Research On Multi-sensor Multi-target Tracking Method Based On Phd Estimating

As one of the research hotpots in the field of multi-sensor information fusion,multi-target tracking technology has a wide range of applications both in military affair as well as civil matters.Traditional multi-target tracking method mainly solves the problem of multi-target data association based on the classical probability theory,however it will be affected by complicated environmental factors,such as unknown target number,dense clutter and low detection rate,and in the problems of poor data correlation and the degeneration in tracking accuracy.In recent years,probability hypothesis density estimation method provides a new idea for the multi-target tracking problem.This method describe the state set of the target and sensor measurement set uniformly in a probability hypothesis density space using the stochastic finite set theory,thus can avoid the data association problem in the traditional target tracking algorithm effectively.However,the current multi-sensor PHD multi-target tracking technology is not yet mature.On the one hand,most multi-target tracking methods based on stochastic finite sets are single-sensor probabilistic hypothetical density filtering method.However,it is difficult to rely on single sensor to maintain a stable and accurate multi-target estimation in the complex environment.In such case,one needs to fusing multi-sensor data to achieve better results.On the other hand,most PHD based multi-sensor fusion tracking techniques are based on the assumption at the data from multiple sensors are synchronous,and there are less methods to cope with the asynchronous fusing problemTo solve the above mentioned problems,in this paper,we research the methods of multi-sensor multi-target tracking based on the PHD estimating technique.Specific content is as follows:Firstly,review the multi-sensor multi-target tracking method based on PHD estimating related theory,including the target tracking system model,PHD estimation theory,data fusion and asynchronous sampling multi-sensor time registration method.Secondly,in order to address the problem of tracking degeneration in single sensor PHD filtering applications with dense clutter environment and low detection rate,the paper proposed a multi-sensor multi-stage fusion framework,a multi-target correlation algorithm and an improved convex combination fusion algorithm.Then combined with the GM-PHD filtering method,a multi-sensor multi-level fusion multi-target tracking algorithm is proposed.Simulation results show that the proposed method is effective.Finally,in order to address the problem of that the asynchronous sampling sensors in clutter environment can't maintain high-quality tracking results.By constructing multi-sensor asynchronous detection tracking frameworks for multi-target tracking,a time-synchronization model based on original sensor measurement and a GM-PHD time synchronization method are proposed.Then combined with Gaussian mixture probability hypothesis density filter,TSBF and FBTS algorithm are proposed.And simulation results show that the proposed method is effective.