Design Of Face Recognition Attendance System Based On Embedded Platform

With the rapid development of artificial intelligence and Internet of things technology,the research on face recognition attendance system has made great progress.Face recognition attendance technology has gradually moved from laboratory research to commercial application.Face recognition attendance machine has the advantages of non-contact,safe,high-efficiency,and easy operation.However,the disadvantages of face recognition,such as high cost,high power consumption,and poor portability of face recognition attendance machines based on PC platform,limit the application range of face recognition attendance machines.And the current embedded processor is in terms of power consumption,size,and cost.All have the advantage that the PC is difficult to match,and they have the data processing ability to execute image processing algorithms in real time.Therefore,the face recognition attendance system based on embedded platform has important application value.In this paper,a real-time face recognition attendance system based on the embedded Cortex-A8 processor is implemented.We design the hardware and software system of face recognition attendance machine under embedded platform,including face detection,face recognition,Qt-based GUI,information storage module based on Sqlite's check-in.In the face recognition module,the face recognition algorithm of multi-dimensional integration of global features and local features is researched and proposed.Firstly,SURF corner features that are easy to use in embedded systems with less computation are extracted in the face,and the SURF features are faced.It is difficult to construct feature vectors with equal dimensions when the number of points is not uniform,and the word bag technology BoW(Bag of Word)is used to build such feature vectors as dimension vectors.Then,in order to reduce the amount of computation to extract the global Gabor features of PCA after dimension reduction and make better use of the two types of features to play a complementary role,the two types of features are merged by constructing multidimensional vectors.Finally,SVM(Support Vector Machines)is used for classification and identification.The fusion algorithm is tested in a variety of face databases to achieve a better recognition effect than a single feature,and the recognition speed was improved by an average of 6.1 ms compared to a single algorithm.The recognition rate is 1.16% and 5.13% higher than the single algorithm(SURF,Gabor).Finally,the face recognition attendance system in this paper is used for real-world scene measurement,realizing the real-time face recognition attendance function based on embedded platform.