Influence of ground control point reliability and distribution on UAV photogrammetric 3D mapping accuracy

ABSTRACT

This study investigates the impacts of Ground Control Points (GCP) reliability, quantity, and distribution on the accuracy of Unmanned Aerial Vehicle (UAV) photogrammetric 3D mapping. Previous studies have predominantly focused on overall accuracy metrics, often overlooking local variations in error distribution and failing to conduct thorough analysis of error sources, potentially resulting in inefficiencies and reduced photogrammetric 3D mapping accuracy due to sub-optimal GCP deployment and processing strategies. By employing a structured framework, this research analyzes GCP roles based on variations in their reliability. Aerial images of undulated terrains were analyzed and fifteen distinct GCP configurations were designed which incorporated various distributions and quantities. Random noise was added to simulate GCP reliability unpredictability, subsequently analyzed through a multifaceted approach. The results indicate that GCP reliability influences the upper limit of photogrammetric 3D mapping accuracy and significantly affects the influence of GCP quantity and distribution. It was also found that: when GCP reliability is high (RMSE within 0.1 m; density more than 10 GCP/k⁢m2), GCP number should prioritize over distribution, ie. accuracy improving as the number of GCPs increases; when GCP reliability is moderated (RMSE within 5 m; density around 5–10 GCP/k⁢m2), each of the uniform and distributed-edge distributions leads in different conditions; when GCP reliability is low (RMSE beyond 5 m; density less than 5 GCP/k⁢m2) it is favorable to use a distributed-edge and center distribution. Errors typically originate in edge areas and regions with significant topographic variations and propagating as GCP reliability decreases. Therefore, additional GCPs should be placed in error-prone regions to improve the accuracy.