A thermal noise extraction method for wide-band cross-polarization SAR images considering the incident angle

Abstract

The invention discloses a method for extracting thermal noise from wide-width cross-polarization SAR images considering the incident angle, comprising the following steps: step 1, obtaining wide-width dual-polarization SAR image data of the sea surface under low wind speed; step 2, preprocessing the SAR image data obtained in step 1, and downsampling the image, thereby obtaining downsampled co-polarization backscattering coefficient images, cross-polarization backscattering coefficient images, and incident angle images; step 3, combining wind direction information, using co-polarization geophysical model functions and the incident angle image in step 2, performing sea surface wind speed inversion on the co-polarization backscattering coefficient image obtained in step 2, and obtaining the co-polarization image wind speed; step 4, obtaining the thermal noise of the cross-polarization backscattering coefficient image; step 5, determining the thermal noise model of each sub-swath according to the thermal noise extracted in step 4, thereby realizing thermal noise extraction from wide-width cross-polarization SAR images. The method is simple to operate and highly practical.

Description

Technical Field

[0001] The invention relates to the technical field of synthetic aperture radar data processing, in particular to a method for extracting thermal noise from wide-band cross-polarization SAR images taking incident angle into consideration. Background Art

[0002] Spaceborne synthetic aperture radar (SAR) has the ability to obtain high spatial resolution sea surface wind field. The sea surface wind field data obtained by SAR have important applications in numerical weather forecasting, maritime transportation, maritime rescue, offshore oil spill monitoring, trajectory prediction of floating ice at sea, typhoon intensity monitoring and high spatial resolution offshore wind energy resource assessment. After 20 years of development, the sea surface wind field inversion method of co-polarization SAR has been used for sea surface wind field inversion in a commercial manner. However, the co-polarization backscatter coefficient will be saturated under high wind speed conditions, which limits its wind field monitoring capability under extreme weather conditions, especially typhoon conditions.

[0003] In recent years, some scholars have found through the study of C-band cross-polarization SAR that under high wind speed conditions, the wind speed inverted by cross-polarization SAR has higher accuracy than that inverted by co-polarization SAR. This is mainly due to the fact that the cross-polarization backscatter coefficient is not affected by the saturation effect under very high wind speed conditions. Based on the fact that the cross-polarization backscatter coefficient is more suitable for the inversion of sea surface wind field under high wind speed conditions than the co-polarization backscatter coefficient, scientists have considered adding cross-polarization channels to the next generation of scatterometer designs.

[0004] In the wide-width working mode, the image width of spaceborne SAR can reach 500km, and the wide-width and high-spatial-resolution sea surface wind field data obtained by inversion have a wider range of applications. Since Germany successfully tested the TerraSAR-X system in 2007, TOPSAR (Terrain Observation with Progressive Scans SAR) technology has not only become the wide-width mapping working mode of the European Space Agency's new generation of spaceborne SAR system Sentinel-1 series satellites, but also my country's subsequent satellites planned for GF-3 will adopt the TOPSAR mode to obtain wide-width remote sensing images. Since the launch of Sentinel-1A on April 3, 2014 and Sentinel-1B on April 25, 2016, the European Space Agency (ESA) has shared the SAR images of the Sentinel-1 series satellites with global users free of charge, and plans to launch Sentinel-1C and Sentinel-1D in 2022 and 2023, respectively. Therefore, studying the wind field inversion method of Sentinel-1TOPSAR wide-mode cross-polarization images is of great significance in future sea surface wind field inversion research and application services, and it is also of reference significance for the design of subsequent synthetic aperture radars in my country.

[0005] In TOPSAR working mode, Sentinel-1 series satellites provide wide-swath images in two measurement modes: Interferometric Wide-swath (IW) and Extra Wide-swath (EW). IW images are composed of 3 sub-bands with a width of 250km, and EW images are composed of 5 sub-bands with a width of 400km. Although TOPSAR images have improved the radiation correction of scalloping and sub-band splicing compared with ScanSAR images, the study based on IW cross-polarization images found that the TOPSAR cross-polarization ocean remote sensing products still have obvious sub-band splicing caused by thermal noise. Therefore, before performing TOPSAR sea surface wind field inversion, it is necessary to remove the thermal noise of the image and correct the radiation energy of different sub-bands to the same level.

