Research And Implementation Of Video Face Recognition Based On The Depth Information

Face recognition research has very important significance in video surveillance. However, most of the research on face recognition are based on static image. Face recognition research in the actual monitoring scenario is less. Face detection and recognition technology based on 2D-camera has entered the bottleneck stage, because it need to scan the entire image in order to find human face in the picture, which is a time-consuming operation. At the same time, Deep Learning(DL) is the hot point of the international research and application. The traditional face recognition method has been a great impact. In the aspects of the speed and accuracy, face recognition methods based on DL have a better perform when compare to traditional method.In order to solve this problem, this paper proposes a method of face recognition based on depth information in surveillance video. In the process of face detection, this paper uses the depth information of depth camera Kinect to locate the face quickly, completely saving the time of scan total picture. In the process of face recognition, this paper analyzes the advantages of DeepID algorithm, and on the basis of this, combined with position of head bone, DeepID network has been improved. We come up with our own network training methods, and establish a face data set. On the premise of the real-time performance of the algorithm,a large number of experimental results show that our method has certain advantages in speed and accuracy. We achieve higher accuracy than the traditional algorithm. Finally, based on the above algorithm, this paper designs the face recognition system under monitoring environment, and tests the whole function of the system under the actual scene.