随机神经网络在机动多目标跟踪中的应用

APPLICATION OF STOCHASTIC NEURAL NETWORKS TO MULTI-MANEUVERING TARGET TRACKING

研究密集多回波环境下的机动多目标跟踪问题。通过对多目标联合概率数据关联方法性能特征的分析，将其归结为一类约束组合优化问题。在此基础上，利用随机神经网络求解组合优化问题的策略，采用改进的增益退火算法，提出了一种新颖的机动多目标快速自适应神经网络跟踪方法。仿真结果表明，该方法不仅具有很高的收敛速度和跟踪精度，而且计算量小，关联效果好。

The multi-maneuvering target tracking(MMTT) in dense multi-return environments is studied .The properties of the joint probabilistic data association(JPDA)are analyzed,and data association is reduced to a sort of constraint combinatorial optimization problem. By using the combinatorial optimization strategy of the stochastic neural networks,a new fast adaptive neural network method for MMTT(FANJPDAF) is presented and its computation burden is re-duced drastically Simulation results show that the FANJPDAF is of high accuracy,good asso-ciation performance and tracking effectiveness.