画像処理と深度情報を用いた滑車位置推定に基づく天井クレーン地切り前位置補正システムの開発

抄録

While the automation of overhead cranes is progressing, the lift-off process is still manually operated. In this paper, we developed a pulley position estimation system for the lift-off process, aiming for stable estimation independent of the environment. The system obtains RGB images and depth information from a depth camera, which can be easily installed on actual equipment, uses semantic segmentation to determine the pully region from the RGB images, and extracts only the depth information of the pulley region. The experimental results using an overhead crane show that the proposed system can estimate a pulley position and the amount of trolley movement, even with unlearned images.