Application Research Of Face Image Recognition Based On PCA,LDA And SVM

In the past two decades,data technology has developed rapidly and artificial intelligence has made great strides,especially in the field of face recognition.With the development of recognition technology at home and abroad,face recognition has been widely used in various fields.However,most face recognition technologies are highly susceptible to interference from the external environment,resulting in low recognition accuracy.The purpose of this paper is to reduce the complexity of face recognition algorithm and improve the accuracy of recognition model.A novel face recognition algorithm based on Radon transform,principal component analysis(PCA),linear discriminant analysis(LDA)and support vector machine(SVM)was proposed.In this paper,python and Matlab software are used to conduct simulation experiments on the commonly used face database.Finally,through verification,it is found that the recognition accuracy of PCA+LDA+SVM fusion method is higher than that of other single methods after Radon transformation.Finally,the recognition rate of the fusion algorithm can reach 99%.Generally,the recognition effect is good if the threshold value of face recognition is above 85%,which proves that this method has a good prediction effect.The research content of this paper mainly includes:(1)The thesis summarizes the commonly used image dimension reduction algorithms,including linear dimension reduction method and nonlinear dimension reduction algorithm,specifically introduces the principal component analysis,linear discriminant analysis and local linear embedding method,and summarizes the advantages and disadvantages of PCA and LDA methods for face recognition.(2)Aiming at the problem that PCA and LDA have low image recognition rate and can’t solve small samples when the illumination is uneven,a face feature extraction algorithm combining the two methods is proposed.PCA extracts the overall face features so as to reduce the correlation between the features of the face image and eliminate the noise.In the experiment,a more suitable projection direction is selected to make the new image retain as much as possible the feature information contained in the original image;LDA is used to extract low-dimensional information with high category discrimination ability in order to solve the problem that the recognition rate decreases with the increase of the types of face library and face template.Finally,a comprehensive face recognition scheme is formed by SVM classifier.As SVM classifier is one of the most common and effective grouping modes at present,its generalization function is good,and the classification process is accurate and efficient.Therefore,it is considered to integrate SVM classifier into the corresponding recognition algorithm.Through comparison and analysis with other single methods,it is found that this fusion algorithm can indeed improve the final recognition accuracy of face.The classification effect of LR is the worst,and the recognition rate of SVM is 6 percentage points higher than KNN.Compared with PCA+LDA method for feature extraction and KNN method for classification and recognition,SVM classifier’s recognition rate is also more effective,reaching 97%.(3)On the basis of the above,Radon transform is used to preprocess the face data in complex situations,including illumination transform,human facial expression and posture transform and the presence of occlusions.The experimental results show that the fusion PCA+LDA+SVM method is still effective in the recognition of face image data with external interference.At the same time,it shows that Radon transform can indeed remove the influence of some subjective and objective conditions,and can effectively solve the problem that face image is affected by the change of facial expression,posture expression and light occlusion.