Civil Engineering September 2021 | Vol 29 No 8

28 September 2021 Civil Engineering bottom. The images captured during the inspections were processed using Pix4D software to create point cloud models. It should be noted that there are free or open-source processing applications available, however Pix4D is the preferred software used by Premier Mapping Africa. A desktop computer with Core i7, 64 GB SSD and 128 GB RAM was required to process the images. The total processing time to create the point cloud models was approximately 4 hours. An accuracy of 2 cm to 4 cm was attained in the genera- tion of point cloud models and average GPS error within 1 m for georeferencing. Railway Bridge The first bridge inspected is a three span railway bridge with an overall length of 51 m, overall width of 8 m and maximum height of 7 m. A licenced pilot manually operated the drone to capture aerial images. Images underneath the bridge were cap- tured by walking on the ground using the drone as a manually operated gimbal. The total duration of the bridge inspection was approximately 45 minutes and the battery was swapped after 30 minutes. A combina- tion of 462 images were processed to create a point cloud model of the inspected bridge. River Bridge The second bridge inspected is a three span river bridge with an overall length of 27 m, overall width of 11 m and maximum height of 5 m. The same approach was followed as with the first bridge to capture the images, but vegetation in the river under the bridge prevented the drone pilot from capturing images underneath the bridge. The total duration of the bridge inspection was approximately 30 minutes and one battery was used. A combination of 359 images were processed to create a point cloud model of the inspected bridge. The flight paths and the position of images captured by the drone for the two bridges are shown in Figure 2 and Figure 3. Results and discussion The same defects identified during the 2016 visual bridge inspections were investigated using the point cloud models shown in Figure 4 and Figure 5. Pix4D (the software used to create the point cloud models) enables one to click on any element of the bridge and displays all the images from which the selected point is visible. This enables the inspector to investigate the defects from different angles and provides clear images of loca- tions not easily reachable by an inspector. Figure 2 Flight path and the position of images captured by the drone (Railway Bridge) Figure 3 Flight paths and the position of images captured by the drone (River Bridge) Figure 4 Point cloud model of the Rail Bridge Figure 5 Point cloud model of the River Bridge