Terrain feature inversion method based on radar backscatter coefficient Oh model

Abstract

The invention relates to a terrain feature inversion method based on the Oh model of radar backscatter coefficient, and relates to the field of radar reconnaissance and detection technology. The invention aims to solve the key problem of existing terrain feature measurement based on synthetic aperture radar, which is the large error caused by surface roughness. The technical solution of the invention includes: step 1, analyzing the correlation between the VV polarization backscatter coefficient of the reconnaissance and detection synthetic aperture radar and the surface parameters; step 2, analyzing the correlation between the VV polarization backscatter coefficient of the reconnaissance and detection synthetic aperture radar and the surface terrain characteristics; step 3, analyzing the correlation between the VH polarization backscatter coefficient of the reconnaissance and detection synthetic aperture radar and the surface parameters; step 4, analyzing the correlation between the VH / VV polarization backscatter coefficient and roughness based on the reconnaissance and detection synthetic aperture radar data; step 5, introducing vegetation degree to improve the water cloud model; step 6, analyzing the vegetation parameters; step 7, using the Oh model to invert soil moisture and obtain the backscatter coefficient of bare soil.

Description

Technical Field

[0001] The present invention belongs to the field of geological and hydrological survey technology, and in particular relates to a terrain feature inversion method based on a radar backscatter coefficient Oh model. Background Art

[0002] The radar scattering coefficient, also known as the backscatter coefficient, refers to the radar reflectivity per unit cross-sectional area of a target in the incident direction. It is a parameter that indicates the scattering intensity in the incident direction or the average radar cross-section per unit area of the target. It is usually expressed in decibels. Surface scattering refers to scattering generated on the surface of a medium. Factors influencing surface scattering include the dielectric constant and surface roughness. Natural surfaces can be decomposed into a series of planar elements with small-scale geometry (i.e., roughness). This small-scale surface geometry can be represented by the statistical standard deviation of height (root mean square height) and the surface correlation length, which describe the roughness in the vertical and horizontal directions, respectively.

[0003] However, one of the key issues in existing synthetic aperture radar-based terrain feature measurements is the large error caused by surface roughness. Summary of the Invention

[0004] The present invention aims to solve the problem of large errors caused by surface roughness in the key problem of inverting terrain features based on synthetic aperture radar remote sensing. The present invention proposes a terrain feature inversion method based on the radar backscatter coefficient Oh model.

[0005] The technical solution of the present invention comprises the following steps:

[0006] Step 1: Analyze the correlation between the VV polarization backscatter coefficient of the reconnaissance synthetic aperture radar and the surface parameters;

[0007] Step 2: Analyze the correlation between the VV polarization backscatter coefficient of the reconnaissance synthetic aperture radar and the surface terrain characteristics;

[0008] Step 3: Analyze the correlation between the VH polarization backscatter coefficient of the reconnaissance synthetic aperture radar and the surface parameters;

[0009] Step 4: Analyze the correlation between the VH / VV polarization backscatter coefficient and roughness based on the reconnaissance synthetic aperture radar data;

[0010] Step 5: Introduce vegetation coverage to improve the water cloud model;

[0011] Step 6: Analyze vegetation parameters;

[0012] Step 7: Use the Oh model to invert soil moisture and obtain the backscattering coefficient of bare soil.

[0013] Preferably, the step 1 includes:

[0014] Step 1.1: Analyze the VV polarization backscatter coefficient of the reconnaissance detection radar The relationship between the correlation length l;

[0015] In the Oh model, the input parameters only change the terrain characteristics and root mean square height of the radar detection surface, and observe the response of the VV polarization backscattering coefficient of the reconnaissance detection radar to the correlation length under different terrain characteristics and different root mean square heights. In the study of different detection surface terrain characteristics Ms, the correlation length l and the reconnaissance detection radar backscattering coefficient When the relationship between the detection surface correlation length l is set to [1cm, 18cm], the step size is 0.3cm, and the terrain feature Ms is set to [0.01cm 3 / cm 3 , 0.31cm 3 / cm 3 ], with an interval of 0.02 cm 3 / cm 3 ;

[0016] Step 1.2: Analyze the VV polarization backscatter coefficient of the reconnaissance detection radar Response to the radar detection of the ground surface root mean square height S;

[0017] In the Oh model, the input parameters only change the radar detection surface terrain features Ms and correlation length L, and observe the VV polarization backscattering coefficient of the reconnaissance detection radar under different terrain features Ms and different correlation lengths L. The response to the root mean square height S, in the study of different terrain features Ms, reconnaissance detection radar VV polarization backscatter coefficient When responding to the root mean square height S, the value range of the root mean square height S is set to [0.001cm, 3cm], and the step size is 0.05cm.

[0018] Preferably, the step 2 includes:

[0019] Step 2.1: Analyze the VV polarization backscatter coefficient of the reconnaissance detection radar The relationship between the radar detection of the surface terrain features Ms;

[0020] In the Oh model, the input parameters only change the radar detection surface root mean square height and correlation length, and observe the VV polarization backscatter coefficient of the reconnaissance detection radar at different root mean square heights and different correlation lengths. The response to the detection of surface terrain features Ms is studied in the study of different correlation lengths L, terrain features Ms and reconnaissance detection radar VV polarization backscattering coefficient When the relationship is, the value range of Ms is [0.01cm 3 / cm 3 , 0.31cm 3 / cm 3 ], with a step length of 0.005 cm 3 / cm 3 .

