This contribution describes the vision module of a 3-D object manipulation system, which was implemented in the framework of the so-called COR project. This large interdisciplinary project was promoted by the Mechatronics working group of the Swiss Federal Institute of Technology. The Mechatronics group comprises the Institutes of Robotics, Electronics, Control Theory, Communication Technology and the Chair for Electrotechnical Constructions. In order to focus the work and to be able to monitor and demonstrate progress, it was decided to define a concrete and challenging task which consisted in building a complete system for clearing cafeteria trays. The vision system was to analyze the constellation of dishes and cutlery pieces on a tray so that it could communicate to the robot the information necessary to grasp a piece and put it aside. When a sequence of these operations had cleared a tray, it was moved away, and the next tray to be cleared moved into position. The vision system used three CCD cameras to get the necessary visual data, and the interpretation of the scene relied fully on the object model based approach. The decision procedure used can be briefly characterized as rule-based evidence accumulation. The project extended over four years, and completed with a successful demonstration of the full system at the exhibition ``Industrial Handling 92'' in Zurich.