Automatic Extraction of Structural and Non-Structural Road Edges from Mobile Laser Scanning Data

Area of Science:

• Geomatics Engineering
• Computer Vision
• Transportation Engineering

Background:

• Accurate road information is crucial for intelligent transportation systems (ITS), road maintenance, and network updates.
• Mobile laser scanning (MLS) is effective for road data extraction, but challenges remain in handling complex road conditions and large datasets.
• Existing methods struggle with diverse road types (structural and non-structural) and require additional data like intensity or images.

Purpose of the Study:

• To develop a robust, automated method for extracting structural and non-structural road edges from large-scale MLS data.
• To overcome limitations of traditional methods by focusing on surface roughness rather than height jumps or density.
• To validate the method's performance across diverse road conditions and datasets.

Main Methods:

• Proposed a novel method utilizing a topological network of laser points between adjacent scan lines and auxiliary surfaces.
• Focused on surface roughness for road and curb point extraction, avoiding reliance on traditional thresholds (height jump, slope, density).
• Evaluated the method on five large-scale road datasets with varied curb types and complex road scenes.

Main Results:

• Achieved high accuracy in road edge extraction, with correctness, completeness, and quality metrics exceeding 95.5%, 91.7%, and 90.9%, respectively.
• Demonstrated the method's independence from auxiliary data such as intensity, images, or geographic information.
• Confirmed the method's effectiveness across various road widths, regularities, and in the presence of pedestrians and vehicles.

Conclusions:

• The proposed method offers a robust and automated solution for road edge extraction from large-scale MLS data.
• It provides a valuable tool for road authorities developing ITS, particularly for applications like self-driving vehicles.
• The technique is practical, stable, and valid for diverse and complex road environments without external data dependencies.