Scholarlink

<p>This study proposes CMUNet, an optimised neural network architecture for monitoring the water body upstream of the Ertan Hydropower Station in Panzhihua. By comparing systematic ablation experiments with 15 CBAM placement combinations, the research results show that adding a single CBAM attention module after the third downsampling layer of the UNet can achieve the best performance. Experimental results on the Sentinel-2A satellite data set show that the CMUNet has an average IoU of 0.752, a Dice coefficient of 0.840, and a pixel accuracy of 0.940, which is better than benchmark models such as FCN and DeepLabV3 + . This model maintains a low computational overhead while exhibiting strong segmentation capabilities under different environmental conditions. At the same time, based on CMUNet, an intelligent hydropower station management system has been designed. This system can automatically generate gate control and unit operation strategies based on changes in the water area, providing reliable technical support for monitoring and managing water resources around the hydropower station.</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

CMUNet neural network-based water body survey management upstream of hydropower station

  • 期刊论文

  • 发表日期：

    2025年6月19日

查看全文
  • Yipeng Qiu,
  • Haicheng Tian,
  • A. P. Xuliang Duan,
  • Ao Feng,
  • Antong Deng

摘要

This study proposes CMUNet, an optimised neural network architecture for monitoring the water body upstream of the Ertan Hydropower Station in Panzhihua. By comparing systematic ablation experiments with 15 CBAM placement combinations, the research results show that adding a single CBAM attention module after the third downsampling layer of the UNet can achieve the best performance. Experimental results on the Sentinel-2A satellite data set show that the CMUNet has an average IoU of 0.752, a Dice coefficient of 0.840, and a pixel accuracy of 0.940, which is better than benchmark models such as FCN and DeepLabV3 + . This model maintains a low computational overhead while exhibiting strong segmentation capabilities under different environmental conditions. At the same time, based on CMUNet, an intelligent hydropower station management system has been designed. This system can automatically generate gate control and unit operation strategies based on changes in the water area, providing reliable technical support for monitoring and managing water resources around the hydropower station.