A Novel Approach to Video Database Indexing via Semantic Classification

本研究项目的目标是开发一个新的框架，以支持更有效和安全的基于内容的大型数据库视频访问。具体而言，该框架包括：(a)分层视频数据库管理和索引结构，以定义相关视觉概念之间的上下文和逻辑关系，从而促进更有效的视频访问；(b)一个概率语义视频分类器，以弥合低级视觉特征和高级视觉概念之间的语义差距；(c)分层视觉摘要和展示技术，以便在IP网络上快速浏览视频；(d)多层视频数据库访问控制方案，以服务不同类别的用户，并维持视频数据的完整性和安全性。广泛的潜在应用可能受益于这项研究，如教育和远程学习、卫生保健和远程医疗、数字图书馆、国家安全和国防、基于网络的媒体和娱乐

The goal of this research project is to develop a novel framework to support more effective and secure content-based video access over large-scale database. Specifically, the framework includes: (a) a hierarchical video database management and indexing structure to define the contextual and logical relationships among the related visual concepts and thus facilitate more efficient video access; (b) a probabilistic semantic video classifier to bridge the semantic gap between the low-level visual features and the high-level visual concepts; (c) a hierarchical visual summarization and presentation technique to enable fast video browsing over IP networks; (d) a multi-level video database access control scheme to serve different classes of users and maintain the video data integrity and security. A wide range of potential applications could benefit from this research, such as education and distance learning, health care and telemedicine, digital library, national security and defense, network-based media, and entertainment