The Research And Implementation Of Radar Target Tracking Algorithm Based On Particle Filter

With the rapid development of military,the mobility of the target continues to increase,which put forward higher requirements to radar target tracking system,including radar target movement model and the filtering effect and real-time performance of the radar target tracking algorithm.In recent years,the superiority of particle filter in nonlinear system has attracted the attention of scholars,but its structure is complex and the calculation is large,which restricts the practical application of particle filter,especially for high real-time system.How to design and implement particle filters that the filtering effect is good and real-time is strong is still the focus of domestic and foreign research.The research topic comes from the needs of national ministries,in this paper,we focus on maneuvering target motion model,particle filter and its hardware implementation method.The traditional maneuvering target motion models include the Singer model,the Jerk model,the alpha-eta model and so on.The shortcomings of the traditional model are analyzed in this paper:Because they are all priori models,which artificially set the estimated parameters according to the prior knowledge,and once set,can no longer be adjusted according to the actual situation.In view of the above shortcomings,the parameter adaptive alpha-eta model?AP-alpha-eta?is presented in this paper,which considers the maneuver frequency and the jerk frequency as state variables and joins the state vector group,thus completing the maneuver frequency and jerk frequency parameters of real-time correction.The simulation results show that the AP-alpha-eta model has a smaller root mean square error than the alpha-eta model for uniform,uniform acceleration and Jerk motion.Considering that the resampling operation needs to be done after determining the likelihood distribution of all the weights of the particles,the time consuming of the resampling algorithm is a key factor in the particle filter run time.Based on this,the particle selecting method and particle replication scheme are optimized in the process of resampling algorithm and double threshold stratified resampling algorithm is proposed in this paper.The algorithm introduces the threshold comparison to complete the particle selection,according to the specific situation of particle selection to take the appropriate replication strategy.The results show that the filtering error of DTSR algorithm is less than54.46%compared with the SR algorithm.Finally,aiming at the problem of particle deficiency,a solution is presented in this paper.Based on the idea of linear optimization,use the weight of small particles to resample,which can effectively avoid the loss of particle diversity.Combined with bearing-only tracking?BOT?,the particle filter algorithm based on double threshold stratified resampling?DTSR-PF?is designed in this paper,including sampling and updating module,weight calculation module,resampling module and state output module.In the aspect of circuit design,aiming at the serious cost of time of resampling,the discriminant mechanism is introduced to avoid the resampling operation at every time;Considering the normalization of weights,the weight expression and the state output expression are simplified,which reduces the computational complexity and improves the real-time performance of the algorithm.In terms of functional verification,MATLAB and Modelsim are used to simulate the results.The relative error of the position dimension is in the order of 10^-2 and the relative error of the velocity dimension is in the order of 10^-3,which verifies the circuit design function correctness.In the performance analysis,using Xilinx ISE for synthesizing,translation,mapping and placement and routing,the result shows the maximum clock frequency can be up to 125MHz,thus DTSR-PF operating frequency can be up to 40KHz.