Object motion estimation and interception based on stereo vision for soccer robots in 3D space

As is the first time that binocular vision has been introduced in the RoboCup Middle Size League competition, the paper utilized a stereo vision system in the robot goalie to realize ball recognition and localization in 3D space, and proposed a technique for object motion estimation including the fitting of the moving trace and the prediction of the touchdown-point. Based on the perception information, finite state machine was applied to design the goalie´s defense behavior to intercept the shooting ball. The experimental results and the actual application show that soccer robots´ ability can be enhanced greatly for object perception and defense in 3D space by the described work.