Pallet Detection and Distance Estimation with YOLO and Fiducial Marker Algorithm in Industrial Forklift Robot
Eric Sean Kesuma, Pranoto Hidaya Rusmin, Devira Anggi Maharani
Abstract
Utilising technology such as artificial intelligence and robotics potentially improves E-Commerce in efficiency. In this trends, the usage of autonomous forklifts in the warehouse to lift and arrange things should be implemented. The picking system in the warehouse needs pallet detection and tracking to carry out the things. This research will find the best performance of the YOLOv5 model and correct the distance estimation model to the fiducial marker. In this paper, we used the ArUco fiducial marker to mark the pallet target and estimate the pose and distance in real time. The insertion points of the pallet were also detected using the YOLOv5 algorithm to validate the pallet and get the coordinate variables of the holes. The YOLOv5n gives the best performance at 24 fps in real-time detection. Distance measurement from the marker detection had an average error of 2.28 cm with linear regression.