A New Content-Based Video Retrieval System - Data Model and Query Processing

Abstract

A video retrieval system manages large scale of video archive and provides content-based accesses to users. In this paper, we propose a video data model with the advantages of flexibility and sharability by grouping different video materials of interest to users into different representation frameworks according to their individual characteristics. Unlike pervious approaches of keywords or free text, we provide a nested annotation language for describing complex scenarios in video data effectively. A SQL-like query language based on the proposed model is presented. With the assistance of domain knowledge and index organizations, we develop algorithms to efficiently process six types of queries about semantic information of video content, including elementary domain query, elementary content query, object-event relation query, object path query, event inference query, and temporal query. Finally, a prototype content-based video retrieval system is implemented.