Research On Image Recognition Method Based On Multi Feature Fusion

Face recognition technology is a very active research field about image recognition technology,which includes many scientific and technological contents such as digital image processing,image recognition,computer vision,convolutional neural network,pattern recognition,mathematics,etc.Because of the simplicity of the traditional method of face recognition,the extracted face features are too simple.In order to overcome the problem of too simple feature extraction,this paper proposes an algorithm that can extract more abstract and complex facial features through local binary patterns(LBP)+(fast Fourier transform: FFT),thus improving the face recognition rate of the system.This face recognition system has a very good performance under the influence of the external environment,not only the recognition speed is fast,but also the accuracy is very high.In the open ORL,Yale and GT face databases,compared with LBP,FFT,(local binary patterns: LBP)-(histogram of original gradient: HOG)and multi-layer lbp-hog,the proposed multi-layer lbp-fft improves the recognition accuracy by 0.8%-41.5%.Experimental results show that multi-layer lbp-fft can extract face features better than all existing methods.In this paper,the recognition system of face feature extraction is designed based on the deep learning theory and sparse automated collaborative representation based classification.At the same time,combined with the proposed(sparse automated collaborative representation based classification: SA-CRC)and(collaborative representation classification: CRC),it can better improve the accuracy of recognition.