A New Method for Automated Monitoring of Road Pavement Aging Conditions Based on Recurrent Neural Network

Xiao Chen, Xianfeng Zhang, Jonathan Li, Miao Ren, Bo ZhouIEEE Trans. ITS 2022 · OpenReview

Abstract

The automated monitoring of road pavement conditions is a challenging subject in intelligent transportation. However, the existing studies mostly focus on extracting pavement damages such as cracks, while the pavement aging conditions are still less investigated. In this paper, a novel method based on a modified recurrent neural network is designed for automated monitoring of asphalt pavement aging phenomena from fine-resolution satellite imagery. A spectral augmentation method is proposed to enhance the spectral details of the road pavements. A novel loss function is also proposed to improve the bi-directional gated recurrent unit (Bi-GRU) network in order to better classify different degrees of road pavement aging and non-pavement objects. In order to demonstrate the outperformance of the modified network Bi-GRU+, the Worldview-2 satellite image (16360*7728) covering 16 asphalt roads in the southwestern suburb of Beijing City is used. The results show that the proposed approach has better performance than existing machine learning methods, with an overall accuracy of 98.16% and a Kappa coefficient of 0.97. The overall processing time of the proposed method is 7836 seconds in our case study. The proposed method is efficient for large-scale monitoring of road health conditions from fine-resolution satellite imagery. It can become a part of intelligent transportation and provide a new foundation for large-range automated monitoring of road pavement aging conditions.

Introduction

Road maintenance is a critical component of urban infrastructure management, yet it remains a significant logistical challenge. Traditionally, road health is monitored through manual field inspections or specialized sensor-equipped vehicles. While effective, these methods are slow, potentially hazardous, and difficult to scale across vast metropolitan areas. Recent advancements in artificial intelligence have focused on detecting visible pavement damage, such as cracks and potholes. However, there is a notable gap in monitoring "pavement aging"—the gradual process of oxidation and loss of asphalt oils that precedes structural failure.

Detecting these subtle chemical and physical changes requires a shift from reactive to proactive maintenance. High-resolution satellite imagery offers a scalable solution, but it presents technical hurdles. Road surfaces in satellite photos are often only a few pixels wide, making it difficult for standard Convolutional Neural Networks (CNNs) to rely on spatial patterns. Furthermore, the spectral differences between various aging stages are extremely subtle, necessitating a more sophisticated approach than traditional image classification.

Methodology

To address these challenges, this study introduces a comprehensive framework centered on a modified Recurrent Neural Network (RNN) architecture, termed Bi-GRU+. The workflow encompasses data preprocessing, network training, and decision support.

Overview of the proposed Bi-GRU+ framework for road pavement aging monitoring

Spectral Augmentation and Bi-GRU+ Architecture

The methodology relies on three key technical innovations:

  1. Spectral Augmentation: A nonlinear mapping function is employed to "stretch" the reflectance values of the road pixels. This process enhances spectral details, making the subtle differences in reflectance across various aging stages more discernible for the model.
  2. Bi-directional Gated Recurrent Unit (Bi-GRU): Instead of treating satellite data as a 2D image, the model treats the eight spectral bands as a sequence. The bi-directional nature of the GRU allows the model to learn dependencies across the spectrum in both directions, effectively capturing the "chemical fingerprint" of asphalt degradation.
  3. Modified Loss Function ($L_{loss}$): To improve classification accuracy for "hard samples"—pixels that are difficult to distinguish or belong to similar classes—the standard cross-entropy loss was modified. This new function forces the network to focus on the subtle boundaries between aging stages.

Dataset and Ground Truth

The model was tested using Worldview-2 imagery of the Fangshan District in Beijing, featuring a spatial resolution of 0.5 meters across eight spectral bands. Ground truth was established through field investigations of 16 asphalt roads, where segments were categorized into three stages: Slightly Aged, Moderately Aged, and Heavily Aged. A total of 66,000 pixels were used for training, validation, and testing.

Results and Discussion

The Bi-GRU+ model demonstrated exceptional performance, achieving an Overall Accuracy (OA) of 98.16% and a Kappa coefficient of 0.97. This significantly outperforms traditional remote sensing methods and standard CNN architectures like ResNet18.

The Graying Effect and Spectral Signatures

As asphalt ages, it undergoes a "graying" process where dark oils are lost, exposing light-colored gravel. This transition is reflected in the spectral signatures: slightly aged pavement typically reflects 5–10% of light, whereas heavily aged pavement reflects between 14% and 35%. The Bi-GRU+ model successfully identifies these signatures even when spatial information is limited.

Visual and spectral characteristics of different pavement aging stages and non-pavement objects

Practical Application and Efficiency

The entire processing pipeline for the study area took approximately 2.2 hours, proving its viability for large-scale urban monitoring. Furthermore, the authors developed a "Pavement Aging Index" (PAI). By weighting classification results, each road segment receives a score from 0 to 1. Roads scoring above 0.5 are flagged as "heavily aged," providing a clear, automated signal for maintenance prioritization.

Limitations

While highly accurate, the method faces challenges such as shadows from trees or buildings, which can alter spectral signatures. Additionally, "salt-and-pepper" noise in pixel-based classification and "mixed pixels" at road edges remain areas for future improvement. Integrating spatial-spectral hybrids or utilizing higher-resolution UAV data could mitigate these issues.

Conclusion

The proposed Bi-GRU+ framework provides a robust and efficient tool for automated road pavement aging monitoring. By treating spectral data as a sequence and enhancing it through targeted augmentation and loss functions, the system enables proactive infrastructure management at a city-wide scale. This research lays the foundation for integrating satellite-based monitoring into intelligent transportation systems, moving toward more sustainable and cost-effective road maintenance strategies.

Summary updated

Abstract

The automated monitoring of road pavement conditions is a challenging subject in intelligent transportation. However, the existing studies mostly focus on extracting pavement damages such as cracks, while the pavement aging conditions are still less investigated. In this paper, a novel method based on a modified recurrent neural network is designed for automated monitoring of asphalt pavement aging phenomena from fine-resolution satellite imagery. A spectral augmentation method is proposed to enhance the spectral details of the road pavements. A novel loss function is also proposed to improve the bi-directional gated recurrent unit (Bi-GRU) network in order to better classify different degrees of road pavement aging and non-pavement objects. In order to demonstrate the outperformance of the modified network Bi-GRU+, the Worldview-2 satellite image (16360*7728) covering 16 asphalt roads in the southwestern suburb of Beijing City is used. The results show that the proposed approach has better performance than existing machine learning methods, with an overall accuracy of 98.16% and a Kappa coefficient of 0.97. The overall processing time of the proposed method is 7836 seconds in our case study. The proposed method is efficient for large-scale monitoring of road health conditions from fine-resolution satellite imagery. It can become a part of intelligent transportation and provide a new foundation for large-range automated monitoring of road pavement aging conditions.
BibTeX
@article{chen2022new,
  author = {Xiao Chen and Xianfeng Zhang and Jonathan Li and Miao Ren and Bo Zhou},
  title = {{A New Method for Automated Monitoring of Road Pavement Aging Conditions Based on Recurrent Neural Network}},
  journal = {IEEE Trans. Intell. Transp. Syst.},
  year = {2022},
  url = {https://openreview.net/forum?id=sH4Yc6-f6Vf},
}
Figure