[0021] Preferably, the step 3 includes:

[0022] Step 3.1: Analyze the VH polarization backscatter coefficient of the reconnaissance detection radar The relationship between the radar detection surface correlation length l;

[0023] In the Oh model, the input parameters only change the terrain characteristics and root mean square height of the radar detection surface, and observe the response of the VH polarization backscattering coefficient of the reconnaissance detection radar to the correlation length under different terrain characteristics and different root mean square heights. In the study of different radar detection surface terrain characteristics Ms, the correlation length l and the reconnaissance detection radar backscattering coefficient When the relationship is established, the correlation length l is set to [1cm, 18cm], the step length is 0.3cm, and the radar detection surface terrain feature Ms is set to [0.01cm 3 / cm 3 , 0.31cm 3 / cm 3 ], with an interval of 0.02 cm 3 / cm 3 ;

[0024] Step 3.2: Analyze the VH polarization backscatter coefficient of the reconnaissance detection radar Response to the root mean square height S of the detected surface;

[0025] In the Oh model, the input parameters only change the terrain characteristics Ms and correlation length L of the detection surface, and observe the VH polarization backscattering coefficient of the reconnaissance detection radar under different detection surface terrain characteristics Ms and different correlation lengths L. The response to the root mean square height S, in the study of different detection surface terrain features Ms, reconnaissance detection radar VH polarization backscatter coefficient When responding to the root mean square height S, the value range of the root mean square height S is set to [0.001cm, 3cm], and the step size is 0.05cm.

[0026] Preferably, the step 4 includes:

[0027] Step 4.1: Analyze the VH / VV polarization backscatter coefficient of the reconnaissance detection radar The relationship between the radar detection surface correlation length l;

[0028] In the Oh model, the input parameters only change the terrain characteristics and root mean square height of the detection surface, and observe the response of the VH / VV polarization backscatter coefficient of the reconnaissance detection radar to the correlation length under different detection surface terrain characteristics Ms and different root mean square heights. The relationship between the different detection surface terrain characteristics Ms, the correlation length l and the reconnaissance detection radar backscatter coefficient is studied. The relationship is consistent with that described in step 3.1;

[0029] Step 1.2: Analyze the VH / VV polarization backscatter coefficient of the reconnaissance detection radar Response to the root mean square height S of the detected surface;

[0030] In the Oh model, the input parameters only change the terrain characteristics Ms and correlation length L of the detection surface, and observe the VH / VV polarization backscattering coefficient of the reconnaissance detection radar under different terrain characteristics Ms and different correlation lengths L. Study the VH / VV polarization backscatter coefficient of reconnaissance detection radar in response to the root mean square height S, different terrain features Ms The response to the RMS height S is the same as described in step 3.2.

[0031] Preferably, in step 5, in order to better estimate the backscatter coefficient value, the vegetation coverage of the detected surface is introduced to improve the water cloud model and calculate the actual backscatter coefficient. The specific form is shown in formula (1):

[0032]

[0033] in, is the backscatter coefficient of the total reconnaissance detection radar; F veg To detect the vegetation coverage of the surface pixels, F veg The value is between (0, 1). When it approaches 1, it means that the vegetation coverage area of the detected surface accounts for a larger proportion. On the contrary, when it approaches 0, the exposed surface area of the detected surface accounts for a larger proportion. veg Sentinel-2 data can be used to calculate vegetation coverage using a pixel binary model, as shown in formula (2):

[0034] F veg =(NDVI-NDVI min ) / (NDVI max -NDVI min ) (2)

[0035] Among them, NDVI is the normalized difference vegetation index value calculated from Sentinel-2 data; NDVI minThe NDVI is the normalized vegetation index value of the exposed area, which is theoretically 0; max is the normalized vegetation index value of the vegetation coverage area; the values corresponding to the 5% and 95% cumulative distribution probability of the NDVI value in the detection area are selected as the NDVI min and NDVI max , combining formula (2) with the water cloud model to obtain formula (3):

[0036]

[0037] Preferably, in step 6, the A and B values under different vegetation types of the detected surface are first determined, the backscattering coefficient of the soil sample point is used, the A and B values are fixed and the value of the other parameter is changed, and the water cloud model improved by the vegetation coverage of the detected surface is run to obtain the backscattering coefficient under different A and B values based on the detected surface.

[0038] Preferably, in step 7, the response of the backscatter coefficient of the reconnaissance detection radar to the detection surface parameters under different polarizations is studied. For the Oh model, the correlation length L is set in the range of 1 to 18 cm with a step length of 3 cm; the root mean square height s is set in the range of 0.4 to 2.2 cm with a step length of 0.1 cm; the terrain feature Ms is set to 0.01 to 0.31 cm 3 / cm 3 , step length is 0.005cm 3 / cm 3 At the same time, the free space wave number and other parameters are input, and the Oh model is run multiple times to obtain a database consisting of the corresponding reconnaissance detection radar backscatter coefficients under different correlation lengths, different root mean square heights and other conditions. The backscatter coefficients of the pixels are obtained from the detection SAR image, and the bare soil backscatter coefficients that remove the influence of vegetation cover in the detection area are found by the lookup table method. The bare soil backscattering coefficient simulated by Oh with the minimum cost function (M) The specific form of the cost function is as follows:

