An STP-HSI index method for urban built-up area extraction based on multi-source remote sensing data. Issue 11 (23rd November 2022)

• Record Type: Journal Article
• Authors: Bu, Lijing; Dai, Dong; Tu, Liying; Zhang, Zhengpeng; Deng, Mingjun; Xie, Xinyu
• Is Part Of: Royal Society open science. Volume 9:Issue 11(2022)
• Publication Date: 2022-11-23
• Subjects: extraction of built-up area -- night-time light remote sensing image -- landsat -- human settlement index -- random forest -- multi-source; Science -- Periodicals; 500
• DOI: 10.1098/rsos.220597
• Languages: English
• ISSNs: 2054-5703

Abstract

Abstract : A change in an urban built-up area can reflect the process of urbanization and the development of a city. At present, multi-source remote sensing data extraction of built-up areas based on the human settlement index (HSI) has achieved relatively good results but the existence of noise, such as light spillover in the night-time light remote sensing data, seriously affects the accuracy of the HSI. In this paper, a high-precision human settlement index (STP-HSI) method based on spatio-temporal remote sensing and point-of-interest (POI) data is presented to improve the classification accuracy in urban built-up areas extractions. First, to correct light spillover, a new night-time light index the fuzzy c-means spatio-temporal point (FCM-STP) based on fuzzy c-means clustering is proposed, which integrates the spatio-temporal characteristics and uses night light video imaging data from Luojia-1 and POI data. Then, based on the FCM-STP index, the HSI is updated to the STP-HSI index. Finally, a random forest algorithm is used to extract the urban built-up areas, and the random forest feature database is composed of normalized difference vegetation index (NDVI), normalized difference built-up index (NDBI) and STP-HSI index features and texture features. To develop and evaluate the accuracy of the new method for built-up areas extraction with multi-source data, three test sites located in the cities of China (Guangzhou, Xiamen and Nanjing) are used. The experimental results show that our method outperforms the single-source multi-spectral (Landsat 8) data extraction results, the overall accuracy is improved by up to 7.52%, and the kappa coefficient is improved by up to 14%. Compared with the HSI index, the maximum contribution rates of the STP-HSI increased by 25.74%. These experimental results show that the method in this paper is feasible.