神经网络在多机动目标跟踪中的应用

为解决多目标跟踪中数据关联与状态估计的问题引入神经网络方法.针对联合概率数据关联(JPDA)存在的计算组合爆炸问题,利用Hopfield网络解决TSP问题的思路找到神经网络联合概率数据关联(NJPDA)方法,并对其进行一定的改进解决其参数确定问题.基于以上关联方法得到的关联概率,利用简化的信息融合自适应滤波算法,实现对目标轨迹的状态滤波与预测.以上综合方法充分利用了神经网络的优点,保证了多目标的跟踪精度及实时性.

The Application of Neural Network in Multiple Target Tracking

To solve the problem of data association and state estimation in multiple targets tracking system, we brought the network into the algorithms. By using the thinking of Hopfield network in TSP problem,NJPDA algorithm avoided the large computational burden in JPDA, and the parameters in NJPDA can be confirmed online. Based on the prime data association received from NJPDA, we used simplified data fusion adaptive filtering algorithm to realize the state estimation and prediction. These algorithms make the best use of the merits of the neural network ensure the accuracy and real-time of multiple targets tracking.