Research On Partcle Filter For Target Tracking And TBD Algorithm

With the increasing complexity of modern combat environment, the traditional tracking and detection methodis more and more difficult to solve the problem oftarget detection with low SNRand target tracking in nonlinear environment.So the new target tracking and detection algorithmsfor solving these problems areurgently needed.TBD(Track Before Detect) algorithm is an effective method to solve the problem of target detection with the low SNR, which detects objectsaccording to certain rules after multiple frame data accumulationand obtains the tracking results in the same time with detecting target.Particle Filter(PF) is a Bayesian recursive filter algorithm which is good at solvingthe problem of nonlinear system, moreover the likelihood ratio can be constructed by the weights of particles. These features make the PF can realizetarget detection and trackingcompletely.Therefore, the target tracking and TBD algorithmbased on PF are studied in this dissertation.1. Two target tracking algorithms based on PF are studied and compared in this dissertation. They are the SPF(Standard PF) algorithm and the IPF(improved PF) algorithm which uses an optimized importance function based on local linear technique and improved resampling to slovetheparticledegeneracy in SPF algorithm. Besides the conventional performance evaluation index, the KLD(Kullback Leibler Divergence) estimationcan evaluate the performance of the two algorithm in the simulation,which can more fully evaluate the performance of PF algorithm.2. Three PF-TBD algorithms are studied and compared in this dissertation. The standard PF-TBD algorithm forms a complete tracking and detection system through dividing the particles into the death and survival particles to constitute a hybrid state estimation. Because of only focusing on survival particles, the efficient PF-TBD algorithm is more effective on usingparticle information than SPF-TBD.The PF-TBD algorithm based on likelihood ratio detection use the likelihood ratio to test, which is different from the former two algorithms andthe former two use the probability of target existing for testing.In simulation, the first work is determining the optimal detection threshold of these algorithms through comparing the averagedetection probability and false alarm probability. Then the detection performance of these algorithms are compared.3. In view of the RAM(Rocket Artillery Mortar) targets, first of all,the target motion model is established.Considering this kind of target motion state can use a single model, therefore, first the SPF algorithm is used for tracking the target. Based on the previous conclusion, the IPF algorithm is further used.The simulation results show that theseRAM target tracking algorithms based on PF can track the RAM target and the IPF’s tracking performance is superior to the SPF.4. In view ofthe radar target, first of all,the basic model of radar applies to TBD processingis established.Fully considering the mobility characteristics of target, the MMPF(Multiple Model PF) algorithm is first studied. Then the TBD algorithm based on MMPF is studied to solve the problem of maneuvering targetTBD provessing. At last these algorithmsare verified in the simulation experiments.