EuroPallet Detection with RGB-D Camera Based on Deep Learning

Abstract

This study related to enhancing processes and production efficiency in industrial settings by opting for autonomous mobile robots dedicated to indoor transportation and logistics. The specific focus of this research is on employing deep learning for the detection of Euro pallet objects, enabling a mobile robot, specifically the Servant-T1500 model by Kar Metal company with natural navigation capabilities, to autonomously dock and handle palletized loads with precision. To accomplish this, an original dataset was curated using an RGB-D camera, and data augmentation techniques were applied to expand this dataset. Subsequently, a deep learning model was trained on this data to detect Euro pallets in images it had not encountered during training. The transfer learning method was applied using the YOLOv5 model on the dataset. The successful outcome of this process demonstrates the autonomous robot's capability to accurately recognize Euro pallet objects even under changing lighting conditions. This achievement marks a significant step toward optimizing industrial processes and improving production efficiency through the integration of autonomous mobile robots in logistics and transportation tasks.