Influence of Terrestrial LiDAR Scanning Conditions on Registration Results for Physical AI Development

Purpose: This study investigates the effects of indoor scanning geometry on target-based registration behavior through variations in target centroid estimation in terrestrial laser scanning (TLS) for physical AI development. The manuscript argues that, in feature-poor interior environments, registration reliability depends more strongly on scan distance and viewing geometry—which govern the consistency of target centroid estimation—than on nominal ranging precision. Method: A controlled lecture-room experiment was conducted using a Leica RTC 360 and four targets mounted on a planar interior wall. Scans were repeatedly collected under a structured matrix comprising two distances (5m and 10m), four scanner heights, and multiple lateral scan positions, with three repetitions per scan configuration. Target-based registration was performed for each repetition, and registration outcomes were analyzed by calculating Euclidean displacements between corresponding target centroids, without using external control points. Result: Scanning distance and lateral scan position were identified as the primary factors influencing target centroid displacement behavior, whereas scanner height showed no systematic effect. Increased scanning distance produced larger displacement magnitudes and greater dispersion, while moderate off-axis positioning was associated with reduced displacement, which is due to the reduced scanner-to-target distance rather than different geometric configurations, leading to increased target centroid consistency. These findings provide practical guidance for scan planning in indoor architectural TLS.