Advanced Technology Research On Face Recognition At Long Distance

Face recognition is a biometric recognition technology based on face feature information.It has been a hot research topic in recent years.At present,the algorithm of face recognition at close distance are well developed,and many theoretical methods and practical systems are becoming more and more perfect.But with the current situation of public security is becoming increasingly serious,existing face recognition systems due to the quality of the image,the influence of environmental change and the uncontrollable factors,mostly unable to meet the remote identification requirements,traditional methods of face recognition were encountered numerous theoretical and technical bottlenecks,combined with monitoring scenarios in reality and public safety maintenance needs,has a very far-reaching significance to actively carry out research on distance theory of face recognition method and system technology.This thesis mainly focuses on long distance face recognition,and analysis thoroughly that the distance,environment and illumination are the main factors with increasing distance.In this paper,we study two key steps in long distance face recognition: feature extraction and image preprocessing.Combined with the more advanced technology,we explore a more efficient long distance face recognition algorithm.The main contributions are as follows:1.This thesis studies a overview of the overall flow of traditional face recognition and each processing stage are analyzed in detail,including the face detection,image preprocessing,face alignment,feature extraction and face matching.Combined with practical instance,this paper selects the most suitable methods and technical proposal to build a basic framework of face recognition system.The experimental results show that the system achieved good results.2.A robust feature extraction method based on convolutional neural network is proposed to solve the problem that the face image resolution is very low because of long distance.Thus most of the traditional effective features and successful processing methods is invalid with low resolution face image in long distance face recognition.This paper inspired by up-to-data deep learning methods,uses the convolution neural network to extract the robust features of the face image,and then combines the SVM classifier to identify the samples.It achieved good performance in practice finally.3.This paper combines with Retinex image enhancement algorithm and guided filter to preprocess the image.By analyzing of the causes of serious degradation of of outdoor long distance face images,we find the fog and haze in the environment makes the color of image “gray-out”,the definition poor and the contrast low,so that the saliency feature is difficult to identify.Using the Retinex image enhancement algorithm can significantly improve the image quality,especially Multi-Scale Retinex with Color Restoration(MSRCR)algorithm.But there is still some noise in the image,the use of guide image filter can relief this problem.The combination of these two methods can improve the image quality and the recognition accuracy.