Abstract
As 3D scanning technology becomes integral to reverse engineering and quality control, the ability to accurately identify geometric primitives—such as planes, cylinders, and spheres—from point clouds is critical. This study investigates the influence of environmental and process parameters on the dimensional accuracy of 3D scanned mechanical parts. Using the Taguchi L9 orthogonal array method, nine experimental configurations were tested to evaluate the impact of three key factors: lighting conditions (Natural, Neon, LED), surface preparation (original, white matte, and white matte with markers), and scanning brightness parameters (10%, 40%, 80%). A mechanical test plate containing varied geometric features was scanned using a handheld laser scanner (Artec Leo), and the resulting models were processed to extract geometric primitives for comparison against nominal CAD dimensions. The statistical analysis, including Signal-to-Noise (S/N) ratios and Analysis of Means, revealed that surface preparation is the most statistically significant factor influencing both the geometric accuracy and the efficiency of the data file size. The results demonstrate that treating the object surface with a matte coating and reference markers drastically reduces noise and improves the precision of primitive identification, significantly outweighing the effects of ambient lighting or software scanning parameters. These findings provide a practical framework for optimizing 3D scanning workflows in industrial metrology.