Under-forest terrain inversion method and device based on area network adjustment and considering penetration depth, equipment and medium

Abstract

The application discloses a kind of forest terrain inversion methods, device, equipment and medium based on area network adjustment and considering penetration depth, method is: obtaining forest area multi-scene double-station TanDEM-X data and carrying out interference and flat processing, the coherence coefficient of image is obtained;Phase geographic geographic coding is carried out to the interference flat after, unwrapping filter and height conversion are obtained initial digital elevation model;Considering the penetration effect of X wave band in forest area and the system error generated by TanDEM-XDEM, a total error model considering penetration is constructed;Select joint ICESat-2 ground height measuring point at connecting point to adjust calculation, solve unknown parameter in model, calculate the entire area of forest terrain data according to total error model using InSARDEM.The application considers penetration depth to carry out area network adjustment and inversion to forest terrain, and it is a kind of robust and feasible area network adjustment method, can be inverted in large scale, and the adaptability is stronger.

Description

Technical Field

[0001] This invention belongs to the field of InSAR image data processing, specifically relating to a method, apparatus, equipment, and medium for forest understory topography inversion based on regional network adjustment and taking into account penetration depth. Background Technology

[0002] A Digital Elevation Model (DEM) is a three-dimensional digital model describing the elevation of the bare Earth. High-resolution, high-precision DEMs play a significant role in economic development and military defense. Large-scale DEM mapping requires acquiring a large number of SAR images to cover the entire area. However, due to inaccuracies in orbital error estimation between different images, the extracted large-scale DEMs cannot be accurately stitched together. Therefore, regional network adjustment is often used for large-scale DEMs. Existing studies on DEM regional network adjustment generally assume that the measured elevation is the surface elevation. However, as an X-band binary system, TanDEM-X, while possessing the advantage of zero-time decorrelation and being less affected by time decorrelation and atmospheric delay, still experiences penetration during DEM generation, leading to a difference between InSAR altimetry and the actual elevation. Existing error equations in TanDEM-X DEM regional network adjustment lack consideration of this factor, resulting in inaccurate estimation of forest understory topography during regional network adjustment in forest areas. Summary of the Invention

[0003] This invention provides a method, apparatus, equipment, and medium for forest understory topography inversion based on regional network adjustment and taking into account penetration depth, which can obtain large-scale, high-precision topographic data in forest areas with incomplete penetration.

[0004] To achieve the above technical objectives, the present invention adopts the following technical solution:

[0005] A method for forest understory topography inversion based on regional network adjustment and taking into account penetration depth includes:

[0006] Step 1: Acquire multi-view dual-station TanDEM-X image data of the forest area to be studied, and perform interferometric processing to obtain the coherence coefficient and phase of the images;

[0007] Step 2: Simulate the terrain phase using parameters from external DEM data and TanDEM-X imagery, and then perform differential interferometric phase analysis with the interferometric phase obtained in Step 1. The differential interferometric phase is then filtered and unwrapped, and finally phase-height conversion is performed to obtain the initial digital elevation model.

[0008] Step 3: Construct an elevation error model that considers the phase center height error caused by systematic errors and incomplete penetration;

[0009] Step 4: Select ICESat-2 ground altimetry points as control points to provide absolute elevation constraints for regional network adjustment, select multi-scene connection points to provide relative elevation constraints for regional network adjustment, and perform adjustment calculations on the forest floor ground elevation based on the initial digital elevation model and elevation error model to obtain the coefficients of the elevation error model.

[0010] Step 5: Based on the initial digital elevation model obtained in Step 2, use the elevation error model with known coefficients to invert the understory topography data of the forest area to be studied.

[0011] Furthermore, the elevation error model constructed in step 3, which considers systematic errors and phase center height errors caused by incomplete penetration, is expressed as follows:

[0012] (1)

[0013] in It is a polynomial characterizing the systematic errors generated when generating a DEM from TanDEM-X data. These represent the column and row numbers of the TanDEM-X data in the SAR coordinate system, respectively. It is a model designed to characterize the relationship between penetration depth and coherence coefficient and vertical wavenumber. The image coherence coefficient is denoted as . Vertical wavenumber; and and vertical wavenumber The expressions are as follows:

[0014] (2)

[0015] (3)

[0016] (4)

[0017] In the formula, For expression The coefficients in For expression The coefficients in For each image scene, there are 8 parameters to be determined; For vertical baseline, For wavelength, Slope distance The angle of incidence is denoted as .

