Full-spectrum in-orbit cross radiation calibration method for micro-nano hyperspectral satellite sensor

Abstract

The invention aims to provide a full-spectrum on-orbit cross radiation calibration method for a micro-nano hyperspectral satellite sensor. The method comprises the following steps of: a, screening a to-be-calibrated satellite-reference satellite image data pair meeting a cross calibration basic condition; b, Gaussian analysis and parameter extraction of the satellite sensor SRF; c, wave band matching and parameter calculation of the sensor SRF; d, matching reference image simulation of a to-be-calibrated satellite sensor spectrum; e, site BRDF simulation and image BRDF correction are carried out; and f, executing cross calibration of the target sensor. The method is applied to the technical field of remote sensing data processing methods and applications.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote sensing data processing methods and application, and particularly relates to a micro-nano hyperspectral satellite sensor full-spectral on-orbit cross-radiometric calibration method, which is suitable for micro-nano hyperspectral satellite on-orbit cross-radiometric calibration and radiometric quality monitoring. BACKGROUND

[0002] Satellite sensor radiometric calibration is an important step in the process of satellite image data preprocessing to establish the mathematical relationship between the digital quantization value (DN) of remote sensing images and the actual physical quantity (TOA radiance), which has a significant impact on the quantitative accuracy of the sensor. According to the satellite launch and operation stage, the sensor radiometric calibration can be divided into pre-launch calibration (such as laboratory calibration, etc.) and post-launch calibration (such as on-board calibration, surrogate calibration, cross calibration, etc.). The latter is a necessary part of satellite sensor radiometric quality monitoring and calibration parameter adjustment. Among them, cross calibration is a method that uses internationally popular high-precision satellite sensors as a reference to shoot stable ground targets (such as PICS) under similar imaging conditions and perform radiometric reference transfer and comparative analysis, thereby realizing sensor quality monitoring and parameter optimization. For the current popular micro-nano satellite sensors, the characteristics of low economic cost and few space-carryable devices make it impossible to effectively deploy on-board calibration devices, while cross calibration can be used as a reference standard for large satellites carrying high-precision imaging instruments and on-board calibration devices, which to some extent makes up for the limitations of micro-nano satellite post-launch calibration, and is more abundant in reference data sources and imaging sites than surrogate calibration.

[0003] Due to the inherent differences in the relative spectral response (SRF) of different satellite sensors, the calculation of the spectral band matching factor (SBAF) between sensors is a common means of simulating the SRF of synchronous sensors in cross calibration. However, for hyperspectral satellite sensors with high spectral resolution, relatively narrow channels, and dimensional differences (such as between dozens of channels and hundreds of channels), the channel settings of the satellite sensor to be calibrated and the reference satellite sensor may not overlap or cover multiple bands, resulting in "single to multiple" (Single to Multiple) situations, etc. Therefore, the traditional "single to single" (Single to Single) band SBAF calculation method for multispectral satellite sensors cannot be used, and a method is needed that can overcome the above problems and achieve synchronization of sensor SRF reference. In addition, spectral synchronization between sensors also needs to consider the influence of the bidirectional reflectance distribution function (BRDF) of the site. Due to site distribution, there are few places where actual BRDF measurements can be carried out, and data applications are not easy. The data accumulation and engineering application effect of cross-radiometric calibration based on this reference are not good, and it is not conducive to the quantitative evaluation of the radiometric quality of the key band.

[0004] Based on the above key technical problems of hyperspectral satellite cross calibration engineering application, a cross calibration technology that can adapt to the spectral channel setting difference between hyperspectral sensors and the TOA BRDF correction of the ground imaging site is crucial for the post-launch radiometric calibration and quality monitoring of micro-nano hyperspectral satellites. SUMMARY

[0005] In view of the cross-radiometric calibration requirements of micro-nano hyperspectral satellites in orbit, the present application provides a micro-nano hyperspectral satellite sensor full-spectral on-orbit cross-radiometric calibration method that meets different dimensional spectral design.

