Extracting built-up land area of airports in China using Sentinel-2 imagery through deep learning

Abstract

In China, airports have a profound impact on people’s lives, and understanding their dimensions has great significance for research and development. However, few existing airport databases contain such details, which can be reflected indirectly by the built-up land in the airport. In this study, a deep learning-based method was used for extraction of built-up land of airports in China using Sentinel-2 imagery and for further estimating their area. Here, a benchmark generation method is introduced by fusing two reference maps and cropping images into patches. Following this, a series of experiments were conducted to evaluate the network architectures and select the positive impact bands in Sentinel-2 imagery. A well-trained model was used to extract the built-up land for China airports, and the relationship between China airports’ built-up land and the carrying capacity of air transportation was further analysed. Results show that ResUNet-a outperformed U-Net, ResUNet, and SegNet, and the B2, B4, B6, B11, and B12 bands of Sentinel-2 had a positive impact on built-up land extraction. A well-trained model with an overall accuracy of 0.9423 and an F₁ score of 0.9041 and 434 China airports’ built-up land was extracted. The four most developed airports are located in Beijing, Shanghai, and Guangzhou, which matches China’s political and economic development. The area of built-up land influenced the passenger throughput and aircraft movements. The total area influenced the cargo throughput, and we found a certain correlation among the built-up land, carrying capacity, and nighttime light.

Keywords:

• Remote sensing
• deep learning
• Sentinel-2
• built-up land
• airports

Acknowledgments

The authors are thankful to the editor and the anonymous reviewers for their valuable comments and insightful ideas, authors also acknowledge European Space Agency, for providing the satellite data and images which we used in this study. This work is supported by the China Scholarship Council.

Disclosure statement

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

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.