Georgia DOT Study Provides Lessons on Environmental Conditions for UAVs Based on Imagery Data from Two Bridges in Georgia.
Douglasville, Georgia
Calhoun, Georgia
UAS-Assisted Inspection of Bridges for Corrosion Effects
Summary Information
Unmanned Aerial Vehicles (UAVs) are becoming more widely used to conduct imagery-based bridge inspections and damage evaluation. However, corrosion detection algorithms require certain environmental conditions to function properly. This study explored the use of UAVs as a proof of concept to automatically detect and characterize corrosion on bridges in Georgia. The study used imagery data from two bridges in Douglasville and Calhoun, GA, collected in December 2022 and February 2023 via the use of a drone, that was then fed into varying machine learning algorithms for corrosion detection. The researchers first implemented texture and color thresholding methods, followed by an unsupervised machine learning clustering algorithm that utilized texture and color features. Finally, deep learning methods were applied with automated feature extraction.
- Optimize UAV-aided bridge Inspections by timing them in Spring or Summer to avoid thin branches that could reduce the UAV's vision-based obstacle avoidance capabilities. More leaves and denser vegetation can reduce the presence of thin branches. This best practice is especially true for bridges in rural areas, but may not be as necessary for bridges in urban areas that do not have many trees nearby.
- Fly the drones on days with weather conditions desirable for UAV-aided bridge inspections. To minimize any problems with reflectance in the images, enhance the uniformity of the lighting, and improve the visibility of corrosion in the images, UAV-aided bridge inspections should be conducted on days that are overcast rather than partially or fully sunny.
- Perform pre-flight planning before the inspection. Pre-flight planning was recommended to assess any vegetation surrounding the bridge that may limit where the drone can fly that would require alternative ways to collect data.
- Utilize the zoom feature of the UAV frequently. This is especially recommended by this study when taking photographs of bridge components that may be more prone to corrosion, such as the bottom of piles. Using the zoom feature would improve the automatic corrosion detection from the UAV-aided bridge inspection.
- Continue to improve automatic corrosion detection algorithms. Despite testing several methodologies, the study showed that no algorithm as of yet works perfectly. Therefore, while this study showed the potential of these methodologies to assist in bridge inspections, it also highlights the necessity for further development.
