Hyperspectral remote sensor cross calibration method based on dynamic Gaussian distribution

Abstract

The invention particularly relates to a hyperspectral remote sensor cross calibration method based on dynamic Gaussian distribution, and solves the technical problems that the accuracy and precision of a spectrum matching factor obtained by an existing spectrum matching factor calculation method are relatively low, and the cross calibration precision of an interference type hyperspectral remote sensor is difficult to improve. The method comprises the following steps: 1, determining a cross calibration test area; 2, calculating the simulated spectral radiance; 3, calculating spectrum matching factors of a plurality of spectrum channels which greatly influence the to-be-calibrated remote sensor A in the to-be-calibrated remote sensor B based on dynamic Gaussian distribution; 4, obtaining a final spectrum matching factor of each spectrum channel of the to-be-calibrated remote sensor A; 5, calculating the real spectral radiance of a plurality of spectral channels which have great influence on the to-be-calibrated remote sensor A in the to-be-calibrated remote sensor B; 6, calculating the real spectral radiance of the remote sensor A to be calibrated; and 7, calculating the radiometric calibration coefficient of each spectrum channel of the to-be-calibrated remote sensor A, and completing the cross calibration of the hyperspectral remote sensor.

Description

Technical Field

[0001] The invention specifically relates to a hyperspectral remote sensor cross-calibration method based on dynamic Gaussian distribution. Background Art

[0002] Interferometric hyperspectral remote sensors have become one of the irreplaceable payloads in the aerospace field with their unique advantages such as multi-channel, high resolution, and high signal-to-noise ratio. The hyperspectral data collected by this type of remote sensor has the characteristics of equal wavenumber distribution in the interferometric inversion process, which also leads to uneven distribution of central wavelengths of each spectral channel and inconsistent spectral resolution of each spectral channel. At present, on-orbit radiation calibration mostly uses a method based on spectral channel matching for cross-calibration, calculating the spectral matching factor between the channels of the remote sensor to be calibrated and the reference remote sensor, so as to obtain the calibration coefficients of each spectral channel of the remote sensor to be calibrated.

[0003] The spectral matching factor is a parameter that quantifies the degree of matching between the spectral channels of two remote sensors. Its accuracy can be used to evaluate the accuracy and effect of cross-calibration. However, the current calculation method of the spectral matching factor still has certain defects, especially for interferometric hyperspectral remote sensors. When it is used as a remote sensor to be calibrated, the central wavelength distribution of each spectral channel is uneven and the range of the spectral response function is inconsistent, resulting in large errors in the acquisition of the spectral matching factor. Therefore, how to improve the accuracy and precision of the spectral matching factor, and then improve the cross-calibration accuracy of interferometric hyperspectral remote sensors, is an urgent problem to be solved in the field of on-orbit radiation calibration. Summary of the invention

[0004] The purpose of the present invention is to solve the technical problem that the spectral matching factors obtained by the existing spectral matching factor calculation method have low accuracy and precision, and it is difficult to improve the cross-calibration accuracy of interferometric hyperspectral remote sensor, and provide a hyperspectral remote sensor cross-calibration method based on dynamic Gaussian distribution.

[0005] In order to achieve the above object, the present invention adopts the following technical solution:

[0006] A hyperspectral remote sensor cross calibration method based on dynamic Gaussian distribution, which is special in that it includes the following steps:

[0007] Step 1: determine the cross-calibration test area, obtain the atmospheric parameters, surface reflectivity, and observation geometry parameters of the to-be-calibrated remote sensor A and the reference remote sensor B in the test area;

[0008] Step 2, obtaining the spectral channel characteristics of the remote sensor A to be calibrated and the reference remote sensor B; the spectral channel characteristics include the central wavelength, the spectral channel and the spectral response function; according to the atmospheric parameters of the test area, the surface reflectivity, and the observation geometric condition parameters of the remote sensor A to be calibrated and the reference remote sensor B obtained in step 1, respectively simulate the apparent radiance of the remote sensor A to be calibrated and the reference remote sensor B, and calculate the simulated spectral radiance of the remote sensor A to be calibrated and the reference remote sensor B after the response of their respective spectral channels according to the apparent radiance and the spectral channel characteristics;

