Face Recognition Algorithms Based On Local Direction Gradient Magnitude And Phase Difference

Automatic face recognition,as a biometrics based on physiological charact eristics,has a great challenge in addition to the theoretical value,and it has a wide application prospect in the field of identity authentication.Face recognition is currently a popular topic in the field of pattern recognition,image proce-sssing and computer vision.After many years of domestic and foreign research scholars of the tireless efforts,and now,in a controlled environment,the face recognition algorithm has been basically able to meet the demand.However,in the light,gesture,expression and occlusion and other uncontrollable complex environment,face recognition still has many problems.Local Direction Pattern(LDP),as an effective face description method,improves the robustness of the local Binary Pattern(LBP)algorithm to some extent,but it also exists inherently defect.Based on the analysis of LDP,LDN and MBP methods,a face recognition algorithm with local gradient and amplitude difference is proposed.The main contents of the thesis include:(1)A face recognition algorithm combining the local direction gradient and phase difference(LDGMPD).First,the gradient amplitude of the extracted image is combined with the phase,the gradient amplitude image and the eight Kirsch templates are convoluted to obtain the eight edge gradient values of each sub neighborhood,and then the phase is localized.LDGMPD uses only the edge gradient value and the phase of the local differential value in the direction of the maximum encoding into a two octal number,resulting LDGMPD value.And then the LDGMPD feature face is weighted and the LDGMPD histogram feature is extracted.Finally,the nearest neighbor classifier is used to classify and identify.Experiments show that the proposed algorithm is simple and effective,and it has better effect on light,expression and attitude.(2)This paper presents and implements the LDGMPD face recognition system based on Qt + OpenCV under ARM platform.Firstly,it builds the embedded development environment,including the configuration of virtual machine,Ubuntu10.0,cross compiler and other development tools.Finally,in the algorithm implementation phase,through the V4L2 acquisition of images,combined with open source OpenCV visual library,in the Qt development environment Qt Creator to achieve the mainframe,the root of the file system,the relevant drive cutting,compiling and transplantation;The face of the detection and identification.