Long-wave band PolInSAR forest-underground terrain inversion method and device based on non-penetration depth correction

Abstract

The application provides a long-wave band PolInSAR under-forest terrain inversion method and equipment based on non-penetration depth correction, acquires two scenes of full polarization SAR image data and pre-processes, generates a polarization interference covariance matrix; carries out white processing on the polarization interference covariance matrix, applies a normal matrix constraint, acquires an optimal normal matrix approximate solution, extracts two end point complex coherence coefficients of a coherent line segment; based on a polarization coherence scattering model, solves model parameters by using the end point complex coherence coefficients, including a ground surface phase, a forest height, an extinction coefficient and a ground body amplitude ratio; combines sparse LiDAR height data, calculates the non-penetration depth between the forest height and the RVoG inversion terrain and the LiDAR reference terrain, constructs a one-dimensional look-up table based on the mapping relationship between the forest height and the non-penetration depth; and corrects the under-forest terrain height deviation of the PolInSAR inversion according to the one-dimensional look-up table. The application can significantly improve the precision of the ground surface terrain inversion in the forest area, and can realize large-scale under-forest terrain mapping by only relying on a single polarization interference.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of radar interferometry and the technical field of forest scene parameter inversion, and proposes a technology scheme for estimating and correcting under-forest terrain based on long-wave PolInSAR data. BACKGROUND

[0002] The under-forest terrain product is a basic data set for monitoring the forest ecosystem. Accurate characterization of bare ground terrain can more accurately estimate the vertical structure and topographic features of the forest. This is of great significance for monitoring geological disasters in forest areas, promoting climate change adaptation, supporting the development of carbon emission reduction strategies, and deepening the understanding of the functions and stability of forest ecosystems. Synthetic Aperture Radar (SAR) interferometry technology (InSAR) has been widely used in large-scale terrain mapping due to its all-weather, all-day imaging capability. However, traditional InSAR technology mostly relies on short-wave radar such as X-band or C-band, which is limited by signal penetration capability and often cannot penetrate the tree canopy to the ground in forest areas. The interference phase center is mostly located in the canopy layer, resulting in terrain products that are closer to Digital Surface Model (DSM) rather than the true "bare ground" terrain model.

[0003] To accurately obtain the "bare ground" terrain, the scientific community has been continuously concerned about the application of long-wave (such as L-band, P-band) SAR systems, especially in high-density forest coverage areas such as tropical rainforests. Compared with short-wave radar, long-wave radar has stronger vegetation penetration capability and can theoretically provide interference information closer to the true terrain. At the same time, in order to further separate the scattering contributions of the ground and vegetation, Polarimetric InSAR (PolInSAR) technology has been proposed and widely studied. Among them, the Random Volume over Ground (RVoG) model is a classic polarization coherence scattering parameterization model in PolInSAR, which establishes a physical relationship between multi-polarization complex coherence coefficients and forest structure parameters (such as forest height, ground phase, ground amplitude ratio, extinction coefficient) by assuming that vegetation is a penetrable random volume layer and an impenetrable ground layer. To some extent, it can extract the ground phase and then invert the under-forest terrain. However, even in the case of using P-band full-polarization SAR data combined with the RVoG model, there may still be incomplete penetration, especially in high forest and high canopy density tropical rainforest areas. At this time, the inverted ground phase will still deviate from the true terrain, resulting in systematic bias in the under-forest terrain.

[0004] Therefore, how to compensate the error of the under-forest terrain caused by the insufficient penetration in the RVoG model by establishing the mapping relationship between the forest height and the non-penetration depth on the basis of the known PolInSAR data becomes a problem to be solved in the current research. In view of this, no related method is found through the search, and the application discloses a long-wave band PolInSAR under-forest terrain inversion method based on non-penetration depth correction.

[0005] When the RVoG model is used to invert the under-forest terrain, the traditional method usually depends on a three-stage algorithm or a nonlinear iterative optimization algorithm to solve the ground phase. However, due to the strong linear correlation between the multi-polarization interference complex coherence coefficients under the same baseline, the rank deficiency problem occurs in the inversion model matrix, so that the parameter solution is not unique or unstable. In order to overcome the rank deficiency problem, the application introduces the normal matrix constraint to extract the two end point complex coherence coefficients of the coherent line segment from the polarization interference complex covariance matrix. SUMMARY

[0006] The purpose of the application is to provide a one-dimensional look-up table terrain correction method based on the mapping relationship between the forest height and the non-penetration depth, which is used for correcting the ground elevation deviation problem caused by the insufficient penetration in the RVoG model inversion of the long-wave band PolInSAR technology.