[0039]

[0040] The beneficial effect of the present invention lies in that it addresses the error caused by surface roughness in the key issue of soil moisture inversion based on multi-source remote sensing. First, the backscattering coefficient is corrected using a water-cloud model to remove the influence of winter wheat cover in farmland, obtaining the contribution of the direct backscattering coefficient of the surface soil, and then the Oh model is used to invert soil moisture. When the water-cloud model is used to remove the influence of crops, it includes the inversion of crop water content, the reference of vegetation cover, and the solution of model coefficients. Finally, the backscattering coefficient of bare soil is obtained, which solves the problem of large errors caused by surface roughness in the key issue of soil moisture inversion based on multi-source remote sensing. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 Schematic diagram of the relationship between correlation length and VV polarization backscattering coefficient at different soil moisture levels;

[0042] Figure 2 Schematic diagram of the relationship between the correlation length and the VV polarization backscattering coefficient at different root mean square heights;

[0043] Figure 3 Schematic diagram of the relationship between root mean square height and VV polarization backscattering coefficient at different soil moisture levels;

[0044] Figure 4 Schematic diagram of the relationship between the RMS height and the VV polarization backscattering coefficient at different correlation lengths;

[0045] Figure 5 Schematic diagram of the relationship between soil moisture and VV polarization backscattering coefficient at different correlation lengths;

[0046] Figure 6 Schematic diagram of the relationship between soil moisture and VV polarization backscattering coefficient at different root mean square heights;

[0047] Figure 7 Schematic diagram of the relationship between root mean square height and VH polarization backscattering coefficient at different soil moisture levels;

[0048] Figure 8 Schematic diagram of the relationship between soil moisture and VH polarized backscattering coefficient at different root mean square heights;

[0049] Figure 9 Schematic diagram of the relationship between the correlation length and the VH / VV polarization backscattering coefficient at different root mean square heights;

[0050] Figure 10 Schematic diagram of the relationship between the root mean square height and the VH / VV polarization backscattering coefficient at different correlation lengths;

[0051] Figure 11 It is a line graph of the backscattering coefficient changes at different B values;

[0052] Figure 12 It is a line graph of the backscattering coefficient changes at different A values;

[0053] Figure 13 This is the flow chart of soil inversion based on Oh model;

[0054] Figure 14 To remove the influence of vegetation on VV backscatter coefficient comparison;

[0055] Figure 15 To remove the influence of vegetation on VH backscatter coefficient comparison;

[0056] Figure 16 Comparison of soil moisture inversion accuracy by removing roughness effects for VV polarization;

[0057] Figure 17 Comparison of soil moisture inversion accuracy after removing roughness effects for VH polarization;

[0058] Figure 18 This is the spatial distribution map of soil moisture in the study area;

[0059] Figure 19 This is the spatial distribution map of the fused vegetation index in the study area. DETAILED DESCRIPTION

[0060] Specific implementation method 1: This implementation method is described in detail with reference to the accompanying drawings. Figure 1-19 As shown, the terrain feature inversion based on the radar backscatter coefficient Oh model described in this embodiment includes the following steps:

[0061] Step 1: Analyze the correlation between the VV polarization backscatter coefficient of the reconnaissance synthetic aperture radar and the surface parameters;

[0062] Step 2: Analyze the correlation between the VV polarization backscatter coefficient of the reconnaissance synthetic aperture radar and the surface terrain characteristics;

[0063] Step 3: Analyze the correlation between the VH polarization backscatter coefficient of the reconnaissance synthetic aperture radar and the surface parameters;

[0064] Step 4: Analyze the correlation between the VH / VV polarization backscatter coefficient and roughness based on the reconnaissance synthetic aperture radar data;

[0065] Step 5: Introduce vegetation coverage to improve the water cloud model;

[0066] Step 6: Analyze vegetation parameters;

[0067] Step 7: Use the Oh model to invert soil moisture and obtain the backscattering coefficient of bare soil.

[0068] Specific embodiment 2: This embodiment is described in detail with reference to the accompanying drawings. Step 1 includes:

[0069] Step 1.1: Analyze the VV polarization backscatter coefficient of the reconnaissance detection radar The relationship between the correlation length l;

[0070] In the Oh model, the input parameters only change the terrain characteristics and root mean square height of the radar detection surface, and observe the response of the VV polarization backscattering coefficient of the reconnaissance detection radar to the correlation length under different terrain characteristics and different root mean square heights. In the study of different detection surface terrain characteristics Ms, the correlation length l and the reconnaissance detection radar backscattering coefficient When the relationship between the detection surface correlation length l is set to [1cm, 18cm], the step size is 0.3cm, and the terrain feature Ms is set to [0.01cm 3 / cm 3 , 0.31cm 3 / cm 3 ], with an interval of 0.02 cm 3 / cm 3 ;

[0071] Step 1.2: Analyze the VV polarization backscatter coefficient of the reconnaissance detection radar Response to the radar detection of the ground surface root mean square height S;

[0072] In the Oh model, the input parameters only change the radar detection surface terrain features Ms and correlation length L, and observe the VV polarization backscattering coefficient of the reconnaissance detection radar under different terrain features Ms and different correlation lengths L. The response to the root mean square height S, in the study of different terrain features Ms, reconnaissance detection radar VV polarization backscatter coefficient When responding to the root mean square height S, the root mean square height S value range is set to [0.001cm, 3cm], and the step size is 0.05cm. The result is as follows Figure 3 shown.

