非约束环境下人脸识别技术的研究  

Research on the face recognition technology under unlimited conditions

作　　者：汤一平[1] 严海东[1] 

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310032

出　　处：《浙江工业大学学报》2010年第2期155-161,共7页Journal of Zhejiang University of Technology

基　　金：国家自然科学基金资助项目(60673177)

摘　　要：设计了一种基于全方位视觉传感器(ODVS)和多个高速快球摄像机(PTZ)相融合的装置.利用ODVS在动态图像序列中进行大范围多人体对象目标的检测跟踪,得到大范围内各跟踪人体对象目标的位置、大小等信息;然后根据检测到的这些信息来控制不同方位上的多个高速快球摄像机对人体对象目标的头部进行抓拍;最后根据人脸和人眼检测算法检测出来的人体对象目标的脸部特征对所拍摄到的图像进行筛选,淘汰一些不能用于识别的图像,并统计可以用于人脸识别的图像数量,如果没有达到一定的数量,则控制高速快球继续对该人体对象目标的头部进行抓拍,直至抓拍到足够的可用于人脸识别的图像为止.实验结果表明,通过ODVS和多个高速快球摄像机的视频图像信息融合,即使在人体非约束的情况下也能很好的获取到可用于人脸识别的图像,同时能满足多目标实时跟踪与识别的要求.该装置在安全监控、门禁检查、身份认证等领域都具有很好的实际应用价值.A device is presented in this paper which bases on a fusion of an Omni-Directional Vision Sensors （ODVS） and several PTZ cameras. The ODVS is used to detect and track the human targets on a large monitoring range to acquire the human information such as location and size. According to this information, the PTZ cameras in different azimuth can be controlled to capture the images of the heads of the humans. Finally, these images will be screened according to the characters detected by the face detection and eye detection algorithm. The images can not be used in recognition will be abandoned. If the number of the images used for face recognition is inadequate, the PTZ cameras will be controlled to continuously capture the images of the human face until the number of these images is adequate. The experimental results show that the images used for face recognition can be acquired well under unlimited conditions by this device. The demands of the real-time of Multi-objective tracking and recognition also can be met. This device has a good value in the practical application fields such as security monitoring, entrance guard checking and identity authentication.

关 键 词：非约束环境 多视觉传感器融合 人体对象目标跟踪 人脸检测 人眼检测 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]