Abstract
The visual sensor is an essential component for agricultural robots which work in green-houses, plant factories and so on. We consider that the visual sensor must have three functions, that is discrimination, recognition and position measurement. In the Part 1 of this study, it was observed that cucumber fruit was discriminated from its leaves and stems and that its shape was recognized even when the fruit was divided into two parts from experimental results. In this paper, an experiment of fruit position measurement was done by the method of binocular stereo vision. Besides, an experimental robot harvested some cucumber fruits with the visual sensor. From the experiments, the following were obtained. 1) It was easy to get the correspondence of fruits in left and right images by the method of binocular stereo vision and the fruit position was detected precisely, since cucumber plants were cultivated in inclined trellis training so that few fruits were hidden by leaves and stems. 2) The developed algorithm for position measurement made the correspondence possible even when fruits were partially hidden by obstacles. 3) It was confirmed that the error of measurement by the visual sensor was reasonable and that it could be compensated with the structure of end-effector. 4) After investigating the roles and the performance of visual sensors in the robotic system, appropriate visual sensors for agricultural use should be developed.
