Geo-Tagged Video Management: Storage, Queries and Streaming

Abstract

In this thesis, we investigate the data storage, query processing and video streaming in geo-tagged video management. Our work focuses on the following key issues: (1) the indexing and querying on the geo-tagged video. We design a multi-level grid-based index structure for effectively and efficiently searching for videos, especially for specific query types(e.g., query with bounded radius, query with direction restriction); (2) the uncertainty of the sensor meta-data. The uncertainty model for a single FOV, and a light-weight approximation of the uncertainty of video segment are modeled with probabilistic theory; (3) the video transcoding strategy for DASH in the cloud environment. We propose a dynamic scheduler that can support near-live DASH streaming, with consideration of user's interaction with the video hosting service.