基于嵌入式平台的运动人脸捕捉系统设计及实现

针对现有智能监控系统难以同时满足清晰、准确、实时、可靠等要求的缺点，提出了一种基于嵌入式视频监控的实时人脸捕捉系统。在ARM Linux操作平台上建立流媒体服务器，实现CCD摄像头的视频采集和传输，采用背景差分法与相邻帧差相结合的方法来实现运动目标的分割与精确定位，提出了一种基于肤色模型的人脸区域分割算法，缩小了运动区域的人脸检测范围，在此区域内，通过Adaboost算法实现了运动目标的实时人脸捕捉。采用ARM1176JZF-S内核的Samsung S3C6410处理器，以Linux2.6.28作为系统的软件开发平台，在基于开源计算机视觉库OpenCV软件工具基础上进行系统开发与测试。结果表明，该系统实现了运动目标的检测以及人脸捕捉，并具有很好的清晰度，从其统计的检测率和运行时间看，有很好的鲁棒性和实时性。

Moving face capturing system design and implementation based on embedded platform

Required by the shortage of clarity, accuracy,real time in the intelligent monitor and control system,the paper proposed a real time face capturing based on embeded video monitor. CCD camera captures and transports frames with the help of streaming media server based on arm linux operating system.Moving target spilit and precise positioning are realized by the combination of background subtraction method and adjacent frame difference.Meanwhile,the paper put forward a model based on skin color of the face regional segmentation algorithm,narrowing the scope of the moving area face detection.In this scope,this achieves moving target real time face Snap shot by using Adaboost algorithm.The system is based on Samsung S3C6410,and the core is ARM1176JZF S.The software platform is linux,the version of linux2628.The development and testing uses OpenCV,which is the open source computer vision library.The results show that the design realizes the detection and face capture having the very good clarity and owning fine robust and real time performance from the statistical detection rates and operation time watching.