Mitsuji Monta, Naoshi Kondo, Seiichi Arima, and Kazuhiko Namba

Abstract:
The vision system is one of the most important external sensors for an agricultural robot because the robot has to find its target among various objects in complicated background. Optical and morphological properties, therefore, should be investigated first to recognize the target object properly, when a visual sensor for agricultural robot is developed. A TV camera is widely used as a vision sensor for agricultural robot. Target image can be easily obtained by using color component images from TV camera, when the target color is different from the colors of the other objects and its background. When the target has a similar color with its background, it is often possible to discriminate objects by a monochrome TV camera whose sensitivity is from visible to infrared region. However, it is not easy to measure the target depth by TV cameras because many objects are sometimes overlapped in a field view. In this paper, robotic vision using TV camera for tomato and cucumber harvesting robots and a depth measurement systems using laser scanner are introduced.

Keywords: agricultural robot, robotic vision, discrimination, recognition, depth measurement, TV camera, laser scanner