Automated Road Crack Localization to Guide Highway Maintenance

TL;DR

The paper presents a data-driven, open-data framework for scalable highway crack localization by fine-tuning YOLOv11 on high-resolution airborne imagery and OpenStreetMap data, producing the RHCD index to guide maintenance. It demonstrates the approach in Switzerland, achieving F1 scores of 0.84 for crack vs. 0.97 for no crack, and showing that RHCD correlates weakly with LT-LST-A and TV but provides added value for maintenance prioritization. The method integrates data retrieval, preprocessing, augmentation, and interpretability (Grad-CAM) to enable large-scale, transparent crack detection. The results highlight urban centers and intersections as high-crack-density areas and suggest that the RHCD index can be transferred to other regions to support proactive infrastructure management with open-source data.

Abstract

Highway networks are crucial for economic prosperity. Climate change-induced temperature fluctuations are exacerbating stress on road pavements, resulting in elevated maintenance costs. This underscores the need for targeted and efficient maintenance strategies. This study investigates the potential of open-source data to guide highway infrastructure maintenance. The proposed framework integrates airborne imagery and OpenStreetMap (OSM) to fine-tune YOLOv11 for highway crack localization. To demonstrate the framework's real-world applicability, a Swiss Relative Highway Crack Density (RHCD) index was calculated to inform nationwide highway maintenance. The crack classification model achieved an F1-score of $0.84$ for the positive class (crack) and $0.97$ for the negative class (no crack). The Swiss RHCD index exhibited weak correlations with Long-term Land Surface Temperature Amplitudes (LT-LST-A) (Pearson's $r\ = -0.05$) and Traffic Volume (TV) (Pearson's $r\ = 0.17$), underlining the added value of this novel index for guiding maintenance over other data. Significantly high RHCD values were observed near urban centers and intersections, providing contextual validation for the predictions. These findings highlight the value of open-source data sharing to drive innovation, ultimately enabling more efficient solutions in the public sector.

Figures (9)

• Figure 1: Schematic representation of the proposed framework for scalable highway crack detection, comprising i) the retrieval of road imagery, ii) the training of a crack localization model, and iii) the generation of a RHCD index.
• Figure 2: Overview of the data retrieval process: downloading, masking, and tiling of highway airborne imagery.
• Figure 3: Overview of the modeling pipeline, including data labeling, augmentation of the training set, YOLOv11 model architecture design, and computation of evaluation metrics.
• Figure 4: Visual inspection of highway crack localization results (red = positive detection; blue = negative detection). Two examples of each True Positives (TP), False Positives (FP), True Negatives (TN), and False Negatives (FN), with Guided Grad-CAM visualizations indicating key regions affecting predictions.
• Figure 5: Panel A shows the Swiss highway crack localization map along with major urban centers in Switzerland and a bounding box indicating the area detailed in Panel B. Panel B presents a zoomed-in view highlighting the classification results on 5 m × 5 m airborne imagery tiles used for calculating the RHCD index.