PSI - Issue 62

Antonio Di Pietro et al. / Procedia Structural Integrity 62 (2024) 755–762 Antonio Di Pietro et al. / Structural Integrity Procedia 00 (2019) 000 – 000

761

7

Figure 4. Flight plan at three different altitudes. Curved lines around the piles represent UAS routes. Circles represent the positioning of GCPs.

Automatic flights prove advantageous, especially in situations where they offer greater safety or accuracy. Some drones use advanced sensors and artificial intelligence to avoid obstacles. Although Google maps are commonly used for planning, they have limited precision. More accurate alternatives exist, such as mission planners integrating LiDAR data. These data are essential for safely guiding the drone, avoiding obstacles such as power lines or dense vegetation. 6. Case study and data elaboration The proposed methodology was applied to the Beroide Bridge. Ten targets for the identification and measurement of GCP, CP, one portable GNSS/RTK measuring, one digital high-definition camera (24 MP) with zoom capability were employed. Several nadiral and lateral data were collected using the automatic nadiral flight plan option with a DJI M300 RTK drone. In particular, a set of lateral overflies was performed, with translation movement of the UAS and the camera set at a 0° elevation angle, focus fixed at 1x, distances from the bridge of 12 and 15 m, allowing an 80  80% lateral and vertical overlap. To capture the detail of the lateral bridge structure, a flight was performed at a height of about 15 m with a depression angle of 45°. The piers were surveyed using a DJI Mavic 3 drone in manual elliptical orbit mode as shown in Figure 3. In order to ensure a robust dataset, the pier structure was also photographed by a standing operator with a digital camera. In the process of the 3D reconstruction of the bridge, the efficiency and precision of the Structure-from-Motion (SfM) process largely hinge on the meticulous organization of the acquired photos. The categorization of the photo dataset was divided in three key groups: Nadiral, capturing top-down imagery; Lateral, focusing on side views; and 4K Movie, providing dynamic visual data. The objective was to demonstrate the SFM software's capability by integrating 4K movie frames into the dataset from other sensors and flights with precise alignment. They also aimed to showcase the DJI Avata's potential for reconstructing specific structural components, opting for it over the larger M300RTK drone in certain scenarios.This organization of the photo database allows for easy handling of the reconstruction process while allowing for possible increase of data collected with additional photos. The creation of a 3D model includes the integration of exact reference coordinates and a scale bar, followed by rigorous quality control measures. These steps ensure that the overall point cloud, 3D model, and orthomosaic meet stringent accuracy requirements, with an allowable margin of error limited to just 2 centimeters. A detailed breakdown of this process performed with Agisoft Metashape (Figure 5) is as follows: a) Project Setup with Georeferencing and Scale Bar to incorporate the precise coordinate system based on the exact reference coordinates of the bridge (e.g., UTM zone 32N for the case study). b) Image Alignment and Data Integration: The scale bar is captured in images and the acquired set of GCP are employed together with the exact reference coordinates for georeferencing and accuracy improvement. c) Point Cloud, Mesh, and Texture Mapping Creation: A dense point cloud and a 3D mesh are created considering the scale information from the scale bar to ensure that the model precisely represents real-world dimensions. d) Data Fusion, Verification, and Error Correction: The combination of the scale information from the scale bar and data from other sources, such as GCPs, allows to enhance precision. The export of the orthomosaic with the