Face Recognition Based On Weber Local Features

Face recognition is a kind of biometric recognition technology.With the development of more than half a century,face recognition has begun to take advantage of social life.With its friendly and easy to collect,it has attracted more and more attention,Currently,face recognition has been successfully applied to many areas of social life,such as entrance inspection of train stations,face shopping,tracking of criminals,etc.However,face recognition still faces many problems,and the recognition effect is easily affected by changes in external complex environments.Extracting the local features of the face for recognition is a prominent way.Weber local descriptor(WLD feature)is a typical local feature extraction operator,which can effectively extract the texture and structure information of the face image,and Illumination has good robustness,but the recognition effect in response to environmental changes such as expressions and occlusions is not ideal.In this paper,the Weber local descriptors are improved to improve their robustness,and combined with other features to form fusion features to improve the ability of the algorithm to cope with complex environmental changes.The main work of the thesis is as follows:First: A face recognition algorithm based on multi-directional Weber gradient histogram is proposed.The algorithm firstly improves the Weber local descriptor.Based on the original differential excitation,the relationship between the neighboring pixels is further considered.The neighboring pixels are summed to take the absolute value and then summed,and the original Weber direction is replaced.The direction of the gradient constructed by the pixels in the diagonal direction.On the other hand,the Weber gradient magnitude is accumulated in the original Weber direction to extract the edge contour feature,and the histogram from the feature is concatenated with the statistical histogram in the improved Weber local descriptor to form a fusion feature histogram.Finally,use the nearest neighbor classifier to classify and identify.Experiments on different face databases show that the algorithm has good robustness in dealing with environmental changes such as expression,illumination and occlusion.Secondly,a face recognition algorithm based on Weber local ternary mode is proposed.Firstly,the Sobel operator is used to replace the Weber direction in the improved Weber local descriptor.In addition,the local ternary mode is used to encode the neighborhood pixel and the central pixel,and the coded feature is used with Sobel.The improved Weber local descriptor sub-block statistical histogram of the operator,the statistical histograms are concatenated to form a combined feature and classified by the nearest neighbor classifier.Experiments on the YALE,ORL,AR and CAS-PEAL-R1 face databases show that the algorithm has high recognition rate and good anti-noise performance.