Scholarlink

<p>Potholes pose a significant threat to road safety and contribute to increased vehicle damage and maintenance costs. This paper is aimed to explore the use of a deep learning-based object detection model to autonomously detect potholes on the road surface using road surface images. In particular, the Faster Region-based Convolutional Neural Network (Faster R-CNN) is used to identify and localize potholes in a wide range of real-world scenarios, such as the changes in lighting, weather, and texture of the road surface. This methodology is proposed, including dataset preparation with annotated bounding boxes, data pre-processing, augmentation, training and testing of one model with a two-stage detection architecture and optimized hyper-parameters. Standard object detection metrics such as accuracy, precision, recall, F1 score and mAP are used to assess the performance of the model on the detection of objects and their localization. The paper proves that Faster R-CNN is an effective model to be used as an automated pothole detector and reveals its prospective in the fields of intelligent road surveillance and road infrastructure repair systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Smarter Roads with Deep Learning: Faster Region-based CNN for Efficient Pothole Detection

  • 期刊论文

  • 发表日期：

    2026年4月18日

查看全文
  • Asad Ullah,
  • Fakhre Alam,
  • Dilawar Shah,
  • Shujaat Ali,
  • Muhammad Tahir

摘要

Potholes pose a significant threat to road safety and contribute to increased vehicle damage and maintenance costs. This paper is aimed to explore the use of a deep learning-based object detection model to autonomously detect potholes on the road surface using road surface images. In particular, the Faster Region-based Convolutional Neural Network (Faster R-CNN) is used to identify and localize potholes in a wide range of real-world scenarios, such as the changes in lighting, weather, and texture of the road surface. This methodology is proposed, including dataset preparation with annotated bounding boxes, data pre-processing, augmentation, training and testing of one model with a two-stage detection architecture and optimized hyper-parameters. Standard object detection metrics such as accuracy, precision, recall, F1 score and mAP are used to assess the performance of the model on the detection of objects and their localization. The paper proves that Faster R-CNN is an effective model to be used as an automated pothole detector and reveals its prospective in the fields of intelligent road surveillance and road infrastructure repair systems.