[0073] In this embodiment, if Figure 1-4 As shown, Figure 1 The difference is, Figure 2 Shown are the VV polarization backscatter coefficients at different RMS heights s. The response change diagram of the correlation length l, at this time the soil moisture Ms is a constant value, the root mean square height s range is [0.01cm, 0.31cm], and the interval is 0.02cm. Figure 4-1 We can see that under the same soil moisture Ms, the VV polarization backscattering coefficient As the correlation length L increases, when L is greater than 3 cm, The rate of increase is slower than that of L less than 3 cm; at the same L, VV polarization backscattering coefficient It increases with the increase of soil moisture Ms, and the growth rate tends to decrease. Figure 4-2 It can be seen that when the root mean square height S is 0.01cm and 0.03cm respectively, the VV polarization backscattering coefficient As the correlation length L increases, the speed of increase is relatively slow, and at the same L, the distance between the two The backscattering coefficient changes by more than 1dB with a large change in S. When S exceeds 0.03cm, The change is within 1dB. As S gradually approaches 0.31cm, under the same L, As S increases, the speed gradually slows down. At the same root mean square height S, and within the relevant length L range, The change is about 1dB.

[0074] Figure 4 The results show that the VV polarization backscattering coefficients at different correlation lengths L are Response change diagram of the root mean square height S, where the value of L ranges from [1cm, 18cm] with an interval of 1cm.

[0075] from Figure 3 It can be seen that at the same soil moisture Ms, VV polarization backscattering coefficient It increases rapidly with the increase of the root mean square height S, and then gradually slows down. At the same root mean square height S, As Ms increases, the growth rate shows a downward trend and gradually becomes flat, and then has a decreasing trend. Under different correlation lengths L, The response to S is Figure 4 It can be seen that, under the same L, As S increases, it first increases and then decreases, and as L increases, The speed of reduction slows down and gradually tends to a horizontal state; within the range of the root mean square height S, The change is about 6dB.

[0076] contrast Figure 1 、 Figure 2 、 Figure 3 and Figure 4 It can be seen that the trends of the first three figures are similar. Figure 1 and Figure 2 , soil moisture from 0.01cm 3 / cm 3 Increased to 0.31cm 3 / cm 3When the correlation length L increases from 1 cm to 18 cm and the root mean square height S increases from 0.01 cm to 0.31 cm The increase is similar, both increasing by about 10dB, indicating that the VV polarization backscattering coefficient in the study area is more sensitive to changes in soil moisture. Figure 3 and Figure 4 When the root mean square height S increases from 0.01cm to 0.31, the soil moisture increases from 0.01 to 0.31. The change amplitude is about 12dB; when the correlation length L increases from 1cm to 18cm The change amplitude is about 9dB. Figure 2 and Figure 4 It can be seen that The response to S is stronger than that to L.

[0077] Specific embodiment 3: This embodiment is described in detail with reference to the accompanying drawings. Step 2 includes:

[0078] Step 2.1: Analyze the VV polarization backscatter coefficient of the reconnaissance detection radar The relationship between the radar detection of the surface terrain features Ms;

[0079] In the Oh model, the input parameters only change the radar detection surface root mean square height and correlation length, and observe the VV polarization backscatter coefficient of the reconnaissance detection radar at different root mean square heights and different correlation lengths. The response to the detection of surface terrain features Ms is studied in the study of different correlation lengths L, terrain features Ms and reconnaissance detection radar VV polarization backscattering coefficient When the relationship is, the value range of Ms is [0.01cm 3 / cm 3 , 0.31cm 3 / cm 3 ], with a step length of 0.005 cm 3 / cm 3 .

[0080] In this embodiment, if Figure 5 and Figure 6 As shown in the figure, under the same correlation length L or the same root mean square height S, the VV polarization backscattering coefficient The change trends of the two figures are similar. Under the same L or the same S, Ms increases from 0.01cm 3 / cm 3 Increased to 0.31cm 3 / cm 3 , The change is about 10dB, which once again proves the sensitivity of radar backscatter coefficient to the change of soil moisture. The magnitude of the change with the increase of S is larger than that with the increase of L.

[0081] Specific embodiment 4: This embodiment is described in detail with reference to the accompanying drawings. Step 3 includes:

[0082] Step 3.1: Analyze the VH polarization backscatter coefficient of the reconnaissance detection radar The relationship between the radar detection surface correlation length l;

[0083] In the Oh model, the input parameters only change the terrain characteristics and root mean square height of the radar detection surface, and observe the response of the VH polarization backscattering coefficient of the reconnaissance detection radar to the correlation length under different terrain characteristics and different root mean square heights. In the study of different radar detection surface terrain characteristics Ms, the correlation length l and the reconnaissance detection radar backscattering coefficient When the relationship is established, the correlation length l is set to [1cm, 18cm], the step length is 0.3cm, and the radar detection surface terrain feature Ms is set to [0.01cm 3 / cm 3 , 0.31cm 3 / cm 3 ], with an interval of 0.02 cm 3 / cm 3 ;

[0084] Step 3.2: Analyze the VH polarization backscatter coefficient of the reconnaissance detection radar Response to the root mean square height S of the detected surface;

[0085] In the Oh model, the input parameters only change the terrain characteristics Ms and correlation length L of the detection surface, and observe the VH polarization backscattering coefficient of the reconnaissance detection radar under different detection surface terrain characteristics Ms and different correlation lengths L. The response to the root mean square height S, in the study of different detection surface terrain features Ms, reconnaissance detection radar VH polarization backscatter coefficient When responding to the root mean square height S, the value range of the root mean square height S is set to [0.001cm, 3cm], and the step size is 0.05cm.

