A Recalibration Method for Historical SAR Data Through Inter-Satellite Alignment

Recalibration of historical Synthetic Aperture Radar (SAR) data facilitates the adjustment of data that have not undergone radiometric calibration or have suboptimal calibration accuracy, ensuring consistency among diverse historical images. This consistency is crucial for long-term sequential remote sensing quantitative applications. Traditional calibration methods based on artificial calibrators require the presence of such calibrators in historical data, a condition often unfulfilled, rendering the recalibration process impractical. In contrast, cross-calibration methods based on natural scenes only necessitate contemporaneous and co-located calibrated satellite data to recalibrate historical datasets. However, existing cross-calibration methods for historical data recalibration typically utilize data from a single SAR sensor. This approach inadequately corrects system errors within the SAR sensor, resulting in low calibration accuracy. To address this issue, this paper proposes a radiometric recalibration method for historical SAR data based on inter-satellite data alignment. Initially, suitable data from different SAR satellite sensors operating in the same frequency band are selected. Subsequently, a resampling method is employed to construct cross-calibration data pairs with consistent resolution. Finally, the recalibration coefficients are fitted using a Gaussian Mixture Model (GMM) to mitigate the impact of anomalous interference values. Experiment results on Sentinel-1A and GF-3A data demonstrated the effectiveness of the proposed method. Compared to traditional fitting methods, the proposed method reduced calibration errors in saline-alkali and bare land areas by 0.82 dB and 0.77 dB, respectively.