Regional urban area extraction using DMSP-OLS data and MODIS data

35th International Symposium on Remote Sensing of Environment

地球科学;生态环境科学

Zhang, X.Y.^1 ; Cai, C.^1 ; Li, P.J.^1

Institute of Remote Sensing and GIS, Peking University, Beijing 100871, China^1

关键词: Classification results;    Defense Meteorological Satellite Programs;    Land surface temperature;    Morphological reconstruction;    One-class support vector machines (OCSVM);    Surface temperatures;    Thresholding techniques;    Vegetation index values;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/17/1/012156/pdf
DOI  :  10.1088/1755-1315/17/1/012156

学科分类：环境科学（综合）

来源: IOP

【 摘 要 】

Stable night lights data from Defense Meteorological Satellite Program (DMSP) Operational Line-scan System (OLS) provide a unique proxy for anthropogenic development. This paper proposed two new methods of extracting regional urban extents using DMSP-OLS data, MODIS NDVI data and Land Surface Temperature (LST) data. MODIS NDVI data were used to reduce the over-glow effect, since urban areas generally have lower vegetation index values than the surrounding areas (e.g. agricultural and forest areas). On the other hand, urban areas generally show higher surface temperatures than the surrounding areas. Since urban area is the only class of interest, a one-class classifier, the One-Class Support Vector Machine (OCSVM), was selected as the classifier. The first method is classification of different data combinations for mapping: (1) OLS data and NDVI data, (2) OLS data and LST data, and (3) OLS data, NDVI data and LST data combined. The second one is a morphological reconstruction based method which combines classification results from OLS plus NDVI data and from OLS plus LST data. In the morphological reconstruction based method, the classification result using OLS and NDVI data was used as a mask image, while the classification result using OLS and LST data was used as a marker image. The north China area covering 14 provinces was selected as study area. Classification results from Landsat TM/ETM+ data from selected areas with different development levels were used as reference data to validate the proposed methods. The results show that the proposed methods effectively reduced the over-glow effect caused by DSMP-OLS data and achieved better results compared to the results from the traditional thresholding technique. The combination of all three datasets produces more accurate results than those of using any two datasets. The proposed morphological reconstruction based method achieves the best result in urban extent mapping.