[0009] Step 3: According to the simulated spectral radiance obtained in step 2, the spectral matching factors SBAF of multiple spectral channels in the remote sensor B to be calibrated that have a greater impact on the remote sensor A to be calibrated are calculated based on the dynamic Gaussian distribution. i,G :

[0010]

[0011] In the formula, R A (λ i ) is the simulated spectral radiance of the remote sensor A to be calibrated after responding to the i-th spectral channel, R B (λ1), R B (λ2),…,R B (λ G ) are the simulated spectral radiances of the reference remote sensor B after responding to the 1st, 2nd, …, G spectral channels respectively;

[0012] Step 4: assign a weight coefficient to the spectral matching factor obtained in step 3, and weight it to obtain the final spectral matching factor of each spectral channel of the remote sensor A to be calibrated;

[0013] Step 5, calculating the true spectral radiance of multiple spectral channels in the remote sensor B to be calibrated that have a greater impact on the remote sensor A to be calibrated;

[0014] Step 6: Calculate the true spectral radiance of each spectral channel of the remote sensor A to be calibrated according to the final spectral matching factor obtained in step 4 and the true spectral radiance obtained in step 5;

[0015] Step 7: According to the real spectral radiance obtained in step 6, the radiation calibration coefficients of each spectral channel of the remote sensor A to be calibrated are calculated, and the spectral matching factor based on the dynamic Gaussian distribution is generated to complete the cross-calibration of the hyperspectral remote sensor.

[0016] Furthermore, step 2 is specifically as follows:

[0017] 2.1. Obtain spectral channel characteristics of the remote sensor A to be calibrated and the reference remote sensor B; the spectral channel characteristics include central wavelength, spectral channel and spectral response function;

[0018] 2.2. Input the atmospheric parameters, surface reflectivity, and observation geometry parameters of the uncalibrated remote sensor A and the reference remote sensor B obtained in step 1 into the MODTRAN model; use the MODTRAN model to simulate the continuous apparent radiance of the uncalibrated remote sensor A and the reference remote sensor B at the top of the atmosphere, and obtain the apparent radiance f A (λ) and f B (λ);

[0019] 2.3. Based on the spectral channel characteristics obtained in step 2.1 and the apparent radiance f obtained in step 2.2 A (λ), the simulated spectral radiance R of the remote sensor A to be calibrated after responding to the i-th spectral channel is calculated by the following formula A (λ i ):

[0020]

[0021] Where a and b are the given band λ i The lower and upper bounds of the wavelength of the i-th spectral channel, i = 1, 2, 3, ..., I, I is the number of spectral channels of the remote sensor A to be calibrated; S A (λ i ) is the spectral response function of the remote sensor A to be calibrated in the i-th spectral channel;

[0022] 2.4. Based on the spectral channel characteristics obtained in step 2.1 and the apparent radiance f obtained in step 2.2 B (λ), the simulated spectral radiance R of the reference remote sensor B after the jth spectral channel response is calculated by the following formula B (λ j ):

[0023]

[0024] Where c and d are the given band λ j The lower and upper bounds of the wavelength of the jth spectral channel, j = 1, 2, 3, ..., J, where J is the number of spectral channels of the reference remote sensor B; S B (λ j ) is the spectral response function of the reference remote sensor B in the jth spectral channel.

