Multi-sensor Multi-target Tracking Algorithm Based On Probability Hypothesis Density Fusion

As one of the important research directions of information fusion,multi-target tracking technology has a wide range of applications in both military and civil fields.The traditional multi-target tracking method is to deal with the data association problem by solving the matching relationship between targets and observations.However,the data association problem becomes extremely difficult due to the complicated environmental factors,such as the increase of the target number,low detection rate and dense clutter.Probability hypothesis density(PHD)filter provides a new solution to multi-target tracking problem,this method describes the target set and measurement set in a probability density space based on the random finite sets,thus the data association problem can be avoided effectively.Nevertheless,the current research on PHD filtering is not yet mature,it is because most of the traditional PHD methods are proposed for single sensor,but it is difficult to rely on single sensor to estimate the targets accurately in complex scenarios,so we can fuse the data from multiple sensors to achieve better results.So based on the Gaussian mixture probability hypothesis density(GMPHD)filter,this paper studies the multi-sensor multi-target tracking problem under synchronous and asynchronous conditions,the main results are as follows:1)As the performance of single sensor GMPHD tracking method will degenerate in complex environment,this paper propose a multi-sensor sequential fusion framework and a multi-sensor fusion method based on the GMPHD filter.Then for the problems of fusion sequence sensitivity and the balance of fusion weight,an adaptive multi-sensor fusion method for the unbalanced weight GMPHD filter is further proposed.The simulation results show that,this algorithm can effectively improve the tracking accuracy compared with the traditional algorithm.2)For the multi-target tracking problem under random finite sets,an adaptive multi-sensor tracking algorithm based on the iterative updating GMPHD filter is proposed.Firstly,we construct a multi-sensor iterative updating GMPHD filter framework.Then,a method to fuse data from multiple sensors based on this framework is proposed.Finally,we study the influence of sensor iteration sequence with different observation quality on the fusion results,and propose an adaptive sorting method.The simulation results show that the proposed algorithm can improve the tracking accuracy when the sensor observation quality is different.3)For the asynchronous sampling problem,an asynchronous multi-sensor fusion algorithm based on GMPHD filter is proposed.Firstly,we construct an asynchronous multi-sensor GMPHD tracking framework.Then,to solve the data synchronization problem,we propose a time registration method based on state extrapolation.At last,an improved covariance intersection algorithm is proposed to fuse the posterior estimates.Simulation results shows that,compared with the single-sensor GMPHD method,the proposed method is more accurate and robust.