1. An on-orbit cross calibration method for a synthetic aperture radar is characterized by comprising the following steps:

step A, data acquisition, which is used for acquiring all used data;

a1, acquiring SAR satellite data to be calibrated, wherein the SAR satellite data to be calibrated needs to irradiate a uniform bare area; acquiring the transit time T1 of the SAR satellite data to be calibrated through a metafile of the SAR satellite data to be calibrated;

a2, selecting two groups of calibrated SAR satellite data; the two groups of calibrated SAR data are required to come from the same satellite, the data transit time difference is more than 90 days, the two groups of calibrated SAR data have the same lifting orbit, the incident angle and the irradiation area, the satellite calibration precision is high, and the two groups of calibrated SAR data irradiate the same area with the SAR satellite data to be calibrated; wherein the difference between the transit time T2 of a group of calibrated SAR satellite data and the transit time T1 of the SAR satellite data to be calibrated is within 20 days;

step B, determining a uniform target range for selecting a cross target position;

b1, obtaining backscattering coefficients of the calibrated SAR satellite data 1 and the calibrated SAR satellite data 2;

preprocessing the image digital values of the two groups of calibrated satellite data, and acquiring backscattering coefficient values of calibrated SAR satellite data 1 and calibrated SAR satellite data 2 through preprocessing;

b2, making data slices;

registering two groups of calibrated SAR satellite data, performing primary registration through longitude and latitude, performing subjective registration judgment through a landmark feature, and verifying the accuracy of the longitude and latitude judgment; searching for landmark ground objects of high buildings and road inflection points, and correcting longitude and latitude registration results; translating the pixel positions of the two images based on the registration result to enable the horizontal and vertical coordinates of the two groups of calibrated satellite data to correspond one to one, and ensuring the image values of the two images to be calculated and compared;

after the registration is finished, segmenting different uniform areas in the calibrated SAR data; finding out specific horizontal and vertical coordinates corresponding to the slicing areas in the calibrated SAR satellite data 1 for slicing, and slicing the calibrated SAR satellite data 2 in the same coordinate area; the slicing requires that the size of the corresponding slicing pixels of the calibrated SAR satellite data 1 and the calibrated SAR satellite data 2 is consistent, the same uniform area is irradiated, and the number of slices is not less than 3 groups; the backscattering coefficient values of the slices are uniformly distributed, and the high backscattering coefficient area and the low backscattering coefficient area are sliced;

b3 comparison of time stability of backscattering coefficient

Because two groups of calibrated SAR satellite data slices may have errors in individual pixel points, cell division is carried out on the two groups of calibrated data after slicing, and systematic errors are reduced; taking the mean value of each grid of data as the backscattering coefficient value of the data block and adding operation;

analyzing the backscattering coefficient change between the two groups of calibrated SAR satellite data; the change of the backscattering coefficient is simultaneously influenced by soil humidity, root-mean-square height of the earth surface, dielectric constant and incident angle; for calibrated data, the incidence angles are consistent, the calibrated data belong to the same SAR satellite data, each group of data blocks irradiates the same target, and the factors influencing the scattering characteristics only comprise soil humidity, surface roughness and soil dielectric constant surface parameters; the smaller the change of the backscattering coefficient between the two groups of data is, the higher the scattering stability of the region is in the transit time of the two groups of data;

for the stability of the backscattering coefficient, the mean square error MSE is used as an evaluation standard; the slice range with the minimum mean square error of the two groups of calibrated SAR satellite data is a uniform area range with the most stable scattering characteristic; the specific calculation formula is as follows:

in the formulaAndthe backscattering coefficients of the data blocks corresponding to the two groups of calibrated data are respectively; n is the number of data blocks of each group of slices; MSE is the mean square error value of each set of slice data;

b4, determining the uniform target range

Taking the MSE value as a scattering stability judgment standard, and selecting a region to which the slice data with the minimum MSE value belongs as a uniform target region;

if the MSE values of all the slices are larger than 1dB, the data slicing fails, the data slicing is carried out again, and the uniform area is selected again;

step C, cross calibration for calculating calibration coefficient of satellite to be calibrated

