Multi-Spectral Satellite Imagery Dataset Based on Sentinel-2 for Change Detection Task

Change detection is a key research topic in the field of remote sensing image analysis. It aims to identify changes on the Earth's surface by comparing images captured at different times over the same geographic area. However, change detection based on visible light images is often susceptible to illumination variations, which can significantly degrade performance. To address this limitation, we construct a Multi-Spectral Change Detection Dataset (MSCD) using Sentinel-2 satellite imagery and provide benchmark results using several deep learning methods. Furthermore, it is feasible to evaluate the effectiveness of incorporating infrared bands by proposing an improved network based on STANet tailored for multi-spectral change detection. Experimental results demonstrate that multi-spectral fusion methods significantly outperform those using only visible or infrared bands individually, highlighting the advantages of multi-spectral data for change detection tasks and offering a reliable solution for analyzing surface changes in complex scenarios. The dataset and code is availabel at https://github.com/kaka0910/MSCD.