结合改进CamShift算法的人脸识别门禁系统  

Access control system based on face recognition and improved CamShift algorithm

作　　者：邹方圆[1] 胡珂立[1] 魏智[1] 谷宇章[1] 

机构地区：[1]中国科学院上海微系统与信息技术研究所无线传感网与通信重点实验室,上海200050

出　　处：《现代电子技术》2013年第3期108-111,共4页Modern Electronics Technique

基　　金：中国科学院战略性先导科技专项:物理信息功能感知系统理论研究与设计验证(XDA06020202)

摘　　要：人脸识别门禁系统已经得到了广泛的应用,但是要求被识别人员靠近图像采集设备,使用起来极不方便。解决此问题的关键是在被识别人员的正常行走过程中完成人脸识别的任务,然而在人的正常行走过程中,因为角度、表情等变化,可能会导致人脸识别产生错误的结果。针对此问题,设计了一种结合CamShift跟踪算法的人脸识别门禁系统,采用人脸检测的结果来改进CamShift跟踪算法,然后再利用这条CamShift跟踪轨迹上人脸识别结果序列来修正错误识别。实验结果表明该系统能够完成正常行走过程中的人脸识别任务,并且有更高的识别率。The access control system based on face recognition has been widely used, but users must be close enough to the image acquisition device so that it is not convenient to use. The key to solve this problem is to comp｜ete the task of face recognition in the normal walking process. However, facial angle and expressions always change during walking, which may result in incorrect face recognition results. To solve this problem, an access control system based on face recognition combined with im- proved CamShift tracking algorithm is designed. More robust face trajectory is achieved by using sparse face detection results. And the wrong face recognition result is corrected with the help of the trajectory got before. Experimental results show that the designed system can complete face recognition task in the normal walking process with higher recognition rate.

关 键 词：人脸识别 CAMSHIFT算法 目标跟踪 门禁系统 

分 类 号：TN911.734[电子电信—通信与信息系统] TP391.4[电子电信—信息与通信工程]