Online Detection of Tomatoes for Robotic Harvesting

To minimize potential damage, it is crucial to carefully harvest greenhouse crops like tomatoes at the optimal time. To improve this process, the use of robotic harvesting methods has been proposed. The robotic harvester consists of important components including a mobile platform with robotics, displacement units that can move linearly or rotationally, a manipulator, a gripper, a camera, an image processing-based fruit detection unit, and a depth sensor. A robotic manipulator with three linear degrees of freedom was created in the Cartesian coordinate system. To enhance its capabilities, a gripper mechanism was incorporated, providing an additional rotational degree of freedom. The primary objective of this robot was to autonomously detect the position of ripe tomatoes. To achieve this, the displacement control of both the robot arms and gripper was executed through commands from the image processing unit. Different channel of some color space was studied. The effectiveness of this channels was assessed by conducting tests in the presence of tomato plants. The accuracy of the system in approaching the crop were thoroughly evaluated. Channels H of HSV color space, Cr of YCrCb color space, and a of Lab color space showed better result. The accuracy of detecting ripe tomatoes in channel H of HSV color space was the highest and 87%.