[0018] Furthermore, step 4 specifically includes the following processes:

[0019] Step 4.1: In the area where regional network adjustment is required, select images with high coherence for each TanDEM-X image acquired in Step 1. Using ICESat-2 ground elevation measurement points as control points, absolute elevation constraints are provided for the regional network adjustment:

[0020] (5)

[0021] in, Indicates the first Elevation of each ICESatt-2 ground elevation measurement point; Indicates the first The initial digital elevation values ​​corresponding to each control point; Indicates the first The elevation error value of each control point, i.e. The matrix expression for each image expanded to n control points is:

[0022] (6)

[0023] Step 4.2: Within the overlapping area of ​​two adjacent TanDEM-X image data sets, uniformly select tie points; based on the principle that tie points at the same geographical location should have the same altitude on different images, provide relative elevation constraints for the regional network adjustment. A connection point in the image and images The constraints that should be satisfied are:

[0024] (7)

[0025] In the formula, Indicates the first Each connection point is represented by an image. The obtained initial digital elevation values, Indicates the first Each connection point is represented by an image. The obtained initial digital elevation values, Indicates the first Each connection point is represented by an image. The obtained elevation error, Indicates the first Each connection point is represented by an image. The obtained elevation error, Representing the first The column and row numbers of each connection point in the image; For the first in the image The coherence coefficient of each connection point For the first Vertical wavenumber of each connection point;

[0026] The matrix expression for m connection points is:

[0027]

[0028] The parameters in the formula with subscripts J and K are respectively related to the image. and images correspond;

[0029] Convert constraint (7) to:

[0030] (8)

[0031] Step 4.3, combining the constraints on control points and connection points, that is, unifying the matrix expressions for the n control points and the m connection points into a matrix representation for the regional network adjustment equation:

[0032] (9)

[0033] Where V is the residual error, B is the global coefficient matrix, X is the parameter matrix to be estimated, and L is the elevation error vector; in the elevation error vector L, the elements corresponding to the control points represent the differences between the measured values ​​and ICESat_2 and the initial DEM, i.e. In the elevation error vector L, the elements at the connection points represent the elevation error values ​​between the two images, i.e. ;

[0034] The unknown parameter X is estimated using least squares estimation, minimizing the residual error V:

[0035]

[0036] in:

[0037] P is the weight matrix;

[0038] ;in Let J be the coefficient matrix consisting of control points on image J, where Let J be the coefficient matrix consisting of control points on image J, where and The coefficient matrices are composed of the connection points on images J and K, respectively. , , and The unified representation is:

[0039]

[0040] In the formula, the subscript in the lower right corner This indicates the number of connection points or control points in the corresponding image;

[0041] Elevation error, where ICESat_2 height measurement data for control points on images J and K; The initial digital elevation values ​​for the control points on images J and K; These are the initial data elevation values ​​of the connection points of the overlapping regions of image J and image K, respectively;

[0042] This represents the matrix of parameters to be determined for images J and K.

[0043] Furthermore, after step 4.3, the process includes: performing a significance estimate on each parameter obtained from the current solution; if the significance requirement is met, proceed to step 5; otherwise, recalculate the adjustment error model, i.e., no longer estimate the parameter with the smallest significance in the parameter matrix to be solved in the regional network adjustment equation, thereby reducing the number of unknowns in the regional network adjustment equation; perform least squares regional network adjustment again to calculate the value of the parameter to be solved, and perform a significance estimate on each parameter obtained from the solution, thereby iteratively estimating the parameters until all parameters meet the significance requirement;

[0044] The significance of each parameter is estimated using a t-distribution:

[0045] (10)

[0046] In the formula, For each parameter to be estimated, The standard deviation is required to be the ratio t of each parameter to be estimated to the standard deviation, which must not be less than a given value.

[0047] Furthermore, after obtaining the coherence coefficient in step 1, the SNR correction method is used to compensate for the coherence coefficient.

[0048] Furthermore, in step 2, when removing the terrain phase, orbital information is used to simulate the terrain phase instead of baseline information.

[0049] Furthermore, in step 2, the phase-high conversion is performed without baseline least squares.

