Optimal photo overlap for road accident documentation and reconstruction applying UAV imagery

This study investigates the optimal photo overlap for documenting and reconstructing road accident sites using drone imagery. While a general recommendation for drone imagery overlap stands at 60–80%, this research aims to determine the minimum acceptable overlap required to generate a precise 3D point cloud suitable for forensic road accident simulation. A DJI Mavic Air 2 drone captured images at 2-meter intervals over a junction and a connecting road segment from varying altitudes, following the same flight path. The experiment systematically excluded images from the original dataset, processing photo sets taken at 2, 4, 6, 8, and 10-meter intervals. The corresponding point clouds were evaluated for accuracy and fragmentation. Comparisons were made regarding the number of images, the size of image sets and processing times. Additionally, 3D mesh surfaces were generated in the Virtual Crash software, and their quality was assessed. Results revealed that a 50% overlap was adequate for generating satisfactory 3D simulation environments, thereby reducing size of the raw data, the point cloud and processing time considerably. This finding is significant for forensic experts seeking efficient methods of road accident scene reconstruction, emphasizing the practicality of lower photo overlap in such scenarios.