Study On The Methods Of Radar Dim Target Detection And Tracking

How to improve the detection performance of low observable targets and promotetheir early warning time is one of the great challenges that modern radar are facing. Thetechnique of long time processing of radar target echo signal is an effective approach topromote the detection performance of radar. As one of the key techniques and researchfocuses of radar signal processing, it has been obtained great concern and extensiveresearch by scholars in various countries. In this paper, the study of radar dim targetdetection and tracking is mainly implemented in the aspect of long time processing ofradar target echos. On the one hand, the algorithms of particle filter basedtrack-before-detect of radar dim target are studied and on the other hand, the algorithmsof radar signal long time integration which are suitable for practical applications arestudied and practised. The specific work is summarized as follows:1. A particle filter based radar extended target detection and tracking algorithm isproposed. In the proposed algorithm, firstly, extended target model of radar isestablished according to the2-D data in radar range and Doppler map. Then, thelikelihood ratio function of the proposed model is deduced. At last, the detection andtracking algorithm of radar extended target is implemented based on particle filter.Simulation results manifest that the algorithm can effectively detect and track radarextended target with low signal to noise ratio (SNR).2. An algorithm based on particle swarm filter for radar multi-target detection andtracking (MTDT) is proposed, which turns the problem of MTDT into the problem ofmultiple independent detection and tracking of a single target. Two kinds of particleswarms, the birth and tracking particle swarms, which are designed to introduce newtargets and track existing targets, respectively, are employed to implement a recursiveBayesian filter in the proposed algorithm. Each particle swarm only detects and tracksone target and deals with its corresponding target independently. In the proposedalgorithm, target number and states are estimated by the particle swarms and theircorresponding probabilities of target existence (PTE). Besides, a probabilistic dataassociation method based on range information is proposed to realize track associationof close distributed target. The proposed method simplifies the complicated problemand reduces the computational complexity of the MTDT. Simulation results show thatthe proposed method can well estimate the target number and states and have highcorrect ratio of data asscociation. 3. A scheme for multi-target detection and tracking based on PHD filter is proposed.The proposed scheme firstly generates particles which represent the target tracks inseveral sequential frames of measurements according to the system dynamic equationand the current particles in PHD filter. Then, the correlation coefficients of the particlesand the estimated target tracks are used to implement the association of particles andtargets. At last, data association is accomplished in the process of estimating the state oftargets by the particles associated with them. Due to the sufficient usage of theinformation of target tracks in several sequential frames of the measurements, thecorrect ratio of data association and the precision of estimated target state areremarkably improved by the proposed methods, and the proposed algorithm can alsotrack the interacted or close spaced targets. Simulation results demonstrate itseffectiveness.4. Radon-Fourier Transform (RFT) integrates radar target energy by integrating thetarget reflected energy in range-slow time domain according to the target motionparameters, which can effectively increase the coherent integration time of radar signal.A fast RFT method based on Chirp-Z Transform (CZT) is proposed to alleviate highcomputation cost often encountered by the RFT and the energy lose caused byquantization error without interpolation. In addition, the proposed method can alsoeliminate the loss of matched filter by compensating the Doppler frequency of targetwithout increasing the computation cost. For the stationary target with near constantacceleration, Dechirping method is adopted to compensate the quadratic phase causedby the acceleration of target, which further increase the effective coherent integrationtime of radar signals. The experimental data processing results show that the integrationperformance of the proposed method can almost achieve the optimal one under idealcondition.5. Methods of radar signal long time processing which are suitable for practicalapplications are investigated. A radar signal long time integration algorithm based onshifting operation among frames is proposed and, its effectiveness is validated by theexperimental data processing results. Through the analysis of real radar data, a modifiedmethod is proposed. In the modified method, firstly, the target track in the3-D space(range-Doppler-time data) is corrected into a certain plane which is perpendicular to theaxes of Doppler frequency by shifting operation among frames. Then, the corrected3-Ddata is decomposed into parallel2-D matrices of range-time by matrices reconstruction.At last, long time integration is fulfilled by integrating the target track whichapproximates to a straight line in the recombined2-D matrices. The method can be implemented by shifting and accumulating operation with low computation cost and berealized in practical application. The experimental results show that the proposedmethod can increase the effective integration time of radar target echo signal andimprove target detection performance of radar systems.