<title>Dynamic workspace monitoring</title>

摘要

To give advanced robotic systems the capability to share their workspace with humans, they must be equipped with cooperative behavior. Efficient cooperation is only possible if the separation surface between the two workspaces can be defined in a very flexible way and as a function of the current action of the robot (dynamically). We use a texture based stereo vision method. In order to test if an object is about to enter the robot's workspace, it is sufficient to know if the object is on the separation surface. In order to detect objects on the separation surface, one camera image is transformed by assuming that every point is on the separation surface, and then compared with the real image. Corresponding regions in both images belong to objects on the separation surface. We present this new method, its algorithms, and a possible implementation suitable for real time processing. The experimental setup and the results of the evaluation of an adequate correlation method are also presented.