Raw RGB-D Point Cloud Scans for Weld Bead Recognition and Robotic Grinding Path Planning
Description
# Raw 3D Point Cloud Dataset of Weld Beads This dataset contains raw 3D point cloud scans (`.ply` format) of various weld bead profiles on cylindrical aluminum tubes, captured using a REVOPOINT 3D Acusense RGB-D camera. For methodology, preprocessing, and model implementations, please refer to the published paper: > Chunhui Chung, Chia-Yuan Wu, "Weld bead recognition and robotic grinding path planning via deep learning point cloud segmentation," *Measurement*, Vol. 284, 122250, 2026. > DOI: https://doi.org/10.1016/j.measurement.2026.122250 --- ### File Naming Convention Files are named according to the workpiece sample and scan index: `<Sample>_<Index>.ply` (e.g., `A_1.ply`, `A_2.ply`, `B_1.ply`). * **Samples A – E:** In-distribution weld geometries (used for model training in the paper). * **Samples F – J:** Unseen weld geometries (used for generalization testing in the paper). --- ### Data Format & Notes * **Format:** Unordered point cloud in Stanford Triangle Format (`.ply`). * **Coordinates:** Point coordinates `(x, y, z)` are defined in the camera coordinate frame. * **Condition:** Unprocessed raw sensor data containing optical noise, minor surface reflections, and occlusion gaps from single-view scanning. --- ### License This dataset is distributed under the **Creative Commons Attribution 4.0 International (CC BY 4.0)** license.
Files
Institutions
- National Cheng Kung UniversityTaiwan, Tainan City