基于嵌入式视频监控的人脸检测系统设计

针对视频监控系统智能化的要求,提出了一种基于嵌入式视频监控的人脸检测设计方案。采用在ARM-Linux操作平台上建立Servfox流媒体服务器,实现USB摄像头的视频采集和传输。在客户端首先采用帧间差分更新背景模型的改进算法获取运动目标,减少视频中运动目标对背景模型的影响,且为人脸目标的检测缩小了范围。在运动目标区域内,通过Ada-Boost人脸检测算法,最终获得人脸位置。结果表明,系统对人脸检测效果良好,准确率可达95.2%,检测时间22～27 ms/frame,满足视频监控的实时性要求。

Design of face detection system based on embedded video monitor

Required by the intelligentizing of video monitor system,this paper proposes a face detection design based on embedded video monitor.With the help of the transaction of Servfox streaming media on ARM-Linux operation platform,the collection and transmission of USB camera are realized.This paper adopts the inter-frame difference to renew the background and that can track the moving target and at the same time to lessen the influence of the moving target on the background model.It narrows the scope of detection of the human face.It locates the human face by AdaBoost human face detection in the target moving area.The system performs well on human face detection and the correctness reaches 95.2%.The detection time is 22~27 ms/frame.It satisfies the real-time requirement of the video monitor.