Research And Application Of Key Technologies In Face Recognition

With the rapid development of artificial intelligence,it becomes possible to recognize a face with the machine instead of the human eye,and the intelligent face recognition technology has become a research focus in the academic and business field.With the introduction of deep learning technology,the accuracy of face detection and recognition has been improved,but the traditional method is still superior to the deep learning technology in term of speed.The paper is aimed to implement a fast and efficient face recognition system on the basis of traditional method.The paper achieves a real-time detection by improving detection algorithm;enhances the recognition accuracy by improving the feature extraction algorithm;further improves the accuracy by proposing a Multi-to-multi face verification strategy.The main work of this paper is as follows:(1)In the face detection,based on the Adaboost algorithm,the paper uses the correlation of face position between two consecutive frames,to narrow the search range and improves the detection speed,which achieves the speed of 51fps.With template matching technology,it solves the missing detection problem when the face inputs at an angle.(2)In the face recognition,the paper deeply studies the key technology of face recognition proposed by Chen et al,which is up to 95.17%,and locates dense facial landmarks by IntraFace.In the part of feature extraction,the paper associates the scale size and the number of landmarks,and uses LBP descriptor to describe the face.By the experimental results on the feature dimension of 75000 show that the recognition accuracy is improved by 0.59%.(3)In order to verify the effects of the improved face recognition algorithm,PCA dimensionality reduction method is adopted,and the joint Bayesian model is trained and tested by the international open LFW facedata sets and the sample pairs.(4)Finally,the face detection module,feature extraction module and joint Bayesian model are combined to be a face recognition system,and a Multi-to-multi strategy is proposed.Then collectes photoes of 50 students by real-time camera as test objects,and get 92%recognition accuracy,compared to one-on-one verification accuracy rate increased by 8%.