Video Retrieval Methods Using Geographic Information in Windows Azure Cloud

摘要

Every day, thousands of videos are uploaded to the web, creating an ever-growing demand for methods to make them easier to retrieve, search, and index. These videos include both spatial and temporal geographic features captured via camera and embedded sensors, e.g., GPS and the digital compass. The current state-of-the-art video retrieval systems are based on content-based or concept-based techniques. In addition, querying massive videos is inherently data intensive, computing intensive and concurrent intensive, thus, the process demands effective computing solutions, such as cloud computing. In this paper, we focus on video retrieval methods using geographic information in the Windows Azure cloud. The main idea is to query videos utilizing the location, trajectory and azimuth information acquired by sensors. The raw spatial information is synthesized to point, line and polygon according to the camcorder parameters. We defined the frame point, video trajectory, field of view polygon and cone, and then used the spatial relationships to retrieve videos. We implemented a framework with the methods in Windows Azure. In addition, we evaluated and analyzed the performance and efficiency. This research illustrates the feasibility and advantages of cloud computing-based video retrieval using geographic information, and reveals important application values in the industry and community.