C1, registering the calibrated SAR satellite data 2 and the SAR satellite data to be calibrated;

selecting a group of calibrated SAR satellite data 2 which is closest to the transit time of the SAR satellite data to be calibrated to perform subsequent cross calibration treatment; registering the image digital value of the calibrated SAR satellite data 2 and the image digital value of the SAR satellite data to be calibrated; adopting a maximum correlation coefficient method to carry out registration; the maximum correlation coefficient method is as follows:

wherein u 'is the frequency shift in the range direction and v' is the shift in the azimuth direction; SAR satellite number to be calibratedAccording to the amplitude M of the calibrated SAR satellite data 2₁(i,j)、M₂(i, j); selecting u 'and v' with the maximum R as a registration result, translating the SAR satellite data to be calibrated, and corresponding the SAR satellite data to be calibrated and the calibrated SAR satellite data in a one-to-one manner;

c2, cutting uniform target area data;

b, data cutting is carried out on the basis of the registration result and the uniform target area selected in the step B; registering the calibrated SAR satellite data 2 and the SAR satellite data to be calibrated to a pixel level; positioning the uniform target position through the horizontal and vertical coordinates when the uniform target area is cut through a step B2 of calibrated SAR satellite data 2; clipping the image digital values of the SAR satellite data to be calibrated and the calibrated SAR satellite data 2 on the same horizontal and vertical coordinates and the backscattering coefficient value of the calibrated SAR satellite data 2; cutting and dividing the cut data according to a unit scale, taking the average value of each unit block data as a cutting result, and setting the unit scale to be 10 multiplied by 10; averaging the data in each unit block to divide the data blocks is used for reducing registration difference caused by difference of satellite system parameters obtained by data with different incidence angles;

c3, obtaining a backscattering coefficient of the SAR satellite data to be calibrated;

obtaining a backscattering coefficient of SAR satellite data to be calibrated through a backscattering model;

in the Oh model, obtaining a backscattering coefficient of SAR satellite data to be calibrated: the Oh model is based on a back scattering coefficient, a homopolarization ratio p, a cross polarization ratio q, a soil root mean square height s and a soil humidity m_vAnd an empirical model established by the incidence angle theta, wherein the expressions are shown in formulas (3) to (5);

because two groups of data slices irradiate the same uniform target area with stable scattering characteristics, the transit time is within 20 days, and the two groups of data are in the same wave band; can be regarded as root mean square s of soil and soil humidity m_vThe wave number k is consistent; the backscattering coefficient value based on the change of the incident angle is derived by an Oh model based on these conditions, and the formula is as follows:

in the formula sigma_vv1Is the backscattering coefficient, theta, of each data block of SAR satellite data to be calibrated₁Is the angle of incidence, σ, of each data block of the SAR satellite data to be calibrated_vv2Is the backscattering coefficient, theta, of each data block of the calibrated SAR satellite data₂Is the incident angle of each data block of the calibrated SAR satellite data; calculating a backscattering coefficient value of SAR satellite data to be calibrated in a uniform target type range;

c4, obtaining the calibration coefficient of the SAR satellite to be calibrated

Performing least square fitting on the backscattering coefficient value and DN value of the SAR satellite data to be calibrated; the least squares method is as follows:

σ_vv＝a·DN_vv+b (7)

in the formula sigma_vvIs the backscattering coefficient of the SAR satellite to be calibrated,is the backscattering coefficient of the SAR satellite to be calibrated by least square fitting, and n is the backscattering of the satellite data to be calibrated calculated in the step C3The total number of coefficient values, E is the sum of the squares of the errors between the true and predicted values; calculating an intercept a that minimizes E; and the intercept a calculated by the least square method is the scaling coefficient K of the SAR satellite data to be scaled.