[0007] The above technical problems of the application are mainly solved by the following technical scheme:

[0008] A long-wave band PolInSAR under-forest terrain inversion method based on non-penetration depth correction comprises the following processes:

[0009] Two scenes of full-polarization SAR image data are acquired and preprocessed to generate a polarization interference complex covariance matrix;

[0010] The polarization interference complex covariance matrix is subjected to whitening processing, the normal matrix constraint is applied, the optimal normal matrix approximate solution is acquired, and the two end point complex coherence coefficients of the coherent line segment are extracted;

[0011] Based on the polarization coherence scattering model, the end point complex coherence coefficients are used to solve the model parameters, and the model parameters include the ground phase, the forest height, the extinction coefficient and the ground amplitude ratio;

[0012] In combination with the sparse LiDAR elevation data, the non-penetration depth between the forest height and the RVoG inversion terrain and the LiDAR reference terrain is calculated, and the one-dimensional look-up table based on the mapping relationship between the forest height and the non-penetration depth is constructed;

[0013] According to the one-dimensional look-up table, the elevation deviation of the PolInSAR inverted under-forest terrain is corrected, and the corrected digital terrain model is output.

[0014] Moreover, the implementation process of the normal matrix constraint includes normal matrix constraint on the whitened interference cross-correlation matrix, generation of an intermediate variable parameter, solution of an optimal normal matrix approximate solution, and ensuring of geometric stability of the coherent line segment in the complex unit circle.

[0015] Moreover, when the model parameter is solved by using the end-point complex coherence coefficient, the ground surface phase is determined by using the interference vertical wave number sign; the end-point complex coherence coefficient is substituted into the polarized coherent scattering model, and the forest height, the extinction coefficient and the ground body amplitude ratio are solved by using the nonlinear iterative optimization.

[0016] Moreover, the implementation of the one-dimensional lookup table includes division of samples according to preset forest height intervals, statistics of the mean forest height and the mean non-penetration depth in each interval, construction of the mapping relationship between the forest height and the non-penetration depth, and obtaining of the sparse lookup table; and the continuous mapping relationship between the forest height and the non-penetration depth is generated by one-dimensional interpolation based on the sparse lookup table.

[0017] Moreover, the implementation process of the correction of the under-forest terrain elevation deviation is,

[0018] The non-penetration depth compensation value is obtained by one-dimensional lookup table interpolation based on the forest height inverted by PolInSAR;

[0019] The compensation value is deducted from the terrain inverted from the RVoG, and the corrected under-forest terrain elevation is obtained.

[0020] Moreover, the interference baseline length of the two scenes of full-polarization SAR image data satisfies that the ambiguity height is greater than half of the maximum forest height.

[0021] Moreover, the method is used for realizing the under-forest terrain inversion in the high forest coverage area.

[0022] On the other hand, the present application also provides an electronic device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor realizes the long-wave band PolInSAR under-forest terrain inversion method based on non-penetration depth correction as described above when executing the program.

[0023] On the other hand, the present application also provides a non-transitory computer readable storage medium, which stores a computer program, and the computer program realizes the long-wave band PolInSAR under-forest terrain inversion method based on non-penetration depth correction as described above when executed by a processor.

[0024] On the other hand, the present application also provides a computer program product, which comprises a computer program, and the computer program realizes the long-wave band PolInSAR under-forest terrain inversion method based on non-penetration depth correction as described above when executed by a processor.

[0025] The innovation of the present application is that the sparse LiDAR ground elevation information and the PolInSAR polarized coherent scattering model are combined, the non-penetrating system elevation deviation existing in the RVoG model inversion of the under-forest terrain is compensated by means of a one-dimensional lookup table, the accuracy of the ground terrain inversion in the tropical rainforest region can be significantly improved, and only a single polarized interference pair can be used to realize large-scale under-forest terrain mapping, thereby assisting the current and future polarized satellite SAR (such as the BIOMASS satellite) under-forest terrain mapping task. Therefore, the present application has strong engineering feasibility and practical application value, and is especially suitable for the under-forest terrain inversion and DEM product correction task of SAR in the high forest coverage area.

