Scholarlink

<p>The classification and localization of road distress play a crucial role in intelligent road health monitoring systems. To address the challenges of complex road backgrounds, diverse shapes of distress objects, and high computational resource requirements, this paper proposes an efficient focusing real-time road distress detection model (EF-RT-DETR). Based on RT-DETR, the model designs a new backbone network aimed at accurately capturing fine features and optimizes the attention mechanism to effectively reduce interference from complex backgrounds while enhancing the processing of detailed information. Additionally, an innovative fusion module is introduced in the feature fusion stage to further enhance the interaction between local and global features, while also reducing computational costs. Experiments conducted on the China_Motorbike subset of the RDD2022 dataset include ablation studies to validate the effectiveness of the proposed modules. The experimental results show that EF-RT-DETR reduced background false positives compared to the baseline model, with <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11554_2025_1641_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="50" /> </InlineMediaObject> <EquationSource Format="TEX">\({\hbox {mAP}}_{50}\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mtext>mAP</mtext> <mn>50</mn> </msub> </math></EquationSource> </InlineEquation> and <InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11554_2025_1641_Article_IEq2.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="65" /> </InlineMediaObject> <EquationSource Format="TEX">\({\hbox {mAP}}_{50:95}\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mtext>mAP</mtext> <mrow> <mn>50</mn> <mo>:</mo> <mn>95</mn> </mrow> </msub> </math></EquationSource> </InlineEquation> improving by 9.5 and 7.1%, respectively, while reducing computation by 35.4% and the number of parameters by 25.7%.</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

EF-RT-DETR: a efficient focused real-time DETR model for pavement distress detection

  • 期刊论文

  • 发表日期：

    2025年2月19日

查看全文
  • Tao Han,
  • Shuainan Hou,
  • Can Gao,
  • Shanyong Xu,
  • Jiale Pang,
  • Hai Gu,
  • Yourui Huang

摘要

The classification and localization of road distress play a crucial role in intelligent road health monitoring systems. To address the challenges of complex road backgrounds, diverse shapes of distress objects, and high computational resource requirements, this paper proposes an efficient focusing real-time road distress detection model (EF-RT-DETR). Based on RT-DETR, the model designs a new backbone network aimed at accurately capturing fine features and optimizes the attention mechanism to effectively reduce interference from complex backgrounds while enhancing the processing of detailed information. Additionally, an innovative fusion module is introduced in the feature fusion stage to further enhance the interaction between local and global features, while also reducing computational costs. Experiments conducted on the China_Motorbike subset of the RDD2022 dataset include ablation studies to validate the effectiveness of the proposed modules. The experimental results show that EF-RT-DETR reduced background false positives compared to the baseline model, with \({\hbox {mAP}}_{50}\) mAP 50 and \({\hbox {mAP}}_{50:95}\) mAP 50 : 95 improving by 9.5 and 7.1%, respectively, while reducing computation by 35.4% and the number of parameters by 25.7%.