effects of missing detections, this paper develops a compensatory measurements generating mechanism and presents a novel measurement compensation based SMC-PHD filter, which can avoid the unreliable clustering. Comparative results verify the proposed filter, indicating good application prospects.