Robust Helmert Variance Component Estimation for InSAR DSM Block Adjustment

Digital surface model (DSM) block adjustment is a crucial step to improve absolute height accuracy and consistency within overlap in large-scale topographic mapping using bistatic interferometric SAR. More objects, fixed error models, and more observations types with varying accuracies pose significant challenges for adjustment. To this end, this letter proposes an integration of iterative method, hypothesis testing, and surveying adjustment into the solution. Preconditioning technology and t-test are combined for adaptive correcting of the error model, while robust Helmert variance component estimation (RHVCE) is employed for dynamic update of the stochastic model. Large-scale spaceborne bistatic SAR data containing diverse land cover types are utilized to validate the method. The 90% linear error (LE90) of DSM height against Ice, Cloud and land Elevation Satellite (ICESat)-2 decreases from 26.27 to 4.90 m, corresponding to a decrease from 27.98 to 3.70 m in LE90 of height difference at tie-points (TPs). The results demonstrate not only the ability to obtain the most suitable height error model for each DSM scene but also the determination of appropriate weighting ratio between control point and TP while resisting gross error.