Research On Retrieval Models And Applications Of Corrosive Material Feature Image Based On Ontology

Nowadays, corrosion characteristic on the surface of material is mainly shared and administer through the method of text retrieval in a China Industries Institute Natural Environmental Testing Center. As text retrieval on corrosion characteristic is influenced by individual differences and experiences, so there are some defects with it. Firstly, it is hard to guarantee the reliability of retrieval. Secondly, the similarity of corrosion conditions makes many similar images in the image library. Finally, in the information age, as the increasing number of corrosion feature images, the keyword retrieval method is also difficult to adapt to customer needs.Ontology theory has been drawing more and more research attention in the field of image retrieval in recent years. In this thesis we firstly introduce ontology theory, and then built a general-purpose ontology model used to describe images comprehensively, with which we can describe different aspects of image information and make five-tuple description of ontology.It is difficult to understand the image directly obtained from the low-level visual features, which requires analyzing the semantics of the image deeply, making the computer can understand the similarity based on human cognition. To this end, we associate the body, proposed framework for ontology-based image: by ontology-based semantic extension, retrieval semantic query process to make up for the lack of information; through the ontology, the semantic definition of the concept of image relationship. Image ontology models include the visual image features, high-level semantic concept, not only take full advantage of the low-level features of image itself, but also consistent with the visual images of people from understanding, thus filling the "semantic gap" between the low-level visual features and high-level semantics ; other studies staff inspired to do this paper with further improvement of watershed algorithm , in ensuring the segmentation results basically unchanged to lower the time complexity. In order to achieve high-level image automatic extraction of semantic features, this paper introduces the basic theory of SVM, support vector machine using the underlying characteristics of the image mapped to the ontology of the high-level semantic concepts, so we will automatically obtain semantic annotation information of images from the underlying characteristics.Finally, we design a Ontology-based corrosive material image retrieval system by the establishment of the tool, the system rule, custom rules and the ontology model of corrosion images that designed by myself. We try to provide new ideas for the application of ontology in information retrieval. As the limited time and many other reasons the study in this thesis is just a beginning step of the research in this field. There are a lot of problems in the model and need a further study.