2. The method of claim 1, wherein the preprocessing comprises: track correction, radiometric calibration and terrain correction; 1) track correction: the orbit state vector provided in the metadata of the synthetic aperture radar product is usually not accurate enough, and the orbit file provided by the satellite within a few days to a few weeks after the data product is generated is used for carrying out refinement correction processing; 2) radiation correction: the reference data used are derived from calibrated satellites, the calibration coefficients being known to describe the value of the relationship between the image digital value DN and the value of the backscatter coefficient; 3) terrain correction: and the distortion is improved by adopting a terrain correction mode.

Background

Synthetic Aperture Radar (SAR) absolute radiometric calibration is a relation for establishing a digital quantization value of an SAR image and a surface feature backscattering coefficient value, is a basis for quantitative remote sensing application, and generally uses a manual scaler with a known Radar Cross Section (RCS) as a reference target for calibration. However, the number of manual scalers is limited, and the manufacturing, maintenance and deployment costs are high, which makes it difficult to perform normalization and high-frequency calibration of the satellite-borne SAR constellation.

The cross calibration method is an absolute radiometric calibration method for a field-free place, and is a method for realizing the calibration of satellite data to be calibrated by comparing numerical values of the satellite data to be calibrated and the satellite data to be calibrated when the same target area is observed by utilizing the satellite data to be calibrated (the calibration result is better). The cross calibration method does not need a ground calibration field calibrator and has the advantages of low calibration cost and strong timeliness.

The cross-calibration method is widely used in the field of optical remote sensing. There are three main methods for cross-calibration in the optical field: 1) the light matching cross calibration method needs the selected sensors to be capable of simultaneously and simultaneously observing, has high requirements on the sensors and is difficult to implement, and the method is less in use at present. 2) The spectrum matching cross calibration method requires two sensors to observe the same region, proportional relation exists between entrance pupil radiance or apparent reflectivity of the two sensors, and the entrance pupil radiance or the apparent reflectivity of the sensor to be calibrated is calculated according to the existing proportional factor. Although the atmospheric conditions and the observation geometry are considered to be the same when acquiring data, there is a proportional relationship between the entrance pupil radiance or the apparent reflectance due to the different spectral responses of the different sensors. 3) A radiation transmission model method is based on a radiation transmission model and is a method which is applied more in the field of optical cross calibration at present. The input values for the radiation transmission model are highly required. The 6s transmission model method is a cross calibration model in the field of optical remote sensing.

The cross-calibration method in optics is not suitable for Synthetic Aperture Radar (SAR) calibration. Because optical remote sensing is passive remote sensing, based on the radiant energy of the sun, the penetrability is not strong, and the radiation calibration in the optical remote sensing needs to consider the geometric relationship among the sun, the ground objects and the sensor, the aerosol, the atmospheric mode, the spectral characteristics of the sensor and the surface reflectivity. The SAR imaging adopts an active remote sensing mode, the penetrability is strong, the influence of atmosphere and aerosol is small, and the radiation energy of the sun does not need to be considered, so that an optical cross calibration method is not suitable for the synthetic aperture radar.

In the SAR calibration field, cross calibration research is less. The only research is also limited to airborne SAR and requires an approximate imaging geometry; the satellite-borne SAR cross calibration method is not reported in the public.

Disclosure of Invention

The invention mainly aims to provide a synthetic aperture radar cross calibration method.

The invention provides a cross calibration method for satellite-borne SAR real data after fully investigating related directions of SAR absolute radiometric calibration and optical cross calibration. Different from the traditional SAR absolute radiometric calibration method, the invention has the specific innovation points that: applying the cross calibration idea in optics to SAR calibration; two groups of calibrated SAR satellite data are used as auxiliary parameters, and a reasonable and uniform target area is selected; and considering the imaging angle difference between the calibrated SAR satellite and the SAR satellite to be calibrated, and acquiring the backscattering coefficient value of the SAR satellite data to be calibrated by using an Oh model. The main implementation object of the invention is the off-scale satellite-borne SAR satellite data. The main task is to perform absolute radiometric calibration on it.

The technical scheme of the invention specifically comprises the following technical contents:

and step A, data acquisition, which is used for acquiring all used data.