[0006] The technical scheme adopted by the present application is a micro-nano hyperspectral satellite sensor full-spectral on-orbit cross-radiometric calibration method, which comprises the following steps: a. Selecting satellite-reference satellite image data pairs to be calibrated that meet the basic conditions for cross calibration; b. Obtaining the spectral response function (SRF) of the satellite sensor to be calibrated and the reference satellite sensor, substituting it into the Gaussian curve model to evaluate the deviation, matching the SRF of the reference satellite sensor to the satellite sensor to be calibrated based on the principle of linear least squares, and resampling the reference satellite image data to adapt to the spectral channel of the satellite sensor to be calibrated; c. Based on the Ross-Li theoretical model, a field TOA BRDF model is constructed based on the MODIS global coverage data, and key band field BRDF data is generated to correct the synthesized image after adaptation; d. Geometric precise correction and spatial resolution adaptation are performed on the reference satellite sensor synthesized image and the image to be calibrated; e. Based on the radiance of the L1C level synthesized image and the DN quantization value of the original image to be calibrated, the calibration coefficient is calculated by substituting it into the linear equation, and the cross calibration is completed.

[0007] Specifically, in step a, the cross calibration basic conditions include: The site is a desert pseudo-invariant calibration site, and the ground objects are stable; The imaging time is close and there is no cloud layer interference; The angle difference between the satellite imaging sun elevation angle and the observation elevation angle is less than 5°.

[0008] Specifically, in step b, the expression of the Gaussian curve model is: , Where x is the curve wavelength, μ is the center wavelength of the current band, σ is the standard deviation, and S is the scaling factor of the satellite sensor SRF and the theoretical amplitude of the Gaussian curve.

[0009] Specifically, in step b, the specific process of matching the SRF of the reference satellite sensor to the satellite sensor to be calibrated is: Detecting the waveband in which the FWHM wave parameter of the reference satellite sensor is located in the waveband range of the waveband channel of the satellite sensor to be calibrated, and constructing the SRF matrix MAT of the reference satellite sensor REF and the SRF matrix MAT of the satellite sensor to be calibrated TAR , and then obtaining the simulated waveband SRF array STsimu based on W, wherein, , , , In the formula, SR and ST are the SRF values of the reference satellite sensor and the satellite sensor to be calibrated respectively, x in the subscript is the SRF response sample point number of each waveband of the reference satellite sensor under the condition that the FWHM wave parameter of the reference satellite sensor is located in the current waveband of the satellite sensor to be calibrated, y is the waveband number of the reference satellite sensor under the above condition, and W is the response weight array of each waveband of the reference satellite sensor calculated by the least square method.

[0010] Specifically, in the b step, when resampling the reference satellite image data, the simulated image matching the radiation reference of the satellite sensor to be calibrated is generated based on the following formula, , wherein b is , i.e. the waveband matching interval, λ is the waveband number, w λ is the response weight value of the reference satellite sensor waveband corresponding to the current waveband of the satellite sensor to be calibrated, and L λ is the radiation brightness value array of the reference satellite sensor waveband.

[0011] Specifically, in the c step, the expression of the Ross-Li model is: , wherein f iso , f geo , and f vol are the coefficients of the isotropic kernel, the L-R geometric scattering kernel, and the R-T volume scattering kernel respectively, K geo , and K vol are the corresponding kernel functions.

[0012] Specifically, in the c step, the basic data source for constructing the site TOA BRDF model is the global observation TOA reflectivity observation product MYD02HKM and the corresponding observation condition parameter record product MYD03 in the MODIS product, and the kernel function coefficients are solved by screening the product, extracting the parameters, establishing a linear equation system, and substituting into the LSF model.

[0013] Specifically, in the b step, when calculating the extraterrestrial solar spectral irradiance ESSI, the ISSSolar ESSI V2.0 version data published by the International Space Station (ISS) is adopted, and the matched SRF data is integrated and sampled to obtain.

[0014] Specifically, in the d step, the spatial resolution adaptation adopts a manner of downsampling the high-resolution image to adapt to the low-resolution image.

