An End-to-End Automated Computer Vision Framework for Pavement Distress Detection, Segmentation, Geometric Quantification, and Condition Assessment

Reliable pavement-distress assessment is essential for road asset management, but manual surveys are labor-intensive, and many vision-based methods do not provide physical measurements. This study presents an end-to-end framework for detecting, segmenting, and quantifying linear cracks, alligator cracking, and potholes, and for assessing pavement condition. A YOLO11m instance-segmentation model is combined with camera calibration, ground-sampling-distance scaling, and Inverse Perspective Mapping (IPM) to estimate distress dimensions in physical units. A Surface Condition Index (SCI) is formulated using area-normalized distress densities and distress-specific weights informed by ASTM D6433. On an independent field test set, the model achieved detection and segmentation [email protected] values of 91.87% and 89.23%, respectively. Inference speeds were 70.4 frames/s on an NVIDIA Tesla P100 and 42.6 frames/s on a Jetson Orin Nano. On nuScenes front-camera images, supervised domain adaptation increased detection [email protected] from 61.17% to 84.60%. Compared with manual measurements, geometric estimates yielded R² values of 0.908–0.963, with a mean absolute error of 0.50 mm for linear crack width. For 50 roadway sections, empirical mapping from SCI to manual ASTM D6433 Pavement Condition Index (PCI) ratings achieved R² = 0.88 and a mean absolute error of 6.98 PCI points. These results show the potential of the framework to integrate distress recognition, metric quantification, and condition scoring for pavement monitoring across dedicated survey systems and vehicle-camera streams.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-18
DOI
https://doi.org/10.21203/rs.3.rs-11487716/v1
Primary Topic
Infrastructure Maintenance and Monitoring
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

An End-to-End Automated Computer Vision Framework for Pavement Distress Detection, Segmentation, Geometric Quantification, and Condition Assessment

Amir Golroo, Fereidoon Moghadas Nejad, Alexandre Alahi, Mahdi Javanmardi et al.
7 links
Research Square
Infrastructure Maintenance and Monitoring
preprint

An End-to-End Automated Computer Vision Framework for Pavement Distress Detection, Segmentation, Geometric Quantification, and Condition Assessment

Amir Golroo, Fereidoon Moghadas Nejad, Alexandre Alahi, Mahdi Javanmardi, Mohammad Sedighian-Fard, Vahid Shahhosseini, Soroush Amiri, Mohammad Amin Talaghat
preprint en

Abstract

Reliable pavement-distress assessment is essential for road asset management, but manual surveys are labor-intensive, and many vision-based methods do not provide physical measurements. This study presents an end-to-end framework for detecting, segmenting, and quantifying linear cracks, alligator cracking, and potholes, and for assessing pavement condition. A YOLO11m instance-segmentation model is combined with camera calibration, ground-sampling-distance scaling, and Inverse Perspective Mapping (IPM) to estimate distress dimensions in physical units. A Surface Condition Index (SCI) is formulated using area-normalized distress densities and distress-specific weights informed by ASTM D6433. On an independent field test set, the model achieved detection and segmentation [email protected] values of 91.87% and 89.23%, respectively. Inference speeds were 70.4 frames/s on an NVIDIA Tesla P100 and 42.6 frames/s on a Jetson Orin Nano. On nuScenes front-camera images, supervised domain adaptation increased detection [email protected] from 61.17% to 84.60%. Compared with manual measurements, geometric estimates yielded R² values of 0.908–0.963, with a mean absolute error of 0.50 mm for linear crack width. For 50 roadway sections, empirical mapping from SCI to manual ASTM D6433 Pavement Condition Index (PCI) ratings achieved R² = 0.88 and a mean absolute error of 6.98 PCI points. These results show the potential of the framework to integrate distress recognition, metric quantification, and condition scoring for pavement monitoring across dedicated survey systems and vehicle-camera streams.

Research Square
Amirkabir University of Technology (IR), École Polytechnique Fédérale de Lausanne (CH)
Industry, innovation and infrastructure, Sustainable cities and communities
Infrastructure Maintenance and Monitoring
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon