The Researchon Reconstructive Representation Algorithm Of Image Classification

In the 21st century,the rapid development of computer technology and internet technology has been brought to us.As the foundation and important subject of the pattern recognition and computer vision,image classification technology has attracted extensive attention from researchers.Among them,face recognition,target detection and scene classification are the most noticeable problems.In recent years,researchers have proposed reconstruction representation algorithm to enhance the robustness of illumination and expression in face recognition technology.Dictionary learning is used to improve the accuracy of target detection and scene classification.It is easy to overlook that,as different categories of objects,the information between classes and the class content cover is more abundant.This paper focuses on applying the global feature between classes and within classes to the reconstruction representation image classification algorithm and makes the following arrangements:Firstly,in the first chapter,the paper introduces the background of image classification technology,its theoretical significance and practical value;moreover,it analyzes the research status of image classification technology,summarizes the existing problems and challenges in the current technology;and introduces the arrangement of this paper.Then,in the second chapter,a collaborative representation based on l2-norm regularization is proposed.The algorithm uses Tikhonov regularization constrains to reconstruct coefficients and adds regularization constraints to enhance the between-class and within-class discriminability of coefficients.Two regularization constraints are introduced into the cooperative representation to improve the classification quality of the collaborative representation without affecting the time complexity.Then,in the third chapter,a discriminative coefficient collaborative representation-based classification algorithm is proposed.This algorithm adds the extraction of reconstruction coefficient information in Fisher's discrimination dictionary learning process and applies the information to the collaborative representation algorithm for image classification.Then,in the fourth chapter,the article proposes Fisher discriminative coupled dictionary learning,and introduces the effective Fisher discrimination in chapter 2 and 3.Because of the application of collaborative representation,the algorithm can improve the testing time greatly and use the coefficient projection dictionary of double dictionary to carry out the final classification.In chapter 5,the algorithm and research topic are summarized and prospected.