Research On Key Technologies Of Face Recognition In WeChat Small Video

With the rapid growing of the number of WeChat users, WeChat small video has become more and more popular because of its short time, low traffic, low cost,real-time and convenience. On the other hand, the content in the WeChat small video is abundant and comprehensive. It also has some shortcomings as it may contain anti-party, reactionary, pornographic and some other negative or unhealthy content. In addition, it may be used as a tool to spread extremist ideology by terrorists, ethnic separatists and religious extremists. Therefore, intelligent analysis and recognition on the figures in WeChat small video is necessary, it can guarantee public safety and the healthy spreading of network information. The research of intelligent analysis and recognition in WeChat small video have been just started.By studying the theories and methods of traditional video target detection and recognition, the paper carried out face detection in WeChat small video by using Adaboost algorithm and the Skin Segmentation. Namely, we segment skin color regions in YCrCb color space and then we remove the non-face regions according to face geometric features. Finally, we use the Adaboost algorithm that is based on Haar-like features to detect the face regions. The paper validates the validity of the algorithm in MATLAB experiment platform, and the results show that the method can effectively detect face target in complex background of WeChat small video. The average test time is 50ms and the detection rate is 90.22%.Then research on face target tracking and recognition in WeChat small video.The paper proposes a new method to face target recognition in WeChat small video for the frame image of WeChat small video by 2DGabor wavelet transform and then reduce features' dimensions through 2DPCA algorithm and after that to classify and target recognition by using SVM algorithm. The MATLAB experimental results shows the method can be used to achieve face target recognition effectively. The recognition rate in ORL face database is 95.65% and in WeChat small video is 91.38% . And using Canshift algorithm carried out real-time tracking of face target in WeChat small video, the algorithm is accurate, steady and effective.Finally, a prototype system of face detection and recognition for WeChat small video has achieved. It can detect, track and recognize the face target in WeChat small video.