A1, obtaining SAR satellite data to be calibrated, wherein the SAR satellite data to be calibrated needs to irradiate a uniform bare area. And acquiring the transit time of the SAR satellite data to be calibrated through the metafile of the SAR satellite data to be calibrated (T1).

And A2, selecting two groups of calibrated SAR satellite data. The two sets of calibrated SAR data are required to come from the same satellite, the data transit time difference is more than 90 days, the data transit time difference has the same lifting orbit, the incident angle and the irradiation area, the satellite calibration precision is high, and the SAR satellite data to be calibrated irradiate the same area. The difference between the transit time (T2) of a set of calibrated SAR satellite data and the transit time (T1) of the SAR satellite data to be calibrated is within 20 days, so that the influence caused by factors such as climate, surface parameter change, environmental change and the like in the subsequent cross calibration step is avoided.

And step B, determining a uniform target range for selecting the position of the crossed target.

And B1, obtaining backscattering coefficients of the calibrated SAR satellite data 1 and the calibrated SAR satellite data 2.

And preprocessing the image digital values of the two groups of calibrated satellite data to obtain backscattering coefficient values of the calibrated SAR satellite data 1 and the calibrated SAR satellite data 2. The pretreatment mainly comprises the following steps: track correction, radiometric calibration, terrain correction. 1) Track correction: orbit state vectors provided in the synthetic aperture radar product metadata are often not accurate enough and refined correction processing is performed using precise orbit files provided by the satellites several days to several weeks after the data product is generated. 2) Radiation correction: the purpose of the radiation correction is to provide a relationship between the image digital value (DN) and the value of the backscatter coefficient, the reference data used by the invention is derived from a calibrated satellite, the calibration coefficient (the value describing the relationship between the image digital value (DN) and the value of the backscatter coefficient) being known. 3) Terrain correction: the range directions in the SAR image are distorted due to changes in the terrain and the tilt of the satellite sensor. These distortions are improved by means of terrain correction.

And B2, making data slices.

Firstly, two groups of calibrated SAR satellite data are registered, and because the two selected groups of calibrated SAR data come from the same satellite and have the same lifting orbit, incident angle and irradiation area, the initial registration can be directly carried out through longitude and latitude. After the preliminary registration is carried out through the longitude and latitude, subjective registration judgment is carried out through the landmark ground object, and the accuracy of the longitude and latitude judgment is verified. And (4) searching landmark ground objects such as high buildings, road inflection points and the like, and correcting the longitude and latitude registration result. And translating the pixel positions of the two images based on the registration result, so that the horizontal coordinates and the vertical coordinates of the two groups of calibrated satellite data are in one-to-one correspondence, and the numerical values of the two images can be directly calculated and compared.

After the registration is completed, different uniform regions in the calibrated SAR data are segmented. And finding out specific horizontal and vertical coordinates corresponding to the slicing area in the calibrated SAR satellite data 1 to perform slicing cutting, and performing slicing processing on the calibrated SAR satellite data 2 in the same coordinate area. The slicing requires that the size of the corresponding slicing pixels of the calibrated SAR satellite data 1 and the calibrated SAR satellite data 2 is consistent, the same uniform area is irradiated, and the number of the slices is not less than 3 groups. The backscattering coefficient values of the slices are uniformly distributed, and the high backscattering coefficient area and the low backscattering coefficient area are sliced.

B3, comparison of backscatter coefficient time stability.

Because two groups of calibrated SAR satellite data slices may have errors in individual pixel points, cell division is performed on the two groups of calibrated data after slicing, and systematic errors are reduced. The mean value of each bin of data is used as the backscattering coefficient value of the data block to add the operation.

The backscattering coefficient variation between the two sets of calibrated SAR satellite data is analyzed. The variation of the backscattering coefficient is simultaneously influenced by soil humidity, root mean square height of the earth surface, dielectric constant, incidence angle and the like. For calibrated data, the incidence angles are consistent, the data belong to the same SAR satellite data, each group of data blocks irradiate the same target, and the factors influencing the scattering characteristics of the data are only surface parameters such as soil humidity, surface roughness and soil dielectric constant. The smaller the change of the backscattering coefficient between the two groups of data is, the higher the scattering stability of the region in the transit time of the two groups of data is.

