A review of research on remote sensing images shadow detection and application to building extraction

ABSTRACT

Buildings are one of the most important habitats for humans, and therefore, accurate identification and extraction of building information in remote sensing images are crucial. Buildings in remote sensing images vary in shape and color due to differences in sensor acquisition methods, geographical location, and other factors. However, they all share a common feature – the presence of shadows. Obtaining accurate data from building shadows can provide a wealth of reliable information for building research. Consequently, it is crucial to review various methods for extracting building shadows, especially deep learning-based methods, to illustrate shadow implementation scenarios in building research: 1) building detection in very high resolution remote sensing images (VHRRSI); 2) building detection in SAR; 3) building change detection; 4) building damage assessment; 5) building height estimation; 6) building shadow removal; 7) other methods (such as building shadow data enhancement, detection of building shadows in ghost images, and conservation of historic buildings). This study discusses the advantages and disadvantages of building shadow detection methods and provides an overview of the datasets and evaluation metrics commonly used in studies of building shadow applications. We hope that this study will serve as a valuable reference for researchers in the field of building shadow studies.

KEYWORDS:

• Building research
• shadow detection
• deep learning
• remote sensing

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Ministry of Natural Resources [2016KCT-23]; National Natural Science Foundation of China [41571346]; Open Fund for Key Laboratory of Degraded and Unused Land Consolidation Engineering [SXDJ2017-10].