基于神经网络的机动多目标数据关联问题研究

INVESTIGATION ON DATA ASSOCIATION PROBLEM OF MULTI-MANEUVERING TARGET BY NEURAL NETWORK

摘要: 本文通过对联合概率数据关联的性能特征的分析，将其归结为一类约束组合优化问题，在此基础上，利用Hopfield神经网络求解典型的约束组合优化问题（旅行推销员问题）的方法，解决了传统的联合概率数据关联中出现的计算量组合爆炸现象，仿真结果表明，该方法效果良好，在密集多回波环境下，其优越性能更为突出．

Abstract: The properties of the joint probabilistic data association (JPDA) in multi-maneuvering target tracking are analyzed in this paper, then reduced to be a sort of constraint combinatorial optimization problem. In the same way used in "traveling salesman problem" by neural network, the combinatorial explosion of computation quantity of JPDA can be solved. Simulations prove the effectiveness of the neural joint probabilistic data association (NJPDA) method. In dense multi-return environments the superiority of NJPDA has been showed more fully.