基于多特征自适应融合的MeanShift目标跟踪方法  

Mean Shift Object Tracking Based on Adaptive Multi-Features Fusion

作　　者：刘苗[1] 黄朝兵[1] 

机构地区：[1]武汉理工大学信息工程学院,武汉430070

出　　处：《现代计算机（中旬刊）》2016年第4期68-72,共5页Modern Computer

摘　　要：针对传统MeanShift算法特征单一,易受相似背景颜色的干扰并且无法处理目标快速运动的情况,提出一种基于多特征自适应融合的MeanShift目标跟踪方法,利用目标的颜色与纹理特征,并使用LBP纹理特征二值图作为掩模,去除图像中灰度值变化较小的区域的干扰。同时,结合Kalman滤波预测目标位置,解决目标的快速运动和遮挡问题。实验结果表明:该算法能满足实时跟踪的要求,并较好地解决相似背景颜色、遮挡、光照变化等问题。The traditional MeanShift algorithm models the objects to be tracked with single color feature, which leads to that it is more prone to be disturbed by the similar color distribution in background, and the algorithm can ＇t deal with the fast moving objects. To address this problem, proposes a new method of Mean Shift object tracking based on adaptive multi-features fusion, which combines color feature and tex-ture feature in Mean Shift, and makes use of the LBP to form a mask to remove the area where the gray value has small changes to void distractions to objects. At the same time, the algorithm adds the Kalman that predict the location of the target to deal with the fast moving objects and occlusion. The experimental results show that the algorithm can meet the requirements of real-time tracking, and it can solve the problem of similar background color distribution, occlusion and illumination change.

关 键 词：目标跟踪 MEANSHIFT 颜色直方图 LBP KALMAN 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]