For the stability of the backscattering coefficient, the Mean Square Error (MSE) is used as an evaluation criterion. The slice range with the minimum mean square error of the two groups of calibrated SAR satellite data is a uniform area range with the most stable scattering characteristics. The specific calculation formula is as follows:

in the formulaAndthe two sets of scaled data correspond to the backscatter coefficients of the data block, respectively. And N is the number of data blocks of each group of slices. MSE is the mean square error value for each set of slice data.

And B4, determining a uniform target range.

And taking the MSE value as a scattering stability judgment standard. The small MSE shows that the surface parameters of the slice irradiation area are stable, and the surface parameters (soil humidity, surface roughness, soil dielectric constant and other values) are difficult to change along with time. And selecting the region to which the slice data with the minimum MSE value belongs as a uniform target region.

And if the MSE values of all the slices are larger than 1dB, the data slicing fails, and the data slicing is carried out again and the uniform area is selected again.

And step C, cross calibration, which is used for calculating a calibration coefficient of the satellite to be calibrated.

C1, and registering the calibrated SAR satellite data 2 and the SAR satellite data to be calibrated.

And selecting a group of calibrated SAR satellite data (calibrated SAR satellite data 2) closest to the transit time of the SAR satellite data to be calibrated to perform subsequent cross calibration processing. And because the subsequent processing has higher requirement on the matching precision of the image, registering the image digital value of the calibrated SAR satellite data 2 and the image digital value of the SAR satellite data to be calibrated. And adopting a maximum correlation coefficient method for registration. The maximum correlation coefficient method is as follows:

where u 'is the frequency shift in the range direction and v' is the shift in the azimuth direction. SAR satellite data to be calibrated and calibrated SAR satelliteThe amplitude of data 2 is M₁(i,j)、M₂(i, j). And selecting u 'and v' with the maximum R as a registration result, translating the SAR satellite data to be calibrated, and corresponding the SAR satellite data to be calibrated and the calibrated SAR satellite data in a one-to-one manner according to the horizontal and vertical coordinates. And errors caused by registration are reduced, and the registration errors are accurate to a pixel level.

And C2, cutting the uniform target area data.

And D, performing data clipping based on the registration result and the uniform target area selected in the step B. Since the SAR satellite data to be calibrated has been translated, the calibrated SAR satellite data 2 and the SAR satellite data to be calibrated are registered to the pixel level. The uniform target location is located by the abscissa and ordinate when this uniform target area is cropped by the scaled SAR satellite data 2 step B2. And (3) cutting the image digital values of the SAR satellite data to be calibrated and the calibrated SAR satellite data 2 with the same horizontal and vertical coordinates and the backscattering coefficient value of the calibrated SAR satellite data 2. And (4) cutting and dividing the cut data according to a unit scale, taking the average value of each unit block data as a cutting result, and setting the unit scale to be 10 multiplied by 10. The data in each unit block is divided into data blocks by averaging, so that the registration difference caused by the difference of the parameters of the data acquisition satellite system due to different incidence angles is reduced.

And C3, obtaining the backscattering coefficient of the SAR satellite data to be calibrated.

And obtaining a backscattering coefficient of SAR satellite data to be calibrated through a backscattering model, such as an Oh model, an IEM model, an AIEM model and the like.

Taking an Oh model as an example, obtaining a backscattering coefficient of SAR satellite data to be calibrated: the Oh model is based on a back scattering coefficient, a homopolarization ratio p, a cross polarization ratio q, a soil root mean square height s and a soil humidity m_vAnd an empirical model established for the incident angle theta, the expressions of which are shown in formulas (3) to (5).

