Video recognition algorithm of fake face based on SVM model

In response to more and more fake face videos in the network, we propose three feature detection methods based on image color histogram, SURF (Speeded Up Robust Features), and ELA (Error Level Analysis), SVM(Support Vector Machine) model training and verification methods, compared with the current common detection methods, this paper starts from the mechanism of fake face tampering, converts the processing of video streams into image features recognition, and analyzes various image features of faces. The recognition rate of face-swapping videos is significantly improved and the machine learning problem in the case of small samples is solved by using svm.