[0050] A forest understory topography inversion device based on regional network adjustment and taking into account penetration depth includes:

[0051] The preprocessing module is used to: perform interferometric processing on the acquired multi-scene dual-station TanDEM-X image data of the forest area to be studied, and obtain the coherence coefficient and phase of the images;

[0052] The initial digital elevation acquisition module is used to: simulate the terrain phase using parameters from external DEM data and TanDEM-X imagery, and obtain a differential interferometric phase by subtracting the interferometric phase obtained from the preprocessing module. Then, the differential interferometric phase is filtered and unwrapped, and finally, a phase-to-height conversion is performed to obtain the initial digital elevation model.

[0053] The height error construction module is used to: construct a height error model that takes into account systematic errors and phase center height errors caused by incomplete penetration;

[0054] The regional network adjustment calculation module is used to: select ICESat-2 ground altimetry points as control points to provide absolute elevation constraints for regional network adjustment; select multiple scene connection points to provide relative elevation constraints for regional network adjustment; and perform adjustment calculations on the forest underground ground elevation based on the initial digital elevation model and elevation error model to obtain the coefficients of the elevation error model.

[0055] The forest understory topography inversion module is used to invert the forest understory topography data of the forest area under study by using the known coefficients of the elevation error model based on the obtained initial digital elevation model.

[0056] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor enables the processor to implement the forest understory inversion method based on regional network adjustment and taking into account penetration depth as described above.

[0057] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the forest understory inversion method based on regional network adjustment and taking into account penetration depth as described above.

[0058] Beneficial effects

[0059] This invention processes multiple SAR images of forest areas to obtain an initial DEM. Penetration depth is characterized by acquiring image coherence and calculating the vertical wavenumber of the SAR images. Then, external ICESat points and tie points of repeating regions are introduced, and an error equation is constructed using a polynomial of the TanDEM-X system error. Regional network adjustment considering penetration depth is then applied to the images, resulting in higher-precision topographic information for forest areas. This invention considers the influence of penetration depth when performing regional network adjustment of elevations, leading to higher accuracy. Furthermore, the data used in this invention is globally covered TanDEM-X data, enabling large-scale inversion and broad applicability. It provides a new approach and method for exploring regional network adjustment considering penetration depth and can be widely applied to research on forest understory topography in areas where penetration occurs. Attached Figure Description

[0060] Figure 1 This is a flowchart of the method described in an embodiment of the present invention.

[0061] Figure 2 The location of the verification area and the coverage area of ​​the SAR image selected for this invention.

[0062] Figure 3 This is a schematic diagram of the topographic inversion results for the experimental area. Among them, Figure 3 (a) represents the DEM obtained by adjustment and inversion using the method of the embodiments of the present invention. Figure 3 (b) Represents the difference between the area network adjustment results (excluding penetration depth) and LiDAR. Figure 3 (c) Represents the difference between the area network adjustment results considering penetration depth in the embodiments of the present invention and LiDAR. Detailed Implementation

[0063] To better illustrate the methods and steps of this invention, TanDEM-X data from the Teruel experimental area in eastern Spain are used to further describe the invention in detail, and GAMMA software released by the Swiss company GAMMA is used to assist in data preprocessing. Note that the specific implementation described herein is only for explaining the invention and is not intended to limit the invention.

[0064] This embodiment provides a regional network adjustment technique that takes into account penetration depth, such as... Figure 1 As shown, it includes the following steps:

[0065] Step 1: Based on the coordinate range of the selected experimental area, obtain the bistatic TanDEM-X data corresponding to the German Aerospace Center. Perform interferometric processing on the obtained bistatic TanDEM-X data to obtain the interferometric phase and coherence coefficient.

[0066] This implementation example uses TanDEM-X data from the Teruel experimental region in eastern Spain. Data from three adjacent orbitals is selected, and then two images with overlapping areas are chosen from each orbit to create a single image. Multiple scenes of the forest area were acquired, and the data acquisition time is shown in Table 1. The location of the experimental area is shown in Table 1. Figure 2 As shown, the data is subjected to interferometric and coherence estimation processing, and the coherence coefficient corresponding to the image is obtained. .