[0025] Furthermore, step 3 is specifically as follows:

[0026] 3.1. According to the spectral resolution FWHM (λ i ), calculate the Gaussian function gaussian(λ i ):

[0027]

[0028] gaussian(λ i ) = exp[(x-μ) 2 / 2σ 2 ]

[0029] Where σ is the Gaussian function of the i-th spectral channel (λ i ), μ is the value at the center of the Gaussian function, which is used as the center wavelength of the i-th spectral channel center_wave(λ i ), x is any wavelength within the response range of the remote sensor A to be calibrated;

[0030] 3.2. According to the Gaussian function of the i-th spectral channel (λ i ), draw a Gaussian function curve, and normalize the area of ​​the function to obtain the g-th spectral channel of the reference remote sensor B that will have a greater impact on the i-th spectral channel of the calibration remote sensor A during the spectral matching process, g = 1, 2…G, G is the number of channels of the reference remote sensor B that will have a greater impact on the i-th spectral channel of the calibration remote sensor A, G∈J; according to the simulated spectral radiance R obtained in step 2.4 B (λ j ) corresponds to obtaining the spectral radiance of G spectral channels of the reference remote sensor B;

[0031] 3.3. According to the simulated spectral radiance R obtained in step 2.3 A (λ i ) and the spectral radiance of the G spectral channels of the reference remote sensor B obtained in step 3.2, calculate the spectral matching factor SBAF of the G spectral channels in the reference remote sensor B that have a greater impact on the remote sensor A to be calibrated i,g :

[0032]

[0033] Furthermore, step 4 is specifically as follows:

[0034] 4.1. The value of the Gaussian function at the central wavelength of the jth spectral channel of the reference remote sensor B is taken as the weight w of the spectral matching factor corresponding to the spectral radiance of the G channels of the reference remote sensor B. i,g ;

[0035] 4.2. The weight coefficient w is calculated by the following formula i,G Normalize and get the weight coefficient w′ i,g :

[0036]

[0037] 4.3. The weight coefficient w′ obtained according to step 4.2i,g The spectral matching factor SBAF obtained in step 3.3 is calculated by the following formula: i,g Weighted to obtain the final spectral matching factor SBAF of each spectral channel of the remote sensor A to be calibrated i ′ ,g :

[0038]

[0039] Furthermore, step 5 is specifically as follows:

[0040] 5.1. Obtain the same-name image pair of the remote sensor A to be calibrated and the reference remote sensor B in the test area;

[0041] 5.2. Extract the DN values ​​of the remote sensor A to be calibrated in the i-th spectral channel and the reference remote sensor B in the 1st, 2nd…Gth spectral channel test area respectively according to the image pairs with the same name;

[0042] 5.3. According to the linear relationship between DN value and spectral radiance, calculate the true spectral radiance L of reference remote sensor B in G spectral channels B,g :

[0043] L B,g =gain g *DN g +offset g

[0044] Where: DN g is the DN value of the reference sensor B in the test area of ​​the gth spectral channel; gain g and offset g are the radiation calibration coefficients of reference remote sensor B in G spectral channels.

[0045] Furthermore, step 6 is specifically as follows:

[0046] The final spectral matching factor SBAF of each spectral channel of the remote sensor A to be calibrated obtained in step 4.3 i ′ ,g The actual spectral radiance L of the reference remote sensor B in G spectral channels obtained in step 5.3 is B,g , calculate the true spectral radiance L of the remote sensor A to be calibrated in the test area in the i-th spectral channel A,i :

[0047]

[0048] Furthermore, step 7 is specifically as follows:

[0049] According to the real spectral radiance L obtained in step 6 A,i, the calibration coefficients of each spectral channel of the remote sensor A to be calibrated are fitted by the following formula to realize the calibration of the remote sensor A to be calibrated and complete the cross calibration of the hyperspectral remote sensor:

[0050] L A,i =gain i *DN i +offset i

[0051] Where, DN i is the DN value of the remote sensor A to be calibrated in the test area of ​​the i-th spectral channel; gain i and offset i are the radiation calibration coefficients of the remote sensor A to be calibrated in the i-th spectral channel.

