Cognitive Autonomous Systems Laboratory

The CASL Video Server

Design and implementation by Björn Husberg


The video server is able to perform the following three processes continuously and concurrently. On one hand, it receives a video stream from the robotic UAV agent, assigns timestamps to each frame, and stores the video for future retrieval. Secondly, it can receive video forwarding or replay requests and honor them by sending timestamped video to the SGUI, either by directly forwarding a copy of what it is presently receiving from the UAV unit, or by retrieving what was previously stored starting at the time that is specified in the request.

Thirdly, the video server is capable of representing video markup besides the video itself. The markup specifies the position of the UAV and the video camera when the video frame was recorded, as well as the position of mobile objects (such as automobiles) in the video field. This information is required for the transformation from positions in the frame, which are obtained when the user points into the video frame, to the corresponding geographical position and timepoint. The video server can therefore receive queries containing a video timestamp and return the markup information allowing the transformation.

Obtaining the markup for the video that is recorded during an actual flight is of course a difficult problem in itself, and a problem that belonged to the UAV Technologies subproject within WITAS. Awaiting a video interpretation facility that can provide us with the markup in real time, we use two intermediate techniques. On one hand, recorded video from actual UAV flights are provided with markup by manual means, after which they can be used for testruns where the video is used in playback mode. Secondly, in a current project, the existing Hazard simulator/visualizer is being extended so that it generates composite-signal video for the animation together with a markup that is obtained directly from the simulator, thereby shortcutting the video interpretation step.

References

2005-013 Erik Sandewall, Hannes Lindblom and Björn Husberg: Integration of Live Video in a System for Natural Language Dialog with a Robot. DIALOR Workshop on Dialogue.


Posted on 2005-05-17 as part of the CAISOR website. [Version history].