[0067]

[0068] Coherence estimation is an important task in InSAR measurements. The magnitude of the coherence coefficient reflects the quality of the InSAR phase. In this invention, it is not only an important parameter of the error model but also one of the important bases for feature point selection. Its formula is as follows:

[0069]

[0070] In the formula, M and N are the size of the data block used to calculate coherence; m and n are the row and column numbers within the data block. The image is a complex value at the coordinate point.

[0071] After obtaining the coherence coefficient, this embodiment uses the SNR correction method to compensate for the coherence coefficient and obtain the corrected coherence coefficient. The SNR correction method is existing technology and will not be described in detail in this embodiment.

[0072] Step 2: Import external DEM data of TanDEM-X90m for terrain removal processing to obtain differential interferometric phase, then perform filtering and unwrapping processing on the differential phase, and finally perform phase height conversion to obtain the initial digital elevation model.

[0073] Because the dual-station TanDEM-X satellite has a "0" time baseline, multiple coverages, and adjustable baseline, it greatly reduces the impact of time decoherence and atmospheric delay. External DEM data is introduced, and GAMMA software is used to remove the topographic phase to obtain differential interferometric phase. Then, the differential phase is filtered and unwrapped. Finally, GAMMA software is used to perform phase-to-height conversion to obtain the initial digital elevation model. Based on a refined lookup table, the DEM in the SAR coordinate system is converted to geographic coordinates.

[0074] In step 2 of this embodiment, when removing terrain phase, orbital information is used to simulate terrain phase instead of baseline information. The expression for using orbital information to simulate terrain phase is: Where B is the baseline length, λ is the wavelength, θ is the incident angle, and α is the baseline dip angle. The expression for simulating terrain phase using baseline information is: r1 and r2 are the distances from the primary satellite and the secondary satellite to the ground point, respectively, and λ is the wavelength.

[0075] Since least squares are required in the subsequent step 4 of the regional network adjustment process, if the baseline least squares are used in the phase height conversion step, it will lead to a more complex error equation, which will cause inconvenience in removal. In order to avoid the error equation becoming more complex during least squares regional network adjustment, this embodiment does not perform least squares processing on the baseline.

[0076] Step 3: Considering the penetration effect of the X-band in the forest area and the systematic error generated by TanDEM-X DEM, a total error model considering penetration is constructed based on the relationship between coherence coefficient and vertical wavenumber and penetration depth, as well as the relationship between image row and column number and orbit in SAR coordinate system.

[0077] The main errors include systematic errors and phase center height errors caused by incomplete penetration. Therefore, an elevation error model can be constructed as follows:

[0078] (1)

[0079] in It is a polynomial characterizing the systematic errors generated when generating a DEM from TanDEM-X data. These represent the column and row numbers of the TanDEM-X data in the SAR coordinate system, respectively. It is a model designed to characterize the relationship between penetration depth and coherence coefficient and vertical wavenumber. The image coherence coefficient is denoted as . Vertical wavenumber; and and vertical wavenumber The expressions are as follows:

[0080] (2)

[0081] (3)

[0082] (4)

[0083] In the formula, For expression The coefficients in For expression The coefficients in For each image scene, there are 8 parameters to be determined; For vertical baseline, For wavelength, Slope distance The angle of incidence is denoted as .

[0084] The final model obtained is:

[0085]

[0086] An error equation will be established for each feature point, and the unknown parameters of the image will be solved in the end.

[0087] Step 4: Select the connection points and ICESat-2 ground altimetry points for adjustment calculation, perform regional network adjustment considering penetration depth, and solve for the unknown parameters in the error model;

[0088] Based on the above equations, this embodiment introduces some ground control points and repeating zone connection points to solve for the parameters to be estimated, specifically:

[0089] Step 4.1: In the area where regional network adjustment is required, select ICESat-2 points with good distribution conditions and high reliability as control points for each image to provide absolute elevation constraints for regional network adjustment.

[0090] Based on the latitude and longitude of the experimental area, download the corresponding ICESat data. Select points with gentle slopes, good coherence, and available forest tree height data as control points. Taking one scene as an example, the expression for the absolute elevation constraint is:

[0091] (5)

[0092] in, Indicates the first Elevation of each ICESat point; express Initial InSAR elevation of the point; The expression representing the elevation error of that point is... :

[0093] Assuming selection If there are 1 control point, the specific matrix expression is:

[0094] (6)

[0095] Eight parameters representing each image scene to be determined, for In terms of landscape images, there are a total of One parameter to be determined.