[0052] Beneficial effects of the present invention:

[0053] 1. The present invention discloses a hyperspectral remote sensor cross-calibration method based on dynamic Gaussian distribution. According to the data characteristics of the interferometric hyperspectral remote sensor, a Gaussian function that meets the spectral channel response characteristics is used to constrain the selection of spectral channels in the reference remote sensor B, and a weight coefficient is assigned to each spectral matching factor using Gaussian distribution to improve the accuracy of the spectral matching factor. The spectral matching factor of each spectral channel of the remote sensor A to be calibrated can also be efficiently calculated, which simplifies the cross-calibration process and improves the cross-calibration accuracy.

[0054] 2. The present invention provides a hyperspectral remote sensor cross-calibration method based on dynamic Gaussian distribution, which fully considers the variability of the spectral channel of the interferometric remote sensor to be calibrated, and uses the dynamic Gaussian function to achieve better utilization of the spectral channel of the reference remote sensor B. Considering that the spectral response function of the remote sensor A to be calibrated is significantly affected by the superposition effect of multiple spectral channels, the dynamic Gaussian distribution can more accurately capture and process these spectral features, thereby achieving more accurate spectral matching.

[0055] 3. The present invention provides a hyperspectral remote sensor cross-calibration method based on dynamic Gaussian distribution, which can improve the accuracy of cross-calibration of interferometric hyperspectral remote sensors, provide ideas for optimizing the spectral matching factors of other hyperspectral remote sensors, and broaden the application scope of cross-calibration. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flow chart of an embodiment of a hyperspectral remote sensor cross-calibration method based on dynamic Gaussian distribution of the present invention;

[0057] Figure 2 This is a comparison chart of processing hyperspectral remote sensor data using a traditional method (blue) and a hyperspectral remote sensor cross-calibration method based on dynamic Gaussian distribution (red) of the present invention. DETAILED DESCRIPTION

[0058] like Figure 1 As shown, a hyperspectral remote sensor cross calibration method based on dynamic Gaussian distribution is provided; the radiance of each spectral channel of two remote sensors is simulated according to the MODTRAN model, and then the spectral channel characteristics of the two remote sensors are combined, and the spectral matching factor group is calculated using the dynamic Gaussian distribution method; then the DN value of the cross region of the two remote sensor image pairs is extracted, and the calibration of the remote sensor A to be calibrated is realized according to the linear relationship between the DN value and the radiance. The method specifically includes the following steps:

[0059] Step 1, determine the cross-calibration test area, obtain the atmospheric parameters, surface reflectivity, and observation geometry parameters of the to-be-calibrated remote sensor A and the reference remote sensor B in the test area;

[0060] Step 2, obtaining the spectral channel characteristics of the remote sensor A to be calibrated and the reference remote sensor B; the spectral channel characteristics include central wavelength, spectral channel and spectral response function; according to the atmospheric parameters of the test area, the surface reflectivity, and the observation geometric condition parameters of the remote sensor A to be calibrated and the reference remote sensor B obtained in step 1, respectively simulate the apparent radiance of the remote sensor A to be calibrated and the reference remote sensor B, and calculate the simulated spectral radiance of the remote sensor A to be calibrated and the reference remote sensor B after the response of their respective spectral channels according to the apparent radiance and the spectral channel characteristics;

[0061] 2.1. Obtain spectral channel characteristics of the remote sensor A to be calibrated and the reference remote sensor B; the spectral channel characteristics include central wavelength, spectral channel and spectral response function;

[0062] 2.2. Input the atmospheric parameters, surface reflectivity, and observation geometry parameters of the uncalibrated remote sensor A and the reference remote sensor B obtained in step 1 into the MODTRAN model; use the MODTRAN model to simulate the continuous apparent radiance of the uncalibrated remote sensor A and the reference remote sensor B at the top of the atmosphere, and obtain the apparent radiance f A (λ) and f B (λ);

[0063] 2.3. Based on the spectral channel characteristics obtained in step 2.1 and the apparent radiance f obtained in step 2.2 A (λ), the simulated spectral radiance R of the remote sensor A to be calibrated after responding to the i-th spectral channel is calculated by the following formula A (λ i ):

