Scholarlink

<p>Our study proposes a method to efficiently reconstruct physical models generating traffic noise maps using artificial intelligence technologies. A typical noise prediction model consists of noise generation and energy propagation models that use traffic and topographical information as input values. We developed a convolutional neural network (CNN) regression model for noise prediction utilizing the input and output data of the physical model. When the CNN model was applied to predict the average noise levels during peak hours in the cities in South Korea, we found a significant improvement in prediction performance and inference speed compared to existing machine learning models. Additionally, the numerical simulation demonstrated that noise maps for other regions can be effectively predicted by domain adaptation with minimal data. This suggests that providing foundational technology for real-time traffic noise monitoring is possible.</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

Convolutional neural network approach for accelerated traffic noise mapping

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月1日

查看全文
  • Sungsu Choi,
  • Taeho Park,
  • Jaesung Lim,
  • Enda Murphy,
  • Jong-June Jeon

摘要

Our study proposes a method to efficiently reconstruct physical models generating traffic noise maps using artificial intelligence technologies. A typical noise prediction model consists of noise generation and energy propagation models that use traffic and topographical information as input values. We developed a convolutional neural network (CNN) regression model for noise prediction utilizing the input and output data of the physical model. When the CNN model was applied to predict the average noise levels during peak hours in the cities in South Korea, we found a significant improvement in prediction performance and inference speed compared to existing machine learning models. Additionally, the numerical simulation demonstrated that noise maps for other regions can be effectively predicted by domain adaptation with minimal data. This suggests that providing foundational technology for real-time traffic noise monitoring is possible.