A Three-Hierarchy Evaluation of Polarimetric Performance of GF-3, Compared with ALOS-2/PALSAR-2 and RADARSAT-2

2019 
GaoFen-3 (GF-3) is the first Chinese civilian multi-polarization synthetic aperture radar (SAR) satellite, launched on 10 August of 2016, and put into operation at the end of January 2017. The polarimetric SAR (PolSAR) system of GF-3 is able to provide quad-polarization (quad-pol) images in a variety of geophysical research and applications. However, this ability increases the complexity of maintaining image quality and calibration. As a result, to evaluate the quality of polarimetric data, polarimetric signatures are necessary to guarantee accuracy. Compared with some other operational space-borne PolSAR systems, such as ALOS-2/PALSAR-2 (ALOS-2) and RADARSAT-2, GF-3 has less reported calibration and image quality files, forcing users to validate the quality of polarimetric imagery of GF-3 before quantitative applications. In this study, without the validation data obtained from a calibration infrastructure, an innovative, three-hierarchy strategy was proposed to assess PolSAR data quality, in which the performance of GF-3 data was evaluated with ALOS-2 and RADARSAT-2 data as references. Experimental results suggested that: (1) PolSAR data of GF-3 satisfied backscatter reciprocity, similar with that of RADARSAT-2; (2) most of the GF-3 PolSAR images had no signs of polarimetric distortion affecting decomposition, and the system of GF-3 may have been improved around May 2017; and (3) the classification accuracy of GF-3 varied from 75.0% to 91.4% because of changing image-acquiring situations. In conclusion, the proposed three-hierarchy approach has the ability to evaluate polarimetric performance. It proved that the residual polarimetric distortion of calibrated GF-3 PolSAR data remained at an insignificant level, with reference to that of ALOS-2 and RADARSAT-2, and imposed no significant impact on the polarimetric decomposition components and classification accuracy.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    40
    References
    2
    Citations
    NaN
    KQI
    []