[0015] Specifically, in the e step, the following formula is used when calculating the calibration coefficient , Wherein, L TOA_ref is the standard image converted by the reference satellite sensor, DN ori is the original DN image of the satellite sensor to be calibrated, Gain correct is the updated radiation calibration parameter of the satellite sensor to be calibrated.

[0016] The beneficial effects of the present application are as follows: The method of the present application aims at the "single-multiple" channel difference problem of hyperspectral sensors (high spectral resolution, relatively narrow channels and dimensional difference (such as between dozens of channels and hundreds of channels)). First, the deviation of the sensor spectral response function (SRF) from the ideal distribution is evaluated through the Gaussian curve model to realize accurate simulation of the SRF of the reference satellite sensor which does not provide detailed SRF data. Then, based on the principle of linear least squares (LSF), the SRF matrix of the reference satellite sensor and the SRF matrix of the satellite sensor to be calibrated are constructed, the response weight array W is calculated, and the simulated SRF matching the waveband of the satellite sensor to be calibrated is generated. Further, the "multi-band weighted fitting" replaces the traditional multi-spectral satellite "single-single" waveband matching, which can adapt multiple hyperspectral wavebands of the reference satellite sensor to a single waveband of the satellite sensor to be calibrated, effectively solve the spectral synchronization problem caused by the non-overlapping or multiple coverage of hyperspectral "single-multiple" channels, and ensure the accurate transfer of the radiation reference between different spectral dimension sensors, and improve the spectral synchronization accuracy.

[0017] To solve the influence of BRDF (Bidirectional Reflectance Distribution Function) on the calibration result, the method of the present application is based on the Ross-Li theoretical model, takes MODIS MYD02HKM (TOA reflectivity) and MYD03 (observation parameters) as the basic data source, extracts the solar / satellite zenith angle, azimuth angle and TOA reflectivity, constructs the kernel function matrix and reflectivity matrix, then uses LSF to solve the kernel function coefficient, and finally generates the field TOA BRDF model; then without relying on the scarce field BRDF measurement data (the traditional method is limited by the field), the BRDF simulation of any PICS field can be realized through the global coverage of MODIS data; by screening the to-be-corrected waveband with FWHM overlapping more than 50%, the image BRDF correction is completed, the radiation distortion caused by the difference of illumination-observation angle is effectively eliminated, the calibration result is more in line with the real object radiation characteristics, and the reliability of the key waveband radiation quality evaluation is improved.

[0018] The method of the present application is based on on-orbit image data processing, and does not rely on the on-board calibration device which is difficult for micro / nano satellites to carry: firstly, the image pair (satisfying the conditions of stable ground objects, similar time / angle) is obtained through screening historical data or cooperative shooting, avoiding hardware modification; then, the complete transfer of the radiation reference is realized through the steps of DN value conversion to radiation brightness, ESSI parameter accurate sampling, TOA reflectivity / radiance conversion, etc.; finally, the calibration coefficient is calculated after ensuring the spatial consistency through geometric fine correction and high-resolution image down-sampling adaptation. The present application solves the hardware limitation of micro / nano satellites "low cost, less carried device" in a targeted manner, without additional investment in on-board calibration equipment cost; at the same time, each step of data processing is designed around "reducing error sources", and the calibration coefficient calculated through linear equation can accurately establish the relationship between DN value and TOA radiance, reducing the implementation threshold while ensuring the calibration accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a simple flowchart of the method of the present application; Figure 2 is a comparison of the spectral response characteristics RSR of multispectral satellite sensors (a, taking Landsat-8 OLI and Sentinel-2A MSI as examples) and hyperspectral satellite sensors (b, taking OHS-3C CMOS1 and EnMAP as examples), and the display range is limited between 400-1000nm and 400-600nm respectively; Figure 3 is a simulated image map generated by matching the wave spectrum channels of the reference satellite sensor with the satellite sensor to be calibrated; Figure 4 is the comparison result of the image and spectral characteristics before and after the simulated image is generated; Figure 5 is a channel setting diagram of the field TOA BRDF simulation solving matrix. Detailed Implementation

[0020] Terminology definition: SRF: Spectral response function of a satellite sensor; LSF: Linear Least Squares; DN: Digital quantization value of remote sensing image; TOAradiance: A real physical quantity (radiance of the top layer of the atmosphere). SBAF: Spectral band matching factor between sensors; BRDF: Site bidirectional reflectance distribution function; MODIS: Medium Resolution Imaging Spectrometer; PICS: Pseudo-invariant calibration field, refers to a natural or artificial target area whose surface reflectance remains highly stable over a long period of time, providing a benchmark for radiometric calibration; ESSI: Exoatmospheric solar spectral irradiance; FWHM: Full Width at Half Height (FWHM) of a band.

