Using Full Polarization Signatures to Support Interpretation of SAR Imagery

• Main Authors: Chao-Wu Liu, 劉肇武
• Other Authors: Chung-Ru Ho
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/54975153132290071658

碩士 === 國立臺灣海洋大學 === 海洋環境資訊學系 === 97 === Fully Polarimetric Synthetic Aperture Radar (FPSAR) is the trend in remote sensing applications because its results are more informative compared with single/multi polarization Radars. Based on the features of polarization, the FPSAR image is used with the polarimetric synthesis technique to generate different polarization SAR images, and then extract the unique polarization features of specified object. The information provided is extremely beneficial to the interpretation of SAR images. The FPSAR image of Advanced Land Observing Satellite (ALOS) taken in Sapporo Japan on May 9, 2006 was used as the test material. By utilizing the Circular Polarization Algorithm to alter parameters of orientation angle and ellipcity angle of ALOS image, it generates different polarization images. In addition, by analyzing the similarity and the difference of polarization characteristics resulted from Stokes matrix, this thesis sets up a reasonable benchmark in order to automatically process supervised classification of targets. The result shows that spots which look similar in the radar image may have completely contrast polarization characteristics. While conducting a specified automatic supervised classification on the target, spots with same characteristics can be picked out automatically. When compare these spots with those on the optical image obtained from Google Earth website, the accuracy reaches 90%.