A classification algorithm based on Cloude decomposition model for fully polarimetric SAR image

6th Digital Earth Summit

地球科学;计算机科学

Xiang, Hongmao^1 ; Liu, Shanwei^2 ; Zhuang, Ziqi^2 ; Zhang, Naixin^2

Shandong Province Institude of Land Surveying and Mapping, 2301 Jingshi Road, Jinan

250102, China^1

China University of Petroleum (East China), 66 Changjiang West Road, Qingdao

266580, China^2

关键词: Classification accuracy;    Classification algorithm;    Classification methods;    Classification results;    Fully polarimetric SAR;    polarimetric MAR;    Supervised classification;    Unsupervised classification;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/46/1/012060/pdf
DOI  :  10.1088/1755-1315/46/1/012060

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

Remote sensing is an important technology for monitoring coastal zone, but it is difficult to get effective optical data in cloudy or rainy weather. SAR is an important data source for monitoring the coastal zone because it cannot be restricted in all-weather. Fully polarimetric SAR data is more abundant than single polarization and multi-polarization SAR data. The experiment selected the fully polarimetric SAR image of Radarsat-2, which covered the Yellow River Estuary. In view of the features of the study area, we carried out the H/ α unsupervised classification, the H/ α -Wishart unsupervised classification and the H/ α -Wishart unsupervised classification based on the results of Cloude decomposition. A new classification method is proposed which used the Wishart supervised classification based on the result of H/ α -Wishart unsupervised classification. The experimental results showed that the new method effectively overcome the shortcoming of unsupervised classification and improved the classification accuracy significantly. It was also shown that the classification result of SAR image had the similar precision with that of Landsat-7 image by the same classification method, SAR image had a better precision of water classification due to its sensitivity for water, and Landsat-7 image had a better precision of vegetation types.