[0006] Unfortunately, although the European Space Agency provides a calibrated range thermal noise vector to remove the contribution of thermal noise to radar backscattering, there is still residual noise interference on the cross-polarization image after subtracting the calibrated range thermal noise, which results in the change of the cross-polarization backscatter coefficient with range after subtracting the thermal noise provided by the image product still showing the shape of thermal noise changing with the incident angle. This shows that the thermal noise of the wide-width image products provided by the Sentinel-1 satellite is not accurate, and a new thermal noise extraction method needs to be proposed to more effectively remove the thermal noise of the wide-width cross-polarization image products. Summary of the invention

[0007] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a method for thermal noise extraction from wide-width cross-polarization SAR images taking the incident angle into consideration. The method does not require thermal noise calibration information provided by the image, has the characteristics of simple operation and strong practicality, and is particularly suitable for thermal noise removal of wide-width cross-polarization SAR image products that do not provide thermal noise information or provide inaccurate thermal noise information.

[0008] The present invention adopts the following technical solutions to solve the above technical problems:

[0009] A method for extracting thermal noise from wide-band cross-polarization SAR images taking the incident angle into consideration according to the present invention comprises the following steps:

[0010] Step 1, obtaining wide-band dual-polarization SAR image data of the sea surface under low wind speed;

[0011] Step 2, preprocessing the SAR image data obtained in step 1, and downsampling the image, so as to obtain a downsampled co-polarization backscattering coefficient image, a cross-polarization backscattering coefficient image, and an incident angle image;

[0012] Step 3, combining the wind direction information, using the co-polarization geophysical model function and the incident angle image of step 2, the co-polarization backscattering coefficient image obtained in step 2 is used to invert the sea surface wind speed to obtain the co-polarization image wind speed;

[0013] Step 4: Using the co-polarization image wind speed obtained in step 3 as the reference wind speed, extract the cross-polarization backscattering coefficient corresponding to the area where the sea surface wind speed is less than 2 m / s from the cross-polarization backscattering coefficient image obtained in step 2 as the thermal noise of the cross-polarization backscattering coefficient image;

[0014] Step 5, divide the thermal noise extracted in step 4 into three sub-crops, and establish a trigonometric function model related to the thermal noise and incident angle of each sub-crop; use the thermal noise and incident angle in each sub-crop to perform fitting calculations, solve the fitting coefficient of the trigonometric function model, and thus determine the thermal noise model of each sub-crop. The thermal noise models of these three sub-crops are the cross-polarization image thermal noise extraction models, thereby realizing the thermal noise extraction of wide-width cross-polarization SAR images.

[0015] As a further optimization scheme of the wide-band cross-polarization SAR image thermal noise extraction method considering the incident angle described in the present invention, in step 1, the low wind speed is a wind speed less than 5 m / s.

[0016] As a further optimization scheme of the wide-band cross-polarization SAR image thermal noise extraction method considering the incident angle described in the present invention, in step 2, the spatial resolution range of the downsampling is 1 km to 5 km.

[0017] As a further optimization scheme of the wide-band cross-polarization SAR image thermal noise extraction method considering the incident angle described in the present invention, step 4 is specifically as follows:

[0018] The areas in the co-polarized wind speed image obtained in step 3 where the wind speed is less than the preset maximum sea surface wind speed are set to 1, and the other areas are set to 0, to obtain a mask layer that retains the low wind speed area; then a dot multiplication operation is performed on the mask layer and the cross-polarized backscattering coefficient image obtained in step 2 to obtain the cross-polarized backscattering coefficient that only retains the low wind speed area, and the cross-polarized backscattering coefficient is regarded as thermal noise.

[0019] As a further optimization scheme of the wide-band cross-polarization SAR image thermal noise extraction method considering the incident angle described in the present invention, in step 5, the cross-polarization image thermal noise extraction model is:

[0020]

[0021] Among them, the three sub-swaths are the first sub-swath, the second sub-swath and the third sub-swath, σ° simulatedNoise is the thermal noise in the form of linear power simulated by the model, θ is the incident angle of the radar, a1, a2 and a3 are the thermal noise amplitude coefficients of the first sub-swath, the second sub-swath and the third sub-swath respectively, b1, b2 and b3 are the incident angle scaling coefficients of the first sub-swath, the second sub-swath and the third sub-swath respectively, c1, c2 and c3 are the incident angle translation coefficients of the first sub-swath, the second sub-swath and the third sub-swath respectively, d1, d2 and d3 are the thermal noise compensation coefficients of the first sub-swath, the second sub-swath and the third sub-swath respectively.

