Roadside Asset Extraction from Mobile LiDAR Point Cloud

Details

• Creators:
  Ahn, Yushin ; Munjy, Riadh ; Choi, Stephen
• Corporate Creators:
  San Jose State University. College of Business. Mineta Transportation Institute
• Corporate Contributors:
  State of California SB1 2017/2018, Trustees of the California State University Sponsored Programs Administration ; United States. Department of Transportation. University Transportation Centers (UTC) Program ; United States. Department of Transportation. Office of the Assistant Secretary for Research and Technology
• Subject/TRT Terms:
  Artificial Intelligence Data Collection Intelligent Transportation Systems Laser Radar Machine Learning Deep Learning Spatial Analysis
• Publication/ Report Number:
  25-15
• Resource Type:
  Tech Report
• Geographical Coverage:
  United States
• Corporate Publisher:
  Mineta Transportation Institute
• Abstract:
  Mobile LiDAR systems are powerful tools that help us map roads and their surroundings in 3D with great speed and precision. The data provided by these systems support urban planning efforts, digital mapping, transportation infrastructure maintenance, and more. This report presents a comprehensive workflow for roadside asset extraction using Mobile Terrestrial Laser Scanning (MTLS) data, focusing on road lane detection, cross-section slope analysis, and point cloud classification. Roadside asset extraction is the identification and classification of roadside features like signs and poles. The dataset, acquired using a high-resolution mobile LiDAR system, contains over 5.7 billion points (pieces of data) across 68 LAS files. Preprocessing of the data involved tiling, merging, and denoising processes to enhance data quality. This research used two types of image-like maps— one showing elevation and one showing how reflective the surface is—to model the road and identify lane markings. Furthermore, the team developed and trained an AI-driven deep learning-based classifier program using a PointNet-style architecture implemented in PyTorch to segment ground and non-ground features. The classifier was trained using over 500 million labeled points and applied to new LAS files for inference. The model achieved effective binary classification performance and produced classified LAS outputs compatible with downstream GIS workflows. This means the program was able to successfully sort features into two categories—like ground vs. non-ground—and produce files that work with common mapping software. This work demonstrates the feasibility of combining traditional feature extraction with modern deep learning approaches to enhance automation and accuracy in infrastructure mapping.
• Format:
  PDF
• Funding:
  SB1-SJAUX_2023-26
• Collection(s):
  University Transportation Centers
• Main Document Checksum:
  urn:sha-512:831dda775c81f0b8c4342586111a93c28518cd743870ef77f21c528de888a9fa05e521ee0782313914e3269af99e4c22cc6cc2670de11042f896981f8ec12760
• Download URL:
  https://rosap.ntl.bts.gov/view/dot/86099/dot_86099_DS1.pdf
• File Type:
  [PDF - 19.81 MB ]