<![CDATA[This study documents and evaluates an end-to-end workflow for producing 30 textured three-dimensional (3D) assets from small-scale objects (maximum dimension 50 cm) using the active near-infrared scanner Creality CR-Scan Ferret inside a mini booth with diffuse lighting. Raw data were processed in Creality Scan and Blender through artefact elimination, remesh/retopology, mesh repairing, texture baking, and export. Digital dimensions were extracted from the model bounding box using Python and Open3D, then compared with reference physical measurements. The mean per-asset RMSE was 2.00 cm (SD = 1.61; median = 1.80; 95% CI = 1.40–2.60; range = 0.02–6.18 cm). Under the internal thresholds of the study, 12 assets were rated Very Good, 11 Good, 5 Fair, and 2 Poor. The mean RMSE for easy, medium, and hard objects was 1.32, 1.31, and 3.36 cm, respectively; an exploratory Kruskal-Wallis test indicated a difference between categories (H = 9.76; p = 0.0076). Mesh analysis was applied to six representative samples only. Two regular bottle models had Non-manifold Edges = 0 and Shells = 1, whereas the Hashmal model had 14,945 non-manifold edges and 7 shells. The workflow is usable as an initial pipeline for digital visualisation, but the results do not prove universal metric accuracy or watertightness because repeated scans, measurement uncertainty, independent scale calibration, and mesh-metric normalisation are not yet available.]]>
Copyrights © 2026