[0086] In this embodiment, if Figure 7 and Figure 8As shown. Similar to the input to the Oh model when observing VV polarization and surface parameters, since only RMS height and soil moisture are involved in the study of VH polarization and surface parameters, this summary only studies the relationship between RMS height and soil moisture and VH polarization backscattering coefficient under different soil moisture and different RMS height. The input range of parameters can be referred to the previous introduction. Figure 7 and Figure 8 It can be seen that under the same Ms, when S increases from 0.001cm to 3cm, Increased by about 32dB; under the use of S, Ms from 0.01cm 3 / cm 3 Increased to 0.31cm 3 / cm 3 hour, This shows that under VH polarization, the VH polarization backscatter coefficient is less sensitive to changes in soil moisture than to changes in RMS height. When using only the VH polarization backscatter coefficient to invert soil moisture, it is necessary to remove the interference of roughness on radar waves.

[0087] Specific embodiment 5: This embodiment is described in detail with reference to the accompanying drawings. Step 4 includes:

[0088] Step 4.1: Analyze the VH / VV polarization backscatter coefficient of the reconnaissance detection radar The relationship between the radar detection surface correlation length l;

[0089] In the Oh model, the input parameters only change the terrain characteristics and root mean square height of the detection surface, and observe the response of the VH / VV polarization backscatter coefficient of the reconnaissance detection radar to the correlation length under different detection surface terrain characteristics Ms and different root mean square heights. The relationship between the different detection surface terrain characteristics Ms, the correlation length l and the reconnaissance detection radar backscatter coefficient is studied. The relationship is consistent with that described in step 3.1;

[0090] Step 1.2: Analyze the VH / VV polarization backscatter coefficient of the reconnaissance detection radar Response to the root mean square height S of the detected surface;

[0091] In the Oh model, the input parameters only change the terrain characteristics Ms and correlation length L of the detection surface, and observe the VH / VV polarization backscattering coefficient of the reconnaissance detection radar under different terrain characteristics Ms and different correlation lengths L. Study the VH / VV polarization backscatter coefficient of reconnaissance detection radar in response to the root mean square height S, different terrain features Ms The response to the RMS height S is the same as described in step 3.2.

[0092] In this embodiment, if Figure 9 and Figure 10 As shown in the figure, in order to simulate the relationship between the VH / VV polarized soil backscattering coefficient and the surface roughness, the input to the Oh model is similar to that when observing the VV polarization and the surface parameters. Since only the roughness parameter is involved in the study of VH / VV polarization and the surface parameters, this summary only studies the relationship between the correlation length and the RMS height and the VH / VV polarized backscattering coefficient at different RMS heights and different correlation lengths. The input range of the parameters can be referred to the previous introduction, from Figure 9 and Figure 10 It can be seen that at the same root mean square height S, the backscattering coefficient VH / VV It decreases with the increase of the correlation length L; within the range of S, as S approaches 0.31, The change is about 2dB. Under the same L, It increases with the increase of S, and when L approaches 3 cm, its value increases by about 32 dB.

[0093] Specific embodiment 6: This embodiment is described in detail with reference to the accompanying drawings. In step 5, in order to better estimate the backscatter coefficient value, the vegetation coverage of the detected surface is introduced to improve the water cloud model and calculate the actual backscatter coefficient. The specific form is shown in formula (1):

[0094]

[0095] in, is the backscatter coefficient of the total reconnaissance detection radar; F veg To detect the vegetation coverage of the surface pixels, F veg The value is between (0, 1). When it approaches 1, it means that the vegetation coverage area of the detected surface accounts for a larger proportion. On the contrary, when it approaches 0, the exposed surface area of the detected surface accounts for a larger proportion. veg Sentinel-2 data can be used to calculate vegetation coverage using a pixel binary model, as shown in formula (2):

[0096] F veg =(NDVI-NDVI min ) / (NDVI max -NDVI min ) (2)

[0097] Among them, NDVI is the normalized difference vegetation index value calculated from Sentinel-2 data; NDVI min The NDVI is the normalized vegetation index value of the exposed area, which is theoretically 0; maxis the normalized vegetation index value of the vegetation coverage area; the values corresponding to the 5% and 95% cumulative distribution probability of the NDVI value in the detection area are selected as the NDVI min and NDVI max , combining formula (2) with the water cloud model to obtain formula (3):

[0098]

[0099] Specific implementation method seven: This implementation method is described in detail with reference to the accompanying drawings. First, the A and B values under different vegetation types on the detected surface are determined. The backscattering coefficient of the soil sample point is used. The A and B values are fixed and the value of the other parameter is changed. The water cloud model with improved vegetation coverage based on the detected surface is run to obtain the backscattering coefficient under different A and B values based on the detected surface.

