United States Department of Transportation
ROSA P serves as an archival repository of USDOT-published products including scientific findings, journal articles, guidelines, recommendations, or other information authored or co-authored by USDOT or funded partners.
As a repository, ROSA P retains documents in their original published format to ensure public access to scientific information.
i
Development of Multi-Rotor-UAV-Based Rail Track Irregularity Monitoring and Measuring Platform with Image and Lidar Sensors
-
2024-09-01
-
[PDF-2.55 MB]
Details:
-
Creators:
-
Corporate Creators:University of Nevada, Las Vegas. Department of Civil and Environmental Engineering ; University of Nevada. University Transportation Center on Improving Rail Transportation Infrastructure Sustainability and Durability ; Wenzhou University ; Minnan Normal University ; Jiangsu University of Science and Technology
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Publication/ Report Number:
-
Resource Type:
-
Geographical Coverage:
-
Edition:Final Report UNLV-5
-
Corporate Publisher:
-
Abstract:This project aimed to develop a multi-rotor UAV-based track geometry measurement system using image and LiDAR sensors that does not require the closure of track during inspection. The research was divided into three stages: exploration of optimal path planning, development of a LiDAR-only track geometry measurement system, and development of a camera-LiDAR track geometry measurement system. In the first stage, the authors identified reinforcement learning as a solution for optimal UAV path planning and proposed a modified Floyd-Warshall (FW) algorithm. Experiments demonstrated that the proposed algorithm effectively handles the exploration-exploitation tradeoff. As a result, the graph-based algorithm finds the shortest path during exploration, leading to higher efficiency and faster convergence compared to the Q-learning algorithm and its variants. However, this improved efficiency and convergence come at the cost of increased computational complexity. In the second stage, the authors developed a UAV-LiDAR-based platform capable of performing track geometry measurements alongside normal rail operations. This system, built on a UAV equipped with a LiDAR sensor, utilizes machine learning for rail point segmentation, LiDAR SLAM for expanding the point cloud's field of view, and regression techniques for outlier removal and geometry calculations. Compared to traditional field measurements using specialized tools, the platform demonstrated high accuracy in gauge measurement, though it performed less effectively in curvature and profile measurements. In the final stage, the authors explored integrating camera data with LiDAR to enhance track geometry measurement. Due to time constraints, a semi-assisted supervised image segmentation approach was used, yielding high accuracy when the rails were vertically aligned in the images. Calibration results were highly accurate with checkerboard data but showed significant errors when applied to rail data. As a result, the current data fusion approach is not yet suitable for the track geometry measurement platform. Future research will aim to achieve more comprehensive image segmentation and improve the accuracy of LiDAR-camera calibration.
-
Format:
-
Funding:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