[0064]

[0065] Where a and b are the given band λ iThe lower and upper bounds of the wavelength of the i-th spectral channel, i = 1, 2, 3, ..., I, I is the number of spectral channels of the remote sensor A to be calibrated; S A (λ i ) is the spectral response function of the remote sensor A to be calibrated in the i-th spectral channel;

[0066] 2.4. Based on the spectral channel characteristics obtained in step 2.1 and the apparent radiance f obtained in step 2.2 B (λ), the simulated spectral radiance R of the reference remote sensor B after the jth spectral channel response is calculated by the following formula B (λ j ):

[0067]

[0068] Where c and d are the given band λ j The lower and upper bounds of the wavelength of the jth spectral channel, j = 1, 2, 3, ..., J, where J is the number of spectral channels of the reference remote sensor B; S B (λ j ) is the spectral response function of the reference remote sensor B in the jth spectral channel;

[0069] Step 3: According to the simulated spectral radiance R A (λ i ) and simulated spectral radiance R B (λ j ), calculating the spectral matching factors of multiple spectral channels of the remote sensor B to be calibrated that have a greater impact on the remote sensor A to be calibrated based on the dynamic Gaussian distribution;

[0070] The remote sensor A to be calibrated is an interferometric hyperspectral remote sensor with a Gaussian spectral response function. The spectral channel bandwidth and spectral resolution are constantly changing. The spectral channel of the reference remote sensor B is retrieved for the i-th spectral channel (any spectral channel) of the remote sensor A to be calibrated using the dynamic Gaussian distribution method.

[0071] 3.1. According to the spectral resolution FWHM (λ i ), calculate the Gaussian function gaussian(λ i ):

[0072]

[0073] gaussian(λ i ) = exp[(x-μ) 2 / 2σ 2 ]

[0074] Where σ is the Gaussian function of the i-th spectral channel (λi ), μ is the value at the center of the Gaussian function, which is used as the center wavelength of the i-th spectral channel center_wave(λ i ), x is any wavelength within the response range of the remote sensor A to be calibrated;

[0075] 3.2. According to the Gaussian function gaussian(λ i ), draw a Gaussian function curve, and perform area normalization on the function;

[0076] When |x-μ|=2σ, the area enclosed by the Gaussian function accounts for 95% of the total area, which can completely cover the influential spectral channels. The retrieval range is [μ-2σ,μ+2σ], that is, [center_wave(λ i )-2σ,center_wave(λ i )+2σ], and obtain the g-th spectral channel of the reference remote sensor B that has a greater impact on the i-th spectral channel of the remote sensor A to be calibrated during the spectral matching process, g = 1, 2…G, G is the number of channels of the reference remote sensor B that has a greater impact on the i-th spectral channel of the remote sensor A to be calibrated, G∈J, according to the simulated spectral radiance R obtained in step 2.4 B (λ j ) corresponds to obtaining the spectral radiance of G spectral channels of the reference remote sensor B;

[0077] 3.3. According to the simulated spectral radiance R obtained in step 2.3 A (λ i ) and the spectral radiance of the G spectral channels of the reference remote sensor B obtained in step 3.2, calculate the spectral matching factor SBAF of the G spectral channels in the reference remote sensor B that have a greater impact on the remote sensor A to be calibrated i,g :

[0078]

[0079] Step 4: SBAF is the spectral matching factor of G spectral channels. i,g Assign weight coefficients and weight them to obtain the final spectral matching factors of each spectral channel of the remote sensor A to be calibrated;

[0080] 4.1. The value of the Gaussian function at the central wavelength of the jth spectral channel of the reference remote sensor B is taken as the weight w of the spectral matching factor corresponding to the spectral radiance of the G channels of the reference remote sensor B. i,g ;

[0081] 4.2. The weight coefficient w is calculated by the following formula i,G Normalize and get the weight coefficient w′ i,g :

[0082]

