【摘要】 介绍一种自动人脸检测与跟踪的智能视觉监控系统,它可以获取进入银行营业场所人员的清晰正面像。结合自适应阈值判定运动像素点分离运动目标、肤色特征、多级Adaboost分类器和Hausdorff距离实现实时人脸定位;用检测到的人脸位置来初始化跟踪算法实现快速人脸跟踪;以及基于对象的视频编码传输以及算法的DSP实现。更多还原

【Abstract】 An intelligent visual surveillance system is introduced in the paper which obtains a clear frontal facial image of the people who enters the bank′s business location by automatic face detection and face tracking. The face can be positioned with splitting object by judging the adaptive threshold of the moving pixel points, multi-layer Adaboost classifier, feature of complexion, and Hausdorff distance. The position of the face can initialize the parameters of face tracker to realize fast face tracking and video object coding. These algorithms can be realized in DSP.更多还原