基于Hopfield神经网络的雷达多目标跟踪

针对海上船舶雷达在多目标跟踪过程中实时性较差和不能快速响应的问题,提出目前密集杂波情况下多目标跟踪中最为有效的数据关联算法——联合概率数据关联(Joint Probabilistic Data Association,JPDA)算法。为解决JPDA随着目标增多的情况会出现的组合"爆炸"及计算量较大导致跟踪实时性较差的问题,从分析联合概率数据关联确认矩阵着手,依据Hopfield神经网络在解决旅行商问题(Travelling Salesman Problem,TSP)时的思路,提出基于Hopfield神经网络联合概率数据关联(Hopfield Neural Network Joint Probability Data Association,H-JPDA)来改进联合概率数据关联算法,通过简化矩阵拆分过程,显著减少计算量,提高跟踪的实时性。基于上述改进的神经网络联合概率数据关联算法,通过MATLAB对多目标跟踪进行仿真,仿真结果表明该算法能提高跟踪的实时性和快速响应能力。

Radar Multi-Target Tracking Enhanced by Hopfield Neural Network

The JPDA(Joint Probabilistic Data Association)has been used for radar data processing and regarded as the most effective data association algorithm in multi-target tracking under dense clutter.But the association algorithm may need excessive processing time or even fail to work properly when the number of targets gets high.In light of the problem,by analyzing the joint probability data association confirmed matrix,the improved algorithm named as H-JPDA(Hopfield Neural Network Joint Probability Data Association)is introduced from the idea of solving TSP(traveling salesman problem)with Hopfield neural network.The matrix splitting process is simplified and so is the computation.The algorithm is verified as better real-time tracking and quick response through MATLAB simulation.