United States Department of Transportation
ROSA P serves as an archival repository of USDOT-published products including scientific findings, journal articles, guidelines, recommendations, or other information authored or co-authored by USDOT or funded partners.
As a repository, ROSA P retains documents in their original published format to ensure public access to scientific information.
i
Determination of Pavement Surface Type
-
2021-12-01
-
[PDF-11.57 MB]
Details:
-
Creators:
-
Corporate Creators:
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Publication/ Report Number:
-
Resource Type:
-
Geographical Coverage:
-
Edition:Technical Report Month Year – June 2024
-
Corporate Publisher:
-
Abstract:This study evaluated and developed machine learning models for pavement surface classification to enhance accuracy and reliability using advanced computational techniques. Initially, a detailed literature review on pavement texture and measurement methods provided a foundation for developing a prototype system. This system, integrating laser scanners, high-speed cameras, and lighting systems, marked a significant advancement in capturing high-resolution texture data at highway speeds. Data collection covered 425.5 miles of pavement across Texas, including 313.7 miles of flexible and 111.8 miles of rigid pavements, resulting in over 50,000 high- resolution images and texture profiles from 15 types of flexible pavements and seven types of rigid pavements. This dataset, essential for training and validating machine learning models, was made available for future research. A hierarchical classification method was developed, with picture-based models (PBC) excelling at predicting flexible or rigid pavements (Level 1) and tining orientation (Level 5), while index-based classifiers (IBCs) performed better for intermediate levels of specificity (Levels 2-4). The best models achieved high F1 scores, confirming their effectiveness. Comprehensive validation using six diverse test sets across 20 test sites confirmed the models’ robustness and applicability. These findings significantly enhance pavement management systems, enabling more accurate and efficient maintenance strategies. Leveraging advanced machine learning techniques, the developed models can improve road safety, optimize resource allocation, extend pavement lifespan, and support future research and practical applications.
-
Format:
-
Funding:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
