Scholarlink

<p>During the service of a bridge, the spatial distribution characteristics of crack diseases have a direct impact on structural safety assessment. However, traditional two-dimensional detection methods are difficult to accurately express cracks in real space. Therefore, this study constructs a three-dimensional reconstruction model of bridge cracks that integrates improved you only look once version 9 (YOLOv9) and UAV photogrammetry. This method is based on UAV multi-view images and completes the construction of a three-dimensional bridge model through photogrammetry. Moreover, it is time to introduce a simple parameter-free attention module in the crack detection stage to enhance the feature expression ability of key crack areas, thereby achieving accurate mapping of crack detection results to the three-dimensional model.Experimental results revealed that the average precision mean @0.5 of the proposed method in the crack detection task reached 0.90, the detection precision and recall rate reached 0.91 and 0.86 respectively, and the false detection rate was reduced to 0.09. Under complex background conditions, its accuracy rate remained stable above 0.88. In the three-dimensional fusion verification, the spatial consistency of this method at a shooting distance of 10&#xa0;m was close to 0.95. It still maintained about 0.75 under the condition of 80&#xa0;m, and the three-dimensional localization error was controlled within 7&#xa0;cm. At the crack scale of 3px, the three-dimensional mapping success rate reached 79.8%, and was stable at over 90% at larger crack scales. The findings of the study demonstrate that the suggested approach may successfully raise the stability and accuracy of the combination of three-dimensional reconstruction and fracture detection. This provides a reliable technical solution for intelligent drone inspection of bridges.</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

Bridge three-dimensional reconstruction model based on improved YOLOv9 and unmanned aerial vehicle photography

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月29日

查看全文
  • Chao Feng

摘要

During the service of a bridge, the spatial distribution characteristics of crack diseases have a direct impact on structural safety assessment. However, traditional two-dimensional detection methods are difficult to accurately express cracks in real space. Therefore, this study constructs a three-dimensional reconstruction model of bridge cracks that integrates improved you only look once version 9 (YOLOv9) and UAV photogrammetry. This method is based on UAV multi-view images and completes the construction of a three-dimensional bridge model through photogrammetry. Moreover, it is time to introduce a simple parameter-free attention module in the crack detection stage to enhance the feature expression ability of key crack areas, thereby achieving accurate mapping of crack detection results to the three-dimensional model.Experimental results revealed that the average precision mean @0.5 of the proposed method in the crack detection task reached 0.90, the detection precision and recall rate reached 0.91 and 0.86 respectively, and the false detection rate was reduced to 0.09. Under complex background conditions, its accuracy rate remained stable above 0.88. In the three-dimensional fusion verification, the spatial consistency of this method at a shooting distance of 10 m was close to 0.95. It still maintained about 0.75 under the condition of 80 m, and the three-dimensional localization error was controlled within 7 cm. At the crack scale of 3px, the three-dimensional mapping success rate reached 79.8%, and was stable at over 90% at larger crack scales. The findings of the study demonstrate that the suggested approach may successfully raise the stability and accuracy of the combination of three-dimensional reconstruction and fracture detection. This provides a reliable technical solution for intelligent drone inspection of bridges.