Research On Radar Target Tracking Method Based On Bayesian Framework

With the development and widespread application of stealth technology,the radar crosssection(RCS)of stealth targets has decreased sharply,which has brought great challenges to radar detection and tracking.The target tracking method based on Bayesian framework can provide priori information for the radar detection system and show excellent performance in the detection and tracking of dim targets.Particularly,the Probabilistic Data Association(PDA)filtering algorithm and Bayesian track-before-detect(TBD)method have gradually become a hot area of research in recent years.Because the PDA filtering algorithm comprehensively considers all candidate echoes in the associated gates when performing state estimation in dense clutter scenes,compared with the Kalman filtering algorithm that only uses the correct observations,the performance is severely lost,which might produce the phenomenon of tracking failure or errors.The Bayesian track-before-detect algorithm models the original observation data through an optical infrared sensor model,and the performance of the algorithm is very sensitive to the blur parameter of the sensor.In the light of the literatures,the blur parameter of the sensor is usually given in simulation experiments,but in engineering practice,the parameter is often unknown or do not match the radar system parameters,making it difficult for Bayesian trackbefore-detect technology to be applied to actual radar detection systems.To address these problems,this thesis proposed improved algorithms for PDA filtering method and Bayesian track-before-detect technology respectively,which can improve the tracking performance on dim targets.The main contents are summarized as follows:1.Focusing on the target tracking method based on the Bayesian framework,this thesis studies the Bayesian estimation and the dynamic space of the system theories,which includes formalizing the target tracking based on the Bayesian framework as a posteriori probability density estimation of the target state,introducing the optimal and suboptimal algorithms of recursive Bayesian estimation.These contents lay the theoretical foundation for subsequent research.2.For the problem of big tracking error caused by PDA filtering algorithm in dense clutter scenes,a smoothing algorithm for PDA with R-T-S(Rauch-Tung-Striebel)smoothing is studied.On the basis of PDA filtering,the algorithm combines with optimal smoothing idea and adopts smoothing technology as a data post-processing technology to improve data processing accuracy.The algorithm has better target tracking performance in sense clutter scenes than PDA filtering.3.For the problems of the sensor blur parameter unknown or mismatch in Bayesian trackbefore-detect technology,an adaptive sensor blur parameter setting method is given.This method is based on the quantile idea of Gaussian distribution and adaptively adjusts the sensor blur parameter according to the current radar system parameters,so that the noise intensity distribution and the target intensity distribution are separated at a given quantile,which improves the robustness of the Bayesian track-before-detect algorithm,and improves the detection and tracking performance of the algorithm.