Abstract:
Roads and highways are the backbone of the US transportation system, allowing Americans to travel more than two trillion miles annually. But the conditions of the roads are deteriorating, as the need for transportation improvements far outpaces the amount of state and federal backlogs of road maintenance and repairs. In addition, it is becoming more expensive to maintain high-quality driving conditions because many of the roadways in the US were constructed in the 1950s and 1960s. Conducting routine maintenance and repairs of roads, such as fixing potholes and repaving roads are one of the most important task and priority for transportation authorities. In order to do so, road condition data is vital in transportation management. Over the last decades, significant advancements have been made and new methods and knowledge have been shared for efficient collection of road pavement data. This thesis investigates the utilization of small unmanned aerial vehicles (UAVs) like a Mavic 2 Pro for surveying and inspecting road surface conditions. The objective of this research is to identify potholes from UAV captured images and create GIS datasets containing key information such as dimensions, severity level, and location of potholes. This paper will introduce five main methods for data collection, UAV image processing, dataset generation, 3D reconstruction, and feature detection. To test the feasibility of these methods, pilot studies are conducted in two different test sites located in Fairfax County, Virginia. Different flight parameters and environmental settings are examined to minimize technical errors and to build an optimal workflow that will provide the most accurate results. By using photogrammetry, GIS, 3D modeling, and image processing software, this workflow can be utilized by local transportation authorities to conduct quick surveys and evaluation of the roads while requiring minimal training and funding.
