Relevance Feedback In The Cg Project Management System

With the multimedia technology and the rapid development of internet network,image source is expanding,High-speed large-capacity storage systems provides the basic protection for mass storage of images.All walks of life extensive use of images more and more.The organization,management and retrieval of image information resources have become increasingly important.In order to enhance the utilization of image resources,retrieval of image information resources has become a hot research.For the semantic feature extraction become more difficult.Image retrieval based on the characteristics of low-level is a method commonly used.Image retrieval based on the characteristics of low-level easily leads to low-level features and characteristics of the gap between high-level.Relevance-feedback is one of the ways of solving the semantic gap problem currently.There are many types of relevance-feedback.Such as move query vector,modify the weight of characteristics, Bayesian,SVM,neural and networks.In this paper,these relevance-feedback algorithms are compared.Relevance-feedback based on algorithm of move query vector and modify the weight of characteristics are be chosen to research and analysis.An algorithm is designed based on the original algorithm. The algorithm combines two kinds of relevance-feedback algorithm.And it is brought into the CGProject.Major jobs are as follows:1.Analyzing the characteristics of the relevance-feedback algorithms,by comparison of these algorithms,select the appropriate algorithm of move query vector and modify the weight of characteristics for CGProject to research.2.Designed an algorithm based on the algorithm of move query vector and modify the weight of characteristics.And it gave a simulation experiment to compare these three algorithms in this paper.3.Apply relevance-feedback algorithm designed in this paper into CGProject,realize a relevance feedback module in the image retrieval system.Through relevance-feedback increased the efficiency and the performance of the image retrieval system in CGProject.