Artificial Intelligence and Mobile Phone-Based Pavement Marking Condition Assessment and Litter Identification

Regular inspection of transportation assets is essential to ensure pavement markings and pavements remain in good, clean, and safe condition. Our previous MPC-funded project (Report No. MPC-668) demonstrated the strong potential of using artificial intelligence (AI) and mobile phone imagery to identify various transportation assets. However, that initial effort was limited in scale, using only ~1,000 images for training and validation. Building upon this foundation, the present project focuses on two targeted assets: pavement marking issues and roadside litter, while expanding the capability of the previously developed AI packages. In this study, the dataset increases to over 6,000 images for each asset type. Using the You Only Look Once (YOLO) deep learning architecture, two detection models were trained and achieved strong accuracy metrics, with F1 scores of 0.88 for pavement marking issues and 0.84 for roadside litter. In addition, counting and geolocation models are developed to quantify detected objects within a road section or video clip and to determine their precise locations by integrating data from a phone-based global positioning system (GPS) tracker. The geolocation model demonstrates high spatial accuracy, achieving an average positional error of only 0.27 meters. To facilitate practical application, an interactive mapping interface is implemented to visualize the geolocation, object class, inspection time, and cropped image of each identified object. This interface enables clear and intuitive assessments of pavement conditions, specifically faded markings and roadside litter. Overall, this project enhances our prior work by extending capabilities in detection, counting, geolocation, and visualization, which supports regular asset inspection, informs maintenance planning, and ultimately improves roadway safety.

• Record URL:
  • https://www.ugpti.org/resources/reports/details.php?id=1269
• Record URL:
  • https://rosap.ntl.bts.gov/view/dot/87790
• Summary URL:
  • https://rosap.ntl.bts.gov/view/dot/87797
• Supplemental Notes:
  • This document was sponsored by the U.S. Department of Transportation, University Transportation Centers Program. The contents of this publication reflect the views of the authors and do not necessarily reflect the official views or policies of the Transportation Research Board or the National Academy of Sciences.
• Corporate Authors:

  University of Utah, Salt Lake City

  Department of Civil and Environmental Engineering
  Salt Lake City, UT  United States  84112

  Center for Transformative Infrastructure Preservation and Sustainability

  North Dakota State University
  Fargo, North Dakota  United States  58108-6050

  Office of the Assistant Secretary for Research and Technology

  University Transportation Centers Program
  Department of Transportation
  Washington, DC  United States  20590
• Authors:
  • Kuang, Biao
  • 0000-0001-5787-9480
  • Chen, Jianli
  • 0000-0001-7106-7833
• Publication Date: 2025-11

Language

• English

Media Info

• Media Type: Digital/other
• Edition: Final Report
• Features: Figures; Maps; Photos; References; Tables;
• Pagination: 44p

Subject/Index Terms

• TRT Terms: Asset management; Deep learning; Global Positioning System; Image analysis; Litter; Object detection; Road markings
• Subject Areas: Data and Information Technology; Highways; Maintenance and Preservation; Pavements;

Filing Info

• Accession Number: 01973897
• Record Type: Publication
• Report/Paper Numbers: CTIPS-25-002, CTIPS-007
• Contract Numbers: 69A3552348308
• Files: UTC, NTL, TRIS, USDOT
• Created Date: Dec 11 2025 9:44AM