目标实时跟踪与预测算法研究  

Algorithm of real-time object tracking and prediction

作　　者：孙红[1,2] 王晓婉[1,2] 吴钱忠[1,2] 秦守文[1,2] 张建宏[1,2] 屠佥炜[1,2] 

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093 [2]上海现代光学系统重点实验室,上海200093

出　　处：《信息技术》2014年第3期55-57,共3页Information Technology

基　　金：国家自然科学基金项目(61170277);上海市教委科研创新重点项目(12zz137);上海市一流学科建设项目(S1201YLXK)

摘　　要：实时成像跟踪系统要求对运动目标能够有较快的响应速度,跟踪的响应时间越短,系统的实时性就越好,从而可靠的跟踪系统显得尤为重要。文中在研究了目前常用几种跟踪算法的基础上,提出一种基于目标特征匹配和Kalman预测相结合的跟踪方法,选取目标的灰度直方图信息做为特征匹配模板,使用Kalman滤波器对目标在下一帧图像中可能出现的位置进行预测,在预测范围内进行搜索及模板匹配,实验结果表明,该跟踪算法能够对目标实现稳定可靠的跟踪。Real-time imaging tracking system requires a quick response speed to the moving object. The shorter response time of the tracking system, the better real-time performance of the system. So the reliable tracking system is very important. After studying the several common tracking methods according to the target characteristic matching and Kalman prediction, this paper chooses the grey level histogram message of the target as the feature matching template, then uses the Kalman filter to predict where the target of the next frame image will be, so it can search the target and match the template. The experimental results show that it can track the target steadily and reliably.

关 键 词：实时跟踪 目标预测 KALMAN预测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]