摘要： 利用帧差法和基于Sampson距离的随机抽样一致性(RANSAC)匹配算法，实现基于时空信息的多运动目标跟踪。采用帧差法获知运动目标大概位置，运用Harris算法提取每帧视频图像中运动目标的角点，并对其进行归一化处理，利用RANSAC算法对角点进行匹配，使用阈值分割法确定运动目标位置。实验结果表明，该方法能准确跟踪相似目标，其实时性较高。

关键词: 视频, 目标跟踪, Harris算法, 角点检测, 随机抽样一致性匹配算法

Abstract: By using the frame differencing algorithm and Random Sampling Consensus(RANSAC) matching algorithm based on Sampson distance, this paper realizes a multiple motion targets tracking based on temporal and spatial information. It gains the approximate position of motion target by using frame differencing algorithm, uses Harris algorithm to pick up the corner points of motion targets in every frame of video image and normalizes the configuration of these corner points. It uses RANSAC algorithm to match the corner points. A thresholding method is adopted to locate the position of the motion target. Experimental results show that this method can track similar targets with high precision and real-time performance.

Key words: video, target tracking, Harris algorithm, corner point detection, Random Sampling Consensus(RANSAC) matching algorithm

中图分类号: 

• TN941.1