Research On Ontology-based Description And Retrieval Of Surveillance Video And Development Of Streaming Media Release Platform

With the rapid development of society and the multimedia technology, surveillance video system is widely used in many fields, such as supermarkets, banks, traffic and so on. As a result, there exists overwhelming amounts video data, which eventually cause great pressure to surveillance video retrieval and other applications. Video descriptions that define low-level features, mid-level information and high-level semantic information are regarded as a feasibility solution to provide fast retrieval.In this paper, through analyzing the application requirements of surveillance system and referring to the MPEG-7 and other standards, we propose a scheme for description and retrieval of surveillance video based on ontology. Proposed scheme describes videos from two aspects, video metadata and video content. We divided the video contents by visual feature layer, video objects layer and high-level semantic information. According to the Ontology, we propose a description scheme for surveillance videos. Then design a semantic-based surveillance video retrieval system, which use the result of our description scheme. During the realization, we choose the widely-used ontology editor, Protégéand the most powerful ontology language, OWL (Web Ontology Language), to realize description of surveillance video. Jena is used to provide reasoning and OWL query API, Semantic-based retrieval system is realized based on Java platform. In the user interface, we set a preprocessing model and output conversion model. Mapping between input language and OWL language is realized. Based on this system, we succeeded in showing the output with readable video clips.In the last part of this paper, introduce the research and development work on Streaming Media Release Platform.