Urban Impervious Surfaces Estimation from RADARSAT-2 Polarimetric Data Using SVM Method
Urban impervious surface area is a key environmental indicator in assessing urbanizations impacts on urban environmental and ecological conditions,and it has therefore recently attracted more interest from the remote sensing community.The main objective of this investigation is to explore the potential to extract impervious surface coverage in dense urban areas from RADARSAT-2 full PolSAR data.To compare the results,SPOT-5 multispectral imagery is also used.A case study of a dense urban area(Beijing,China)is conducted by applying a support vector machine(SVM)algorithm to SPOT-5 imagery and RADARSAT-2 full PolSAR data.An accuracy assessment is performed using high-resolution WorldView images with a spatial resolution of 0.5 m.The root mean square error(RMSE),the mean absolute error(MAE),and the coefficient of determination(R2)are calculated to validate the accuracy of impervious surfaces derived from the SPOT-5 image and RADARSAT-2 PolSAR data.For the SPOT-5 imagery,the RMSE,MAE,and R2 are 17.03%,12.99%,and 0.7641,respectively.The RADARSAT-2 full PolSAR data yield an RMSE of 13.18%,an MAE of 10.12%,and an R2 of 0.8173.The results indicate that RADARSAT-2 data allow for more accurate impervious surface estimation than SPOT-5.The results also demonstrate that bare soil or water can be easily separated from buildings or asphalt roads using RADARSAT-2 data,which is a difficult task for estimating impervious surfaces with optical remote sensing data.In addition,compared with optical images,impervious surfaces covered by tree crowns can be easily extracted from the RADARSAT-2 PolSAR data in sparse tree cover.Overall,our results demonstrate that PolSAR imagery can provide more efficient and useful polarimetric information and has enormous potential for extracting impervious surfaces.
Xinwu Li Huadong Guo Zhongchang Sun Guozhuang Shen
Center for Earth Observation and Digital Earth,Chinese Academy of Sciences,Beijing 100094,China
国际会议
Progress in Electromagnetics Research Symposium 2011(2011年电磁学研究新进展学术研讨会)
苏州
英文
807-812
2011-09-01（万方平台首次上网日期，不代表论文的发表时间）