Here is a nice image giving an overview of how video information servers work.
Camera servers collect and analyze video data in real time, then provide the results
to clients. The client application uses the data to present the scene to a user. The system off loads the hard work of video analysis to the servers, making complex, useful clients easy to build.
Max Sayles created a nice graphical editor to configure the Video Information Servers. Here is a screen shot of the editor in action.
This client tracked vehicle traffic outside the ICT building at the University of Calgary. It operated reliably for a couple of years before we shut it down to use the equipment elsewhere.
The client uses the calibration data from the server to map the video trajectories
onto an aerial photo of the campus.
This client collected statistical data about regional transitions of vehicles for use in statistical tracking. The user specifies a Voronoi diagram to create regions. The client then uses trajectory data from the server to detect when a vehicle crosses from on region to another.
We wanted to play with tracking athletes in sporting events. It is a major step to move into a sports venue, so we started with a scale model hockey rink.
After we had the bugs out with the scale rink, we had the opportunity to install cameras in the Olympic Oval at the University of Calgary. This gave us a chance to track hockey players on the Oval rinks.
We need lots of cameras to cover a hockey game, plus it takes some smarts to keep track of the game state. Here is the architecture we came up with to build a server to provide game information. The game server is essential the same software in the Video Information Servers (the processing cameras), but with a different set of modules running.
Here is a client that assists with the camera calibration step that is necessary to establish a common coordinate system for all the Video Information Servers.
An operator client interacts with the game server to manage the game information.
Here is a client tracking players in a practice in the Oval.
This is a sample mobile client that we built for following the game.
We had fun with the Video Information Servers when we helped to create some interactive Swarm Art. The Video Information Server send information about the movements of people observing the art back to a swarm simulation so that the swarm can respond to the observers.
This caused us to embark on a journey into the work of interactive installation art, guided by friend and colleague, Jerry Hushlak. The journey continues.
Here is still frame from the very first Swarm Art. Now on the cover of an issue of Leonardo.
The second Swarm Art used the Video Information Server in a different way. The camera was up in the ceiling looking down on a table. The server tracked the colored balls on the table to control swarm projected onto a wall.
Some photos of the second Swarm Art installed in the Nickle Gallery.