[0100] In this implementation, it is known that the water cloud model with improved vegetation coverage needs to obtain the empirical constants A and B values related to crops. Some researchers obtain the backscatter coefficient of vegetation through costly field measurements; some researchers also use Bindlish

[55] Table 4-1 shows empirical values obtained experimentally for different vegetation types. In practice, vegetation types are diverse, and relatively uniform empirical values can lead to deviations in soil moisture inversion. This paper combines the empirical values provided by Bindlish with measured soil moisture and employs the optimal precision theory to obtain the optimal A and B values for different polarizations.

[0101] Table 4-1 A and B values under different vegetation types

[0102]

[0103] As can be seen from Table 4-1, this type of input requires a high level of fine classification of radar images. In actual ground objects, multiple pixels may appear in the same pixel. In this case, using fixed parameters will increase the moisture inversion error. Using the backscatter coefficient of the soil sample point, fix the A and B values and change the value of the other parameter. Run the water cloud model after the vegetation coverage is improved, and obtain the backscatter coefficient under different A and B values. Plot it as shown below. Figure 11 、 12 The line chart shown.

[0104] contrast Figure 11 and Figure 12 It can be seen that when A is constant and B is different, the VV polarization backscatter coefficient changes by about 2 dB. Compared with the case when B is constant and A is different, the VV polarization backscatter coefficient changes by a larger magnitude, and the magnitude of the change in A is larger than that of the change in B (the difference in the value of A is 10-4 The magnitude of B is 10 -2 Order of magnitude). It can be seen that the value of A has little effect on the results, so this paper selects an A value of 0.0018 to input into the improved water cloud model. The present invention proposes to simplify the input of vegetation parameters based on the idea of optimization theory. Compared with the use of unified A and B parameters, it can improve the accuracy of the simulated soil backscatter coefficient and more accurately describe the attenuation degree of vegetation on the radar backscatter coefficient. The root mean square error and mean absolute error of the measured soil moisture and the inverted soil moisture are used as the evaluation criteria to obtain the B value suitable for the study area. After calculation, the A and B values of different polarizations are shown in Table 4-2:

[0105] Table 4-2 A and B values under different polarizations

[0106]

[0107]

[0108] Specific embodiment eight: This embodiment is described in detail with reference to the accompanying drawings. In step 7, the response of the backscatter coefficient of the reconnaissance detection radar to the detection surface parameters under different polarizations is studied. For the Oh model, the correlation length L is set in the range of 1 to 18 cm, with a step size of 3 cm; the root mean square height s is in the range of 0.4 to 2.2 cm, with a step size of 0.1 cm; the terrain feature Ms is set to 0.01 to 0.31 cm 3 / cm 3 , step length is 0.005cm 3 / cm 3 At the same time, the free space wave number and other parameters are input, and the Oh model is run multiple times to obtain a database consisting of the corresponding reconnaissance detection radar backscatter coefficients under different correlation lengths, different root mean square heights and other conditions. The backscatter coefficients of the pixels are obtained from the detection SAR image, and the bare soil backscatter coefficients that remove the influence of vegetation cover in the detection area are found by the lookup table method. The bare soil backscattering coefficient simulated by Oh with the minimum cost function (M) The specific form of the cost function is as follows:

[0109]

[0110] In this implementation, soil moisture inversion based on the Oh model primarily involves simulating the nonlinear relationship between roughness, backscattering coefficient, and soil moisture using the Oh model, then constructing a lookup table of radar backscattering coefficients for different surface roughness and soil moisture levels. This method bypasses the input of measured roughness data and has minimal impact on the real-time performance of the inversion results. By leveraging the relationship between surface roughness, soil moisture, and radar backscattering coefficient in the Oh model, the interference of roughness on radar waves can be eliminated. In the Oh model, when the root mean square height and correlation length take certain values, the correlation between the correlation length, root mean square height, and backscattering coefficient is high. At this point, the radar backscattering coefficient simulated using the Oh model is closest to the true value. When establishing the LUT table, it is necessary to determine the range and step size of the surface parameters. Then, parameters such as the radar incident angle and wavelength are input into the Oh model to simulate the backscattering coefficients for different surface parameters and store them in a database, thereby establishing the LUT table. Finally, the backscatter coefficient of the input pixel or the backscatter coefficient without vegetation coverage is used to find the soil moisture corresponding to the minimum value of the cost function as the inverted soil moisture value.

[0111] The inversion process can be roughly divided into three parts: model improvement, soil moisture inversion, and accuracy evaluation. First, we obtain preprocessed parameters such as the radar incident angle, backscatter coefficient, and vegetation index. Unlike in Section 3, this section introduces vegetation coverage as an input to the water cloud model to better represent vegetation information. Secondly, we construct a cost function, find its minimum value, and finally invert soil moisture. Finally, we evaluate the inversion results against the measured results using the three indicators introduced in Section 3, and obtain a spatial distribution map of the study area.

