Scholarlink

<p>This letter proposes a method based on fusion difference images (DIs) for the task of unsupervised synthetic aperture radar (SAR) image change detection. Firstly, three simple DIs are generated by applying the&#xa0;subtraction operator, mean-ratio operator and log-ratio operator. Secondly, a fusion method which combines the Gaussian weighted function and local energy is proposed to generate a better DI. The fusion method can adaptively allocate the weight to the three simple DIs. The local energy operation fully considers the attributes of adjacent pixels, which can more completely preserve the details of texture regions. The Gaussian weighted with the local energy of pixels as an exponential can highlight the changed regions and edge features of the three simple images respectively. Finally, a hierarchical clustering algorithm (HFCM) and the convolutional wavelet neural network are used to classify the fused DI and get the change detection map (CM). The results of the&#xa0;experiment indicate that the proposed method significantly enhances the precision of SAR image change detection.</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

Unsupervised SAR Image Change Detection Method Based on Fused Difference Images

  • 期刊论文

  • 发表日期：

    2025年7月16日

查看全文
  • Shaona Wang,
  • Di Wang,
  • Jia Shi,
  • Zhenghua Zhang,
  • Haoyu Wang,
  • Xiang Li,
  • Yanmiao Guo

摘要

This letter proposes a method based on fusion difference images (DIs) for the task of unsupervised synthetic aperture radar (SAR) image change detection. Firstly, three simple DIs are generated by applying the subtraction operator, mean-ratio operator and log-ratio operator. Secondly, a fusion method which combines the Gaussian weighted function and local energy is proposed to generate a better DI. The fusion method can adaptively allocate the weight to the three simple DIs. The local energy operation fully considers the attributes of adjacent pixels, which can more completely preserve the details of texture regions. The Gaussian weighted with the local energy of pixels as an exponential can highlight the changed regions and edge features of the three simple images respectively. Finally, a hierarchical clustering algorithm (HFCM) and the convolutional wavelet neural network are used to classify the fused DI and get the change detection map (CM). The results of the experiment indicate that the proposed method significantly enhances the precision of SAR image change detection.