Georgia Institute of Technology
LiDAR-based Pothole Detection and Volume Estimation Pipeline Creation and Trade Space Exploration
Abstract
dc:description.abstractPhysical infrastructure, especially well-maintained roads, is essential for both economic growth and quality of life, as it enhances transport efficiency and safety. Potholes, however, present a serious safety risk, making regular road maintenance critical. Traditional pothole detection methods rely on labor-intensive manual surveys, but UAVs equipped with LiDAR provide a more automated and efficient alternative. This thesis explores UAV-based LiDAR scanning for pothole detection and volume estimation by creating roadway Digital Elevation Models, using HELIOS++ for laser scanning simulation, and creating a process for pothole detection and volume estimation to develop an efficient and cost-effective pipeline for road maintenance and to investigate how flight and LiDAR parameters impact pothole detection rates and volume estimation accuracy. The study found that low-severity potholes (7.5 cm radius, -2.5 cm depth) were detected at least 88\% of the time when the UAV velocity was 1 m/s and 75\% of the time at 2 m/s across 1σ accuracy levels up to 2.4 cm using a rotating VLP16-based scanner at an altitude of 10 m. At both 1 m/s and 2 m/s, no false positives occurred when LiDAR accuracy was below 1 cm. At accuracies up to 1.4 cm, the false alarm rate remained low at 9\%. For moderate-severity potholes (20 cm radius, -4 cm depth), their volume was estimated at an average error of less than 10\% at UAV speeds through 6 m/s. A key finding was a relationship between speed, altitude, and scan frequency that leads to a reduction in the probability of detection and increase in volume estimation error due to subsequent head rotation and channels aligning, creating gaps in the point cloud. Results show that potholes with square cross-sections are detected more often than circular ones of the same dimensions, due to their larger cross-sectional area. Potholes with flat indentations are detected at higher rates and have volume estimated at higher accuracy than those with spherical or conic indentations due to their larger detectable area and relative indentation shape.
Degree
thesis:*- Level thesis:degree_level
- Masters
- Department dc:contributor.department
- Electrical and Computer Engineering
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Damon, Eva
- Advisor dc:contributor.advisor
-
- Valenta, Christopher R.
- Committee members dc:contributor.committeemember
-
- Anderson, David V.
- Christian, John A.
Subjects
dc:subject × 14Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/80109