Real-time tracking of moving objects on a water surface

This paper presents a real-time tracking system for moving objects on a water surface. This tracking system is used to construct a fish-like robot water polo competition platform. Similar to the land robot soccer game, the robot water polo game provides a real-time multi-agent environment from the viewpoint of distributed artificial intelligence and multi-robot cooperation research within water circumstance. For doing this, a camera is set upright over the water surface. A video stream from the camera containing the complete scene on the water surface is transmitted into a computer in real-time. In the computer, the tracking system deals with this real-time video and provides all the posture information of all moving objects(including robots and ball) on the water surface. The specular reflections when the object moving in the water is the main source of noise in the video stream. A robust tracking algorithm based on template-matching method is proposed to overcome this disturbance. Experimental results illustrate the effectiveness of our tracking algorithm.