NCHRP Web-Only Document 476: Implementing an Algorithm to Detect Asphalt Pavement Raveling
Raveling, or the loss of pavement surface aggregates, can reduce ride quality, compromise pavement performance, and lead to costly repairs if not detected and treated promptly. Current detection methods rely largely on labor-intensive, subjective visual inspections that often fail to identify localized areas where targeted preservation treatments could be applied. Although previous research has demonstrated the feasibility of using machine learning and high-resolution 3D laser imaging to automate raveling detection and severity classification, widespread implementation by transportation agencies remains limited.
NCHRP Web-Only Document 476: Implementing an Algorithm to Detect Asphalt Pavement Raveling, produced by TRB’s National Cooperative Highway Research Program, presents a framework for implementing automated raveling detection. It includes procedures for selecting pavement images, preparing datasets, assigning consistent severity classifications, training machine learning models, conducting quality assurance, and integrating results into maintenance and rehabilitation decisions. The report demonstrates the implementation process through a real-world case study and presents findings from three pilot studies conducted by state departments of transportation (DOTs).
Infrastructure Assets: Highway Assets, Pavement
Resource Types: Guide/Manual
Capabilities: Tools & Technology
Management Processes: Strategic Direction