用于运动目标跟踪的改进Camshift算法的研究  

Research on Improving Camshift Algorithm about Target Tracking

作　　者：秦勇[1] 商萌萌[1] 赵晓红 盖建新[1] QIN Yong;SHANG Meng-meng;ZHAO Xiao-hong;GAI Jian-xin(School of Measurement and Control Technology and Communications Engineering,Harbin University of Science and Technology,Harbin 150080,China)

机构地区：[1]哈尔滨理工大学测控技术与通信工程学院,黑龙江哈尔滨150080

出　　处：《哈尔滨理工大学学报》2018年第3期34-38,共5页Journal of Harbin University of Science and Technology

基　　金：国家自然科学基金(61501150);黑龙江省自然科学基金(QC2014C074)

摘　　要：Camshift是一种应用颜色特性来对目标实施跟踪的算法,所以当背景过于复杂、目标被遮挡或做加速运动时很容易发生目标跟丢的情况。针对以上缺点,提出首先采用帧间差分法自动的选取搜索窗口,然后引入加权颜色概率直方图排除相似背景颜色的干扰,最后采用Kalman滤波器预测和Camshift算法相结合的方法准确跟踪做加速运动的物体。实验表明,此方法能够准确定位做加速运动或被遮挡的目标。The target is easily lost by using Camshift when the target is blocked or accelerated in complex background. The reason for this is Camshift is an algorithm that uses color characteristics to track the target. For solving above shortcomings,the improving Camshit algorithm is proposed in this paper. Firstly,the search window is automatically selected using frame difference method. Then,the weighted color histogram is introduced to eliminate the interference of similar background colors. Finally,the accelerating object can be tracked accurately based on Camshift algorithm combined with Kalman filter prediction. It is shown with experimental data that the blocked or accelerating target can be accurately located by used the method proposed in this paper.

关 键 词：CAMSHIFT 帧间差分 加速运动 KALMAN滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]