From Medical Imaging to Robotic Perception: Real-Time Depalletising with Instance Segmentation

DOI

10.2991/978-94-6239-823-8_11How to use a DOI?

Keywords

Instance segmentation; robotic depalletising; YOLOv8-seg; RGB-D sensing; real-time 3D localisation

Abstract

This paper presents a real-time robotic depalletising system that combines YOLOv8-seg-based instance segmentation with RGB-D depth sensing. Motivated by the successful transfer of segmentation paradigms from medical imaging to industrial robotics, the proposed system performs pixel-accurate object detection and three-dimensional localisation of cardboard boxes on the top layers of pallets without requiring a predefined pallet model. The software architecture, implemented in C++17 with OpenCV, ONNX Runtime, and the Orbbec SDK, consists of four modular components: image acquisition, instance segmentation, 3D coordinate calculation, and TCP/IP-based robot communication. Experimental evaluation in a real industrial environment demonstrated a detection rate of 98%, a localisation accuracy of ±3 mm, and a processing throughput of 10 FPS. The results confirm that instance segmentation significantly outperforms traditional object detection methods in providing geometric information for robotic grasping, and that architectures originally developed for medical image analysis can be effectively deployed in logistics automation.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - József Lénárt
AU  - László Rónai
AU  - György Hegedűs
PY  - 2026
DA  - 2026/11/04
TI  - From Medical Imaging to Robotic Perception: Real-Time Depalletising with Instance Segmentation
BT  - Proceedings of the Innovation in Engineering Education 2026 Conference (IEE 2026)
PB  - Atlantis Press
SP  - 126
EP  - 133
SN  - 2667-128X
UR  - https://doi.org/10.2991/978-94-6239-823-8_11
DO  - 10.2991/978-94-6239-823-8_11
ID  - Lénárt2026
ER  -