Multi-target Tracking Based On Box Particle Probability Hypothesis Density Filter

Target tracking is the process of using multi-source heterogeneous sensors to estimate the state of selected moving targets.Whether it is civilian use to meet people's daily travel needs,or military use,as a key technology in modern high-tech warfare and aerospace,has always been a research focus of domestic and foreign experts and scholars.In this paper,based on the PHD and CPHD filters under the framework of random finite set theory,combined with the box particle filter method,the problems of multi-target tracking and multi-extended target tracking are studied.The main work is as follows:1.Aiming at the existing multi-target tracking problem,the idea of Interacting Multiple Model(IMM)is combined with the Box Probability Hypothesis Density Filter(Box-PHD).In the environment,the problem of redundancy and inaccurate target position estimation caused by the large box body in the environment,a partitioned interacting multiple model box particle probability hypothesis density filter(PIMM-Box-PHD)algorithm is proposed.The algorithm first introduces IMM prediction for the maneuvering problem of multiple targets in the prediction stage,and uses the multi-model interaction method to solve the model mismatch problem when the target is moving;secondly,the box partition technology is used to divide the predicted box particles into the same size and weight.In order to improve the accuracy of target position estimation;Finally,Box-PHD is used to update the interval measurement of the divided small box particle set.Experiments are used to verify the good performance of the PIMM-Box-PHD algorithm proposed in this paper in elliptical multi-maneuvering target tracking,and its advantages compared with the IMM-Box-PHD algorithm in target position estimation.2.Aiming at the problems of excessive number of divisions and high computational complexity in the current extended target tracking measurement division method,the density peak fast clustering algorithm CFSFDP(Clustering by fast search and find of density peaks)is combined with the box particle potential probability assumption.Combining Box Cardinalized Probability Hypothesis Density Filter(Box-CPHD),a box particle CPHD extended target filtering algorithm based on CFSFDP algorithm is proposed.The algorithm uses CFSFDP to divide and expand the target measurement set.Based on the difference of the measurement information density,it can effectively divide the interval measurement and eliminate the clutter measurement.Then Box-CPHD is used for prediction update and target state estimation.The simulation experiment shows that compared with the classic distance division method,CFSFDP can significantly reduce the running time under the premise of achieving the same effect;in the high clutter environment after removing the clutter,the change of the clutter only affects the calculation time of the distance division.Influencing the division of CFSFDP,the use of CFSFDP to process measurement information can effectively improve the operating efficiency and real-time performance of the algorithm,and improve the accuracy of target position estimation to a certain extent after removing clutter.