[0083] 4.3. The weight coefficient w′ obtained according to step 4.2 i,g The spectral matching factor SBAF obtained in step 3.3 is calculated by the following formula: i,g Weighted to obtain the final spectral matching factor SBAF of each spectral channel of the remote sensor A to be calibrated i ′ ,g :

[0084]

[0085] Step 5, calculating the true spectral radiance of the reference remote sensor B in G spectral channels;

[0086] 5.1. Obtain the same-name image pair of the remote sensor A to be calibrated and the reference remote sensor B in the test area;

[0087] 5.2. Extract the DN (Digital Number) values ​​of the remote sensor A to be calibrated in the i-th spectral channel and the reference remote sensor B in the 1st, 2nd…Gth spectral channel test area respectively according to the image pairs with the same name;

[0088] 5.3. According to the linear relationship between DN value and spectral radiance, calculate the true spectral radiance L of reference remote sensor B in G spectral channels B,g :

[0089] L B,g =gain g *DN g +offset g

[0090] Where: DN g is the DN value of the reference sensor B in the test area of ​​the gth spectral channel; gain g and offset g They are the radiation calibration coefficients of reference remote sensor B in G spectral channels;

[0091] Step 6: The final spectral matching factor SBAF of each spectral channel of the remote sensor A to be calibrated obtained in step 4.3 i ′ ,g The actual spectral radiance L of the reference remote sensor B in G spectral channels obtained in step 5.3 is B,g , calculate the true spectral radiance L of the remote sensor A to be calibrated in the test area in the i-th spectral channel A,i :

[0092]

[0093] Step 7: According to the real spectral radiance L obtained in step 6 A,i , combined with cross-image DN i The calibration coefficients of each spectral channel of the remote sensor A to be calibrated are fitted by the following formula to achieve the calibration of the remote sensor A to be calibrated:

[0094] L A,i =gain i *DN i +offset i .

[0096] The present invention calculates the spectral matching factor based on the data characteristics and spectral channel characteristics of the interferometric hyperspectral remote sensor, combines multiple spectral channels of the reference remote sensor B under Gaussian distribution, and dynamically distributes weights according to the distribution characteristics of the spectral channels, thereby improving the accuracy of the spectral matching factor and thus improving the cross-calibration accuracy. The method can optimize the spectral matching factor generation method, improve the reliability of the cross-calibration result, and solve the problem of uneven distribution of the central wavelength of the spectral channel and inconsistent spectral resolution in the cross-calibration of the interferometric hyperspectral remote sensor, resulting in low accuracy of the spectral matching factor.

[0097] like Figure 2 As shown, in the visible light and near infrared (VNIR) band, the spectral response function of the spectral channel in the second half of the range is significantly affected by the superposition effect of multiple spectral channels. Compared with the traditional method, the cross-calibration method of the hyperspectral remote sensor based on dynamic Gaussian distribution of the present invention can be scientifically weighted according to the distribution characteristics of the spectral channel, thereby achieving more accurate spectral matching and obtaining a more accurate cross-calibration result. In this embodiment, the remote sensor A to be calibrated selects the hyperspectral remote sensor (interference type) of the Environmental 2 satellite, and the reference remote sensor B selects the hyperspectral remote sensor of the Gaofen 5 satellite.

