Research On Probability Hypothesis Density Filter Algorithm Under The Influence Of Relevant Parameters

Multi- target tracking technology traces and locates targets in the target observation area through process integrated measurement information provided by the multi-sensor, paid close attention by some relevant fields, such as civilian, military, aerospace and so on. Breaking the bondage of data association, Multi-target tracking problem based on random set theory has been a hot topic in the field of information fusion in recent years. Based on random set theory, the paper focuses on Multi- target tracking problem under the influence of relevant parameters.Firstly, Gaussian mixture probability hypothesis density(GM-PHD) filter under high clutter rate and low detection rate has been researched. In this environment, the number of targets will be false by using standard GM-PHD filter and seriously affects the accuracy of Multi-target tracking. For this problem, an improved GM-PHD filter algorithm proposed in this paper, a track-estimate association is implemented in the filtering process and a numerical interpolation technique is used on each target track after association. Simulation results show that the proposed GM-PHD algorithm can accurately track the target.Secondly, this paper studies Gaussian mixture cardinality probability hypothesis density(GM-CPHD) filter at the presence of low detection rate. Compared with GM-PHD filter, GM-CPHD filter can simultaneously estimate the intensity and cardinality probability distribution of targets, having a higher filtering performance. However, in the low detection rate environment, there is still an erroneous number estimation of targets. In order to improve the filtering accuracy further, we propose to make a post-processing algorithm for GM-CPHD filter. The experimental results show that the post-processing algorithm can improve the accuracy of the estimated target number effectively.Finally, this paper studies the influence of Elementary Symmetric Function(ESF) on the implementing Effic iency of GM-CPHD Filter. The time complexity of GM-CPHD filter is high, especially in the case of a high noise rate, the filtering time is too long. The main reason is that the high computational complexity of ESF in the update step of the algorithm. To solve this problem, we replace the definition method with a recursive method to calculate ESF, improving the computational efficiency. Simulation results show that the proposed method can reduce the time complexity of the algorithm, and do not affect the accuracy of the filter.