基于Sentinel-2A与NPP-VIIRS夜间灯光数据的城市建成区提取  

Extraction of urban built-up areas based on Sentinel-2A and NPP-VIIRS nighttime light data

作　　者：刘智丽 张启斌[1] 岳德鹏[1] 郝玉光[2] 苏凯 LIU Zhili;ZHANG Qibin;YUE Depeng;HAO Yuguang;SU Kai(College of Forestry,Beijing Forestry University,Beijing 100083,China;Experimental Center of Desert Forestry,Chinese Academy of Forestry,Bayannur 015200,China)

机构地区：[1]北京林业大学林学院,北京100083 [2]中国林业科学研究院沙漠林业实验中心,巴彦淖尔015200

出　　处：《国土资源遥感》2019年第4期227-234,共8页Remote Sensing for Land & Resources

基　　金：中央级公益性科研院所基本科研业务费专项资金项目“干旱区荒漠化治理效益与生态安全格局构建技术研究”(编号:CAFYBB2017MB026);国家自然科学基金项目“荒漠绿洲区景观格局与生态水文耦合及调控”(编号:41371189)共同资助

摘　　要：利用夜间灯光数据(nighttime light data,NTL)与光学遥感影像提取城市建成区是当今的一个研究热点,其中基于植被校正的城市夜间灯光指数(vegetation adjusted NTL urban index,VANUI)被学者广泛利用,但它容易混淆城市边缘的建筑、水体,空间分辨率较低。对VANUI做出改进,提出基于建筑校正的城市夜间灯光指数(building adjusted NTL urban index,BANUI)。利用该指数对包头市南部的城市建成区进行提取,首先,借助Sentinel-2A遥感影像数据提取研究区的归一化建筑指数;然后,将其与NTL数据结合得到BANUI(空间分辨率为20 m),并由此得到空间分辨率更高、建筑信息更丰富的BANUI图像;最后,利用分水岭分割算法从BANUI,VANUI和NTL中提取出城市建成区并进行对比。结果表明,由BANUI提取的城市建成区总体精度可达93.61%,Kappa系数为0.7934,用户精度为81.34%,生产者精度为85.34%,提取结果与实际城市建成区的分布较吻合、提取精度较高,且优于另外2种数据。此方法可为NTL在城市建成区提取的研究中提供参考意见,也可用于对城市规划发展的监测。Recently,the utilization of nighttime light data and optical remote sensing images to extract urban built-uPareas has become a research hotspot,and the vegetation adjusted nighttime light data(NTL)urban index(VANUI)is widely used.However,it may easily lead to confusion of buildings and water bodies at the edge of the city,and the spatial resolution is relatively low.Therefore,some improvements were made on this index in this paper,and the building adjusted NTL urban index was proposed.The means was used to extract urban built-uPareas in Baotou City in this paper.Firstly,normalized difference build-uPindex(NDBI)was extracted from Sentinel-2A image data and it was combined with NTL to obtain building adjusted NTL urban index BANUI with the spatial resolution of 20 m,which has higher spatial resolution and more information about the building.Finally,the watershed segmentation algorithm was applied to the extraction of urban built-uParea of Baotou City from BANUI,VANUI and NTL,and the results were comparatively studied.The extraction results show that the overall precision of the urban built-uParea extracted by BANUI could reach 93.61%,the Kappa coefficient is 0.7934,the user accuracy is 81.34%,and the producer accuracy is 85.34%.The extraction results are consistent with the distribution of actual urban built-uParea,and the accuracy is high.The result is better than the area extracted by the other two kinds of data.This method could provide some reference for the study of the extraction of urban built-uParea from NTL,and could also be used to monitor the development of urban planning.

关 键 词：夜间灯光数据 Sentinel-2A数据 城市建成区 BANUI 分水岭分割算法 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]