Combined Sentinel-1 and Sentinel-2 Imagery for Destroyed Building Classification in Gaza Strip With Random Forest

Airspace control in war zones poses a significant barrier to the acquisition of high-quality high-resolution remote sensing imagery, which is the prerequisite for analyzing the destruction and damages caused by bombarding and missile attacks and assessing the need for humanistic aids for affected civilians. This lack of information is also an obstacle to plan logistics of rescue operations. In this article, we investigate the use of coarser resolution civilian Sentinel-2 multispectral (MS) images and Sentinel-1 synthetic aperture radar (SAR) data to achieve Random Forest-based classification of destroyed buildings in the Gaza Strip. For input features, we utilize preprocessed MS and SAR images, texture features extracted from MS, and polarization features decomposed from SAR. Additionally, the bi-temporal differences and the fusion of MS and SAR data are assessed for their classification accuracy. The classification results demonstrate the potential 10 m MS and SAR for recognizing destroyed buildings. Furthermore, we conduct the classification and analysis of monthly changes in destroyed buildings in the Gaza Strip. Figures and facts from official agencies and social media confirm the good consistency between remote sensing-based change analysis and the current situation.