Research On Feature Extraction Algorithm Of Face Image Based On Cooperative Representation

Facial recognition technology has a wide range of applications in various fields of identity recognition,including collecting images from high-definition cameras,preprocessing the images after obtaining them,and using various facial recognition algorithms to process the images into data and execute image features Collect,and finally perform feature matching with the face in the database.At present,the main research hotspot is how to optimize the feature extraction algorithm.The main research contents of this paper are as follows:Aiming at the problem that the single feature extracted by the image feature extraction method cannot represent the image well,the idea of feature fusion is introduced.The uniqueness of this algorithm is that a variety of image features can be extracted for classification using one method.A facial recognition algorithm based on two-dimensional linear discriminant analysis and collaborative representation is proposed.Firstly,the between-class and intra-class features extracted by the two-dimensional linear discriminant analysis are reconstructed,and the inter-class virtual image and the intra-class virtual image are obtained.Secondly,the collaborative representation is used to obtain the scores of the inter-class virtual image,the intra-class virtual image,and the original image respectively and perform weighted fusion.Finally,the final score is used for image recognition.This method not only effectively suppresses the influence of illumination and expression on facial recognition,but also effectively improves the performance of facial image recognition based on the complementarity of the obtained inter-class and intra-class virtual images with the original image.Aiming at a problem that the robustness of image features is not good enough,a collaborative representation classification algorithm based on image multi-feature fusion is proposed.First,the two-dimensional fast Fourier transform(FFT)is used to extract the spectral features of the original image.Secondly,two-dimensional principal component analysis(2DPCA)is used to calculate the covariance matrix,then feature extraction is performed,and then the extracted features are reconstructed.The spectral features extracted by FFT and the virtual image obtained by 2DPCA are scored using the collaborative representation method again.Finally,the adaptive weighted fusion mechanism is used to fuse the obtained spectral feature scores with the virtual image scores,and the new scores obtained after the fusion are used for face image classification.Aiming at the problem that 2DPCA cannot use the category information of training samples,a collaborative representation classification algorithm based on two-dimensional linear discriminant analysis of kernel norm is proposed.First,the kernel norm is used to measure the inter-class scatter matrix and the intra-class scatter matrix,because the kernel norm can well represent the spatial relationship of pixels,and can effectively reduce the influence of the external environment and the performance of the algorithm.Secondly,the extracted features are reconstructed.This process is also called obtaining the virtual image of the original face image.Finally,using the collaborative representation method to fuse the inter-class virtual image,the intra-class virtual image,and the original image can make full use of the complementarity between the features,and use the fused facial features to classify the face image.