基于自适应聚概率矩阵的JPDA算法研究

摘要

A novel JPDA method for data association on multi-target tracking system was presented for reducing the class of JPDA algorithm computational complexity and solving the problem of coalesce neighboring tracks.To improve the computational complexity,the joint association event probabilities were calculated with Cheap JPDA algorithm,then the cluster probability matrix was reconstructed by thresholding method to further optimize the computational complexity.Finally,the measurement prone to make wrong association were eliminated by measurement adaptive cancellation method to avoid the track coalescence problem for neighboring tracks.Theoretical analysis and simulation results showed that the proposed algorithm was able to reduce the complexity of the algorithm and improve the timeliness on the basis of preserving the tracking accuracy,and it was also capable of avoiding track coalescence with less errors when tracking the neighboring tracks and cross tracks,comparing with the standard JPDA and Scaled JPDA algorithm.%为了降低联合概率数据关联(joint probabilispic data association,JPDA)算法的计算复杂度,解决跟踪临近目标时出现的航迹合并问题,基于量测自适应消除方法,提出了一种改进JPDA算法.该算法首先通过Cheap JPDA算法计算互联概率,降低算法计算量;其次对聚概率矩阵加以阈值处理,通过重建确认矩阵,进一步优化算法复杂度;最后采用自适应消除方法,去掉聚概率矩阵中易引起错误关联的量测,减小JPDA算法在关联临近目标时的误差.仿真实验结果表明:相较于JPDA算法及Scaled JPDA(SJPDA)算法,本文算法在保证跟踪精度的前提下,降低了算法复杂度,提高了时效性;在跟踪临近目标及交叉目标时,改进算法能避免航迹合并现象及跟错目标情况的发生.