Research on improved box-particle filter in CPHD multi-target tracking

Abstract

The Cardinalized probability hypothesis density(CPHD) filter based on box-particle(BP) technology can effectively enhance the tracking accuracy of numbers of the multi-target tracking. And it also solved the tracking problem with unknown number of targets and missing targets. Point the problems of lack of tracking accuracy and missing tracking targets in the BP-CPHD algorithm, this paper presents some detailed measures about improving tracking. Based on the particle filter theory, we found that in the process of resampling, the particle size distribution cannot be optimized. So, in this paper, we replaced the traditional random sub sampling resampling method with the partition resampling method to improve the method of box-particle filter. Then we simulate and experiment the improved algorithm combined with CPHD filtering. The experimental results demonstrate that the improved method of tracking multi-target with the box-particle CPHD filter is more accurate than the original algorithm and the OSPA distance is smaller, so the effect of the improved tracking method is better.