Road Marking Distress Detection and Assessment Based on UAV Imagery

Area of Science:

• Civil Engineering
• Computer Vision
• Transportation Engineering

Background:

• Autonomous driving relies heavily on accurate lane marking perception.
• Road markings degrade over time due to vehicle loads, impacting safety.
• Existing inspection methods are inefficient and lack precision.

Purpose of the Study:

• To develop an integrated framework for road marking detection and evaluation using UAV imagery.
• To improve the efficiency and accuracy of road marking quality inspection.
• To provide a technical reference for intelligent road maintenance.

Main Methods:

• Lightweight data acquisition using optimized UAV flight parameters.
• Efficient marking extraction via YOLOv8-MEB model with instance segmentation and local image optimization.
• Accurate distress assessment using RANSAC-based template matching and contour correction.

Main Results:

• Lane segmentation precision and recall exceed 90% with FPS > 60.
• Restoration of intact marking shapes using template matching and affine transformation.
• Approximately 10% error in distress ratio calculation for non-severe damage.

Conclusions:

• The proposed UAV-based framework offers a low-cost, flexible, and safe solution for road marking assessment.
• The integrated approach significantly improves detection and evaluation accuracy compared to traditional methods.
• This framework provides a practical technical reference for intelligent road maintenance and enhances driving safety.