[0096] Step 4.2: Within adjacent DEM overlapping areas, select tie points evenly. Based on the principle that tie points at the same geographical location should have the same height on different images, provide relative elevation constraints for regional network adjustment.

[0097] The latitude and longitude of each image in the experimental area were determined. Based on these coordinates, duplicate regions were calculated pairwise, ignoring areas less than 3 km away. Connecting points were selected from images with larger duplicate regions. To ensure uniform selection, duplicate regions were cropped, and each window was 50 pixels x 50 pixels. Null values ​​were not considered in each window. Besides identifying unreliable pixels, thresholds were set for coherence and slope to remove excessively steep or low-coherence pixels. If more than half of the pixels in two images met the criteria, a histogram was used to calculate only the median value of that region. This median was selected as the connecting point. For connecting points that met the requirements, a small window was opened to calculate the standard deviation. Points with smaller standard deviations indicated closer terrain similarity. A 3x3 mean filter was applied to these points to ensure accuracy. One indicator of connecting point reliability is that the two images have the same standard deviation. A smaller standard deviation indicates similar terrain at the same location.

[0098] Taking two images with overlapping areas as an example, the first... A connection point in the image and images The constraints that should be satisfied are:

[0099] (7)

[0100] Right now:

[0101] (8)

[0102] Step 4.3, combining the constraints on control points and connection points, that is, unifying the matrix expressions for the n control points and the m connection points into a matrix representation for the regional network adjustment equation:

[0103] (9)

[0104] Where V is the residual error, B is the global coefficient matrix, X is the parameter matrix to be estimated, and L is the elevation error vector; in the elevation error vector L, the elements corresponding to the control points represent the differences between the measured values ​​and ICESat_2 and the initial DEM, i.e. In the elevation error vector L, the elements at the connection points represent the elevation error values ​​between the two images, i.e. ;

[0105] The unknown parameter X is estimated using least squares estimation, minimizing the residual error V:

[0106]

[0107] in:

[0108] P is the weight matrix;

[0109] ;in Let J be the coefficient matrix consisting of control points on image J, where Let J be the coefficient matrix consisting of control points on image J, where and The coefficient matrices are composed of the connection points on images J and K, respectively. , , and The unified representation is:

[0110]

[0111] In the formula, the subscript in the lower right corner This indicates the number of connection points or control points in the corresponding image;

[0112] Elevation error, where ICESat_2 height measurement data for control points on images J and K; The initial digital elevation values ​​for the control points on images J and K; These are the initial data elevation values ​​of the connection points of the overlapping regions of image J and image K, respectively;

[0113] This represents the matrix of parameters to be determined for images J and K.

[0114] Note that all points originally in the geographic coordinate system need to be transformed to the SAR coordinate system using a refined lookup table, and equations need to be established and solved in the SAR coordinate system.

[0115] Step 4.4: The significance of the eight parameters to be determined for each image scene is estimated, with the significance of the parameters estimated using a t-distribution.

[0116] (10)

[0117] in To estimate the parameters, The standard deviation is required to be the ratio of all parameters to the standard deviation, which is not less than a given value. In this embodiment, the given value is 1 (a value with 1 degree of freedom and a confidence interval of 0.75).

[0118] If the ratio of any unknown parameter is less than 1, the parameter with the smallest ratio is removed and the calculation is repeated until all parameters are greater than 1. It is worth noting that the number of unknown parameters for each image may be different due to this step of calculation.

[0119] Step 5: Based on the initial InSAR DEM obtained in Step 2, use the error model of the estimated parameters to invert the forest understory topography data for the entire area. .

[0120]

[0121] Figure 3 This is a comparison chart of the area network adjustment results considering penetration depth and those not considering penetration depth, according to an embodiment of the present invention. To more clearly demonstrate the advantages of considering penetration depth, a LiDAR DEM was used to assess the accuracy of the data. Figure 3 (a) is the topographic map derived from the inversion. Figure 3 (b) is a plot showing the difference between the adjusted regional network without considering penetration depth and LiDARDEM. Figure 3 (c is a difference map between the regional network adjustment considering penetration depth and the LiDAR DEM, showing that...) Figure 3 (b) and Figure 3 (c) The RMSE increased from 3.34 to 2.83, with an accuracy improvement of 15%, indicating that the regional network adjustment accuracy is higher when the penetration depth is taken into account.

