As the reference radiometric calibration standard of sensors on the Haiyang-1C (HY-1C) satellite platform, the satellite calibration spectrometer (SCS) is equipped with an onboard calibration system composed of double solar diffusers and an erbium-doped diffuser to monitor the postlaunch radiometric response change. Herein, through onboard calibration data analysis, the calibration diffuser performance remains stable without degradation, and the Moderate Resolution Imaging Spectroradiometer (MODIS) on Terra is adopted as a reference to repeatedly verify onboard radiometric calibration results by selecting different dates and reflectance scenes. The SCS equivalent reflectance is obtained by combining the mean digital number (DN) of the SCS crossing area image with the radiometric calibration coefficient. The spectral reflectance is obtained via interpolation and iteration, which is adopted as the actual MODIS incident pupil spectral reflectance because the small imaging time interval can be ignored and almost vertically observed, and it is convoluted with the MODIS spectral response function to obtain the predicted equivalent reflectance. Validation is completed by comparing the predicted MODIS equivalent reflectance to the measured value based on the onboard calibration coefficient. The results show that (1) the difference between the measured and predicted MODIS band equivalent reflectance is between -0.00466 and 0.0039, and (2) the percentage difference between the measured and predicted MODIS band equivalent reflectance ranges from 4.17% and 1.24%, indicating that the calibration system carried on HY-1C can perform high-precision SCS radiometric calibration, meeting the cross-calibration accuracy requirements of other loads on the same platform.

To monitor the spectral position drift, expansion and contraction of the full width at half maximum (FWHM) of the satellite calibration spectrometer (SCS) of the HY-1C satellite during on-orbit operation, an onboard spectral calibration method based on a wavelength diffuser is proposed in this paper. This method uses the wavelength diffuser reflectance measured prelaunch as the standard spectrum, convolves it with the spectral response function of the SCS to obtain a reference spectrum, uses the measured data of the onboard SCS as the measured spectrum, and obtains the spectral drift and variation of the FWHM through spectral line matching. Generally, the spectral response function of a hyperspectral remote sensor follows a Gaussian model, and so does that of the SCS. The spectral calibration results obtained based on the onboard wavelength diffuser are validated and evaluated in comparison to calibration based on an oxygen absorption line. Preliminary results show that (1) the SCS spectral drift is negative, indicating a shift in the shortwave direction, and its absolute value is gradually decreasing with increasing on-orbit operation time; (2) the mean values of the central wavelength and FWHM errors between the two calibration methods are 0.08 nm and 0.20 nm, respectively, indicating that the spectral calibration method based on the wavelength diffuser has high accuracy and reliability; and (3) the SCS spaceborne spectral calibration error has the greatest impact on radiometric calibration in Band 18, with an uncertainty of 0.99%, while the uncertainty in the other bands is less than 0.33%, indicating that the spectral calibration uncertainty meets radiometric calibration accuracy requirements.

<p style='text-indent:20px;'>Satellite networking, as the future development direction of aero-space, requires high-precision autonomous fault diagnosis capability for a single satellite. In this paper, aiming at the characteristics of closed-loop fault propagation and high data dimensionality of spacecraft control system, neural network algorithms are conducted to study the fault diagnosis of spacecraft high-dimensional coupled data. Based on the ground test data of a certain spacecraft, this paper converts the high-dimensional sequence data into grayscale images, and then uses Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) to diagnose them respectively. The effectiveness of the methods in this paper is illustrated by comparing and validating them with three non-image-based machine learning algorithms, namely, K-NearestNeighbor, Bayesian classifier, and K-NearestNeighbor based on Principal Component Analysis.</p>