Research On OTHR Multi-target Tracking Algorithm Based On Probability Hypothesis Density

The problem of strong nonlinearity and non-Gaussian,multi-mode multi-path,low probability of target detection,low data rate,low measurement accuracy and heavy clutters in over-the-horizon radar(OTHR)poses a severe challenge to multi-target tracking(MTT)technology.In recent years,as a novel MTT technology,the probability hypothesis density(PHD)filter which is based on random finite set(RFS)theory is capable of avoiding the data association in the process of state estimation,and is suitable for dense target tracking under non-conventional conditions.This article we do some basic research about multi-target tracking in the complex environment of OTHR based on PHD filtering.The main contributions are as follows:1?Collaborative GMPHD filter for fast multi-target tracking: For the problem of large detection area and dense echoes in OTHR,we consider the difference of dynamic evolution between the survival target and birth target,according to the predictive information to adaptively partition the measurement set into two parts: survival target measurement set and birth target measurement set,which are used to update survival PHD and birth PHD,respectively.This method can reduce the computation cost caused by processing the global measurements in the standard PHD.There may exist the classification error problem in measurement partition,so the extraction of the invalid measurements,and the collaborative interaction mechanism of the survival and birth PHD is necessary.Simulation results show that the proposed algorithm has greatlyreduced the computation cost and significantly improved the peak extraction accuracy.2?Design of EMDs-PHD for Multi-target Tracking in OTHR: For the problem of multi-path and low probability of target detection in over-the-horizon radar(OTHR),PHD filter is started-up for each path and filtering independently.In each path,the survival PHD temporarily retain PHD which is not updated effectively and can avoid target loss due to the loss of target measurements.The information of the multi-path is regarded as an effective source,and then interestedly use the updated PHD of each path to judge the birth or death of target.The exponential mixture densities(EMDs)is used to fuse the results of each path to realize the information interaction and effective fusion.Simulation results show that in OTHR multi-target tracking scenario,the proposed algorithm is able to achieve more accurate estimation of target state,and form the continuous and stable tracks.3?A PHD frame of partition tracking: Many radar systems,including the OTHR,have the characteristics of partitioning.The detection area of this system is divided into multiple connected tracking areas in time and space.The traditional global tracking algorithms need to wait for a full scan and then obtain the measurements of the whole region to estimate the target state.The instantaneity of the traditional method is poor and the computational efficiency is low.This article proposes a PHD frame of partition tracking,which can instantaneously process the data of each area and display the processing results,which greatly reduce the data quantity for per processing step and significantly improve the computational efficiency.For the target moving across sectors,the algorithm of PHD with area measurements transfer and feedback mechanism is designed to ensure the track continuity of multi-target.Simulation results show that in the tracking scenarios of mechanical scanning radar and OTHR the proposed frame have better tracking accuracy and effectively reduce the computation cost.