[0112] The experimental results are as follows Figure 14-19 As shown in Figure 14, through previous research, the required vegetation parameters have been obtained as input, along with the radar incident angle and the total backscatter coefficient of the SAR image as input to the improved water cloud model. Then, by running the model, the contribution of soil to the backscatter coefficient can be obtained. The changes in the backscatter coefficients of 50 VV polarization and VH polarization sampling points before and after removing the vegetation contribution are shown in Figures 14 and 15. Figure 14 and Figure 15 It can be seen that after removing the contribution of vegetation to the backscatter coefficient, its value decreases slightly. The magnitude of the reduction is related to vegetation coverage, vegetation type, and the total SAR backscatter coefficient. Specifically, VV polarization decreases by approximately 2dB, and VH polarization decreases by approximately 5dB. Comparing the changes in VV and VH polarization values again shows that VH polarization is significantly affected by vegetation.

[0113] This paper uses Matlab to estimate soil moisture in the study area using two polarization SAR images before and after removing the influence of winter wheat cover. A comparison chart of the accuracy of measured soil moisture and inverted values is drawn, as shown in the figure below. Figure 16 、 17 As shown. Figure 16 、 17 It is not difficult to see that the accuracy of soil moisture inversion using VV polarization data is better than that using VH polarization data. 2 The difference is 0.2578; the RMSE difference is 0.0031; and the MAE difference is 0.003. This also shows that cross-polarization VH is weaker than co-polarization VV data in soil moisture inversion applications.

[0114] The soil moisture in the study area was inverted based on the LUT table method using VV polarized SAR data after removing vegetation images. Taking March 22, 2020 as an example, the results are as follows: Figure 18-19 shown; contrast Figure 18 and Figure 19 It can be seen that the overall soil moisture in the area is at a relatively low level, which is consistent with the actual situation that the soil is relatively dry after winter wheat overwintering; the vegetation index value in the middle part is relatively high, which indirectly indicates that the vegetation water content in this area is relatively high, which is consistent with the actual winter wheat distribution area.

[0115] The present invention first analyzes the correlation between VV polarization and VH polarization backscattering coefficients and roughness-related parameters and soil moisture, and determines the parameter step size when constructing the lookup table. The water cloud model with improved vegetation coverage is combined with the Oh model to invert soil moisture, and the radar backscattering coefficients before and after removing crop cover are compared. The soil moisture inversion part includes the correlation analysis between surface parameters and radar backscattering coefficients, the construction of the lookup table, and the calculation of the cost function. Finally, the spatial distribution map of soil moisture in the study area is obtained, and the measured data of the sampling points and the inversion results are used to use R 2 , RMSE and MAE were used to evaluate the inversion results.

