Design And Implementation Of Video Interception System Based On Face Recognition

Since the 21 st century,along with the rapid development of society,a variety of unsafe factors have come into being.Although the video monitoring system has been widely used in our life,it is time-consuming and labor-intensive to use human resources to monitor or monitor video in real time to collect key video segment information of key people.Compared with other biometric recognition technologies,face recognition has the advantages of noninvasive,rapid and accurate recognition,etc.With the rapid development of face recognition technology,good research results have been achieved.In this paper,a video interception system based on face recognition technology is designed and implemented.In real-time monitoring or monitoring video,face recognition technology is used to interception the video segment of the target character,so as to reduce the consumption of human resources to find the key information of video segment.This system is used for real-time monitoring or existing monitoring video.The main problems solved are as follows:(1)Image preprocessing.Because of the existence of illumination,the quality of video image will be reduced,so histogram equalization is adopted to preprocess the image.Because of the transmission media and other problems,video image may have noise problems.Gaussian filter and median filter are used to de-noise the image.Image preprocessing can make the detection and recognition module behind more accurate.(2)Face detection.In this system,Haar features of human face are extracted and calculated by integral graph method.Ada Boost algorithm is used for training and learning.By increasing the weight of classified wrong samples and reducing the weight of correctly classified samples,the final classifier has strong classification ability.The multi-classifier cascade structure is used again,that is,multiple classifiers are used to eliminate the non-face region in the first few classifiers,which can effectively improve the detection efficiency.The results show that this system can meet the needs of the system.(3)Face recognition.This system use GBDT based algorithm for the extraction of facial feature points,unable to deeper training network for convolution neural network model of the problem,adopted the Res Net-34 network model based on residual unit face image feature vector calculation,by comparing the target detection and face feature vector character of facial feature vector,for face recognition.Finally,the video interception system based on face recognition is implemented and tested.This system has complete functions and good interface.It can effectively intercept the video stream of monitoring video through the face detection module and face recognition module,which has certain usability.