[0026] The technical core of the present application is to construct a one-dimensional lookup table mapping model between forest height and non-penetrating depth by using a statistical regression method, which is beneficial to realize rapid and accurate correction of under-forest terrain in a wide range. The SAR image observation has the characteristics of large-scale and surface continuous observation, and the LiDAR observation has the characteristics of point observation and high terrain precision. The method fully combines the technical advantages of SAR and LiDAR, and on the basis of the parameter inversion of the polarized coherent scattering model (namely the RVoG model), the local region or the sparse laser radar terrain information is combined to be extended to a large range of regions which only have full polarized interference SAR image coverage but lack high-density laser radar data, and the method has good practicability and popularization potential.

[0027] The present application has the advantages of simple and convenient implementation, strong practicability, solution to the problems of low practicability and inconvenience in actual application of the related art, improvement of user experience, and important market value. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 The figure is a flow chart of the long-wave band PolInSAR under-forest terrain inversion method based on non-penetrating depth correction of the embodiment of the present application.

[0029] Figure 2 The figure is a one-dimensional lookup table and an interpolation version between the non-penetrating depth and the forest height of the embodiment of the present application.

[0030] Figure 3 The figure is a comparison chart of the method of the embodiment of the present application and DTM products obtained by different methods: (a) TanDEM-X DEM; (b) DTM obtained by RVoG; (c) DTM corrected by using the present application.

[0031] Figure 4 The figure is a DTM precision comparison chart of the method of the embodiment of the present application and different methods: (a) TanDEM-X DEM; (b) DTM obtained by RVoG; (c) DTM corrected by using the present application. DETAILED DESCRIPTION

[0032] The concept, specific structure and generated technical effects of the present application will be further described in combination with the drawings and embodiments to fully understand the purposes, features and effects of the present application.

[0033] In view of the current situation that there is a systematic error in the terrain obtained by the RVoG model inversion based on long-wave PolInSAR data in high-density forest areas, the present application defines the difference between the terrain estimated by the RVoG model and the reference terrain of sparse LiDAR as the 'non-penetration depth', and further establishes a one-dimensional lookup table mapping relationship between the forest height and the non-penetration depth. The forest height estimated by the RVoG model is extended to the entire area covered by the polarimetric interferometric SAR image, so as to realize high-precision mapping of the terrain under the forest in a large range. The present application is particularly suitable for improving the surface terrain mapping and inversion precision in tropical rainforest coverage areas.

[0034] Referring to Figure 1 The long-wave PolInSAR under-forest terrain inversion method based on non-penetration depth correction provided by the embodiment of the present application comprises the following steps:

[0035] Step one, data preprocessing:

[0036] It is preferably suggested that the interference baseline length is ensured to be greater than half of the maximum forest height.

[0037] In the embodiment, two scenes of full-polarization SAR images covering the study area are obtained, the interference baseline length is 40 m, a suitable SAR image is selected as the main image, and then the SAR image is interfered, the ground phase is removed, the reference terrain phase is removed, and multiple data processing steps such as multi-view filtering are performed to estimate the polarimetric interferometric covariance matrix , which can be expressed as

[0038]

[0039] wherein, and are the polarimetric covariance matrices of the main image and the slave image respectively, is the interferometric cross-correlation matrix, and the superscript represents the conjugate transpose operation.

[0040] In this example, a region on the west coast of a country in West-Central Africa is selected as the study area, with geographic coordinates ranging from 1.88°S to 1.92°S and 10.17°E to 10.22°E. This region is one of the four typical tropical rainforest study areas in the European Space Agency's AfriSAR airborne SAR experiment project. The project organized an airborne full-polarimetric P-band SAR flight mission in 2016, and the data obtained cover an area of about 60 km2, with good polarization configuration and interferometric baseline design, providing favorable conditions for high-quality polarimetric interferometric imaging analysis. The region is mainly hilly terrain, with a surface elevation change range of about 10 m to 200 m, and the relief is relatively flat. The vegetation type belongs to the typical tropical rainforest ecosystem, mainly natural forest with high density and high canopy closure, with an average forest height of about 30 m. The data processing process of the de-terrain phase and geographic coding uses the TanDEM-X DEM elevation product. At the same time of the project, in February and March 2016, the European Space Agency, the German Aerospace Center, the French Space Agency, and the United States Space Agency used the full-waveform LiDAR (Land, Vegetation and Ice Sensor, LVIS) carried on the NASA Lengley B200 aircraft to obtain 20 m resolution LiDAR data for the experimental region.

[0041] Step two, after whitening the complex covariance matrix, the double phase center normal matrix constraint method is applied to the polarimetric interferometric complex covariance matrix to obtain the optimal normal matrix approximation solution, and the two end points of the visible coherence segment corresponding to the maximum separation of the complex coherence coefficients are extracted.

