Research Of Image Retrieval Based On Shape And Pose In World Wide Web

With the development of network technology and multimedia technology,people have come into contact with more and more image information.But the information distributes around various resource database of image without effective organization,that brings difficult to image retrieval.Content-based image retrieval (CBIR) methods is one of most active points in the research fields of multimedia retrieval.lt is retrieval technology based on image content,such as color,shape and texture, etc. Content-based image retrieval analyses the content of image directly and describes the features to make the course of retrieval more effective and more fit the character of human vision.Shape is one of most prominent character of image. Shape-based image retrieval is one of most important part of content-based image retrieval,it extracts and describes shape features and realize retrieval by the similarity of shape features.The extraction and description of shape features are two key issues of shape-based image retrieval,the paper studies them and specifies the ways of describing shape semantic knowledge in the current research fields.We improve the methods of getting shape features and give the corresponding arithmetic. On the basic of above,we take out some features as feature vector which would be standardized to recognise shape by BP neural network.In this paper,the descriptions of semantic knowledge of shape are discussed.We introduce some descriptions of semantic knowledge and describes semantic knowledge of shape and pose by key words of naturallanguage.Various methods are used to recognise shape.We recognise the shape which has obvious features by shape parameter.As for the shape which has no obvious features,we identify it intangibly by BP neural network.Pose is nautral form of the main body.For example,The pose of birds is flying and dwelling.Pose is identified based on categories of mian body after the main body is confirmed. The recognition of pose is made by BP neural network.We have a test on the recognition of pose,and obtain satisfying results.Last,we have a test on the shape retrieval of leaves on the basic of recognition of shape,and obtain satisfying results.The flat of test used in this paper is developed by Vc++ and Matlab based on Win2000 operating system.