Multitemporal impervious surface estimation via an optimized stable/change pixel detection approach

作     者：Fan, Wei Chen, Jinsong Li, Xiaoli Tarolli, Paolo Wang, Jin 

作者机构：South China Normal Univ Fac Engn Beidou Res Inst Foshan Peoples R China Univ Wisconsin Dept Geog Milwaukee WI 53201 USA Chinese Acad Sci Shenzhen Inst Adv Technol Shenzhen Peoples R China Shenzhen Engn Lab Ocean Environm Big Data Anal & Shenzhen Peoples R China Univ Padua Dept Land Environm Agr & Forestry Legnaro PD Italy 

出 版 物：《GISCIENCE & REMOTE SENSING》 (地理信息系统科学与遥感)

年 卷 期：2022年第59卷第1期

页      面：1406-1425页

核心收录：

学科分类：070801[理学-固体地球物理学] 07[理学] 08[工学] 0708[理学-地球物理学] 0705[理学-地理学] 0816[工学-测绘科学与技术] 

基　　金：National Natural Science Foundation of China Fundamental Research Foundation of Shenzhen Technology and Innovation Council [KCXFZ202002011006298] 

主　　题：Impervious surface change detection classification and regression tree remote sensing 

摘      要：Remote sensing techniques have proved its efficacy for the impervious surface mapping, which is a significant indicator of urbanization process and environmental status. However, systematic and random errors in the existing methods still impact the reliability of subpixel impervious surface estimation, generating compounded errors when conducting multitemporal monitoring. The compounded errors of the conventional methods often significantly impact the temporal consistency of the results. In this study, a novel method based on a straightforward pixel change detection approach was put forward to improve the estimation of multitemporal impervious surface area. Two experimental areas located in Rome in Italy and Shenzhen in China were chosen to testify the generality of the proposed method to estimate different types of impervious surfaces worldwide. By reducing the compounded errors, the proposed method demonstrated its efficiency in achieving higher accuracy in both study areas without involving extensive data sources and intensive manual tasks. Compared with the conventional classification and regression tree algorithm, the overall mean average error and root mean square error of this study declined by more than 15.55% and 8.63%, respectively, and R-2 increased from approximately 0.93 to 0.96. The proposed method also drastically reduced the standard deviation of the multitemporal percent ISA of the stable pixels. The accurate change estimation of percent ISA has been a fundamental but challenging issue associated with monitoring and understanding the urban environment. Therefore, our proposed method, with its improved ability to estimate impervious surface change both spatially and temporally, can provide accurate information required for urban environment research.