人体运动目标检测与跟踪方法的研究与实现  

Research and Realization on Moving Object Detection and Tracking Methods of Human

作　　者：徐雷[1] 裴海龙[1] 

机构地区：[1]华南理工大学自动化科学与工程学院,广东广州510640

出　　处：《现代电子技术》2010年第4期128-131,共4页Modern Electronics Technique

基　　金：国家自然科学基金资助重点项目(60736024);国家自然科学基金资助项目(60574004);高校博士点基金资助项目(708069)

摘　　要：通过摄像机拍摄到的一段视频,对其中运动的人进行检测与跟踪。在目标检测方面,获取运动行人图像序列,利用自适应背景提取方法得到背景,根据将目标与背景进行分离的分割阈值,对差分图像进行分割,提取运动区域,检测出人体运动目标;在目标跟踪方面,选用基于Kalman滤波的运动跟踪系统,通过估计出运动目标的下一位置,对运动目标进行实时跟踪。这里的实验是在Linux平台上利用Intel的开源OpenCV函数库建立起模型并实现算法。经过多次仿真测试表明,使用Kalman滤波可以很好地解决实验中两个人握手时运动目标间遮挡的问题,并跟踪运动目标,而且处理速度比较快,能够满足一般情况下的实时性要求。Through a section of camera video, the moving human bodies are detected and tracked. In target detection,it obtains image sequence, gets the background using adaptive background extraction method. In accordance to the segmentation threshold of the partition separating the objects and images,it segments the images,extracts moving field,and detects the human bodies＇ moving targets. In the target tracking, it chooses the movement tracking system based on Kalman filtering. By estimating the next position of moving targets,it tracks moving targets with real -time. In this paper,it builds the model and algorithm on the experimental platform in the Linux open - source use of Intel＇s OpenCV library. Several rounds of simulation tests show that the use of Kalman filter can be a very good solution to the experiment. When two people shake hands with the issue of intershading,it solves the problem well,gets very good tracking of moving targets and fast processing speed. So,the system meets the real - time requirements under normal circumstances.

关 键 词：图像序列 目标检测 KALMAN滤波 实时跟踪 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]