一种改进Camshift算法及其ARM实现

基于视频序列的运动目标跟踪在安防、军事等领域用途广泛。针对传统Camshift算法易受颜色相近物体的干扰,丢失目标的情况,提出了一种改进的Camshift算法。该算法检测SIFT特征点并进行FREAK特征匹配,通过判断每一帧跟踪结果的跟踪精度修正跟踪矩形框,从而改善跟踪精度。为便于工程应用,在Linux系统上进行了算法移植,实现了基于ARM的运动目标跟踪系统。实验结果证实改进算法对部分遮挡、颜色相近干扰等情况具有稳定性,能够实现对运动目标的准确跟踪。

An Improved Camshift Algorithm and Its ARM Implementation

Moving object tracking based on video sequences is widely used in security, military and other fields. In view of the fact that the traditional Camshift algorithm is vulnerable to interference from similar objects and the loss of target, an improved Camshift algorithm is proposed. The algorithm detects the SIFT feature points and performs FREAK feature matching. The tracking rectangle box is modified by judging the accuracy of each tracking result, thus improving the tracking accuracy. In order to facilitate the engineering application, the algorithm is transplanted on the Linux system, and the moving object tracking system based on ARM is implemented. The experimental results show that the improved algorithm is stable to some occlusion and similar color interference, and can achieve accurate tracking of moving objects.