Multi-source jitter information fusion method of high-resolution optical remote sensing satellite images based on H-infinity filtering

Satellite platform jitter is a key factor that causes distortion in high-resolution satellite remote sensing imagery, which adversely affects applications such as high-precision surveying, mapping and target recognition. Jitter error compensation in remote sensing imagery can be achieved by detection and modelling. This paper proposes a multi-source jitter information fusion method using H-infinity filtering to address uncertainties in inverting absolute jitter error in multispectral Time Delay Integration (TDI) imaging systems and limitations in jitter detection modelling based on reference images. This method effectively combines high-precision jitter error detection results from multispectral TDI imaging with reliable absolute jitter error modelling outcomes from reference images. Unlike the Kalman filter that relies on precise noise statistical assumptions, the H-infinity filtering achieves robust estimation without strict noise distribution requirements, and its inherent robustness enables optimal estimation of time-varying jitter errors. The results of experiments using data from the GF-1 satellite demonstrate that the H-infinity filtering fusion method notably improves the accuracy of jitter detection, with the RMSE of jitter error reduced by up to 69.9%, and the MAE decreased by 55.7% compared with the results of the Kalman filter. Furthermore, the filtering extrapolation model demonstrated robust stability and applicability in predicting long-term time-series jitter, offering a viable solution to enhance geometric quality in high-resolution optical remote sensing imagery.