Claims

1. A hyperspectral remote sensor cross-calibration method based on dynamic Gaussian distribution; characterized in that: The following steps are involved: Step 1: determine the cross-calibration test area, obtain the atmospheric parameters, surface reflectivity, and observation geometry parameters of the to-be-calibrated remote sensor A and the reference remote sensor B in the test area; Step 2, obtaining the spectral channel characteristics of the remote sensor A to be calibrated and the reference remote sensor B; The spectral channel characteristics include central wavelength, spectral channel and spectral response function; according to the atmospheric parameters of the test area, the surface reflectivity, and the observation geometric condition parameters of the remote sensor A to be calibrated and the reference remote sensor B obtained in step 1, the apparent radiance of the remote sensor A to be calibrated and the reference remote sensor B are simulated respectively, and according to the apparent radiance and spectral channel characteristics, the simulated spectral radiance of the remote sensor A to be calibrated and the reference remote sensor B after the response of their respective spectral channels is calculated; Step 3: According to the simulated spectral radiance obtained in step 2, the spectral matching factors SBAF of multiple spectral channels in the remote sensor B to be calibrated that have a greater impact on the remote sensor A to be calibrated are calculated based on the dynamic Gaussian distribution. i,G : In the formula, R A (λ i ) is the simulated spectral radiance of the remote sensor A to be calibrated after responding to the i-th spectral channel, R B (λ1), R B (λ2),…,R B (λ G ) are the simulated spectral radiances of the reference remote sensor B after responding to the 1st, 2nd, …, G spectral channels respectively; Step 4: assign a weight coefficient to the spectral matching factor obtained in step 3, and weight it to obtain the final spectral matching factor of each spectral channel of the remote sensor A to be calibrated; Step 5, calculating the true spectral radiance of multiple spectral channels in the remote sensor B to be calibrated that have a greater impact on the remote sensor A to be calibrated; Step 6: Calculate the true spectral radiance of each spectral channel of the remote sensor A to be calibrated according to the final spectral matching factor obtained in step 4 and the true spectral radiance obtained in step 5; Step 7: According to the real spectral radiance obtained in step 6, the radiation calibration coefficients of each spectral channel of the remote sensor A to be calibrated are calculated, and the spectral matching factor based on the dynamic Gaussian distribution is generated to complete the cross-calibration of the hyperspectral remote sensor.

2. According to claim 1, a hyperspectral remote sensor cross-calibration method based on dynamic Gaussian distribution is characterized in that: Step 2 is as follows: 2.

1. Obtain the spectral channel characteristics of the remote sensor A to be calibrated and the reference remote sensor B; The spectral channel characteristics include central wavelength, spectral channel and spectral response function; 2.

2. Input the atmospheric parameters, surface reflectivity, and observation geometry parameters of the uncalibrated remote sensor A and the reference remote sensor B obtained in step 1 into the MODTRAN model; use the MODTRAN model to simulate the continuous apparent radiance of the uncalibrated remote sensor A and the reference remote sensor B at the top of the atmosphere, and obtain the apparent radiance f A (λ) and f B (λ); 2.

3. Based on the spectral channel characteristics obtained in step 2.1 and the apparent radiance f obtained in step 2.2 A (λ), the simulated spectral radiance R of the remote sensor A to be calibrated after responding to the i-th spectral channel is calculated by the following formula A (λ i ): Where a and b are the given band λ i The lower and upper bounds of the wavelength of the i-th spectral channel, i = 1, 2, 3, ..., I, I is the number of spectral channels of the remote sensor A to be calibrated; S A (λ i ) is the spectral response function of the remote sensor A to be calibrated in the i-th spectral channel; 2.

4. Based on the spectral channel characteristics obtained in step 2.1 and the apparent radiance f obtained in step 2.2 B (λ), the simulated spectral radiance R of the reference remote sensor B after the jth spectral channel response is calculated by the following formula B (λ j ): Where c and d are the given band λ j The lower and upper bounds of the wavelength of the jth spectral channel, j = 1, 2, 3, ..., J, where J is the number of spectral channels of the reference remote sensor B; S B (λ j ) is the spectral response function of the reference remote sensor B in the jth spectral channel.

3. According to claim 2, a hyperspectral remote sensor cross-calibration method based on dynamic Gaussian distribution is characterized in that: Step 3 is as follows: 3.

1. According to the spectral resolution FWHM (λ i ), calculate the Gaussian function gaussian(λ i ): gaussian(λ i )=exp[(x-μ) 2 / 2σ 2 ] Where σ is the Gaussian function of the i-th spectral channel (λ i ), μ is the value at the center of the Gaussian function, which is used as the center wavelength of the i-th spectral channel center_wave(λ i ), x is any wavelength within the response range of the remote sensor A to be calibrated; 3.