[0022] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0023] (1) The present invention does not require ground observation data and can directly use dual-polarization SAR image data to extract cross-polarization thermal noise, which can effectively reduce the observation cost;

[0024] (2) Since dual-polarization data has almost no time difference in imaging time, the use of dual-polarization data to extract cross-polarization thermal noise can reduce the error caused by the time difference;

[0025] (3) Without the need for known thermal noise calibration information, a thermal noise extraction model for cross-SAR data can be automatically established, which is particularly suitable for thermal noise removal of cross-polarization SAR image products that do not provide thermal noise information or provide inaccurate thermal noise information. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of the present invention.

[0027] Figure 2 It is the product thermal noise of Sentinel-1 interferometric wide-swath mode cross-polarization SAR image.

[0028] Figure 3 The cross-polarization backscatter coefficient after subtracting the product thermal noise varies with the incident angle under different wind speed conditions.

[0029] Figure 4 The figure shows the variation of the cross-polarization backscattering coefficient with the incident angle under different wind speed conditions after subtracting the simulated thermal noise obtained by the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] like Figure 1 As shown: The present invention is a method for extracting thermal noise from wide-band cross-polarization SAR images taking the incident angle into consideration. The specific calculation method includes the following steps:

[0032] Step 1: Using external wind speed data as a reference, obtain the sea surface Sentinel-1 interferometric wide-band dual-polarization SAR image data under low wind speed conditions, which includes both co-polarization and cross-polarization images;

[0033] Specifically, the external wind speed data can use existing public data, such as buoy wind speed data, scatterometer inverted wind speed data, radiometer inverted wind speed data and reanalysis wind field wind speed data. Obtain wide-band dual-polarization SAR images containing both co-polarization and cross-polarization under low wind speed conditions. The present invention takes the interferometric wide-band SAR images of the Sentinel-1 satellite recently launched by the European Space Agency as an example, and obtains a large number of interferometric wide-band images of the Sentinel-1 satellite under low wind speed conditions with VV+VH polarization or HH+HV polarization.

[0034] Step 2: preprocessing the acquired images to obtain co-polarization and cross-polarization backscatter coefficient images with a spatial resolution of 5 km, and an incident angle image with a spatial resolution of 5 km;

[0035] Specifically, after acquiring the SAR image, the image needs to be interpreted first to obtain the polarization mode, incident angle information, original DN value image, backscatter coefficient calibration parameters and thermal noise calibration parameters, and then the image is subjected to radiation calibration, geometric correction, land mask, ship filtering and other steps, and then the pre-processed image is resampled to 5km to eliminate the interference of speckle noise on the image, and the co-polarization and cross-polarization backscatter coefficient images with a spatial resolution of 5km are obtained. Finally, the incident angle information is spatially interpolated by the cubic spline interpolation method to obtain an incident angle image that matches the backscatter coefficient image with a spatial resolution of 5km.

[0036] Step 3: Combined with the wind direction information of the external wind field data, the co-polarization geophysical model function is used to invert the sea surface wind speed of the co-polarization backscattering coefficient image obtained in step 2 to obtain the co-polarization wind speed image at a spatial resolution of 5 km;

[0037] In the implementation of the present invention, the ERA5 wind direction data is used as the external wind direction of the co-polarization model function, and is spatially interpolated to obtain an ERA5 wind direction layer with the same spatial resolution as the co-polarization backscatter coefficient image in step 2. Then, the ERA5 wind direction layer, the co-polarization backscatter coefficient image obtained in step 2, and the incident angle image are input into the co-polarization geophysical model function, and then a wind speed image with a spatial resolution of 5 km is obtained through a minimum cost function.

[0038] For C-band VV polarization SAR data, the C-band VV polarization geophysical model function CMOD4 is used as the model function for wind field inversion. The CMOD4 model is shown as follows:

[0039]

[0040] Among them, σ° VV is the VV polarization radar backscatter coefficient, is the angle between the wind direction and the horizontal direction of the radar incident wind direction, v is the wind speed, θ is the incident angle of the radar, and B0, B1, B2 and B3 are functions of wind speed (v) and incident angle (θ).

[0041] For C-band HH polarimetric SAR data, the C-band HH polarimetric geophysical model function CMODH is used as the model function for wind field inversion. The CMODH model is shown as follows:

[0042]

[0043] Among them, σ° HH is the HH polarization radar backscatter coefficient, is the angle between the wind direction and the horizontal direction of the radar incident wind direction, v is the wind speed, θ is the incident angle of the radar, and B0, B1 and B2 are functions of wind speed (v) and incident angle (θ).