[0021] like Figures 1-4 As shown, this invention provides an on-orbit cross-radiometric calibration method for the entire spectrum of a micro / nano hyperspectral satellite sensor, which includes the following steps: a. Select satellite-reference satellite imagery data pairs that meet the basic conditions for cross-calibration; b. Obtain the spectral response function (SRF) of the satellite sensor to be calibrated and the reference satellite sensor, substitute it into the Gaussian curve model to evaluate the deviation, match the SRF of the reference satellite sensor to the satellite sensor to be calibrated based on the principle of linear least squares, and resample the reference satellite image data to the spectral channel of the satellite sensor to be calibrated. c. Based on MODIS global coverage data, construct a site TOA BRDF model based on the Ross-Li theoretical model, generate key band site BRDF data, and perform BRDF correction on the adapted synthetic image. d. Perform geometrical fine correction and spatial resolution adaptation on the composite image from the reference satellite sensor and the image to be calibrated; e. Based on the radiance of the L1C-level composite image and the DN quantization value of the original image to be calibrated, the calibration coefficients are calculated by substituting them into the linear equation to complete the cross-calibration.

[0022] Specifically, the present application takes China OHS-3B CMOS2 (400-1000 nm, 32 bands) and Germany EnMAP HSI sensor (400-2500 nm, 224 bands, actually intercept 400-1000 nm, i.e. 101 bands) as examples of the satellite sensor to be calibrated and the reference satellite sensor, respectively, to implement and test the above-mentioned spectral matching resampling method and related processing, and the specific steps are as follows: (I) Construction of satellite inter-orbit imaging data pairs Step S1: The following requirements are imposed on the imaging conditions of the cross-calibration of the in-orbit satellites: ① The PICS stable ground target should be selected, such as a desert, to eliminate the influence of changes in the properties of ground objects; ② The imaging time and weather should be close to each other, and cloud interference should be excluded to ensure that the ground target receives close-to-constant solar energy; and ③ The imaging geometric angles should be close to each other, and the solar zenith angle (SZA) and the viewing zenith angle (VZA) of satellite imaging should be close to each other, with a difference of less than 5°.

[0023] Since some satellites do not continuously maintain the orbit according to the standard orbit model, the historical data of the satellite to be calibrated and the reference satellite at the PICS or other sites can be selected, or satellite image data pairs can be obtained through collaborative shooting.

[0024] (II) Gaussian analysis of satellite sensor SRF and parameter extraction Step S2: Preprocessing of sensor SRF data. For satellite sensors that have provided detailed SRF data and center wavelength and FWHM parameters, the simulation process in this step can be skipped. For satellite sensors such as the EnMAP satellite sensor, which do not directly provide detailed SRF data but provide basic parameters, the basic data need to be input to simulate the SRF curve under the premise of verifying the imaging quality and stability advantage of the sensor. The theoretical formula of the Gaussian curve distribution of the satellite sensor is as follows: , where x is the curve wavelength, μ is the center wavelength of the current band, σ is the standard deviation, and S is the scaling factor of the satellite sensor SRF and the theoretical amplitude of the Gaussian curve.

