Global Vision Media Focus

The company was featured on CCTV a total of 15 times, and its reach through Party media, central media, and local official media reached a total of 50 million people.

0101

News Categories

In-orbit cross-calibration based on Xiguang-1 01 and Xiguang-1 05 (Tianxianpei) hyperspectral satellites

2025-07-18

Principle introduction

Cross-calibration uses a sensor with higher calibration accuracy as a reference to calibrate the sensor to be calibrated. The principle is to select synchronous or near-synchronous image pairs imaging the same target, and based on the analysis of the matching of the two sensors' spectral responses, observation geometry, and atmospheric parameters, establish the relationship between the digital count values of the two sensor images. The calibration coefficients of the sensor to be calibrated are then solved using the known radiometric calibration coefficients of the reference sensor. Typically, the DN value of a satellite remote sensor has a linear relationship with its entrance pupil radiance (apparent radiance, Top-of-Atmosphere Radiance, TOA): L = gain·DN + offset, where gain and offset are the gain and intercept of the calibration coefficients, respectively, L represents the sensor's entrance pupil radiance, and DN represents the digital count value of the image.

data

	Nishiko No. 1 01 Star	Nishiko No. 1 05 stars (a perfect match)
type	Reference data	Data to be calibrated
Wavelength range (nm)	441-854	431-853
Number of bands	150	72
FWHM（nm）	1.945	9.16
GSD（m）	40	40
Orbital altitude (km)	484	492
Width (km) (@500km)	80	80

process

1. Image selection

1 (2).jpg

Constraints are imposed based on the area range, imaging time, and observation angle. Specifically, the center point distance of each image is within 50km, ensuring a large overlap area. For imaging time differences within one day, the observation angle difference is less than 2°.

2. Region Selection

1 (3).png

Example of sample region of Xiguang-1 01 satellite

1 (4).png

Example of sample area for Xiguang-1 05 satellite (Tianxianpeihao)

1 (5).png

Example of the average radiance curve of the sample area of Xiguang-1 01 satellite

1 (6).png

Example of the average DN value curve of the sample area of Xiguang-1 05 satellite (Tianxianpeihao)

The uniform surface area in the eastern part of the Qaidam Basin was selected as the calibration target area. A total of 16 data sets were ultimately selected. Within each data set, the same region was selected as a sample from the data of Xiguang-1 01 and Xiguang-1 05 satellites (the "Tianxianpei" satellite).

3. Spectral correction of Xiguang-1 01 satellite

The Xiguang-1 01 satellite was spectrally corrected using the 760nm oxygen absorption valley to correct the center wavelength position for subsequent spectral resampling and spectral matching factor (SMF) calculation.

4. Radiation correction of Xiguang-1 01 satellite

Radiometric correction was performed using site calibration coefficients to obtain apparent radiance data for the sample area, and the average radiance value within the area was calculated for each sample.

5. Dark current correction for Xiguang-1 05 satellite (a satellite with a celestial match).

Dark current was calculated using nighttime images of desert areas captured by Xiguang-1 05 satellite (Tianxianpei). The dark current was calculated band by band to correct the original image and obtain the corrected DN value of the sample area. The average DN value of each sample area was calculated.

6. Center wavelength radiance fitting of Xiguang-1 01 satellite to Xiguang-1 05 satellite (a perfect match)

Both the reference and the data to be corrected are hyperspectral data, and their spectral response functions (SRFs) are Gaussian, but their full width at half maximum (FWHM) differ significantly, which can lead to radiometric bias, especially in bands with prominent spectral characteristics (such as absorption bands). Therefore, spectral resampling and spectral matching factor correction are required to compensate for the FWHM discrepancy.

SRF and Spectral Resampling: Gaussian convolution was used to simulate the 9.16 nm SRF of Xiguang-1 05 satellite (also known as Xiguang-1-05) using 1.945 nm data from Xiguang-1 01 satellite, reducing spectral detail loss. Cubic interpolation was also used to match the center wavelength of Xiguang-1 05 satellite (also known as Xiguang-1-05). Next, the SRFs of both satellites were acquired, and different standard ground object spectra were selected. A radiative transfer model was used to perform TOA simulations to calculate SMF and correct for radiative bias.

7. Apparent radiance matching of Xiguang-1 01 satellite and Xiguang-1 05 satellite (Tianxianpei) DN value

1 (7).png

Apparent radiance statistics table

1 (8).png

DN value statistics table

Using SMF-corrected TOA data from Xiguang-1 01 satellite and DN value data from Xiguang-1 05 satellite (also known as Tianxianpei) as the dependent and independent variables for linear fitting, a linear fit was performed, and gain and bias were obtained band by band, totaling 120 samples. The average gain and bias of all samples for each band were calculated as the final gain and bias for that band.

Calibration coefficient verification

1 (9).png

1 (10).png

1 (11).png

1 (12).png

The calibration coefficients of Xiguang-1 05 satellite (Tianxianpeihao) were fitted to perform radiometric and atmospheric corrections, and compared with the TOA and surface reflectance of Xiguang-1 01 satellite to verify the calibration coefficients.

1 (13).png

Comparison of radiance between Xiguang-1 01 and Xiguang-1 05 (the "Tianxianpei" satellite), in W/m²/sr/μm.

1 (14).png

Comparison of reflectivity between Xiguang-1 01 and Xiguang-1 05 (a perfect match) satellites (magnified 10,000 times).

By comparing with the Xiguang-1 01 satellite, the apparent radiance and surface reflectance show the same trend, indicating that the calibration coefficients after cross-calibration are highly feasible.

Example of cross-calibration results image

1 (15).jpg