3D Mapping of the UD Campus via Aerial LiDAR and Photogrammetry

• Logan Blackburn, Computer Engineering, University of Delaware
• Alyssa Mohammed, Electrical Engineering, University of Delaware

• Kenneth Barner, Electrical and Computer Engineering, University of Delaware
• Mohammad Baksh, Electrical and Computer Engineering, University of Delaware

Using the DJI Matrice 350 RTK drone, this study explores the capabilities of photogrammetry and LiDAR to create a comprehensive 3D model of the University of Delaware’s campus. Photogrammetry, which employs photographs from various angles, triangulates common points to construct detailed 3D models. Conversely, LiDAR emits near-infrared laser pulses to measure reflections from terrain and buildings, capturing areas beneath vegetation as the pulses penetrate through foliage to the ground below. In our initial experimentation, we tried individually to model Mitchell Hall, a relatively small building with dense vegetation nearby, using photogrammetry and LiDAR. However, we found that neither method could create a high-quality model. Photogrammetry struggled to capture areas obstructed by vegetation, and LiDAR created noisy and imprecise models due to instability when mounted on the drone. Therefore, we conducted a second set of experiments aiming to combine the two methods and improve our models. We generated three models using various levels of combination: A control made solely with photogrammetry, a model with photogrammetry using LiDAR only over dense vegetation, and a model with photogrammetry and LiDAR scans of the entire area. In areas where photogrammetry captured a clear landscape view, such as buildings and roads, the addition of LiDAR showed no significant improvement. Furthermore, LiDAR increased the face density of the model up to 10 times that of the photogrammetry model, significantly increasing processing time. On the other hand, in areas of dense vegetation, LiDAR excelled in creating more accurate models, with the combination of LiDAR and photogrammetry creating an accurate and precise model. With these results, it is clear that to generate a comprehensive 3D map of campus using aerial imaging, applying photogrammetry in combination with LiDAR strictly around vegetation creates both accurate and precise models while maintaining processing efficiency.