Multi-Source Comprehensive Program to Assess, Monitor, and Report Retroreflectivity of Pavement Markings

Details

• Creators:
  Mohammadi, Abbas ; Medina, Juan C. ; Rashidi, Abbas
• Corporate Creators:
  University of Utah. Department of Civil and Environmental Engineering
• Corporate Contributors:
  Utah. Department of Transportation. Research and Innovation Division ; United States. Department of Transportation. Federal Highway Administration
• Subject/TRT Terms:
  Convolutional Neural Network Data Fusion Feasibility Analysis Image Processing Laser Radar LiDAR Intensity Machine Learning Monitoring Neural Networks Pavement Marking Retroreflectivity Pavements Random Forest Classifier Retroreflectivity Road Markings

  More +
• Publication/ Report Number:
  UT-26.16
• Resource Type:
  Tech Report
• Geographical Coverage:
  United States ; Utah
• Edition:
  Final Report: Jul 2024 to Dec 2025
• Corporate Publisher:
  Utah Department of Transportation
• Abstract:
  This research investigates the feasibility of estimating retroreflectivity of durable longitudinal pavement markings using multi-source data to reduce UDOT's reliance on costly specialized field inspections. The study uses a combination of existing datasets, mobile LiDAR point clouds, street-level roadway imagery, and historical retroreflectivity measurements to develop robust predictive models for assessing pavement marking condition. The research uses existing field retroreflectivity data across major Utah roadways, including I-15, I-80, I-84, and I-215. The study corridors are dominated by durable pavement markings, so the model development and validation presented in this report primarily reflect the performance of durable marking materials. LiDAR data provided geometry and intensity of light reflected back to the sensing unit, while roadway images were processed to extract visual features and assess the condition of lane markings. Both data streams (i.e., LiDAR and imagery) were analyzed independently and also in combination using machine learning methods, including Random Forest classifiers and convolutional neural networks (CNNs). A fusion model was developed to integrate LiDAR and image-based features for enhanced performance. Results demonstrated that image- and LiDAR-only models can accurately classify pavement markings as compliant or non-compliant with federal minimum retroreflectivity standards. The fusion model outperformed individual models, achieving over 95% accuracy and strong agreement with ground truth data at both segment and corridor levels. Regression models were also developed to predict continuous retroreflectivity values directly from image features. The methods developed in this research were implemented in a portable, user-friendly computer tool that enables UDOT to estimate and monitor pavement marking retroreflectivity via data fusion, LiDAR-only, or image-only models. This approach enhances operational efficiency, supports monitoring for federal compliance, and provides a scalable framework for statewide deployment. This study builds on a previous UDOT research phase that evaluated LiDAR‑only retroreflectivity estimation and extends that work by incorporating street‑level imagery and multi‑source data fusion.
• Format:
  PDF
• Funding:
  25-8077
• Collection(s):
  US Transportation Collection
• Main Document Checksum:
  urn:sha-512:be79cbb4178480faf401c7838c8f2605291f742c821a9c35ae106c6b519a5d09afe01eb6676a364090f9b3a3ec9edd9a5b7d7264c5cacfa8fd59813880434b81
• Download URL:
  https://rosap.ntl.bts.gov/view/dot/92743/dot_92743_DS1.pdf
• File Type:
  [PDF - 1.38 MB ]