抄録
In order to automate production and physical distribution systems involving the handling of a variety of products, it is necessary to develop visually sensitive systems that can “see” objects and adapt to the changes of their forms and positions. To meet this requirement, the authors developed a visual device around a TV camera and a simple servo-controlled handling device connected to a control computer. These devices can automatically sort objects flowing along a belt conveyor by visual information transmitted by the TV camera.
The object recognition process of the device involves the following steps: (1) Image input, (2) Image positioning, (3) Transforming into polar coordinates, (4) Matching process utilizing registered object images and a rotational operation.
The computer program for these devices is about 12 kilowords, including a visual data area. From the experiments, it is found that the maximum recognition time is 1 second, and the positioning error in stacking objects is within ±1mm in the X-and Y-directions and less than 10 degrees in the angular direction.
