ESR-DMNet: Enhanced Super-Resolution-Based Dual-Path Metric Change Detection Network for Remote Sensing Images With Different Resolutions

Remote sensing change detection (RSCD) has always been one of the research hot issues in remote sensing. Current research focuses on studying deep learning (DL) change detection (CD) methods for remote sensing images with the same resolution. With the prevalence of multiresolution remote sensing images, how to effectively utilize remote sensing images with different resolutions for CD is a key issue. To solve this problem, this article proposes an enhanced super-resolution-based dual-path metric CD network (ESR-DMNet) to realize high-accuracy and high-efficiency end-to-end CD of remote sensing images with different resolutions. ESR-DMNet provides a new enhanced super-resolution (SR) module for the CD of remote sensing images with different resolutions, which can perceptively reconstruct low-resolution (LR) images into more realistic high-resolution (HR) images. ESR-DMNet proposes an effective and efficient dual-path metric CD network, which processes shallow spatial details information and deep semantic information separately to achieve high-accuracy and high-efficiency CD. Compared with nine state-of-the-art methods, our method shows good performance at three resolutions on three datasets, Sun Yat-Sen University (SYSU), season-varying CD dataset (CDD), and cropland CD dataset (CLCD), confirming its potential for CD tasks in remote sensing images with different resolutions.