[0025] (III) Band matching and parameter calculation of sensor SRF Step S3: Reference satellite sensor SRF synthesis matching. Taking the band channels of the satellite sensor to be calibrated as the benchmark, the bands in which the FWHM of the reference satellite sensor is located within the full wavelength range of each band are detected, and the SRF data MAT TAR of the current band of the satellite sensor to be calibrated is resampled based on the spectral resolution. After resampling, the SRF matrix MAT REFThe present application adopts a SRF synthesis synchronization method based on the least square method to fit the SRF of the current waveband reference satellite sensor and the satellite sensor to be calibrated. The detailed formula of this step is as follows: , , , Wherein, SR, ST are the SRF values of the reference satellite sensor and the satellite sensor to be calibrated respectively, the subscript x is the SRF response sample point number of each waveband of the reference satellite sensor under the condition that the FWHM of the reference satellite sensor is located in the current waveband of the satellite sensor to be calibrated, the subscript y is the waveband number of the reference satellite sensor meeting the above condition, and W is the response weight array of each waveband of the reference satellite sensor calculated by the least square method. As shown in the following formula: Figure 2 In actual execution, the FWHM of several high spectral resolution continuous wavebands of the reference satellite sensor should be completely contained in the FWHM spectrum of the corresponding interval waveband of the satellite sensor to be calibrated to be matched as a matching waveband. The formula is as follows:

[0026] Wherein, FWHM represents the full width of the half maximum of the waveband, B ref represents the reference waveband, B tar represents the target waveband, b represents the matched waveband, and Band valid represents the matched waveband set.

[0027] After obtaining W, the SRF array ST of the reference satellite sensor corresponding to the current waveband of the satellite sensor to be calibrated is calculated according to the following formula: simu .

[0028]

[0029] Step S4: Sensor waveband ESSI parameter calculation. After the spectrum matching of the reference satellite sensor SRF in the foregoing steps is completed, the atmospheric layer outside solar spectral irradiance (ESSI, unit W / (m 2 * μm) ) of each waveband needs to be calculated for the subsequent calculation of the atmospheric top layer reflectivity (TOA reflectance). According to the ESSI data collected and used by the international industry in recent years, the present application selects the ISS Solar ESSI V2.0 version of ESSI data collected by the international space station (ISS) and released in 2020 as the basic data for the sampling calculation of the satellite sensor waveband ESSI, and based on the SRF data integral sampling after the initial SRF spectrum matching of the reference satellite sensor, the ESSI parameters before and after the matching of each waveband are obtained. The calculation formula is as follows:

[0030] where SRF(λ) is the SRF response data of the current band, λ1, λ2 are the start and end positions of the current band spectral range, ESSI value is the value of ISS Solar ESSI V2.0 data within its range.

[0031] (Four) Reference image simulation matching the spectrum of the satellite sensor to be calibrated Step S5: L1 level preprocessing of reference satellite sensor image data. For the reference satellite sensor image whose DN value has been assigned, it is necessary to convert the L1 level image data into radiance (L λ , unit W / (m 2 *μm*sr), and the calculation process is shown in the following formula:

[0032] where Gain and Offset are the gain value and offset value of the band respectively (the offset value of OHS sensor image data is 0). If the initial data product of the reference satellite sensor has been assigned a radiance value, the above calculation is not required. In addition, this step extracts the imaging time, imaging range and central coordinates, solar elevation angle / azimuth angle, satellite observation elevation angle / azimuth angle and other parameters from the metafile of the reference satellite sensor data product, which are used for subsequent TOA reflectance calculation and cross-calibration processing.

[0033] Step S6: Generation of simulated images matching the satellite sensor to be calibrated. After performing the least squares fitting calculation in step S3, the effective band number and response weight value of the reference satellite sensor corresponding to each band of the satellite sensor to be calibrated are obtained, which are used as the basis for matching and simulating the L1 level image data of the reference satellite sensor. Based on the L1 level radiance image of the reference satellite sensor as the basic data, the simulated image of the reference satellite sensor matching the radiometric reference of the satellite sensor to be calibrated is calculated and generated according to the following formula.

[0034] , where b is , i.e. the band matching interval, λ is the band number, w λ is the contribution response weight value of the reference satellite sensor band corresponding to the current band of the satellite sensor to be calibrated, L λ is the radiance value array of the reference satellite sensor band.

