Design And Implementation Of Manual Annotation Video Retrieval System

With the rapid development of multimedia and data compression technology, the video amount has been increasing exponentially and the content is booming vastly. In the field of mass video archiving and management, the demand for video retrieval has shifted to a variety of traditional organization methods, which are based on the whole video as a unit for marking and archiving. While, there is still no mature solution for the needs for the fragment of search, which requires fast and accurate search for the specific content of the video fragments. Under this background, this paper designs and implements video retrieval system based on manual annotation. Introducing with Semantic Association Professional thesaurus aided annotation, and providing entry selection convenient, using existing terms to describe video, semantic video can be extracted accurately, and the standardization of annotation can be guaranteed. What’s more, it can improve the tagging efficiency.This paper introduces the current development status of content-based video retrieval technology, analyzes and compares the existing technical solutions, and expounds the advantages of the rule based video annotation method. Secondly, based on the analysis of system requirements, the overall design idea is put forward, and the selection of key technology is carried out. Then, the system is designed and divided into four modules:video management, label management, video annotation and video retrieval. Details of the implementation of the module are described in implementation chapter. Video management module completes the maintenance of video source data and the establishment of video directory. Annotation management module is responsible for generating annotation words according to the rules and the lexicon, dynamic maintenance. Video retrieval module implements various video content retrieval functions. Referring to the technology selection, this paper makes use of knowledge representation method to generate annotation lexicon, uses the list structure preserved tree data and saves label information with the document database. The author independently completed the system’s needs analysis, preliminary design and detailed design of each module, and implemented a prototype system by encoding.At present, the system is still in trial operation stage and operates well after being tested. Basically achieved the functional requirements and the desired objectives...