基于人脸检测和特征的跟踪方法

A tracking method-based on face detection and feature

在智能视频监控中,常常会发生对同一人脸重复定位的问题.通过分析人脸的非刚体特征和人脸监控视频的特点,提出了一种基于帧间人脸位置匹配的思想和基于颜色直方图的Meanshift算法二者相结合的方法.实验结果表明,提出的方法具有一定的实时性,并能够降低对同一人脸的重复识别率.

A face could be detected and labeled more than once when analysising the video. Through analyzing norigid features as well as the face-videors characters, the paper proposed a method combined with the position matching and the Meanshift tracking. The experimental result shows that the propeosed method can reduce the ratio that a face is labeled more than once, and meanwhile, enjoys a real-time requirement.