[0035] Step S7: Calculation of TOA reflectance of the simulated image. Since TOA reflectance can eliminate the influence of the cosine of the solar zenith angle, it is usually converted into TOA reflectance in cross-calibration. The calculation method is shown in the following formula: , Where L is a reference satellite sensor radiance value array (unit: W / (m²)). 2 *μm*sr), ESSI is the exoatmospheric solar irradiance parameter for the current wavelength band (unit: W / (m²)). 2 *μm)), θ ele d is the solar altitude angle corresponding to the satellite sensor imaging time, and d is the Earth-Sun distance (in AU) corresponding to the imaging time. This invention calculates the Earth-Sun distance with reference to the following formula: , , , In this context, Year, Month, and Date represent the year, month, and date of the imaging, respectively, while DOY is the sequence number of the satellite sensor imaging date within the entire year.

[0036] Furthermore, the ESSI parameters and L calculated in step S5 before spectral matching of the reference satellite sensor are... λ Substituting into formula (9), the initial TOA reflectivity image data of the reference satellite sensor is calculated.

[0037] Taking the area surrounding the La Crau calibration field in France from the RadCalNet global calibration dataset as an example, a comparison is made between the images after matching simulation and the images before unmatched simulation, as well as the site-measured TOA reflectance. Figure 3 As shown, the results indicate that the simulated images of different types of ground features maintain the same reflectance curve characteristics as the initial TOA images. Moreover, the quantitative differences in reflectance values ​​at similar wavelength positions are within the expected range of sensor differences. In other words, the spectral matching and simulated images achieve radiation reference transmission and spectral matching conversion from the reference satellite sensor to the satellite sensor to be calibrated while maintaining the reflectance characteristics of ground features.

[0038] (v) Site BRDF simulation and image BRDF correction Step S8: Fitting the site BRDF model parameters. BRDF correction helps eliminate the influence of illumination-observed angle deviation on radiometric distortion in cross-calibrated image data, improving calibration accuracy. However, since it is difficult to conduct routine field BRDF measurements at the global PICS calibration sites selected for cross-calibration, a method is needed to simulate the site BRDF model using existing data instead of relying on field measurements. This invention adopts a method based on the Ross-Li model to construct a site BRDF model within a specified time period at the TOA level, thereby reducing the impact of radiative transfer model errors while fitting the BRDF model of the cross-calibrated image for the corresponding time period.

[0039] The Ross-Li model, also known as the MODIS BRDF model, consists of three kernel functions: , Where iso, geo, and vol represent the isotropic kernel (default is 1), LR geometric scattering kernel, and RT volumetric scattering kernel of the Ross-Li model, respectively; f corresponds to the coefficients of the three kernel functions, which are the model parameters to be fitted; and K... geo K vol These are the corresponding kernel functions; ρ BRDF This represents a composite / uncalibrated image that has completed BRDF correction.

[0040] The data source selected for fitting in this invention is the global TOA reflectance observation result product MYD02HKM and its corresponding observation condition parameter record product MYD03 (spatial resolution of 500 m, hereinafter referred to as "MYDTOA reflectance product") from the MODIS series products. The actually usable bands are the RGB and NIR bands. For the remaining bands where BRDF correction cannot be performed, on the one hand, when selecting image data pairs of reference satellite sensors and satellite sensors to be calibrated, we can try to ensure that the solar / satellite geometry is close. On the other hand, we can choose PICS with small ground undulations as the calibration field to eliminate the influence of BRDF as much as possible. The bands for which BRDF correction is performed can be used for quantitative assessment of radiation quality.

[0041] First, based on the input time, latitude, and longitude, the corresponding product list is filtered, downloaded, and stored locally, and a local search index function is developed and deployed. Then, parameters such as solar zenith angle / azimuth angle, satellite zenith angle / azimuth angle, and TOA reflectivity required for fitting the Ross-Li model are extracted from the corresponding products. A system of linear equations for that time period is established and then substituted into the LSF (least squares) model to solve for f. iso f geo and f vol Solving the matrix channel structure is as follows: Figure 4 As shown, the mathematical process is illustrated in the following formula: , , , , where MAT Kernel is the kernel matrix calculated based on the illumination-geometry parameters in the cross-calibration area extracted from N MYD TOA reflectance products, K geo represents the geometric optical kernel value calculated by the sun and satellite angles, K vol represents the volume scattering kernel value calculated by the observation geometry; MAT TOA is the reflectance result matrix of N MYD TOA reflectance products, Ref TOA represents the reflectance value; F local is the LSF fitting result matrix, f iso represents the isotropic scattering component, f geo represents the geometric optical scattering component, f vol represents the volume scattering component. The above process is batch calculated for each pixel in the cross-calibration area, and the spatial resolution is 500 m.

