UAV photogrammetry and lidar integration for high-fidelity 3D campus mapping at KFUPM

Accurate, photorealistic, and operationally actionable 3D campus mapping is a key enabler for smart educational environments, which refers to digitally enabled campuses that integrate spatial data to support navigation, facility management, and infrastructure monitoring. This paper presents an end-to-end, replicable UAV workflow that combines RGB photogrammetry from multi-view imagery and UAV LiDAR point clouds to construct a high-fidelity 3D campus model of King Fahd University of Petroleum and Minerals (KFUPM) as a case study. Data were collected using a DJI Matrice 300 RTK equipped with a Zenmuse P1 (45 MP) camera and Zenmuse L2 LiDAR payload, with nadir grid flights (80%/70% overlap) and oblique orbits (~ 45°) at 60 m altitude (RGB GSD ≈ 2.5 cm/pixel; LiDAR mean spacing ≈ 5 cm). LiDAR scans were georeferenced and cleaned, then co-registered with the photogrammetric reconstruction in a common RTK frame. To improve visual realism without altering metric geometry, a lightweight 2× U-Net super-resolution module (U-NetSR) was applied only to the RGB textures used for mesh texturing. Experiments show that combining nadir and oblique views improves facade completeness and reduces surface deviation by ~ 30% relative to nadir-only acquisition, while super-resolved textures increase SSIM (0.88→0.93) and edge sharpness (~ 28%) at a modest post-processing cost. Finally, the model is exported to a WebGIS environment for interactive 3D exploration and campus-operations integration.