[0044] Step 4: Using the co-polarized wind speed obtained in step 3 as the reference wind speed, the cross-polarized backscattering coefficient corresponding to the area where the sea surface wind speed is less than 2 m / s is extracted from the cross-polarized backscattering coefficient image obtained in step 2 as the thermal noise of the cross-polarized image;

[0045] Specifically, the sea surface in the area with wind speed less than 2m / s is regarded as a calm sea surface, so based on the co-polarized wind speed image obtained in step 3, the area with wind speed less than 2m / s is set to 1, and the others are set to 0, to obtain a mask layer that retains the low wind speed area. Then, the mask layer is multiplied by the cross-polarized backscatter coefficient image obtained in step 2 to obtain the cross-polarized backscatter coefficient that only retains the low wind speed area, so this part of the backscatter coefficient can be regarded as thermal noise.

[0046] Step 5, dividing the thermal noise extracted in step 4 into sub-swaths, respectively establishing a trigonometric function model of the thermal noise of each sub-swath and the incident angle, performing fitting calculation using the thermal noise in each sub-swath and the incident angle, solving the fitting coefficient of the trigonometric function, and thus determining the thermal noise model of each sub-swath;

[0047] Specifically, since the wide-width image is composed of multiple sub-slices, the wide-width image must first be divided into different sub-slices. This step can be done by using the sub-slice mark information of the image as a reference to divide the wide-width image into different slices. Then, the thermal noise and incident angle in each sub-slice are used for fitting calculation to solve the fitting coefficient of the trigonometric function. Taking the Sentinel-1 interferometric wide-width image as an example, it is composed of 3 different sub-slices. The trigonometric function model of the cross-polarization backscattering coefficient of different sub-slices with respect to the incident angle is established respectively, and then the least squares method is used to fit the coefficients of the trigonometric function, and the thermal noise extraction model based on the Sentinel-1 interferometric wide-width SAR data is obtained as follows:

[0048]

[0049] Among them, σ° simulatedNoise is the thermal noise in the form of linear power simulated by the model, and a1, b1, c1, d1, a2, b2, c2, d2, a3, b3, c3, d3 are the fitting coefficients.

[0050] Step 6: Verify the effect of the thermal noise extraction model of the cross-polarization image. First, subtract the product thermal noise calculated according to the thermal noise calculation model in the Sentinel-1 product manual from the cross-polarization backscatter coefficient image obtained in step 2 to obtain the cross-polarization backscatter coefficient with the product thermal noise removed as reference data; then, subtract the simulated thermal noise obtained in step 5 from the cross-polarization backscatter coefficient image obtained in step 2 to obtain the cross-polarization backscatter coefficient with the simulated thermal noise removed; finally, according to the wind speed range of 5m / s, plot the changes of the cross-polarization backscatter coefficient with the simulated thermal noise removed and the product thermal noise removed in different wind speed ranges with the incident angle, so as to evaluate the effect of the thermal noise extraction model.

[0051] Specifically, the calculation formula for the cross-polarization backscatter coefficient after removing the product thermal noise is as follows:

[0052]

[0053] in, is the cross-polarization backscatter coefficient with thermal noise removed in dB, σ° cross is the cross-polarization backscatter coefficient in linear power form obtained in step 2, η is the product thermal noise calibration parameter, and A is the backscatter coefficient calibration factor. Among them, The product thermal noise in the form of linear power, η and A can be extracted from the Sentinel-1 data product. Taking the Sentinel-1 interferometric wide-band cross-polarization SAR data as an example, the extracted product thermal noise in dB is as follows: Figure 2 The black curve shows the product thermal noise in dB. The calculation method is as follows:

[0054]

[0055] The formula for the cross-polarization backscatter coefficient after removing the simulated thermal noise is as follows:

[0056]

[0057] in, is the cross-polarization backscatter coefficient in dB after removing the simulated thermal noise, σ°cross is the cross-polarization backscatter coefficient in linear power form obtained in step 2, σ° simulatedNoise There is thermal noise in the form of linear power obtained in step 2.

