Document Type : Original Article
Authors
1
M.Sc. student, Department of Geomatics, University of Tabriz, Tabriz, Iran
2
Department of Geomatics, University of Tabriz, Tabriz, Iran.
10.22059/eoge.2025.392733.1175
Abstract
The objective of this research is to examine and model the influence of ground control point (GCP) configurations, quantity, spacing, and spatial distribution on the accuracy of 3D reconstruction in UAV-based photogrammetry. Four GCP patterns were evaluated: Mode A (minimal corner placement), Mode B (perimeter distribution), Mode C (combined perimeter and central), and Mode D (central-only), across three scenarios with increasing GCP spacing in urban and non-urban areas. The total GCPs ranged from 4 to 42, with distances tested at 100m (1D), 200m (2D), and 300m (3D), corresponding to multiples of 30, 60, and 90 times the GSD. Local accuracy was assessed using 30 random checkpoints, while global accuracy was analyzed via the M3C2 algorithm. Scenario 1 (1D spacing) revealed Mode B achieved the highest local accuracy, with RMSE values of 0.10 m (urban) and 1.06 m (non-urban). Scenario 2 (2D spacing) showed slight accuracy reductions but maintained comparable performance. In Scenario 3 (3D spacing), Mode C outperformed others, yielding an RMSE of 0.17 m (urban) and 0.80 m (non-urban), with errors concentrated at block edges. Global M3C2 analysis confirmed Mode C’s superiority in Scenario 3, demonstrating that central GCP placement becomes critical when spacing exceeds 90×GSD. Results indicate that perimeter-based configurations (Mode B) suffice for smaller intervals (≤30×GSD), but larger spacings (>90×GSD) necessitate combined perimeter and central GCPs (Mode C) to mitigate accuracy degradation. Mode C is recommended for large-scale projects with sparse GCP networks, balancing efficiency and reliability.
Keywords
Main Subjects
)