Abstract: This paper uses dimensionality reduction signal processing technology to get all targets’ reliable initial information based on TBD detection method in the image sequences.And in order to improve the real time quality of detection algorithm，single-frame detection technology is used：Targets tracks are not crossing in the tracking window（the area targets may appear in single frame），PDA is used to track every target individually，or else the improved hierarchical-search JPDA algorithm is used to split the target’s tracks.This paper gives the MATLAB simulation test and the results.The analysis of results shows that the combination of PDA and JPDA has good real-time quality and high-degree accuracy in tracking of multiple point targets.

Key words: Probability Data Association（PDA）, hierarchical-search, Joint Probabilistic Data Association（JPDA）, multiple point targets’ tracking, cross moving

摘要： 在序列图像情况下，采用降维处理技术在二维空间中通过基于TBD的多帧检测算法获得了众多目标可靠的初始信息。而后，为了提高检测算法的实时性，采用了单帧检测技术：当跟踪窗（单帧图像中目标可能的出现区域）内目标轨迹未交叉时，对每个目标分别用PDA算法进行跟踪；当跟踪窗内出现轨迹交叉时，则使用改进的分层搜索JPDA算法进行跟踪。给出了MATLAB仿真试验及其结果。结果分析表明，PDA和JPDA这两种算法的联合使用在点状多目标跟踪领域具有良好的实时性和很高的跟踪性能。

关键词: 概率数据关联, 分层搜索, 联合概率数据关联, 点状多目标跟踪, 交叉运动