Scholarlink

<p>The advancement of remote sensing technology has significantly enhanced the extraction and dynamic monitoring of coastal aquaculture areas. High-resolution remote sensing imagery plays a crucial role in accurately identifying these areas. However, variations in spectral characteristics due to complex environmental conditions, particularly conditions influenced by sun glint interference, can affect the performance of remote sensing methods. This study investigates the effectiveness of traditional supervised classification and deep learning methods in extracting aquaculture areas using Gaofen-2 remote sensing images captured under both normal and challenging imaging conditions caused by sun glint. Our findings indicate that for images captured under normal imaging conditions, the support vector machine (SVM) method and the U-Net deep learning method achieved F1 scores of 96.20% and 95.90%, respectively. For images compromised by sun glint, the U-Net, LinkNet, and DeepLabv3 methods yielded F1 scores of 77.51%, 78.45%, and 78.63%, respectively. Overall, the U-Net deep learning method demonstrated high applicability and robustness across the varying imaging conditions, especially when dealing with sun glint interference.</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

Comparative analysis of remote sensing techniques for coastal aquaculture area extraction from gaofen-2 imagery in the presence of sun glint

  • 期刊论文

  • 发表日期：

    2025年7月22日

查看全文
  • Yangdong Li,
  • Tian Dong,
  • Juntai Leng

摘要

The advancement of remote sensing technology has significantly enhanced the extraction and dynamic monitoring of coastal aquaculture areas. High-resolution remote sensing imagery plays a crucial role in accurately identifying these areas. However, variations in spectral characteristics due to complex environmental conditions, particularly conditions influenced by sun glint interference, can affect the performance of remote sensing methods. This study investigates the effectiveness of traditional supervised classification and deep learning methods in extracting aquaculture areas using Gaofen-2 remote sensing images captured under both normal and challenging imaging conditions caused by sun glint. Our findings indicate that for images captured under normal imaging conditions, the support vector machine (SVM) method and the U-Net deep learning method achieved F1 scores of 96.20% and 95.90%, respectively. For images compromised by sun glint, the U-Net, LinkNet, and DeepLabv3 methods yielded F1 scores of 77.51%, 78.45%, and 78.63%, respectively. Overall, the U-Net deep learning method demonstrated high applicability and robustness across the varying imaging conditions, especially when dealing with sun glint interference.