Research On Semantic Extraction Of Content-based Video Retrieval

Video information classification and retrieval (VICR) is becoming an important issue in the field of multimedia with the increasing demand of customers to acquire needed information quickly by effective organization, representation, management, query and retrieval of the mass unstructurized video data, while traditional video information retrieval program fails to meet such a requirement in abstracting video content automatically, objectively and generally. Thus researches have begun to study content-based video retrieval (CBVR). However, CBVR is far from successful in bridging the semantic gap between low level features and high level semantics of video. That is to say, a lot of problems concerning semantic processing theory and technology of CBVR need a further study.Therefore, the dissertation addresses several problems about semantic information abstraction of CBVR. The overall scheme is to design and analyze the theory framework of video semantic processing first, then to carry out a detailed research on semantic processing technology of video, to present several semantic abstracting algorithms about various video contents, and finally to design and realize a prototype system for video analysis and retrieval based on semantics.Main contributions of the dissertation include:1. A multi-layers object-oriented representation model of the video semantics is presented to narrow the semantic gap, which is critical and fundamental. And then a semantic extraction framework on the basis of multi-semantics sources is constructed, which provides a theoretic foundation for the succeeding research.2. The semantic processing techniques of video based on pattern classification are explored for sports videos from multi-layer and multi-aspect(, which are detailed as follows).(1) Important features are extracted by domain color determination and field segmentation, including sports field color, texture, spatial ratio, motion texture and camera shot-time motion. And then, various ball games are classified based on ICA and SVM. Experiment has proved the efficiency of this method.(2) According to features in the making of football video, the following algorithms are presented: shot classification algorithm based on color and space feature combined with prior knowledge, slow-motion shot detection algorithm based on its making model and frame-difference sequence analysis, and penalty area recognition algorithm based on...