Abstract: In this paper, an improved moving target tracking algorithm with multiple camera based on FOV line is presented. Firstly, it uses SIFT algorithm to select matching points automatically and generate FOV line. When the target passes the line, its color histogram is compared against the target’s real color histogram to identify the best matching, namely, the handover is achieved. Then it uses the Mean Shift algorithm to track the subsequent target. A real-time tracking is carried out, and the experimental results show that the algorithm can achieve continuous tracking effectively among multiple cameras.

Key words: Field of View（FOV） line, Scale-Invariant Feature Transform（SIFT）, object tracking, multiple camera

摘要： 提出了一种改进的基于视野分界线的多摄像机运动目标跟踪算法。该算法通过采用SIFT算法来自动获取特征点，生成视野分界线。当目标穿过分界线后，将待确定目标颜色直方图与目标颜色直方图相匹配，匹配度最好的赋予标识，即完成目标的交接，用Mean Shift算法完成后续目标的跟踪。在搭建的实验平台上进行了实时跟踪，实验结果表明该算法能有效实现多摄机间的目标连续跟踪。

关键词: 视野分界线, 尺度不变特征变换（SIFT）, 目标跟踪, 多摄像机