Because two groups of data slices irradiate the same uniform target area with stable scattering characteristics, the transit time is within 20 days, and the two groups of data are in the same wave band. Can be regarded as root mean square s of soil and soil humidity m_vThe wave number k is the same. Based on these conditions, the backscattering coefficient value based on the change of the incident angle can be derived through an Oh model, and the formula is as follows:

in the formula sigma_vv1Is the backscattering coefficient, theta, of each data block of SAR satellite data to be calibrated₁Is the angle of incidence, σ, of each data block of the SAR satellite data to be calibrated_vv2Is the backscattering coefficient, theta, of each data block of the calibrated SAR satellite data₂Is the angle of incidence for each data block of the scaled SAR satellite data. And calculating the backscattering coefficient value of the SAR satellite data to be calibrated in a uniform target type range through a derived formula.

And C4, acquiring a calibration coefficient of the SAR satellite to be calibrated.

And performing least square fitting on the backscattering coefficient value and DN value of the SAR satellite data to be calibrated. The least squares method is as follows:

σ_vv＝a·DN_vv+b (7)

in the formula sigma_vvIs the backscattering coefficient of the SAR satellite to be calibrated,is passed throughAnd D, fitting the backscattering coefficient of the SAR satellite to be calibrated by the product of the two-decimal multiplication, wherein n is the total number of backscattering coefficient values of the satellite data to be calibrated calculated in the step C3, and E is the sum of squares of errors between a real value and a predicted value. The intercept a that minimizes E is calculated. And the intercept a calculated by the least square method is the scaling coefficient K of the SAR satellite data to be scaled.

Drawings

FIG. 1 is a flow chart of the present invention.

FIG. 2 is a flow chart of uniform target selection.

FIG. 3 is a cross-scaling flow chart

FIG. 4 is a schematic diagram comparing the method of the present invention and the conventional calibration method

FIG. 5 is a schematic view of a uniform target location.

Detailed Description

The basic flow of the synthetic aperture radar cross calibration method based on the cross calibration idea is shown in fig. 1, and the method specifically comprises the following steps:

and step A, data acquisition, which is used for acquiring all used data.

Sample data used in the practice of the present invention is a Sentinel-1 satellite image. The Sentinel-1 (Sentinel one) satellite consists of two Sentinel-1A, Sentinel-1B satellites, and is an earth observation satellite in the European space agency Columbus program (GMES). The satellite launch of the Sentinel-1A in 4/3/2014 is the first environmental monitoring satellite in the European space agency Cobriy program (GMES), and the revisit period is 12 days. The Sentinel-1B satellite No. 25 of 4 months in 2016 successfully launches to the air, two satellites run simultaneously, the observation efficiency is doubled, and the revisit time is shortened to 6 days. Compared with other on-orbit radar satellites, the Sentinel-1 satellite has higher radiation precision, and the absolute radiation precision is 1 dB. Compared with other satellite-borne SAR satellites, the Sentinel-1 satellite data better meets the requirements of absolute radiometric calibration on SAR scattering stability characteristics, and is convenient for further exploration on SAR data radiation precision.

The Sentinel-1 satellite data can be used freely, the absolute radiometric calibration precision is high, the public data volume is large, and the method well meets the calibrated data selection requirement. The SAR satellite data to be calibrated adopts Sentinel-1B satellite data. And selecting a set of Sentinel-1A satellite data of which the phase difference between the set of calibrated SAR satellite data and the SAR satellite image to be calibrated is 11 days. Another set of scaled SAR satellite data is also from a Sentinel-1A satellite and the transit time of this scaled SAR satellite data is greater than 90 days. The specific selected data parameters are shown in table 1:

TABLE 1 SAR satellite data to be calibrated and specific information of calibrated SAR satellite data

Data name	Satellite	Polarization mode	Frequency (GHz)	Obtaining time
					Scaled SAR satellite data 1	Sentinel-1A	VV、VH	5.405	2017/06/01
Scaled SAR satellite data 2	Sentinel-1A	VV、VH	5.405	2017/02/25
					SAR satellite data to be calibrated	Sentinel-1B	VV	5.405	2017/02/14

And step B, determining a uniform target range for selecting the position of the crossed target.

