Video Object Forgery Detection Algorithm Based on VGG-11 Convolutional Neural Network
Yanfen Gan, Jixiang Yang, Wenda Lai · 2019
In recent years, with the development of computer multimedia technology, video forgery has become more and more common. For the video object forgery can cover up some key evidence and it is hard to identify by the experts, the forgery detection technology for this class had always been a research hotspot. However, researchers mostly pay attention to traditional methods such as image processing and classifiers and rarely combine deep learning theory to the research. This paper proposes a video intra-frame forgery forensics algorithm based on the VGG-11 convolutional neural network, which can automatically detect video forgery frames. The algorithm first decompresses the video into a series of frames, calculates the motion residual map of each frame, and extracts the steganographic features. Then four different steganographic feature sample sets are used to construct as the training set and the test set to train and test model. The best-performing feature was selected by the comparison experiment. Finally, the forged frame was marked from the forgery video successfully. A series of experiments show that the proposed algorithm based on the VGG-11 convolutional neural network can automatically identify original or forgery frames in forgery video.