Face Recognition Fusion Multi-feature And Histograms Of Oriented Gradient

Face recognition has been widely used in the fields of finance,justice and security.Face recognition has become a hotspot in the fields of biometrics and machine vision in recent years because of its friendly and non-infringement.At present,many commercial face recognition systems have been introduced,but there are many uncontrollable factors in practical application,which will affect the performance of face recognition system to a great extent.Gestures,expressions,illumination and other factors on the face image have a great impact on the face recognition system,so the extensive application of face recognition is also facing a serious challenge.The main difficulty of face recognition technology lies in feature extraction,feature transformation and classifier design.The extracted feature is the key of the face recognition algorithm.The method of feature extraction can be divided into global feature extraction and local feature extraction.The method based on local feature extraction has better robustness to face features such as gestures,expression and illumination.Histogram of Oriented Gradient(HOG)is an effective local feature extraction operator,but its extracted features cannot fully describe the facial structure information.Therefore,by studying the HOG algorithm and making corresponding improvements algorithm.The research contents are as follows:(1)A method of face recognition based on Histograms of Weber Oriented Gradient is proposed.The differential excitation in the algorithm can extract the structural information of the image and the complete texture information,and the extracted texture information is more in line with the human visual perception,and HOG can describe the image gradient direction distribution information well.Experimental results on YALE,ORL and CAS-PEAL-R1 face database demonstrate that the HWOG algorithm is not only higher recognition rate than other face recognition algorithms,but it is less time-consuming than other algorithms.(2)A method of face recognition based Bidirectional Gradient Difference and HOG Weighted Fusion(BGDHWF)is proposed.First,the gradient components of face images in both horizontal and vertical directions are obtained,and the gradient differential operator is used to encode them.Then,the edge feature of the image is divided into the statistical information of the histogram.The both horizontal and vertical gradient difference and the HOG feature are weighted merged.Finally the recognition is performed by using the nearest neighbor classifier.Experimental results on YALE and ORL face database demonstrate that the BGDHWF algorithm is effective.(3)A method of face recognition based Multi-directional Histograms of Orthogonal Oriented Gradient(MHOOG)is proposed,which is applied to face image recognition.The orthogonal gradient can characterize the illumination invariant feature of the face image,and the multi-directional gradient information can keep the facial image structure change information and the local easy classification information.Experimental results on YALE,ORL and CAS-PEAL-R1 face database demonstrate that the MHOOG algorithm not only has higher recognition rate than other face recognition algorithms,but also robust to light.