Scholarlink

<p>To enhance road safety and reduce vehicle accidents, the timely identification of potential potholes is essential. This paper introduces an intelligent computer vision system for the early prediction and detection of pothole formation on road surfaces. Using deep learning from VGG16 and the real-time object detection capabilities of YOLOv7, the model classifies road segments into three categories: normal roads, rough roads, and potholes. The improved feature extraction algorithms of VGG16 enabled it to achieve a precision of 94%, outperforming YOLOv7’s 65.8%. YOLOv7 speed enables real-time detection, but its lower accuracy can lead to missed potholes, increasing long-term repair costs. The higher accuracy of VGG16 ensures better detection, preventing road deterioration, and optimizing resource allocation. The study has been extended to include advanced versions of the models, namely VGG19 &amp; YOLOv8, for a more comprehensive performance evaluation. These models have been incorporated into the evaluation framework of the study. In conclusion, early detection supports preventive maintenance, cost-effectiveness, and public safety. Therefore, the proposed model demonstrates promising accuracy and efficiency, making it suitable for smart city infrastructure and automated road inspection 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

Automated detection of road surface anomalies using image processing techniques

  • 期刊论文

  • 发表日期：

    2026年5月11日

查看全文
  • Sanskriti Pandey,
  • Shalinee Shukla,
  • Divya Kumar

摘要

To enhance road safety and reduce vehicle accidents, the timely identification of potential potholes is essential. This paper introduces an intelligent computer vision system for the early prediction and detection of pothole formation on road surfaces. Using deep learning from VGG16 and the real-time object detection capabilities of YOLOv7, the model classifies road segments into three categories: normal roads, rough roads, and potholes. The improved feature extraction algorithms of VGG16 enabled it to achieve a precision of 94%, outperforming YOLOv7’s 65.8%. YOLOv7 speed enables real-time detection, but its lower accuracy can lead to missed potholes, increasing long-term repair costs. The higher accuracy of VGG16 ensures better detection, preventing road deterioration, and optimizing resource allocation. The study has been extended to include advanced versions of the models, namely VGG19 & YOLOv8, for a more comprehensive performance evaluation. These models have been incorporated into the evaluation framework of the study. In conclusion, early detection supports preventive maintenance, cost-effectiveness, and public safety. Therefore, the proposed model demonstrates promising accuracy and efficiency, making it suitable for smart city infrastructure and automated road inspection systems.