And B1, obtaining backscattering coefficients of the calibrated SAR satellite data 1 and the calibrated SAR satellite data 2.

Preprocessing image digital values of two groups of calibrated SAR satellite data, wherein the preprocessing mainly comprises the following steps: track correction, radiometric calibration, terrain correction. The radiometric calibration of the Sentinel-1 is mature, and the radiometric calibration operation can be directly carried out through Snap software. Due to cloud layer penetrability of SAR data, atmospheric correction operation of optical images is not needed in the Sentinel-1 data, and therefore radiometric calibration operation is directly carried out through Snap. Range-doppler terrain correction is used on the scaled SAR satellite data images. And carrying out geocoding and topographic radiation correction, and finally obtaining the backscattering coefficient of the calibrated SAR satellite data.

And B2, making data slices.

And slicing the two groups of the processed calibrated SAR satellite data. And selecting a uniform area for slicing, and judging whether the actual ground surface is the uniform area or not by referring to the actual ground surface information of a Google Earth map (Google Earth). Scribing is carried out according to Earth surface information provided by Google Earth, a uniform area with a large range on the Google Earth is selected, because the resolution of the selected Sentinel-1 satellite image is 10m, if the uniform area is too small, the size of an actual data slice is influenced, and the data of the area on the satellite image is too little, so that the purpose of an experiment is difficult to achieve.

And slicing two groups of calibrated SAR satellite data irradiating the same position, subjectively judging the slicing position, and preliminarily matching the two groups of data through longitude and latitude. Two sets of backscattering coefficient slices of the scaled SAR satellite data are divided into 90 x 90 slice data.

B3, comparison of backscatter coefficient time stability.

The variation of the backscattering coefficient is simultaneously influenced by soil humidity, surface roughness, dielectric constant, incidence angle, the performance of the sensor, and the like. Can be expressed as:

σ_vv＝f(θ，r)·g(s,m_v,ε) (9)

where f (θ, r) represents the influence of the incident angle θ and the performance r of the sensor itself on the backscattering coefficient, and g (s, m)_vε) represents the soil moisture m_vThe root mean square height of the earth s, and the influence of the dielectric constant epsilon on the backscattering coefficient. In the invention, two groups of calibrated SAR satellite data come from the same sensor, have the same incidence angle and can be regarded as f (theta, r) being the same.

Backscattering coefficient sigma of same target through two groups of calibrated SAR satellite data_vvDifference, analysis of soil moisture m_vThe root-mean-square height s of the earth surface and the stability of the dielectric constant epsilon along with the change of time. The soil humidity m_vThe stability of the root mean square height s of the earth surface, the dielectric constant epsilon is considered as the stability of the backscattering coefficient. The backscattering coefficient values for each 10 x 10 cell of the two sets of scaled SAR satellite data were averaged. And (3) obtaining a backscattering coefficient data set with the size of 9 multiplied by 9 in all the slice data, and calculating the Mean Square Error (MSE) between backscattering coefficient data sets of the two sets of data.

And B4, determining a uniform target range.

And taking the MSE value as a scattering stability judgment standard. The small MSE shows that the surface parameters of the slice irradiation area are stable, and the surface parameters (soil humidity, surface roughness, soil dielectric constant and other values) are difficult to change along with time. And selecting the region to which the slice data with the minimum MSE value belongs as a uniform target region.

The MSE of the selected slice in the experiment is 0.7204dB, and the slice meeting the judgment condition can be used as a uniform target.

And step C, cross calibration, which is used for calculating a calibration coefficient of the SAR satellite to be calibrated.

C1, and registering the calibrated SAR satellite data 2 and the SAR satellite data to be calibrated.

And selecting a group of calibrated SAR satellite data (calibrated SAR satellite data 2) closest to the transit time of the SAR satellite data to be calibrated to perform subsequent cross calibration processing. The difference between the transit time of the calibrated SAR satellite data 2 and the SAR satellite data to be calibrated meets the requirement of 11 days. And registering the DN values of the calibrated SAR satellite data 2 and the SAR satellite data to be calibrated by a maximum correlation coefficient method.

