Abstract/Summary

Synthetic aperture radar (SAR) image change detection (CD) focuses on identifying the change between two images at different times for the same geographical region. SAR offers advantages over optical sensors for disaster-related change detection in remote sensing due to its all-weather capability and ability to penetrate clouds and darkness. The performance of change detection methods is affected by several challenges. Deep learning methods, such as convolutional neural networks (CNNs), have shown promising performance in dealing with these challenges. However, CNN methods still suffer from speckle noise, adversely impacting the change detection performance F1 score. To tackle this challenge, we propose a CNN model that despeckles the noise prior to applying change detection methods. We extensively evaluate the performance of our method on three SAR datasets, and the results of our proposed method demonstrate superior performance compared to state-of-the-art methods such as DDNet and LANTNet performance. Our method significantly increased the change detection accuracy from a baseline of 86.65% up to 90.79% for DDNet and from 87.16% to 91.1% for LANTNet in the Yellow River dataset.