[0042] Step S9: Synthesis of the image and the BRDF correction of the image to be calibrated. The synthesis of the image and the image to be calibrated is converted into an array, and is spatially matched with the BRDF model array. Since the spatial resolution of the synthesis of the image and the image to be calibrated adopted by the present application is better than that of the BRDF model array, the assignment result of the corresponding BRDF correction value of the pixel is realized by the method of positioning the image main body in the BRDF array pixel. The matching of the BRDF to be corrected band of the synthesis of the image and the image to be calibrated adopts the method of selecting the FWHM overlap more than one half (as shown in the following formula), , where FWHM represents the band full width at half maximum, B Sat represents the synthesis of the image band, B BRDF represents the BRDF model array, b represents the matched band, and Band Sat represents the matched band set.

[0043] Then, the correction is completed according to the following (six).

[0044] (Six) Cross-calibration execution of the target sensor Step S10: geometric matching between image data pairs. After the simulation of spectral matching and the conventional band BRDF correction, the TOA image data of the reference satellite sensor is converted into TOA radiance image data, and different data may have spatial bias, so the geospatial matching under visual conditions should be performed first, so that the reference satellite sensor image and the satellite sensor image to be calibrated are basically overlapped in space, and then the high-resolution image is aligned and valued based on the low-resolution image data pixel. For OHS and similar sensors, the following formula is substituted:

[0045] wherein, L TOA_ref is the standard image converted from the reference satellite sensor, DN ori is the original DN image of the satellite sensor to be calibrated, Gain correct is the updated radiometric calibration parameter of the satellite sensor to be calibrated.

[0046] The application solves the technical problems of the adaptation of the hyperspectral sensor'single-multi' channel and the dependence of the BRDF measured data, improves the accuracy and engineering applicability of the micro-nano hyperspectral satellite in-orbit radiometric calibration, and can be widely applied to the in-orbit radiometric calibration and radiometric quality monitoring of the micro-nano hyperspectral satellite.

[0047] Finally, it should be emphasized that the above description is only the preferred embodiment of the application and is not used to limit the application. For those skilled in the art, the application can have various changes and modifications, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application should be included in the protection scope of the application.

Claims

1. A method for on-orbit cross-radiometric calibration of a micro / nano hyperspectral satellite sensor across the entire spectrum, characterized in that, The method includes the following steps: a. Select satellite-reference satellite imagery data pairs that meet the basic conditions for cross-calibration; b. Obtain the spectral response function (SRF) of the satellite sensor to be calibrated and the reference satellite sensor, substitute it into the Gaussian curve model to evaluate the deviation, match the SRF of the reference satellite sensor to the satellite sensor to be calibrated based on the principle of linear least squares, and resample the reference satellite image data to the spectral channel of the satellite sensor to be calibrated. c. Based on MODIS global coverage data, construct a site TOA BRDF model based on the Ross-Li theoretical model, generate key band site BRDF data, and perform BRDF correction on the adapted synthetic image. d. Perform geometrical fine correction and spatial resolution adaptation on the composite image from the reference satellite sensor and the image to be calibrated; e. Based on the radiance of the L1C-level composite image and the DN quantization value of the original image to be calibrated, the calibration coefficients are calculated by substituting them into the linear equation to complete the cross-calibration.

2. The method for full-spectrum on-orbit cross-radiometric calibration of a micro / nano hyperspectral satellite sensor according to claim 1, characterized in that, In step a, the basic conditions for cross-calibration include: The site is a desert-type pseudo-invariant calibration field with stable terrain features; The imaging times were similar and there was no cloud interference; The difference between the solar elevation angle and the observation elevation angle in satellite imaging is less than 5°.

