| Title |
Cylinder Pose Estimation Methods |
| DOI |
https://doi.org/10.5573/ieie.2025.62.7.85 |
| Keywords |
Automated guided vehicle; Vision-based alignment; Cylindrical object pose estimation |
| Abstract |
The role of automated guided vehicles (AGVs) in the manufacturing and logistics industries has expanded significantly in recent years, making accurate alignment and precise position estimation essential. Traditional approaches based on line following, QR codes, and AR markers have limitations such as difficulty in unstructured environments and poor performance when markers are damaged. In this paper, we propose a vision-based alignment system that does not use additional markers by utilizing cylinder-shaped structures carried by AGVs. The proposed method integrates lateral-edge-based cost and segmentation-based cost to realize an alignment system that is robust to environmental changes and occlusions. The lateral-edge-based cost is computed from lateral boundary information extracted within a region of interest (ROI), while the segmentation-based cost is derived from mask data segmented from the same ROI. By minimizing an optimization objective function that integrates these two cost functions, the three-dimensional position and pose (including pan/tilt angle) of the cylinder can be estimated in real-time. This work is different from existing approaches in that it does not require the attachment of markers and provides alignment performance that is robust to noise and illumination variations. Furthermore, it uses a lightweight optimization algorithm that can be applied in real-time, and experiments under various pan/tilt conditions and AGV positions verify the efficiency and reliability of the proposed method. This work presents a precise and time-efficient solution for alignment tasks in the dynamic environment of AGVs. |