Research On Face Recognition Technology Based On Improved SIFT Algorithm

With the rapid development of social economy and the improvement of people's living standards,more and more people realize the importance of information security.In recent years,with the high-speed iterative updating of computer hardware and software,many target human identification techniques have appeared,and face recognition technology has been widely used in people's daily life.However,in practical application,the accuracy of face recognition will be affected by non-ideal factors such as occlusion,illumination and scale transformation,and how to realize the high accuracy of face recognition under complex and non-ideal conditions has become an expert in the research of experts and scholars.The feature descriptor generated by SIFT algorithm has the characteristics of insensitivity to illumination and occlusion,affine,rotation and scale invariant,because of its high unique and robust characteristics,which are widely used in image recognition and classification.Therefore,this paper makes a related research and exploration on the application of SIFT algorithm in face recognition.Firstly,this paper introduces the basic structure model of face recognition system,describes its working principle through its model,and comprehensively analyses the main factors affecting the accuracy of face recognition.On this basis,several common face recognition methods are introduced.Secondly,the basic theory of SIFT is expounded,the Gaussian kernel and Gaussian scale space,which can only realize scale change,are introduced,and the detailed steps of SIFT algorithm are deduced theoretically in order to carry out the follow-up work smoothly.Then,the improved SIFT algorithm is used to extract the feature vectors of human faces and realize face recognition in the word bag model.The improvement focuses on two aspects.One is to detect candidate feature points in a wide range during the extraction of feature points;the other is to reduce the dimension of 128-dimensional feature vectors by using the improved dimension reduction method.Finally,a comparative experiment is designed to verify the effectiveness of the improved algorithm.Finally,three improvements are made to overcome the shortcomings of SIFT algorithm in face recognition,such as background interference.Viola-Jones algorithm is used to extract the face image from complex background in the process of face preprocessing.Secondly,Sobel filter is used to define the gradient and direction of feature points.Finally,a simplified circular neighborhood is used to describe the generated SIFT feature descriptor.