[0122] The above embodiments are preferred embodiments of this application. Those skilled in the art can make various changes or improvements based on them. Without departing from the overall concept of this application, these changes or improvements should fall within the scope of protection claimed in this application.

Claims

1. A method for forest understory topography inversion based on regional network adjustment and taking into account penetration depth, characterized in that, include: Step 1: Acquire multi-view dual-station TanDEM-X image data of the forest area to be studied, and perform interferometric processing to obtain the coherence coefficient and phase of the images; Step 2: Simulate the terrain phase using parameters from external DEM data and TanDEM-X imagery, and then perform differential interferometric phase analysis with the interferometric phase obtained in Step 1. The differential interferometric phase is then filtered and unwrapped, and finally phase-height conversion is performed to obtain the initial digital elevation model. Step 3: Construct an elevation error model that considers the phase center height error caused by systematic errors and incomplete penetration; Step 4: Select ICESat-2 ground altimetry points as control points to provide absolute elevation constraints for regional network adjustment, select multi-scene connection points to provide relative elevation constraints for regional network adjustment, and perform adjustment calculations on the forest floor ground elevation based on the initial digital elevation model and elevation error model to obtain the coefficients of the elevation error model. Step 5: Based on the initial digital elevation model obtained in Step 2, use the elevation error model with known coefficients to invert the understory topography data of the forest area to be studied. The elevation error model constructed in step 3, which considers systematic errors and phase center height errors caused by incomplete penetration, is expressed as follows: （1） in It is a polynomial characterizing the systematic errors generated when generating a DEM from TanDEM-X data. These represent the column and row numbers of the TanDEM-X data in the SAR coordinate system, respectively. It is a model designed to characterize the relationship between penetration depth and coherence coefficient and vertical wavenumber. The image coherence coefficient is denoted as . Vertical wavenumber; and and vertical wavenumber The expressions are as follows: （2） （3） （4） In the formula, For expression The coefficients in For expression The coefficients in For each image scene, there are 8 parameters to be determined; For vertical baseline, For wavelength, Slope distance The angle of incidence is denoted as .

2. The forest understory topography inversion method based on regional network adjustment and taking into account penetration depth as described in claim 1, characterized in that, Step 4 includes the following specific steps: Step 4.1: In the area where regional network adjustment is required, uniformly select images with high coherence from each TanDEM-X image acquired in Step 1. Using ICESat-2 ground elevation measurement points as control points, absolute elevation constraints are provided for the regional network adjustment: （5） in, Indicates the first Elevation of each ICESat-2 ground elevation measurement point; Indicates the first The initial digital elevation values ​​corresponding to each control point; Indicates the first The elevation error values ​​of each control point; the matrix expression for each image expanded to n control points is: （6） Step 4.2: Within the overlapping area of ​​two adjacent TanDEM-X image data sets, uniformly select tie points; based on the principle that tie points at the same geographical location should have the same altitude on different images, provide relative elevation constraints for the regional network adjustment. A connection point in the image and images The constraints that should be satisfied are: （7） In the formula, Indicates the first Each connection point is represented by an image. The obtained initial digital elevation values, Indicates the first Each connection point is represented by an image. The obtained initial digital elevation values, Indicates the first Each connection point is represented by an image. The obtained elevation error, Indicates the first Each connection point is represented by an image. The obtained elevation error, Representing the first The column and row numbers of each connection point in the image; For the first in the image The coherence coefficient of each connection point For the first Vertical wavenumber of each connection point; The matrix expression for m connection points is: ； The parameters in the formula with subscripts J and K are respectively related to the image. and images correspond; Convert constraint (7) to: （8） Step 4.3, combining the constraints on control points and connection points, that is, unifying the matrix expressions for the n control points and the m connection points into a matrix representation for the regional network adjustment equation: （9） Where V is the residual error, B is the overall coefficient matrix, X is the parameter matrix to be estimated, and L is the elevation error vector; in the elevation error vector L, the elements corresponding to the control points represent the difference between the ICESat_2 measurement value and the initial DEM, i.e. In the elevation error vector L, the elements at the connection points represent the elevation error values ​​between the two images, i.e. ； The unknown parameter X is estimated using least squares estimation, minimizing the residual error V: ； in: P is the weight matrix; ;in Let J be the coefficient matrix consisting of control points on image J, where Let K be the coefficient matrix consisting of control points on image K, where and The coefficient matrices are composed of the connection points on images J and K, respectively. , , and The unified representation is: ； In the formula, the subscript in the lower right corner This indicates the number of connection points or control points in the corresponding image; Elevation error, where ICESat_2 height measurement data for control points on images J and K; The initial digital elevation values ​​for the control points on images J and K; These are the initial data elevation values ​​of the connection points of the overlapping regions of image J and image K, respectively; This represents the matrix of parameters to be determined for images J and K.

