Reseach On Algorithms Of Tracking And Track Maintenance For Passive Multi-sensor Systems

The techniques of target tracking and track maintenance for passive multi-sensorare important topics in multi-sensor data fusion systems, and have wide applications inboth military and civil areas. Therefore, their developments have attracted worldwideattention from researchers. Sponsored by the National Natural Science Foundation ofChina, the dissertation mainly investigates the target tracking and track maintenancetechniques. The main contributions of the dissertation are as follows:1. Considering that the modified input estimation (MIE) method cannot be used totrack high maneuvering targets, fuzzy reasoning technique is introduced, which caneffectively improve the adaptive tracking capacity of the MIE method for highmaneuvering target tracking. However, the fuzzy MIE algorithm demonstrates a highcomputational cost, so we further propose a strong tracking MIE algorithm. In thismethod, strong tracking filter multiple fading factors are introduced according to thetheory of strong tracking filter, which has a real-time performance and can enhance thetracking accuracy of the MIE method for high maneuvering target tracking byadjusting the filter gains adaptively. For the disturbance among the models in theinteracting multiple model (IMM) method, a switched IMM algorithm based on theViterbi technique is proposed, which can improve the utilization rate of each model,consequently enhancing the performance of maneuvering target tracking.2. Considering the heavy computational load of the standard particle filteralgorithm combined with the joint probability data association (JPDA) technique, anindependent particle filter (iPF) algorithm based on the Fuzzy clustering JPDAtechnique is proposed. In the proposed method, the targets can be associated with themeasurements through the fuzzy clustering technique, which can effectively reduce thecomputational cost of data association. Moreover, in order to remove the particledegeneracy and alleviate the particle impoverishment, the particle swarm optimization(PSO) technique is introduced, which can effectively enhance the performance of theiPF algorithm. Finally, for the closely spaced target tracking, a joint particle filteralgorithm based on fussy clustering technique is proposed, which has a goodperformance in multi-target tracking with a parallel flight.3. Taking into account the shortcomings of the traditional particle probabilityhypothesis density (PHD) algorithm for passive multi-target tracking, an improved particle PHD algorithm based on the Gaussian Hermit (GH) technique is proposed. Inthe method, the better importance density function is approximated with some newGaussian distributions produced by a bunch of GH filters, and the latest measurementsare fully utilized to effectively improve the accuracy of the multi-target tracking. Inaddition, due to the advantages of the strong tracking MIE method for the singlemaneuvering target tracking, this method is extended to PHD filter framework in orderto achieve the multiple maneuvering target tracking, which has a better performancethan the PHD method based on the current statistic model (CSM).4. For the track maintenance of multi-target tracking, a novel algorithm based onthe cross entropy (CE) technique and the PHD method is proposed. Firstly, aconnectivity graph with associated weights is constructed according to the output ofthe PF-PHD filter, and the weights are modified by the motion directions of theestimated targets. Then the CE technique is employed as a global optimization schemeto calculate the optimal feasible associated events, which can effectively achieve thetrack continuity even when the targets are close to or cross each other. Furthermore,due to the advantages of the CPHD filter and the Rao-Blackwellized particle filter(RBPF), we propose another track maintenance algorithm based on the CE technique,named the RBPF-CPHD tracker. This algorithm can further improve the trackmaintenance performance with a stronger robustness and a greater anti-jammingcapability.5. For the multiple maneuvering target tracking, taking into account theadvantages of the cardinality-balanced MeMBer (CBMeMBer) filter, a novel multiplemaneuvering target tracking algorithm is proposed by extending multiple model (MM)method to the CBMeMBer filter and then using the sequential Monte Carlo (SMC)implementation. Simulations show that the proposed algorithm has a higher accuracyof state estimates and a better computational efficiency than the MMP-PHD/CPHDalgorithm. Moreover, since the MM particle CBMeMBer algorithm cannot give thetracks, a novel track maintenance algorithm is proposed by introducing the particlelabeling method, which can effectively achieve the track continuity for the multiplemaneuvering target tracking.