And C2, cutting the uniform target area data.

And D, performing data clipping based on the registration result and the uniform target area selected in the step B. And finally, the SAR satellite data to be calibrated and the calibrated SAR satellite data 2 are cut to be consistent in size. And simultaneously cutting the SAR satellite data to be calibrated, the DN value of the calibrated SAR satellite data 2 and the backscattering coefficient value of the calibrated SAR satellite data 2. The two sets of data are divided into blocks according to a unit scale, and the unit scale is set to 10 × 10. The data parameters after clipping are shown in table 2:

TABLE 2 data post-crop specific parameters

And C3, obtaining the backscattering coefficient of the SAR satellite data to be calibrated.

The backscattering coefficient value of the scaled SAR satellite data 2, and the angle of incidence of the SAR satellite data to be scaled are substituted into equation (10).

In the formula sigma_vv1Is the backscattering coefficient, theta, of each data block of SAR satellite data to be calibrated₁Is the angle of incidence, σ, of each data block of the SAR satellite data to be calibrated_vv2Is the backscattering coefficient, theta, of each data block of the scaled SAR satellite data 2₂Is the angle of incidence for each data block of the scaled SAR satellite data 2. And calculating a backscattering coefficient of the SAR satellite data to be calibrated under the corresponding incident angle.

And C4, acquiring a calibration coefficient of the SAR satellite to be calibrated.

And calculating a scaling coefficient K of the SAR satellite data to be scaled. And performing least square fitting on the backscattering coefficient value of the SAR satellite data to be calibrated and the DN value of the SAR satellite data to be calibrated. And the intercept calculated by the least square method is the scaling coefficient K of the SAR satellite data to be scaled.

In SAR absolute radiometric calibration, currently, a manual calibrator with a known Radar Cross Section (RCS) is usually used as a reference target to construct DN-RCS data pairs to obtain calibration parameters. However, the radiometric calibration method using the artificial calibrator as the reference target is gradually insufficient in meeting the requirements of the normalized high-frequency calibration of the serial satellites/constellation satellites due to the difficulties in the artificial calibration manufacture and deployment.

The invention mainly aims to provide a satellite-borne SAR absolute radiometric calibration method based on a natural boundary distributed target and a cross calibration idea, aiming at the problem that the calibration period of the traditional satellite-borne SAR absolute radiometric calibration method is too long. The method comprises the steps of taking a natural-boundary distributed target as a calibration reference target, and acquiring a reference target truth value by a calibrated SAR satellite in a crossed mode, so that calibration of the SAR satellite to be calibrated is realized. The method can solve the problems that the traditional calibration method is limited by a satellite revisiting period and has insufficient calibration timeliness, and the time required by calibration is shortened.

The main innovation points of the invention comprise the following 2:

1. and judging the scattering stability of the scene. The important coefficient of absolute radiometric calibration, the backscattering coefficient, is influenced by factors such as soil moisture, root mean square height of the earth's surface, dielectric constant, and angle of incidence. The method judges whether the factors influencing the backscattering coefficient in the scene are stable or not through two groups of reference data. For two groups of reference data, the incidence angles of the two groups of reference data are consistent and belong to the same satellite data, and the only factors influencing the backward scattering coefficients of the two groups of reference data are the surface parameters such as soil humidity, surface roughness and soil dielectric constant. The stability of different surface parameters changing along with time is analyzed, errors caused by surface parameter changes are avoided, and cross calibration precision is improved.

2. Cross-scaling based on the Oh model. By analyzing the Oh model, the relation between the incidence angle and the backscattering coefficient is deduced under the condition that the scene scattering characteristics are stable. The backscattering coefficient of the satellite data to be calibrated is obtained by referring to the satellite data, and the incident angle information is corrected, so that the cross calibration precision is improved, and the calculated backscattering coefficient value is closer to the real backscattering coefficient of the actual data to be calibrated.

TABLE 3 comparison table of two absolute radiometric calibration technique approaches