Development of an individual weed treatment system using a robotic arm

A robotic weed control system was developed to mechanically cut a targeted weed apart while applying a minute amount of chemical to the cutting surface of the weed. The developed system consisted of an USB video camera, robotic arm with a duel weed control mechanism end effector, soil pads, and artificial weeds contained in an aluminum structure. A modified excessivegreen method provided a good threshold image, and noise was well filtered out using continuity of pixel property, comparing the gray level of the pixel in the threshold image over neighboring pixels. The center coordinates of the detected objects in the image were determined by using the geological moment. The use of the objects geological moment obtained from the image was used to successfully determine the center of the objects with Total Error (TE) ranging from 0.0 to 0.3 cm. Inverse kinematics of the arm were solved with the given center coordinates of the weeds, and solving given inverse kinematics obtained an angle set of the arm joints moving the end effector to the desired point. A program written in Visual Basic was used for both image analysis and the control of the robotic arm. The robotic arm has a recommended workspace of 11 to 14 cm in longitudinal direction (X) with - 6 to 3 cm in the horizontal direction (Y) from the world origin. The developed robotic weed control system has an 83 % rate of success in the image area; however, a success rate of 100 % was observed within the recommended workspace. The dual-mechanism end effector showed potential for controlling weeds using both mechanisms; the mechanical weed control layer of the end effector showed feasibility to remove the top portion of the targeted weed, and the chemical insertion mechanism of the end effector provided an acceptable chemical applicability to a target.