Deep Learning in Damage Assessment with Remote Sensing Data: A Review

Over the years, deep learning (DL) algorithms have been employed in a variety of applications, including damage assessment utilizing remote sensing data. More than 70 articles in the subject, the majority of which were published during the last five years, are evaluated, and analyzed in this study, which introduces various significant DL approaches that use remote sensing data. This meta-analysis examines a variety of major DL concepts in relation to a variety of disaster types, focal objects, spatial resolutions, research regions, and classification accuracy. A thorough examination of how DL has been employed for damage assessment analytical tasks is undertaken, covering the dataset used, image preprocessing with fusion and registration, image/semantic segmentation, and object recognition and classification. Finally, a conclusion on current approaches, a critical conclusion on open challenges, and research directions are offered.