带宽自适应的人体运动跟踪均值漂移算法  

Automatic Searching of Optimal Bandwidth for Human Motion Tracking of Mean Shift Algorithm

作　　者：李颖[1] 张大波[1] 王蓓蕾[1] 李智慧[1] 韩静晶[1] 

机构地区：[1]辽宁大学信息学院,沈阳110036

出　　处：《小型微型计算机系统》2013年第8期1913-1916,共4页Journal of Chinese Computer Systems

基　　金：辽宁省教育厅项目(L2010153)资助

摘　　要：传统Mean shift算法带宽固定的问题,导致了人体逐渐远离或者逐渐靠近摄像头运动时跟踪效果不好甚至失败.考虑到人体运动目标特征提取后的二值图像中,头部区域的图像呈圆形形状,并且可以用Hough变换检测出圆半径的大小.本文提出根据Hough变换检测出的圆半径在人体视频图像序列中的缩放比例,对带宽进行相应调整的算法.实验和分析表明本算法有很好的准确性和鲁棒性.The traditional Mean shift algorithm is a fixed bandwidth which leads to poor tracking performance even failure when the body is moving away from or gradually close to the camera movement. Taking into account the binary image after the body moving target feature extraction, the head area of the image presents a circular shape, and Hough transform can detect the size of the radius. In this paper, proposing an automatic bandwidth selection algorithm, which is according to the scaling of the radius based on the Hough transformation in the video sequences of the human body. The experimental results and analysis show that the algorithm has stron~ accuracy and robustness.

关 键 词：Mean SHIFT算法 HOUGH变换 带宽 运动跟踪 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]