Vision-centric 3D point cloud technique and custom gripper process for parcel depalletisation

Abstract

Vision-based in-truck parcel recognition plays a key role in providing picking guidance for automated robotic in-truck parcel-unloading systems. The complexity of the parcel system and the variety of colours and shapes of the target objects significantly affect the quality of the results. To establish an effective in-truck parcel depalletisation system, it is crucial to develop a method that can automatically recognise parcels in a 3D environment and guide robots during unloading tasks. To address these requirements, this study proposes a system for detecting geometric point clouds in parcels that uses regression knn to find the nearest pick-up point of a detected parcel box by calculating the minimum Euclidean distance, thereby improving detection accuracy. The validation of the robotic system underlines its practical utility, demonstrating its potential to replace humans and reduce labour costs in factory environments. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.