基于NPP-VIIRS及Landsat-8的多时相城市建成区提取方法探讨

Discussion on Extraction Method of Multi-temporal Urban Built-up Areas Based on NPP-VIIRS and Landsat-8 data

随着遥感技术的发展,由夜间灯光数据提取城市建成区已成为主要方法之一。传统阈值提取方法在小尺度的研究中存在较明显误差,以三明市三元区为研究区,基于2013—2021年NPP-VIIRS夜间灯光数据和Landsat-8遥感数据,利用夜间灯光数据的空间自相关特性,在扣除植被和水体后向前迭代提取2013—2021年多时相城市建成区。研究表明,改良方法总体分类精度在2013—2021年间不低于99%,Kappa系数不低于0.88,较传统的阈值法提取结果在精度和稳定性上有明显提升。

With the development of remote sensing technology,extracting urban built-up areas from nighttime lighting data has become one of the main methods.The traditional threshold method has obvious error when doing small-scale research.Taking Sanyuan District of Sanming City as the research area,based on the NPP-VIIRS night light data and Landsat-8 remote sensing data from 2013 to 2021,the candidate areas of built-up areas are extracted by using the spatial autocorrelation of night light data,and the built-up areas are extracted the multi-time urban built-up areas of 2013-2021 year after deducting vegetation and water.The results indicate that:the overall classification accuracy of the improved method in this paper is not less than 99%between 2013 and 2021,and the Kappa coefficient is not less than 0.88.Compared with thetraditional threshold method,the accuracy of the extracted results is significantly improved.