| Face recognition is one of the important fields of biometrics. It has a very wide application prospect in man-machine interaction, authentication, social security and so on.At present, based on the depth of learning to extract facial features for face static image recognition method, the correct rate in the Labeled Faces in the Wild (LFW) data set and other standard sets is almost close to the artificial recognition rate. However, in the video stream, such as face in the monitoring system, due to the movement of the human body and attitude deviation and other issues, resulting in the detected part of the face area is seriously blurred and incomplete. in this case, simply use image-based face recognition method, the accuracy rate will be seriously reduced.The video has a large number of face area image information that can be used for recognition. However, not all of the face images in the frame are suitable for image recognition. Therefore, the use of all the frames in the video recognition does not necessarily improve performance, but significantly increases the recognition of the calculation time.In this paper, in consideration of the posture of the face in the video, the blurring of the image, the large redundancy of the face image and other issues, the face image selection algorithm based on the feature autocorrelation coefficient and cross-correlation coefficient of the face region and the face recognition method of weighted voting desicions are designed and implemented. The system can effectively select the high-quality face pictures for face recognition and reduce the redundancy of the same face picture in the continuous video frame. Through the recognition of a few face images and weighted voting decisions,the number of calculations of the face recognition is reduced, and the accuracy, real-time and stability of the identification system are improved.The main work of this paper is as follows:1. An index of measuring the characteristic quality of face image is proposed, which is the characteristic autocorrelation coefficient of face image. Through this index,we can filter blurred and posture tilted face images, and select the high-quality face images that are suitable for recognition, so as to improve the face recognition rate.2. An index of measuring the redundancy of the multiple face images is proposed,which is the characteristic cross-correlation coefficient of the image. Through this index, we can reduce the number of image recognition and the amount of calculation, and provide the possibility for real-time video face recognition.3. Using K-NN classification face recognition algorithm based on the depth learning to realize the recognition of single image, and proposing a weighted voting face decision algorithm based on the cross-correlation coefficient.4. Achieving the video stream face recognition system. And through the version control of the category of classification system and the characteristics of category,it’s able to remotely update the recognition capability of system online. |
