A Time-Series Cross-Calibration Method for SAR Data Using Iterative Robust Linear Regression

Time-series calibration is essential for ensuring consistency and comparability across different Synthetic Aperture Radar (SAR) images, which is critical for long-term sequence analysis tasks. However, traditional time-series calibration methods that rely on artificial calibrators encounter significant challenges due to the labor-intensive and time-consuming nature of deploying these calibrators. Cross-calibration offers a promising alternative by using calibrated satellites to calibrate uncalibrated satellites, providing substantial advantages in calibration frequency and timeliness. This paper proposes an advanced time-series calibration method based on the cross-calibration approach. Moreover, since traditional ordinary least squares (OLS) methods for determining calibration coefficients often struggle with outlier interference in scatter-stable pixels, this paper introduces a calibration coefficient computation method based on iterative robust linear regression (IRLR). It iteratively calculates residual control intervals, excludes outliers, and performs robust regression to derive accurate radiometric calibration coefficients. Experimental results from Sentinel-1 data demonstrate the effectiveness of the proposed method. Compared to traditional OLS methods, it offers superior accuracy, and 95.6% of the time-series images show a reduction in Root Mean Square Error (RMSE) by 0.01 to 0.156 dB.