[0042] When using the RVoG model to invert the terrain under the forest, the traditional method usually relies on a three-stage algorithm or a nonlinear iterative optimization algorithm to solve the surface phase. However, due to the strong linear correlation between the multi-polarization interferometric complex coherence coefficients under the same baseline, the rank deficiency problem occurs in the inversion model matrix, which makes the parameter solution not unique or unstable. To overcome the rank deficiency problem of the RVoG model parameter solution, the present invention applies a normal matrix constraint to the whitened interferometric cross-correlation matrix to obtain the approximate solution of the best approximation normal matrix, ensuring that all polarization channels of the complex coherence coefficients present geometrically stable coherence segments in the complex unit circle plane, thereby more reliably extracting the complex coherence coefficients of the two end points of the coherence segment.

[0043] First, the polarimetric interferometric complex covariance matrix is whitened to obtain the whitened interferometric cross-correlation matrix :

[0044]

[0045] wherein, The mean of the polarized covariance matrix of the primary and secondary images.

[0046] Under the regular matrix constraint condition of the two-phase center, the intermediate variable parameters are calculated:

[0047]

[0048]

[0049]

[0050]

[0051]

[0052]

[0053]

[0054] wherein, , , , , , , , , , , , , are intermediate variable parameters; is the imaginary unit, is the trace of the matrix, is the inverse tangent function.

[0055] Then, the optimal regular matrix approximate solution can be expressed as:

[0056]

[0057] wherein, is the identity matrix, is the conjugate of .

[0058] The above optimal regular matrix approximate solution ensures the geometric stability of the complex coherent straight line in the complex unit circle plane, so that the complex coherent coefficients corresponding to the two endpoints of the coherent line segment can be more reliably extracted.

[0059] Step three, based on the polarized coherent scattering model RVoG, the complex coherent coefficients corresponding to the two endpoints of the coherent line segment are taken as the observation, and the model parameters including the ground surface phase, forest height, extinction coefficient and two ground body amplitude ratios are solved by nonlinear iterative optimization.

[0060] Firstly, based on the two complex coherence observations (corresponding to the two endpoints of the coherence line segment) obtained in step two, the two intersection points of the corresponding coherence straight line and the unit complex coherence circle are calculated (i.e. and ), and the sign of the interference vertical wave number is adaptively determined to determine the ground phase.

[0061] For a single station system, the formula for calculating the interference vertical wave number is as follows:

[0062]

[0063] wherein, represents the radar wavelength, represents the vertical baseline length, represents the slant range of the sensor and the ground target.

[0064] According to the interference geometric relationship, the criterion for determining the ground phase by the sign of the interference vertical wave number is:

[0065] For :

[0066]

[0067]

[0068] For :

[0069]

[0070]

[0071] wherein, represents the conjugate of , represents the conjugate of , represents the exponential function, ( ) represents the amplitude operation.

[0072] After obtaining the ground phase by the above-mentioned method, the ground phase can be taken as a known value and substituted into the following polarized coherent scattering model formula (i.e. RVoG model).

[0073]

[0074] wherein, represents a mathematical constant, represents the complex coherence coefficient observation of the polarization mode , denotes the pure body de-coherence coefficient, denotes the extinction coefficient, denotes the polarization mode denotes the ground body amplitude ratio, denotes the forest height.

[0075] The complex coherence coefficients corresponding to the two end points of the coherent segment obtained by the optimal normal matrix approximation solution are taken as the observation of the RVoG model, and substituted into the nonlinear iterative optimizer, so that the forest height can be calculated, that is:

[0076]

[0077] wherein, denotes the RVoG model, denotes the norm of the Euclidean vector, denotes the minimum value of the function.

[0078] Step four, based on the external TanDEM-X DEM elevation product, the ground phase is converted by the phase height conversion factor to obtain the ground elevation inverted by the RVoG model; combined with the accurate elevation information of the sparse airborne LiDAR, the non-penetration depth is obtained and a one-dimensional mapping lookup table of forest height and non-penetration depth is constructed.

[0079] Based on the ground phase obtained in step three, the phase height conversion operator and the TanDEM-X DEM elevation product used in step one , the ground elevation based on the RVoG model inversion is obtained :

[0080]

[0081] wherein, is the RVoG under-forest terrain product.