3. The method for full-spectrum on-orbit cross-radiometric calibration of a micro / nano hyperspectral satellite sensor according to claim 1, characterized in that, In step b, the expression for the Gaussian curve model is: ， Where x is the curve wavelength, μ is the center wavelength of the current band, σ is the standard deviation, and S is the scaling factor of the satellite sensor SRF and the theoretical amplitude of the Gaussian curve.

4. The method for full-spectrum on-orbit cross-radiometric calibration of a micro / nano hyperspectral satellite sensor according to claim 1, characterized in that, In step b, the specific process of matching the SRF of the reference satellite sensor to the satellite sensor to be calibrated is as follows: Using the band channels of the satellite sensor to be calibrated as a reference, the bands in the reference satellite sensor whose overall FWHM spectral parameters fall within that band range are detected, and the SRF matrix MAT of the reference satellite sensor is constructed. REF SRF matrix MAT of satellite sensors to be calibrated TAR The response weight array W is calculated using the following formula, and then the simulated band SRF array STsimu is obtained based on W, where, ， ， ， In the formula, SR and ST are the SRF values ​​of the reference satellite sensor and the satellite sensor to be calibrated, respectively. The subscript x is the number of SRF response sample points of each band of the reference satellite sensor under the condition that the overall FWHM spectral parameters of the reference satellite sensor are within the current band of the satellite sensor to be calibrated. The subscript y is the number of bands of the reference satellite sensor that meet the above conditions. W is the response weight array of each band of the reference satellite sensor calculated by the least squares method.

5. The method for full-spectrum on-orbit cross-radiometric calibration of a micro / nano hyperspectral satellite sensor according to claim 1, characterized in that, In step b, when reacquiring reference satellite imagery data, a simulated image matching the radiometric reference of the sensor to be calibrated is generated based on the following formula. ， Where b is That is, the band matching interval, where λ is the band number and w λ L represents the weighted value of the contribution response of the reference satellite sensor band corresponding to the current band of the satellite sensor to be calibrated. λ This is an array of radiance values ​​for the reference satellite sensor band.

6. The full-spectrum on-orbit cross-radiometric calibration method for a micro / nano hyperspectral satellite sensor according to claim 1, characterized in that, In step c, the expression of the Ross-Li theoretical model is: ， Where f iso f geo f vol K represents the coefficients of the isotropic kernel, the LR geometric scattering kernel, and the RT volumetric scattering kernel, respectively. geo K vol These are the corresponding kernel functions.

7. The method for full-spectrum on-orbit cross-radiometric calibration of a micro / nano hyperspectral satellite sensor according to claim 1, characterized in that, In step c, the basic data source for constructing the site TOA BRDF model is the global observation TOA reflectance observation result product MYD02HKM and its corresponding observation condition parameter record product MYD03 in MODIS products. The kernel function coefficients are solved by screening products, extracting parameters, establishing a system of linear equations, and substituting them into the LSF model.

8. The method for full-spectrum on-orbit cross-radiometric calibration of a micro / nano hyperspectral satellite sensor according to claim 1, characterized in that, In step b, when calculating the extra-atmospheric solar spectral irradiance (ESSI), the data from the International Space Station (ISS) ISSolarESSIV 2.0 version is used, and the data is obtained by integral sampling based on the matched SRF data.

9. The method for full-spectrum on-orbit cross-radiometric calibration of a micro / nano hyperspectral satellite sensor according to claim 1, characterized in that, In step d, spatial resolution adaptation is achieved by downsampling the high-resolution image to adapt to the low-resolution image.

10. The method for full-spectrum on-orbit cross-radiometric calibration of a micro / nano hyperspectral satellite sensor according to claim 1, characterized in that, In step e, the following formula is used to calculate the scaling factors. ， Among them, L TOA_ref Standard imagery derived from reference satellite sensors, DN ori These are raw DN images from satellite sensors awaiting calibration, Gain correct For updating the radiometric calibration parameters of the satellite sensors to be calibrated.

Images