基于视频监控的人脸识别方法

Face Recognition Method Based on Video Surveillance

摘要:

目前人脸识别方法主要针对静态图像进行识别，而在监控视频中，不同视频帧人脸具有相关性且只有部分人脸能够有效反映人脸信息。根据监控视频中人脸图像变化特性，提出了一种基于监控视频的人脸识别方法。首先通过结合人脸检测与跟踪技术获得视频人脸序列，然后以视频人脸序列中部分人脸图像识别结果为导向选取全部人脸序列图像中的代表人脸图像进行识别，最后根据选取的全部人脸图像识别结果综合反映人脸信息。实验结果表明，该方法能够在确保识别率和误识率的前提下有效提升监控视频中人脸识别的实时性。

Abstract:

At present, face recognition methods mainly target at static images. In video surveillance, faces of different video frames are correlated and only part of the faces can effectively reflect face information. A face recognition method based on surveillance video is proposed according to the changing characteristics of face images in surveillance video. Firstly, video face sequence is obtained by combining face detection and tracking technology, and then face images representing all face sequence images are selected for recognition based on partial face image recognition results of video face sequence. Finally, face information is comprehensively reflected according to all face image recognition results selected. Experimental results show that this method can effectively improve the real-time performance of face recognition in video surveillance while ensuring the recognition rate and error rate.