Privacy Preserving Juveniles Video Based On Face Recognition

With the arrival of the era of big data,the growth of information and data in geometric dimension,especially the rise of network self-Media and various short video applications,makes all kinds of images and videos flooding the whole network,bringing convenience and entertainment to people,but also brings many problems,such as the leakage of confidential information,personal privacy rights and interests damaged.Especially for minors,their minds are not yet mature,and their right to privacy is more vulnerable to violation.According to the relevant regulations in China's Law on the Protection of Minors,the privacy rights and interests of minors should also be reflected in the image and video.Therefore,on the premise of guaranteeing the quality of video and the naturalness of content,it is a new research problem that how to automatically recognize minors and fuzzify their faces to prevent their identity information from leaking.This paper focuses on Video Privacy Protection for minors.Based on deep convolution neural network,face detection and image blurring algorithm,a video privacy protection mechanism for minors is proposed.This mechanism can realize computer automatic recognition of minors' faces and fuzzy processing to protect minors' privacy.Therefore,the main work of this paper is as follows:Firstly,a minor face recognition model is built based on the residual network.Aiming at the problem of lacking training and testing data set in the process of modeling,this paper chooses the appropriate age-divided facial age data set as the basis,and reconstructs a real data set which can be used to train and test the minor face recognition model by using the method of data expansion and data annotation conversion.Based on the data set,the model parameters are adjusted and optimized many times through experiments,and the optimal face recognition model for minors under certain conditions is obtained.Then the model is applied to video processing.Video subframes are processed and face detection is carried out in each frame of video using HOG feature and regression tree algorithm.The detected face is sent to the minor face recognition model for recognition.The recognized minor face image is fuzzified by mosaic algorithm.Finally,each frame is fuzzified by a mosaic algorithm.A frame is merged and restored to video.Compared with other research at present,the innovation of this paper is to consider the application of in-depth learning in the research of Video Privacy Protection for minors for the first time,design a video privacy protection mechanism for minors,establish a minor face recognition model based on residual network,and improve the accuracy of model recognition and verify the feasibility of the mechanism through experiments,and finally realize it.In order to protect the video privacy of the minors,the computer automatically recognizes the minors in the video and fuzzifies their faces.