一种基于SIFT的遮挡目标跟踪算法

摘要

To resolve moving objects sheltering, a new object tracking algorithm is presented. The first three frames are used to establish the background model. Then, contiguous-frame subtracting and background subtracting are used to separate the moving objects automatically. SIFT feature is computed for each moving object. When sheltering is detected, SIFT features are used to match the sheltered region with the single moving object. The location of single moving objects in the sheltered region is located by the coordinate of the feature points. Then, RANSAC is used to optimize the SIFT features matching. Experiments show that our algorithm can track the sheltered objects accurately. It can effectively resolve moving objects sheltering and track moving objects continuously and stably.%针对运动目标遮挡的难题,提出一种新的遮挡目标跟踪算法.采用三帧取均值进行背景建模,采用相邻帧差法和背景差分结合自动提取出运动目标,对单运动目标生成SIFI(scale invariant feature transform)向量.当运动目标处于遮挡状态时,将遮挡区域与单运动目标进行SIFT特征匹配,通过特征匹配点的坐标,找出单运动目标在遮挡区域中的位置,并对SIFT特征匹配运用RANSAC算法进行优化,实现遮挡情况下目标的有效跟踪.实验表明,该算法能准确地跟踪处于遮挡中的目标,实现运动目标跟踪的连续性和稳定性.