Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 28, 2023 - July 01, 2023
When weeding with a robotic hand, part of a weed may remain depending on the grasping position. Therefore, it is necessary to detect a weed stem as an appropriate weeding position. A method that uses deep learning to segment weeds and detect stem positions requires pixel-wise labeling of the images in the dataset, which takes a lot of time and effort. This study proposes a method that detects a weed picking position using YOLOv7 and image processing. The proposed method identifies weeds using YOLOv7, and then detects the stem position only by combining simple image processing. This simplifies the preparation of the dataset compared to deep learning-based methods for detecting stem positions.
