Abstract
This paper reports an approach to accomplish tasks assigned in a competition of automated or remote controlled robots harvesting tomatoes. For automation of harvesting, we use RGB-D images to estimate position of tomatoes. With an assumption in the competition that a large black wall stands behind tomatoes, we can recognize the wall by calculating a plane equation fitting retrieved point clouds, and identify a tomato in front of the wall from red color regions of it. And then position of the tomato can be also obtained with approximating it by a sphere. Moreover our proposed approach to tomato recognition can be applied to a bunch of tomatoes using a presumed size of tomatoes.
