摘要
提出一种多方位运动人体识别算法,首先利用混合高斯模型检测出运动物体,对检测出的运动区域利用线形变换后的YCbCr颜色空间建立头发模型,对侧面和背面人体进行识别,正面人体采用Adaboost人脸检测算法进行识别,实验结果表明该算法能识别出运动人体。
A new people recognition algorithm that is capable of handling significantly variable conditions in terms of viewpoint (i.e. frontal, profile, back view) is proposed. A video object segmentation algorithm based on Gaussian mixture model was used first. Then, the Adaboost face detection method is used to recognize frontal people. In order to overcome the drawbacks of Adaboost, a new hair model for the moving regions to recognize the profile and back views is built. Experimental results demonstrate that the proposed method performs well.
作者
李争名
LI Zheng-ming (GuangDong Polytechnic Normal University GuangDong Industry Training Center, Guangzhou 510665, China)
出处
《电脑知识与技术（过刊）》
2010年第15期4016-4018,共3页
Computer Knowledge and Technology
基金
广东技术师范学院校级科研项目(08kjy16)
