Generation of video metadata supporting video-gis integration

Abstract

Geospatial information expressed by video can provide more realistic and comprehensible information that cannot be obtained from digital map. We introduce video and integrate it to CIS by supporting video-map cross reference and bidirectional search. To support content-based search of geospatial objects appearing on video, we generate video metadata that has necessary information of geospatial objects. The most important part of the video metadata is outline of object on video frame. The outline of an object varies frame to frame, and thus should be generated for every frame in which the object appears. Because the object moves rapidly between video frames collected with 1-second interval, object tracking cannot be easily done by simple algorithm. In this paper, we devise a semi-automatic object tracking method by combining photogrammetric solution with image-based method.