An Iterative Adaptive Polarization Calibration Method Independent of Corner Reflectors

Polarimetric calibration (PolCal) is essential for the quantitative processing of polarimetric synthetic aperture radar data. Traditional distributed target methods typically require at least one corner reflector (CR) to determine the copolarization (co-pol) channel imbalance. However, the deployment of CRs is costly and impractical in challenging areas, and therefore, advanced PolCal methods that do not rely on CRs have been developed. One widely used method estimates crosstalk and cross-polarization channel imbalance based on volume-dominated pixels. This method estimates co-pol channel imbalance based on Bragg-like pixels, using the Gauss−Newton method according to the unitary zero helix (UZH) constraint and obtains the final co-pol channel imbalance through fitting. However, the selection of initial values and step sizes affects the convergence of the Gauss−Newton method, while the fixed threshold used in fitting impacts the accuracy of sample selection. These issues collectively influence the precision of the calibration results. Therefore, we proposed an iterative adaptive PolCal method to improve the UZH method from a computational perspective. This method first uses an iterative process to obtain a more accurate global initial value for each block's initial input. It then combined the Levenberg−Marquardt algorithm to adjust step sizes and solve for each block. In addition, the variation coefficient was introduced to achieve adaptive sample selection and enhance fitting accuracy. The effectiveness of our proposed method was validated using GF3 02 satellite two-scene calibration field images. The experiments demonstrated that the proposed method not only ensured convergence but also improved the accuracy and stability of PolCal methods without CRs.