Pallet Detection and Estimation for Fork Insertion with RGB-D Camera
Ryosuke Iinuma, Yusuke Hori, Hiroyuki Onoyama, Takanori Fukao, Yukihiro Kubo
Abstract
Forklifts are necessary to carry pallets in agricultural work; however, a high level of skill is required to steer and control forks. Inserting the forks into a hole of the pallet is difficult for operators to control both the height and angle of the forks to avoid damaging the pallets because the size of the hole is small. The slight inclination of the ground could lead to the collision of the pallet and the forks. We describe an approach to automatically insert forks into a hole of a pallet on a truck under a tilted ground. Our approach for the safe fork insertion of an autonomous forklift uses an RGB-D (RGB, Depth) camera. The pallet is detected by semantic segmentation with the RGB image during insertion. The system calculates the position and orientation of the pallet with the result of semantic segmentation and depth. We verify the effectiveness of the approach with an experiment with the forklift in an outdoor environment assuming the real task.