3. The forest understory topography inversion method based on regional network adjustment and taking into account penetration depth as described in claim 2, characterized in that, After step 4.3, the following steps are also included: performing a significance estimate on each parameter obtained from the current solution. If the significance requirement is met, step 5 is continued; otherwise, the adjustment error model is recalculated. That is, the parameter with the smallest significance in the parameter matrix to be solved in the regional network adjustment equation is no longer estimated, thereby reducing the number of unknowns in the regional network adjustment equation. The least squares regional network adjustment is performed again to calculate the value of the parameter to be solved, and a significance estimate is performed on each parameter obtained from the solution. The parameters are iteratively estimated until all parameters meet the significance requirement. The significance of each parameter is estimated using a t-distribution: （10） In the formula, For each parameter to be estimated, The standard deviation is required to be the ratio t of each parameter to be estimated to the standard deviation, which must not be less than a given value.

4. The forest understory topography inversion method based on regional network adjustment and taking into account penetration depth as described in claim 1, characterized in that, After obtaining the coherence coefficient in step 1, the SNR correction method is further used to compensate for the coherence coefficient.

5. The forest understory topography inversion method based on regional network adjustment and taking into account penetration depth as described in claim 1, characterized in that, In step 2, when removing terrain phase, orbital information is used to simulate terrain phase instead of baseline information.

6. The forest understory topography inversion method based on regional network adjustment and taking into account penetration depth as described in claim 1, characterized in that, Step 2 involves performing phase-to-height conversion without baseline least squares operation.

7. A forest understory topography inversion device based on regional network adjustment and taking into account penetration depth, characterized in that, include: The preprocessing module is used to: perform interferometric processing on the acquired multi-scene dual-station TanDEM-X image data of the forest area to be studied, and obtain the coherence coefficient and phase of the images; The initial digital elevation acquisition module is used to: simulate the terrain phase using parameters from external DEM data and TanDEM-X imagery, and obtain a differential interferometric phase by subtracting the interferometric phase obtained from the preprocessing module. Then, the differential interferometric phase is filtered and unwrapped, and finally, a phase-to-height conversion is performed to obtain the initial digital elevation model. The elevation error construction module is used to: construct an elevation error model that considers systematic errors and phase center elevation errors caused by incomplete penetration; represented as: （1） in It is a polynomial characterizing the systematic errors generated when generating a DEM from TanDEM-X data. These represent the column and row numbers of the TanDEM-X data in the SAR coordinate system, respectively. It is a model designed to characterize the relationship between penetration depth and coherence coefficient and vertical wavenumber. The image coherence coefficient is denoted as . Vertical wavenumber; and and vertical wavenumber The expressions are as follows: （2） （3） （4） In the formula, For expression The coefficients in For expression The coefficients in For each image scene, there are 8 parameters to be determined; For vertical baseline, For wavelength, Slope distance Angle of incidence; The regional network adjustment calculation module is used to: select ICESat-2 ground altimetry points as control points to provide absolute elevation constraints for regional network adjustment; select multiple scene connection points to provide relative elevation constraints for regional network adjustment; and perform adjustment calculations on the forest underground ground elevation based on the initial digital elevation model and elevation error model to obtain the coefficients of the elevation error model. The forest understory topography inversion module is used to invert the forest understory topography data of the forest area under study by using the known coefficients of the elevation error model based on the obtained initial digital elevation model.

8. An electronic device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the computer program is executed by the processor, the processor causes the processor to implement the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.

Images