Face Detection And Recognition Based On CNN And Feature Point Blocking

Artificial Intelligence(AI)is an important branch of computer science and has attracted a lot of attention from the academic community.It is a challenging science.Face detection and recognition is an important application in the field of artificial intelligence,giving computing intelligence,analysis and understanding of faces.Face detection and recognition is a technique for recognizing facial feature information.A related technique for identifying a face that is detected.Face images are subject to environmental factors such as illumination and occlusion,as well as age and posture changes,which can present significant challenges for face detection analysis and recognition techniques.The environmental factors of unrestricted interference have higher requirements on the robustness of face detection and recognition,and the differences within the face class are large and the differences between classes are small.Therefore,it is becoming more and more important to design a robust detection and recognition technology against interference.This paper mainly studies the research and application of deep convolutional neural network in face detection,face feature segmentation,face feature point location and face recognition.The specific research is as follows:1.The facial image collected by the camera device has illumination blur,occlusion,etc.The local Binary Patterns(LBP)feature pre-processing images are proposed to eliminate the influence of illumination on the image,and the key feature points of the face image are Positioning and segmentation are performed to reduce the influence of occlusion on recognition,and then the convolutional network is used to extract facial features.Through the combination of two features and multi-scale detection,the faces of different sizes and denseness in the photographs should be dealt with.The effectiveness of the method is verified based on the experimental results of the AR face dataset identification flag library and the collection of daily social dataset tests.2.The face detection algorithm based on candidate frame fast convolutional neural network(Faster Region-CNN,Faster R-CNN)target detection model is studied.The feature regions are extracted by using the Regions of Interest(ROI)of the selective search mapping in the Faster R-CNN,and the face image features are extracted more finely using the multi-scale feature extraction method.Replace the original VGG16 network of Faster R-CNN with the network residual network(Res Net)with better classification effect,deepen the network depth and reduce the parameter size,and improve the detection speed without losing accuracy.3.The face image is extracted from the face image after the feature point is segmented.The Convolutional Neural Network(CNN)is a VGG13 network structure designed and improved on VGG16.The task for face recognition is the target.Optimize network parameters.4.Finally,the trained face detection and recognition model integrates the above two algorithm models to realize the 1:N face verification of the face image in the video image.