Accuracy of positioning spherical objects with a stereo camera system

The usage of sensory information for automation systems like autonomous robots is rising. For these applications the accuracy of the sensory data is of high importance. This is also valid for the task of transport by throwing where a thrown object has to be caught based on trajectory predicting using input data of a stereo camera system. Here the positioning accuracy of such a system is examined under varied illumination circumstances for a spherical object that is stationary. Both test cases with external illumination (susceptible to reflexion and other effects) and with internal illumination (a self-lightning sphere) show that the vision system is capable to determine the sphere's center position with an accuracy in the magnitude of single digit millimeters. When considering the required prediction accuracy for a catching robot, which is in the range of single digit centimeters, the vision system is a suitable sensor system for the specified task. Future work will target the accuracy determination of moving objects and the utilization of the vision system for the task for transport by throwing.