An End-to-End Deep Learning System for Pavement Distress Detection, Severity Estimation, and Condition Reporting [Research Brief]

Details

• Creators:
  Aboah, Armstrong ; Tolliver, Denver ; Kyem, Blessing Agyei ; Denteh, Eugene ; Asamoah, Joshua Kofi ; Danyo, Andrews ; Rattai, Dylan
• Corporate Creators:
  North Dakota State University ; Center for Transformative Infrastructure Preservation and Sustainability (CTIPS) Region 8 University Transportation Center (UTC) ; Upper Great Plains Transportation Institute
• Corporate Publisher:
  Center for Transformative Infrastructure Preservation and Sustainability (CTIPS) Region 8 University Transportation Center (UTC)
• Corporate Contributors:
  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:
  • Pavement condition
  • Deep learning
  • Risk assessment
  • Estimation theory
  • Numerical analysis
  • Decision making
  • Datasets
  • Potholes
  • Artificial intelligence
  • Pavement Condition Index
  More +
• Publication/ Report Number:
  CTIPS-26-010
• Resource Type:
  Brief
• Geographical Coverage:
  United States
• Edition:
  Research Brief
• Description:
  The researchers developed and evaluated PaveCap, an artificial intelligence system that examines roadway photographs to assess pavement conditions and describe visible damage. Researchers used 7,158 pavement images collected from Jefferson City, Missouri; Peoria, Illinois; and Washington, Illinois. Pavement experts provided condition ratings and detailed descriptions of the damage visible in each image. The dataset included several common types of roadway damage such as cracks, potholes, and pavement patches. The system was developed with two main capabilities. First, it identifies damaged pavement areas and estimates an overall condition score on a scale of 0 to 100. Second, it automatically generates written descriptions identifying the types and severity of damage present. Researchers trained the system using approximately 80% of the images and evaluated its performance using the remaining 20%. They compared the system's condition estimates with expert-assigned ratings and assessed how closely its written descriptions matched those prepared by pavement experts. Additional experiments examined how different image-processing approaches influenced performance. The results helped identify the system's strengths, limitations, and opportunities for improvement before potential use by transportation agencies.
• Format:
  PDF
• Funding:
  CTIPS-051
• Download URL:
  https://rosap.ntl.bts.gov/view/dot/94247/dot_94247_DS1.pdf
• File Type:
  [PDF - 1.90 MB]
• Collection(s):
  University Transportation Centers
• Main Document Checksum:
  urn:sha-512:fa8bb2e10e375246a682faaa787c63d6c4ea4efd835b5a70e6cc1f6f6f9c1ea60e41f88529a387d43e0c93961eb0f5bcb228109ae594e8fb5a2b1ef5b5c9f61c