A Novel Indexing Method for Digital Video Contents Using a 3-Dimensional City Map

摘要

This paper presents a novel indexing method for digital video contents. The method automatically identifies city buildings captured by digital video camera. This is done by extracting the objects from candidate objects using GPS location-and-time data of the video shooter, video camera posture data taken by a Gyro attached to the camera, and a 3-dimensional city map stored in a database. The automatic calculation identifies the start and end video frames for each building object captured in a video unit. An index is created to refer to the set of all video units in which a specified building is really captured. A concrete experiment implementing the proposed algorithm in Ginza Area, Tokyo demonstrated that the algorithm works as designed.