Research On Data Association Method Of Group Target

In the maritime surveillance application based on satellite reconnaissance, the interval of obtaining data from the spaceborne sensor is relatively long because of the influence of motion period. It is difficult to track every target steadily and effectively when the target state may change greatly. Consequently, it is the effective approach to the problem that the group targets are regard as the study object and they are tracked through taking full advantage of their stable features. Group target data association (GTDA) is core content in solving the problem. The dissertation studies the basic flow and the correlative model for GTDA, and intensively studies the two primary aspects of GTDA, including the problem of target information clustering and group target data association. The main research results are summarized as follows:1. The basic flow and the correlative model of GTDA are established respectively. Firstly, the basic flow of GTDA is proposed, which includes target information clustering and group target data association. Secondly, aiming at the two problems, the resolvent is presented respectively in the background of the low measurement data rate. And finally, the model of GTDA based on multiple hypothesis clustering is established.2. A multiple hypothesis clustering method is proposed. The clustering problem is described as multiple hypothesis problem firstly. And the method is proposed based on genetic algorithm. The two dimensions encoding scheme and the fit measure calculation method based on template matching are put forward. In the way of genetic operation, the betting by turn method, the longitudinal multi-point crossover, the longitudinal random mutation is proposed respectively. And some conditions are set, including the initial population, the constrain condition of clustering, etc al. The simulation has proved the astringency and effectiveness of the method.3. A group target data association algorithm based on composition and array feature is proposed. Firstly, the mathematics models of composition and array feature are established, and their basic probability assignment functions (BPAF) are calculated respectively. Secondly, the two functions are synthesized based on D-S evidence theory and the synthetical BPAF is established. Finally, the better clustering hypothesis combination is found by adopting the comprehensive evaluation, and the association result is obtained after making use of one-zero integer programming. The simulation results show that the algorithm can reduce the influence of the low measurement data rate, and has more probability of correct association than the existing algorithms.