Research And Design Of Face Recognition Intelligent Access Control System Based On Embedded Technology

In recent years,with the rapid development of information technology,face recognition as a convenient and safe biometric technology has been more and more used in access control system.However,at present,the access control system based on face recognition often uses PC as the carrier and needs to establish a server,which has high cost and is limited by the network.Therefore,it is of great significance to study the face recognition access control system based on embedded technology.However,it still has some problems such as slow face detection speed,low recognition rate with fewer samples,and no face liveness detection and so on.In view of the above problems,this paper studies and optimizes the access control system of face recognition under the embedded system of raspberry PI.Firstly,the Ada Boost face detection algorithm was studied.In view of the disadvantages of haar-like feature description method,which requires a large amount of computation,the improved LBPH feature description method was used to replace it,and the detection time was reduced by about 90% through experiments.Secondly,the face recognition algorithm based on PCA was studied,and according to the disadvantage of losing structure information,face enhancement is carried out for legitimate face data through perspective transformation to simulate the posture of head up,head down and head turning,so that the recognition rate could be improved by about 10% in the case of a small number of single face samples,and better than other improved algorithms.Thirdly,the algorithm of face liveness detection was studied.The convolutional neural network can automatically extract the different features of real faces and fake faces according to the data samples,and the accuracy is higher,but the convolutional neural network has a large amount of calculation and a large number of parameters,not suitable for operation in embedded devices.In order to solve this problem,this paper proposed a novel convolution kernel and convolution method,which make the parameters of the convolution neural network greatly reduced in the case of small loss of accuracy,and then a convolution neural network SDChannel Nets is designed.At the same time,face liveness detection function was realized by training SDChannel Nets,which was verified by experiments to reduce the calculation time of the baseline model by 45%,the number of parameters by 84.2%,and the accuracy loss was small.Finally,the face recognition intelligent access control system based on embedded technology is designed,which is a completely offline face recognition access control system integrating face detection,face recognition and face liveness detection.In addition,the functions and performance indexes of the system were tested through the data set composed of students in the laboratory,which achieved the expected goal and had a good application prospect.