一种用于城市信息提取的改进居民地指数

A study of urban area extraction with the modified human settlement index

区域或者全球尺度上的城市分布信息提取是目前研究的热点与难点。采用DMSP-OLS夜晚灯光数据直接提取城市信息会受到灯光溢出问题的影响,且溢出问题因灯光光斑大小而异,不易定量分析。采用可见光-近红外遥感影像提取城市信息时,多选取植被丰富的地区,避免了裸土对城市信息提取造成的影响,但限制了研究区域的选择。为了解决以上问题,应用DMSP-OLS夜晚灯光数据和可见光-近红外遥感影像,对居民地指数(human settlement index,HSI)进行改进,构建了改进居民地指数(modified human settlement index,MHSI)。采用MHSI对中国和美国的城市进行了提取实验,并利用中国历年城市统计数据和美国NLCD土地覆盖数据集对提取结果进行验证。实验结果表明,MHSI在解决灯光溢出问题的同时,避免了其他地物类型(裸土、水体和植被)对城市信息提取的影响,一次性实现了区域或者全球城市信息的提取,提取精度优于HSI和MODIS土地覆盖数据集。

Urban areas extraction at regional and global scales remains a challenge. To map urban areas using DMSP - OLS nighttime light data is limited due to the saturation of data values, especially in urban cores. Different nighttime facula sizes lead to different degrees of light overflow, which causes difficulty for quantitative analysis. Vegetation - rich areas are selected to avoid the impact of bare soil when visible - near infrared image is used to map urban. To solve the problems above, this paper proposes modified human settlement index （MHSI） on the basis of human settlement index （ HSI）, which is composed of DMSP - OLS nighttime light data and visible - near infrared image. The MHSI has been tested in China and USA and testified by using the China city statistical data and USA NLCD land cover data. The results indicate that MHSI can overcome the overflow problem effectively and discriminate urban areas from other feature types such as bare soil, water and vegetable. MHSI can extract the regional or global city areas completely, and the accuracy is better than that of HSI and MODIS land cover data sets.