[0082] With the help of the laser radar (LiDAR) ground elevation data obtained by airborne LVIS, the non-penetration depth of SAR signal can be obtained, that is:

[0083]

[0084] With the forest height obtained in step three, a series of sampling intervals of forest height are preset, such as 0-5 m, 5-10 m, 10-15 m, 15-20 m, 20-25 m, 25-30 m, 30-35 m, 35-40 m, 40-45 m, 45-50 m, 50-55 m, 55-60 m, etc., and the mean value of forest height and the mean value of non-penetration depth of all sample points in each forest height interval are counted, so as to construct a one-dimensional lookup table mapping relationship between the forest height and the non-penetration depth That is,

[0085]

[0086] Figure 2 The one-dimensional lookup table and the interpolation version between the non-penetration depth and the forest height are shown. The P-band SAR signal has complete penetration in the tropical rainforest low forest area, and there is non-penetration in the high forest area; with the increase of the forest height, the non-penetration depth also gradually increases.

[0087] In step five, based on the forest height result obtained by the PolInSAR image inversion, the height deviation contained in the RVoG under-forest terrain product is corrected through the one-dimensional lookup table, so as to realize the purpose of large-scale high-precision under-forest terrain mapping.

[0088] Through the sparse lookup table obtained in step four, the continuous mapping relationship between the forest height and the non-penetration depth is obtained through one-dimensional interpolation This process can also be completed in advance in step four; then, the forest height obtained by the PolInSAR surface observation and continuous coverage inversion is used to correct the height deviation in the RVoG under-forest terrain product caused by the insufficient SAR penetration, so as to obtain a large-scale high-precision under-forest terrain product That is,

[0089]

[0090] Among them, is the non-penetration depth mapped from the forest height according to the one-dimensional lookup table interpolation version.

[0091] Figure 3 And Figure 4 The DTM product inverted by different methods and the height accuracy graph taking the LiDAR DTM as a reference are shown in (a) TanDEM-X DEM; (b) DTM obtained by RVoG; (c) DTM corrected by using the application. As Figure 4(a) The TanDEM-X DEM product shows a significant systematic overestimation of the terrain, with a root mean square error (RMSE) of 17.14 m and a bias of 15.95 m. This result highlights the limited penetration capability of X-band SAR signals, with the interferometric phase center usually located in the upper part of the canopy rather than near the actual ground, which will lead to a severe overestimation in densely forested areas. Figure 4 (b) The linear fitting method based on the RVoG model shows a terrain result with a RMSE of 9.57 m and a bias of 6.46 m compared with the LiDAR DTM. This result shows that even if the full polarimetric P-band PolInSAR data and the parametric RVoG model inversion method are used, the incomplete penetration of the radar signal will still cause the terrain to be overestimated; because even if the polarimetric coherent scattering model is used for ground phase compensation, the P-band SAR signal still cannot fully penetrate to the ground. As shown in Figure 4 (c) The corrected terrain result of the present application further reduces the RMSE to 5.90 m and the height bias to -1.12 m; the bias of the corrected DTM product of the present application is basically close to zero, which verifies that the proposed method can solve the problem of terrain height overestimation caused by insufficient SAR signal penetration, and shows its great potential in high-precision mapping of under-forest terrain in dense forest areas.

[0092] At this point, the long-wave band PolInSAR under-forest terrain inversion method based on non-penetration depth correction is completed and implemented. Through the above process, based on the polarimetric coherent scattering RVoG model, with the help of local area or sparse LiDAR ground elevation information, through a one-dimensional lookup table from tree height to non-penetration depth, the task of PolInSAR technology for large-scale under-forest terrain mapping can be achieved, especially for tropical rainforest ground terrain correction.

[0093] In specific implementation, the method proposed by the technical scheme of the present application can be automatically run by a person skilled in the art using computer software technology, and the system device of the method, such as a computer readable storage medium storing the corresponding computer program of the technical scheme of the present application and a computer device including a computer program running device, should also be within the protection scope of the present application.

[0094] The following describes the long-wave band PolInSAR under-forest terrain inversion electronic device based on non-penetration depth correction provided by the present application, and the long-wave band PolInSAR under-forest terrain inversion electronic device based on non-penetration depth correction described below can be mutually corresponding to the long-wave band PolInSAR under-forest terrain inversion method based on non-penetration depth correction described above.

[0095] The electronic device can include a processor, a communications interface, a memory and a communications bus, wherein the processor, the communications interface and the memory complete mutual communication through the communications bus. The processor can invoke a logic instruction in the memory to execute the long-wave band PolInSAR under-forest terrain inversion method based on non-penetration depth correction, mainly including the software processing part in the above steps.

