Review of cloud removal methods for aerial images

DOI:	10.11809/bqzbgcxb2023.07.008
Keywords:	aerial images; cloud removal method; cloud region detection; deep learning; assessment criteria
Abstract:	The existence of cloud regions seriously affects the later interpretation of aerial images. How to remove cloud area interference with high quality is an important component of the current aerial image research. With the mature application of deep learning and the emergence of special datasets, cloud removal for aerial images has attracted extensive attention and also achieved phased results. Firstly, several commonly used datasets are introduced in this paper, and their number, resolution, acquisition media and characteristics are compared and analyzed. The download links are also given. Then, according to whether the cloud removal model uses deep learning theories, the cloud removal methods are divided into classical theory methods and deep learning methods. The methods based on deep learning are highlighted, all methods are compared and analyzed according to complementary data sources, application scenarios, datasets and their own characteristics, and the description of relevant models is conducted. After that, several common assessment criteria are introduced. Finally, the shortcomings of the existing methods are discussed, and the future development direction is prospected.
Published:	2023-07-28
PDF HTML