Point Cloud Segmentation Dataset of a 1:10 Scale CFST Arch Rib Model

This dataset accompanies the paper "Automated Point Cloud Segmentation of Long-Span Arch Bridges Using Growth Inertia Derived from the Scanning Sequence" and provides the point cloud data and ground truth annotations of a 1:10 scale concrete-filled steel tubular (CFST) arch rib model. The model comprises 4 chord members, 16 web members, and 10 cross-braces (30 individual members in total). The point cloud was acquired from a single station using a Leica P50 terrestrial laser scanner with the scanning range set to 120 m and a resolution of 1.6 mm @ 10 m. The release includes: raw_point_cloud.ptx: Raw point cloud retaining the original row-column raster topology. preprocessed_point_cloud.ptx: Point cloud after background removal and statistical outlier removal, containing 5,763,993 points. ground_truth_chord_members.txt: Ground truth point cloud of all chord members. ground_truth_web_members.txt: Ground truth point cloud of all web members. ground_truth_cross_braces.txt: Ground truth point cloud of all cross-braces. ground_truth_labeled.txt: Merged ground truth with per-point semantic labels (0 = chord members, 1 = web members, 2 = cross-braces). For full details on data acquisition, preprocessing, and format specifications, please refer to the accompanying paper and the README file included in the archive.