False face video detection method and system based on face cross-parallel ratio under trust mechanism

Abstract

The invention discloses a false face video detection method and system based on a face intersection ratio under a trust mechanism. The method comprises a network training step and a sample testing step, wherein the network training step comprises model training and model verification. In the model training step, a segmentation network is trained, and a segmentation network model and a weight valueare stored; in the model verification step, a trained segmentation network is used to predict a mask, and a binarization threshold and a decision threshold are determined by using a grid search method, wherein the sample test comprises the steps of performing video framing preprocessing to obtain an input picture, determining a to-be-detected area by utilizing a segmentation network, performing smooth denoising, binarizing the to-be-detected area, determining a face area and an undetermined tampered area, calculating an intersection ratio of the face area and the undetermined tampered area under a trust mechanism, and finally performing true and false face judgment according to a judgment threshold value. The false face video generated by the deep face changing tool is detected, high accuracy is obtained in different databases, the cross-database test performance is obviously improved, and an effective way is provided for false face video detection.

Description

technical field [0001] The invention relates to the technical field of tampering detection of digital video, in particular to a fake face video detection method and system based on facial cross-merge ratio under a trust mechanism. Background technique [0002] Among the many biological features, the human face is one of the most representative features, with a high degree of recognizability. Therefore, with the rapid development of face recognition technology, the security threat brought by face tampering is increasing, especially in the contemporary era when mobile phones are highly popular and social networks are becoming more and more mature. Deep face swapping tools mainly use deep neural networks such as autoencoders or adversarial generative networks to generate fake faces and then replace the faces of the original video. According to the features used, existing fake face video detection technologies can be roughly divided into three types: Major categories: based on ...

AI Technical Summary

Problems solved by technology

The above method can identify fake face videos to a certain extent, especially in the library, and the test can achieve a high accuracy rate. However, the accuracy rate drops sharply in the cross-database test, that is, there is a problem of insufficient generalization ability.

Images