Face Recognition Research Based On Local Binary And Direction Pattern

With the development of biometric technology,face recognition technology has been widely used in various fields.Face recognition technology is an important biological identification technology and plays an important role in public safety and national security.Compared with other biometrics,face recognition technology has the advantages of non-contact,easy collection and high recognition efficiency.It has attracted more and more attention and achieved a good recognition effect.Feature extraction,feature dimension reduction and the design of classifier are key steps in face recognition.How to design an effective feature extraction method is especially important in face recognition.Local binary pattern and local direction pattern are two excellent local feature extraction methods.However,the method of extracting features is relatively simple and can't effectively extract the texture information of the face,through the study of local binary pattern and local direction pattern,we proposed a corresponding face recognition algorithm.The main contents of the dissertation are as follows:1)Aiming at the problems of insufficient sampling and sensitivity to random noise and illumination change of local binary model,an improved gradient local binary pattern(IGLBP)is proposed.Firstly,two groups of 3×3 neighborhood are sampled by multi-radius and multi-direction sampling mode.Then,the two 3×3 neighborhood are extracted by the gradient local binary pattern(GLBP),and then the two sets of features are encoded to produce IGLBP.Finally,the IGLBP feature is divided according to the block histogram to get the feature vector of the face,and the histogram is used for classification and recognition.The experimental results of CAS-PEAL and AR face database show that algorithm can effectively extract the feature information and is robust to variations of the illumination,expression,partial occlusion and noise in face recognition.2)To solve the problem that the local direction pattern can only extract face image information within a fixed radius scale,resulting in insufficient extraction of facial image information,a face recognition algorithm based on asymmetric local orientation pattern(ARLDP)is proposed.Firstly,we use Kirsch template to convolve with 3×3 sub-neighborhoods and asymmetrical 5×5 sub-neighborhoods to obtain two sets of edge response values,and each set of edge response values has eight directions;Then,the directions of the respective maximum value are encoded into a two-bit octal number to generate an ARLDP value;Finally,the obtained ARLDP features are divided into blocks and statistical histograms to obtain the feature vectors of the face,and face recognition is performed using the histogram intersection method.Experimental results on CAS-PEAL and AR face databases show that the algorithm can extract effective information of face images and has good robustness to the changes of illumination,expression,and partial occlusion.