2. According to the Gaussian function of the i-th spectral channel (λ i ), draw a Gaussian function curve, and normalize the area of ​​the function to obtain the g-th spectral channel of the reference remote sensor B that will have a greater impact on the i-th spectral channel of the calibration remote sensor A during the spectral matching process, g = 1, 2…G, G is the number of channels of the reference remote sensor B that will have a greater impact on the i-th spectral channel of the calibration remote sensor A, G∈J; according to the simulated spectral radiance R obtained in step 2.4 B (λ j ) corresponds to obtaining the spectral radiance of G spectral channels of the reference remote sensor B; 3.

3. According to the simulated spectral radiance R obtained in step 2.3 A (λ i ) and the spectral radiance of the G spectral channels of the reference remote sensor B obtained in step 3.2, calculate the spectral matching factor SBAF of the G spectral channels in the reference remote sensor B that have a greater impact on the remote sensor A to be calibrated i,g :

4. According to claim 3, a hyperspectral remote sensor cross-calibration method based on dynamic Gaussian distribution is characterized in that: Step 4 is as follows: 4.

1. The value of the Gaussian function at the central wavelength of the jth spectral channel of the reference remote sensor B is taken as the weight w of the spectral matching factor corresponding to the spectral radiance of the G channels of the reference remote sensor B. i,g ; 4.

2. The weight coefficient w is calculated by the following formula i,g Normalize and get the weight coefficient w′ i,g : 4.

3. The weight coefficient w′ obtained according to step 4.2 i,g The spectral matching factor SBAF obtained in step 3.3 is calculated by the following formula: i,g Weighted to obtain the final spectral matching factor SBAF of each spectral channel of the remote sensor A to be calibrated i ′ ,g :

5. The method for cross-calibration of a hyperspectral remote sensor based on dynamic Gaussian distribution according to claim 4, characterized in that: Step 5 is as follows: 5.

1. Obtain the same-name image pair of the remote sensor A to be calibrated and the reference remote sensor B in the test area; 5.

2. Extract the DN values ​​of the remote sensor A to be calibrated in the i-th spectral channel and the reference remote sensor B in the 1st, 2nd, ..., G-th spectral channel test area respectively according to the image pairs with the same name; 5.

3. According to the linear relationship between DN value and spectral radiance, calculate the true spectral radiance L of reference remote sensor B in G spectral channels B,g : L B,g =gain g *DN g +offset g Where: DN g is the DN value of the reference sensor B in the test area of ​​the gth spectral channel; gain g and offset g are the radiation calibration coefficients of reference remote sensor B in G spectral channels.

6. The method for cross-calibration of a hyperspectral remote sensor based on dynamic Gaussian distribution according to claim 5, characterized in that: Step 6 is as follows: The final spectral matching factor SBAF of each spectral channel of the remote sensor A to be calibrated obtained in step 4.3 i ′ ,g The actual spectral radiance L of the reference remote sensor B in G spectral channels obtained in step 5.3 is B,g , calculate the true spectral radiance L of the remote sensor A to be calibrated in the test area in the i-th spectral channel A,i :

7. The method for cross-calibration of a hyperspectral remote sensor based on dynamic Gaussian distribution according to claim 6, characterized in that: Step 7 is as follows: According to the real spectral radiance L obtained in step 6 A,i , the calibration coefficients of each spectral channel of the remote sensor A to be calibrated are fitted by the following formula to realize the calibration of the remote sensor A to be calibrated and complete the cross calibration of the hyperspectral remote sensor: L A,i =gain i *DN i +offset i Where, DN i is the DN value of the remote sensor A to be calibrated in the test area of ​​the i-th spectral channel; gain i and offset i are the radiation calibration coefficients of the remote sensor A to be calibrated in the i-th spectral channel.

Images