Face Recognition Based On Feature Fusion And Deep Learning

As a hot research topic in the field of computer vision and pattern recognition,face recognition technology has been widely concermed by academia and industry.At present,face recognition technology has been widely applied in various fields of daily life,however,in the realistic scene,illumination and occlusion will seriously interfere with the performance of face recognition algorithm.For this reason,this paper improves the face recognition technology on the basis of the existing research results.Face recognition includes three steps:image feature extraction,data dimensionality reduction and feature matching.In this paper,face image feature extraction method is mainly studied.In this paper,the shortcomings of face recognition technology are studied.When multiple interference factors exist,a single facial feature can not get good recognition results.Therefore,the effect of various feature fusion methods on the performance of face recognition system is analyzed.In the aspect of feature fusion,the composite invariant moment and the gray level co-occurrence matrix are weighted together in this paper.Firstly,the image processing technology is used to divide the face region into the small size image,which is regarded as the basic operation of the later feature fusion.Secondly,the Hu moment feature of each sub block 1s extracted,and the weight is determined by the method of information theory,and the weight coefficients are used to combine the characteristics of these sub blocks.Then,the GLCM features of the face image are extracted.Finally,we use the support vector machine to achieve the weight fusion of two features in the decision-making layer.The experimental results of ORL and Yale face database show that the feature fusion method proposed in this paper can improve the face recognition rate significantly.Face recognition based on CRC is directly combined with all face images into over complete dictionaries.The dictionary constructed in this way can only describe the global characteristics of face images,and can not represent the local features of face samples very well.Therefore,in order to improve the recognition performance of CRC algorithm,we use Gist operator to extract local features of human faces.Based on that,we propose a new face recognition model GL-PCRC.The final experimental results show that the face recognition model GL-PCRC proposed in this paper can further improve the recognition performance of the CRC algorithm.