[0096] In addition, the logic instruction in the memory described above can be realized in the form of a software function unit and sold or used as a stand-alone product, and can be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0097] On the other hand, the present application also provides a computer program product, which comprises a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and the computer program is executed by a processor, and the computer can execute the software processing part in the long-wave band PolInSAR under-forest terrain inversion method based on non-penetration depth correction provided by the above-mentioned methods.

[0098] In yet another aspect, the present application also provides a non-transitory computer readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the software processing part in the long-wave band PolInSAR under-forest terrain inversion method based on non-penetration depth correction provided by the above-mentioned methods.

[0099] The device embodiments described above are only schematic, wherein the units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e. they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the present embodiment scheme according to actual needs. Those skilled in the art can understand and implement it without creative labor.

[0100] Those skilled in the art can clearly understand the implementation of the embodiments by means of software and necessary general hardware platforms through the description of the above embodiments, and of course, the embodiments can also be implemented by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, and the computer software product can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0101] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A long-wave band PolInSAR under-forest terrain inversion method based on non-penetration depth correction, characterized in that, The method comprises the following steps: acquire two scenes of full-polarization SAR image data and pre-process the data to generate a polarimetric interferometric covariance matrix; whiten the polarimetric interferometric covariance matrix, apply a normal matrix constraint, obtain an optimal normal matrix approximate solution, and extract two end-point complex coherence coefficients of a coherent line segment; based on a polarimetric coherent scattering model, solve model parameters using the end-point complex coherence coefficients, wherein the model parameters include a surface phase, a forest height, an extinction coefficient, and a ground volume amplitude ratio; combine sparse LiDAR elevation data, calculate the forest height and the non-penetrating depth between the RVoG inverted terrain and the LiDAR reference terrain, and construct a one-dimensional lookup table based on the mapping relationship between the forest height and the non-penetrating depth; correct the deviation of the PolInSAR-inverted under-forest terrain elevation according to the one-dimensional lookup table, and output a corrected digital terrain model.

2. The method of claim 1, wherein: The implementation process of applying the normal matrix constraint comprises applying a normal matrix constraint to the whitened interferometric cross-correlation matrix, generating an intermediate variable parameter, solving the optimal normal matrix approximate solution, and ensuring the geometric stability of the coherent line segment within the complex unit circle.

3. The method of claim 1, wherein: When solving the model parameters using the end-point complex coherence coefficients, the surface phase is determined using the sign of the interferometric vertical wave number; the end-point complex coherence coefficients are substituted into the polarimetric coherent scattering model, and the forest height, the extinction coefficient, and the ground volume amplitude ratio are solved through nonlinear iterative optimization.

4. The method of claim 1, wherein: The construction implementation of the one-dimensional lookup table comprises dividing samples according to preset forest height intervals, and statistically calculating the mean forest height and the mean non-penetrating depth in each interval, thereby constructing the mapping relationship between the forest height and the non-penetrating depth, and obtaining a sparse lookup table; based on the sparse lookup table, a continuous forest height and non-penetrating depth mapping relationship is generated through one-dimensional interpolation.

5. The method of claim 1, wherein: The implementation process of correcting the deviation of the under-forest terrain elevation is as follows: based on the PolInSAR-inverted forest height, the non-penetrating depth compensation value is obtained through one-dimensional lookup table interpolation; the compensation value is deducted from the RVoG-inverted terrain to obtain the corrected under-forest terrain elevation.

6. The method of claim 1, wherein: The length of the interferometric baseline of the two scenes of full-polarization SAR image data satisfies the condition that the ambiguity height is greater than half of the maximum forest height.

7. The method of claim 1, wherein: The method is used for realizing the inversion of under-forest terrain in a high-forest coverage area.

8. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that: The processor executes the program to realize the long-wave band PolInSAR under-forest terrain inversion method based on non-penetrating depth correction according to any one of claims 1 to 7. 9.A non-transitory computer-readable storage medium having stored thereon a computer program. The computer program is executed by the processor to realize the long-wave band PolInSAR under-forest terrain inversion method based on non-penetrating depth correction according to any one of claims 1 to 7.

10. A computer program product comprising a computer program, characterized in that: The computer program is executed by the processor to realize the long-wave band PolInSAR under-forest terrain inversion method based on non-penetrating depth correction according to any one of claims 1 to 7.

Images