[0058] Then, plot the wind speed under different wind speed conditions in every 5m / s wind speed interval. As the incident angle changes, Figure 3 Similarly, plot the wind speed under different conditions As the incident angle changes, Figure 4 As shown. Figure 3 It can be seen that in the case of low wind speed, As the incident angle changes, the Figure 2 The variation trend of the product thermal noise with the incident angle indicates that the thermal noise provided by the product is not very good for removing the thermal noise of the cross-polarization image. Figure 4 It can be seen that at low wind speeds It is basically invisible as the incident angle changes Figure 2 The variation trend of the thermal noise of the product with the incident angle shows that the simulated thermal noise proposed in the present invention is better than the thermal noise provided by the product in removing the thermal noise of the cross-polarization image, which indicates that the thermal noise extraction method proposed in the present invention is feasible.

[0059] Before using wide-width cross-polarization SAR images to invert the sea surface wind field, it is first necessary to remove the interference of thermal noise on the image. However, the thermal noise correction methods provided by some current cross-polarization SAR image products cannot effectively remove the interference of thermal noise on the image, such as Sentinel-1 wide-width cross-polarization image products. In view of the problem that the thermal noise provided by SAR image products cannot be effectively used to remove the interference of thermal noise on cross-polarization images, the present invention takes Sentinel-1 interferometric wide-width (IW) cross-polarization images as an example to provide a wide-width cross-polarization SAR image thermal noise extraction method. The method does not require thermal noise calibration information provided by the image, has the characteristics of simple operation and strong practicality, and is particularly suitable for thermal noise removal of wide-width cross-polarization SAR image products that do not provide thermal noise information or provide inaccurate thermal noise information.

[0060] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a technician familiar with the technical field within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for thermal noise extraction from wide-band cross-polarization SAR images considering the incident angle, characterized in that: The following steps are involved: Step 1, obtaining wide-band dual-polarization SAR image data of the sea surface under low wind speed; Step 2, preprocessing the SAR image data obtained in step 1, and downsampling the image, so as to obtain a downsampled co-polarization backscattering coefficient image, a cross-polarization backscattering coefficient image, and an incident angle image; Step 3, combining the wind direction information, using the co-polarization geophysical model function and the incident angle image of step 2, the co-polarization backscattering coefficient image obtained in step 2 is used to invert the sea surface wind speed to obtain the co-polarization image wind speed; Step 4: Using the co-polarization image wind speed obtained in step 3 as the reference wind speed, extract the cross-polarization backscattering coefficient corresponding to the area where the sea surface wind speed is less than 2 m / s from the cross-polarization backscattering coefficient image obtained in step 2 as the thermal noise of the cross-polarization backscattering coefficient image; Step 5, dividing the thermal noise extracted in step 4 into three sub-crops, and establishing a trigonometric function model related to the thermal noise of each sub-crop and the incident angle; performing fitting calculation using the thermal noise in each sub-crop and the incident angle, solving the fitting coefficient of the trigonometric function model, thereby determining the thermal noise model of each sub-crop, and the thermal noise models of these three sub-crops are the cross-polarization image thermal noise extraction models, thereby realizing wide-width cross-polarization SAR image thermal noise extraction; Step 4 is as follows: The areas in the co-polarized wind speed image obtained in step 3 where the wind speed is less than the preset maximum sea surface wind speed are set to 1, and the other areas are set to 0, to obtain a mask layer that retains the low wind speed area; then a dot multiplication operation is performed on the mask layer and the cross-polarized backscattering coefficient image obtained in step 2 to obtain a cross-polarized backscattering coefficient that only retains the low wind speed area, and the cross-polarized backscattering coefficient is regarded as thermal noise; In step 5, the cross-polarization image thermal noise extraction model is: Among them, the three sub-swaths are the first sub-swath, the second sub-swath and the third sub-swath, σ° simulatedNoise is the thermal noise in the form of linear power simulated by the model, θ is the incident angle of the radar, a1, a2 and a3 are the thermal noise amplitude coefficients of the first sub-swath, the second sub-swath and the third sub-swath respectively, b1, b2 and b3 are the incident angle scaling coefficients of the first sub-swath, the second sub-swath and the third sub-swath respectively, c1, c2 and c3 are the incident angle translation coefficients of the first sub-swath, the second sub-swath and the third sub-swath respectively, d1, d2 and d3 are the thermal noise compensation coefficients of the first sub-swath, the second sub-swath and the third sub-swath respectively.

2. The method for extracting thermal noise from wide-band cross-polarization SAR images considering the incident angle according to claim 1, characterized in that: In step 1, the low wind speed is a wind speed less than 5 m / s.

3. The method for extracting thermal noise from wide-band cross-polarization SAR images considering the incident angle according to claim 1, characterized in that: In step 2, the spatial resolution of the downsampling ranges from 1 km to 5 km.

Images