A Vision System for RoboCup: Diploma Thesis

This diploma-thesis presents a vision system for robotic soccer which was developed and implemented by the author and tested on Sony’s four legged robot Aibo. The input for the vision system are images of the camera and the sensor readings of the robot’s head joints, the output is a set of percepts, where each percept describes the position of a recognized object in relation to the robot. There are two main features of the vision system: it is fast and it needs no manual color calibration. The high processing speed is reached by an attention based distribution of scan lines over the image which leads to a reduced number of image pixels that have to be processed. During the normal operation of the robot the colors are calibrated by the vision system itself using knowledge about the environment of the robot. The adaptation of colors is based on statistics which are computed when recognizing objects. Three different levels of color representation are used to refine the color calibration while the robot explores more and more parts of its environment. Experiments on Sony’s four-legged robot Aibo show that the method is able to recognize and distinguish objects under a wide range of different lighting conditions and is able to recalibrate in appropriate time.