Claims

1. A terrain feature inversion method based on the radar backscatter coefficient Oh model is characterized by: The following steps are involved: Step 1: Analyze the correlation between the VV polarization backscatter coefficient of the reconnaissance synthetic aperture radar and the surface parameters; Step 2: Analyze the correlation between the VV polarization backscatter coefficient of the reconnaissance synthetic aperture radar and the surface terrain characteristics; Step 3: Analyze the correlation between the VH polarization backscatter coefficient of the reconnaissance synthetic aperture radar and the surface parameters; Step 4: Analyze the correlation between the VH / VV polarization backscatter coefficient and roughness based on the reconnaissance synthetic aperture radar data; Step 5: Introduce vegetation coverage to improve the water cloud model; Step 6: Analyze vegetation parameters; Step 7: Use the Oh model to invert soil moisture and obtain the backscattering coefficient of bare soil; The step 1 includes: Step 1.1: Analyze the VV polarization backscatter coefficient of the reconnaissance detection radar The relationship between the correlation length l; In the Oh model, the input parameters only change the terrain characteristics and root mean square height of the radar detection surface, and observe the response of the VV polarization backscattering coefficient of the reconnaissance detection radar to the correlation length under different terrain characteristics and different root mean square heights. In the study of different detection surface terrain characteristics Ms, the correlation length l and the reconnaissance detection radar backscattering coefficient When the relationship is established, the value range of the detection surface correlation length l is set to [1cm, 18cm], the step size is 0.3cm, and the value range of the terrain feature Ms is [0.01cm 3 / cm 3 , 0.31cm 3 / cm 3 ], with an interval of 0.02 cm 3 / cm 3 ; Step 1.2: Analyze the VV polarization backscatter coefficient of the reconnaissance detection radar Response to the radar detection of the ground surface root mean square height S; In the Oh model, the input parameters only change the radar detection surface terrain features Ms and correlation length L, and observe the VV polarization backscattering coefficient of the reconnaissance detection radar under different terrain features Ms and different correlation lengths L. The response to the root mean square height S, in the study of different terrain features Ms, reconnaissance detection radar VV polarization backscatter coefficient When responding to the root mean square height S, the value range of the root mean square height S is set to [0.001cm, 3cm], and the step size is 0.05cm; The step 2 includes: Step 2.1: Analyze the VV polarization backscatter coefficient of the reconnaissance detection radar The relationship between the radar detection of the surface terrain features Ms; In the Oh model, the input parameters only change the radar detection surface root mean square height and correlation length, and observe the VV polarization backscatter coefficient of the reconnaissance detection radar at different root mean square heights and different correlation lengths. The response to the detection of surface terrain features Ms is studied in the study of different correlation lengths L, terrain features Ms and reconnaissance detection radar VV polarization backscattering coefficient When the relationship is, the value range of Ms is [0.01cm 3 / cm 3 , 0.31cm 3 / cm 3 ], with a step length of 0.005 cm 3 / cm 3 ; The step 3 includes: Step 3.1: Analyze the VH polarization backscatter coefficient of the reconnaissance detection radar The relationship between the radar detection surface correlation length l; In the Oh model, the input parameters only change the terrain characteristics and root mean square height of the radar detection surface, and observe the response of the VH polarization backscattering coefficient of the reconnaissance detection radar to the correlation length under different terrain characteristics and different root mean square heights. In the study of different radar detection surface terrain characteristics Ms, the correlation length l and the reconnaissance detection radar backscattering coefficient When the relationship is established, the correlation length l is set to [1cm, 18cm], the step length is 0.3cm, and the radar detection surface terrain feature Ms is set to [0.01cm 3 / cm 3 , 0.31cm 3 / cm 3 ], with an interval of 0.02 cm 3 / cm 3 ; Step 3.2: Analyze the VH polarization backscatter coefficient of the reconnaissance detection radar Response to the root mean square height S of the detected surface; In the Oh model, the input parameters only change the terrain characteristics Ms and correlation length L of the detection surface, and observe the VH polarization backscattering coefficient of the reconnaissance detection radar under different detection surface terrain characteristics Ms and different correlation lengths L. The response to the root mean square height S, in the study of different detection surface terrain features Ms, reconnaissance detection radar VH polarization backscatter coefficient When responding to the root mean square height S, the value range of the root mean square height S is set to [0.001cm, 3cm], and the step size is 0.05cm; The step 4 includes: Step 4.1: Analyze the VH / VV polarization backscatter coefficient of the reconnaissance detection radar The relationship between the radar detection surface correlation length l; In the Oh model, the input parameters only change the terrain characteristics and root mean square height of the detection surface, and observe the response of the VH / VV polarization backscatter coefficient of the reconnaissance detection radar to the correlation length under different detection surface terrain characteristics Ms and different root mean square heights. The relationship between the different detection surface terrain characteristics Ms, the correlation length l and the reconnaissance detection radar backscatter coefficient is studied. The relationship is consistent with step 3.1; Step 4.2: Analyze the VH / VV polarization backscatter coefficient of the reconnaissance detection radar Response to the root mean square height S of the detected surface; In the Oh model, the input parameters only change the terrain characteristics Ms and correlation length L of the detection surface, and observe the VH / VV polarization backscattering coefficient of the reconnaissance detection radar under different terrain characteristics Ms and different correlation lengths L. Study the VH / VV polarization backscatter coefficient of reconnaissance detection radar in response to the root mean square height S, different terrain features Ms The response to the root mean square height S is the same as in step 3.2; The step 5 includes: in order to better estimate the backscatter coefficient value, introducing the vegetation coverage of the detected surface to improve the water cloud model, and calculating the actual backscatter coefficient, the specific form is shown in formula (1): in, is the backscatter coefficient of the total reconnaissance detection radar; F veg To detect the vegetation coverage of the surface pixels, F veg The value is between (0, 1). When it approaches 1, it means that the vegetation coverage area of the detected surface accounts for a larger proportion. On the contrary, when it approaches 0, the exposed surface area of the detected surface accounts for a larger proportion. veg Sentinel-2 data can be used to calculate vegetation coverage using a pixel binary model, as shown in formula (2): F veg =(NDVI-NDVI min ) / (NDVI max -NDVI min ) (2) Among them, NDVI is the normalized difference vegetation index value calculated from Sentinel-2 data; NDVI min The NDVI is the normalized vegetation index value of the exposed area, which is theoretically 0; max is the normalized vegetation index value of the vegetation coverage area; the values corresponding to the 5% and 95% cumulative distribution probability of the NDVI value in the detection area are selected as the NDVI min and NDVI max , combining formula (2) with the water cloud model to obtain formula (3):

2. The terrain feature inversion method based on the radar backscatter coefficient Oh model according to claim 1, characterized in that: In step 6, the A and B values under different vegetation types of the detected surface are first determined, and the backscattering coefficient of the soil sample point is used. The A and B values are fixed and the value of the other parameter is changed. The water cloud model improved by the vegetation coverage of the detected surface is run to obtain the backscattering coefficient under different A and B values based on the detected surface.

3. The terrain feature inversion method based on the radar backscatter coefficient Oh model according to claim 1, characterized in that: In step 7, the response of the backscatter coefficient of the reconnaissance detection radar to the detected surface parameters under different polarizations is studied. For the Oh model, the correlation length L is set to a range of 1 to 18 cm with a step size of 3 cm; the root mean square height s is set to a range of 0.4 to 2.2 cm with a step size of 0.1 cm; and the terrain feature Ms is set to 0.01 to 0.31 cm. 3 / cm 3 , step length is 0.005cm 3 / cm 3 At the same time, the free space wave number parameters are input, and the Oh model is run multiple times to obtain a database consisting of the corresponding reconnaissance detection radar backscatter coefficients under different correlation lengths and different root mean square height conditions. The backscatter coefficients of the pixels are obtained from the detection SAR image, and the bare soil backscatter coefficients that remove the influence of vegetation cover in the detection area are found by the lookup table method. The bare soil backscattering coefficient simulated by Oh with the minimum cost function (M) The specific form of the cost function is as follows:

Images