基于带约束可能性聚类的多目标跟踪新算法

A New Multiple Target Tracking Algorithm Based on the Constrained Possibilistic Clustering

针对密集杂波环境下的多目标跟踪问题,提出了一种基于可能性聚类的联合概率数据关联滤波算法。在提出算法中,分析了传统FCM数据关联算法在噪声抑制方面的不足;利用可能性聚类能够有效抑制噪声的优势,同时结合多目标跟踪中,聚类中心应该在目标预测位置或者在其附近的特点,提出了一种以目标预测位置为约束条件的可能性聚类新目标函数,通过对目标函数进行优化得到目标观测的数据关联矩阵,有效减少由杂波引起的错误关联,实现对多目标与观测的准确关联。实验结果表明,提出的方法能够有效解决多目标与观测的关联问题,关联准确率要高于传统的Fitzgerald’JPDAF、MEF-JPDAF算法和IF-JPDAF算法。

According to multiple target tracking problem in dense clutter environment,a new joint probabilistic data association filter based on probabilistic clustering is proposed(PC-JPDAF).In the proposed algorithm,firstly,the shortcomings of traditional FCM data association algorithm in noise suppression are analyzed;the probabilistic clustering is used to effectively suppress noise,and the characteristics that the clustering center should be at or near the target prediction position in multi-target tracking is combined,a new probabilistic clustering objective function with the target predicted states as the constraint condition is proposed.By optimizing the objective function,the data association matrix of observed and measured targets can be obtained,which can effectively reduce the incorrect associations caused by clutter and can realize the accurate correlation between multi-target and observation and measurement.The experimental results show that the proposed method can effectively solve the association problem between multi-targets and observation and measurement,and the association accuracy is higher than that of traditional Fitzgerald fuzzy joint probabilistic data association filter(Fitzgerald’JPDAF),maximum entropy JPDAF(MEF-JPDAF)and IF-JPDAF(intuitionistic fuzzy JPDAF)algorithm.