Scholarlink

<p>The detection of semantic changes in satellite images over time is known as Urban Land Change Detection (ULCD). The existing techniques for change detection struggled to&#xa0;detect missing changes or minor changes. To address these challenges, this research proposes a novel Dual Contextual Multi Head-Self Attention (DCMH-SA)&#xa0;approach that accurately detects changes in satellite images without missing any small changes with high accuracy and low computational time and cost. This ensures its efficiency in performing change detection tasks. Moreover, the Dung Beetle Optimization Algorithm (DBOA) is introduced to optimize the loss function of the proposed method, thereby enhancing the entire performance of the ULCD. Experimental evaluations demonstrate that the proposed method obtains 99.26% accuracy in the one satellite change detection dataset and 98.04% accuracy in the semantic change detection dataset, thus highlighting the robustness and effectiveness of the proposed framework.</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

Dung Dual Beetle Contextual Multi-Head Self-Attention Based Urban Land Cover Change Detection Using Satellite Images

  • 期刊论文

  • 发表日期：

    2025年7月4日

查看全文
  • Jambukeshwar S. Pujari,
  • Aprna Tripathi,
  • Javed Wasim

摘要

The detection of semantic changes in satellite images over time is known as Urban Land Change Detection (ULCD). The existing techniques for change detection struggled to detect missing changes or minor changes. To address these challenges, this research proposes a novel Dual Contextual Multi Head-Self Attention (DCMH-SA) approach that accurately detects changes in satellite images without missing any small changes with high accuracy and low computational time and cost. This ensures its efficiency in performing change detection tasks. Moreover, the Dung Beetle Optimization Algorithm (DBOA) is introduced to optimize the loss function of the proposed method, thereby enhancing the entire performance of the ULCD. Experimental evaluations demonstrate that the proposed method obtains 99.26% accuracy in the one satellite change detection dataset and 98.04% accuracy in the semantic change detection dataset, thus highlighting the robustness and effectiveness of the proposed framework.