Forest height inversion method and device for steep mountainous area based on fused ascending-descending track short wave InSAR data

Abstract

The application discloses a kind of steep mountain forest height retrieval method and device of fusion lifting track short wave InSAR data, to solve the problem of insufficient observation information when InSAR is retrieved, steep mountain slope decorrelation, including: obtaining lifting track short wave InSAR data and sparse LiDAR data and preprocessing;Utilize digital elevation model to calculate slope, and carry out topographic effect correction to effective vertical wave number;Determine the overlapping area of lifting track data, establish semi-empirical SINC model to retrieve forest height in non-overlapping area;For overlapping area, construct double baseline forest height retrieval model considering coherence phase and amplitude simultaneously, determine optimal model parameters with sparse LiDAR forest height as control, and retrieve forest height;Finally, mosaic forest height in overlapping area and non-overlapping area forms complete continuous product.The application realizes high-precision forest height retrieval in steep mountain by fusing lifting track InSAR data, with good stability and popularization value.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of quantitative remote sensing inversion of forest attribute parameters, and particularly relates to a steep mountainous forest height inversion method and device fusing ascending and descending track short-wave InSAR data. BACKGROUND

[0002] Forest height, as a key parameter reflecting the growth state and vertical structure of forests, is closely related to biomass accumulation and carbon sink capacity. Accurate measurement of forest height is of great significance for quantitative estimation, evaluation and change monitoring of aboveground biomass and terrestrial carbon storage. Interferometric Synthetic Aperture Radar (InSAR) technology has the advantages of strong penetration, sensitivity to forest height, vertical structure and dielectric properties in forest areas, and is widely used in forest height inversion. However, in complex mountainous areas with steep terrain, single baseline and single polarization InSAR data face the following two problems when inverting forest height: 1) InSAR complex coherence only contains two observations (phase and amplitude), and the observation information is insufficient, which leads to an underdetermined problem when using existing models for inversion; 2) Due to the slope decorrelation, the corresponding relationship between coherent amplitude and forest height is distorted in complex mountainous areas, and the slope will cause the penetration depth of InSAR signal to change. Although some methods have been developed to correct the slope effect, the model solving is limited by various assumptions, and lacks universality. Therefore, it is necessary to design an inversion method that can solve the above two problems, and use single baseline and single polarization short-wave InSAR data to realize large-scale and high-precision forest height inversion. SUMMARY

[0003] The application provides a steep mountainous forest height inversion method and device fusing ascending and descending track short-wave InSAR data, which can realize high-precision forest height inversion in steep terrain areas under the condition of using only ascending and descending track InSAR data and a small amount of spaceborne sparse LiDAR forest height as control.

[0004] To achieve the above technical purposes, the application adopts the following technical solutions:

[0005] A steep mountainous forest height inversion method fusing ascending and descending track short-wave InSAR data, comprising:

[0006] Step 1: Obtain ascending and descending track short-wave bistatic InSAR data and sparse LiDAR data in the forest coverage area, and pre-process the ascending and descending track InSAR data respectively to obtain complex coherence, a digital elevation model and an effective vertical wave number;

[0007] Step 2, calculate the range slope based on the digital elevation model, and use the range slope to correct the effective vertical wavenumber for the terrain effect;

[0008] Step 3, determine the overlapping area of the elevation InSAR data according to the latitude and longitude range; for the non-overlapping area, based on the semi-empirical SINC model, determine the semi-empirical parameters by taking the spaceborne sparse LiDAR as the control point and the minimum root mean square error as the criterion, and then use the semi-empirical SINC model to perform forest height inversion for each pixel in the non-overlapping area;

[0009] Step 4, for the overlapping area, first, calculate the elevation difference using the digital elevation model obtained by the elevation InSAR; then, construct a double-baseline forest height inversion model considering both the coherent phase and the amplitude information; then, determine the optimal model parameters of the double-baseline forest height inversion model by taking the spaceborne sparse LiDAR forest height as the control; finally, use the double-baseline forest height inversion model to perform forest height inversion for each pixel in the overlapping area;

[0010] Step 5, mosaic the forest height of the overlapping area and the forest height of the non-overlapping area to generate a complete and continuous forest height inversion product.

[0011] Further, the InSAR data is preprocessed, including: SAR image registration, spectrum filtering, interference, baseline estimation, flat phase removal, coherence estimation, phase adaptive filtering, phase unwrapping, phase-height conversion and geographic coding.

[0012] Further, the effective vertical wavenumber calculated in the preprocessing is:

[0013] (1);

[0014] In the formula, indicates the effective vertical wavenumber calculated by preprocessing, is the vertical baseline, is the radar wavelength, is the slant range, is the incidence angle.

[0015] Further, the process of terrain effect correction in step 2 is:

[0016] First, calculate the range slope in the radar coordinate system according to the digital elevation model generated in step 1:

[0017] (2);

[0018] In the formula, indicates the range slope, indicates the digital elevation model, and respectively represent the range and azimuth coordinates of a pixel, represent the ground range corresponding to a single pixel;

[0019] Then, the local incidence angle is calculated:

[0020] (3) ;

[0021] where, is the incidence angle;

[0022] Finally, the effective vertical wavenumber calculated in the preprocessing is corrected using the local incidence angle :

[0023] (4) ;

[0024] where, represents the effective vertical wavenumber calculated in the preprocessing, is the corrected effective vertical wavenumber.

[0025] Further, the specific process of step 3 includes:

[0026] Step 3.1, determine the spatial overlap area of ascending and descending InSAR data according to the latitude and longitude ranges of the data; let the latitude and longitude ranges of the ascending data be: , and the latitude and longitude ranges of the descending data be: , then the overlapping area of the ascending and descending InSAR data is:

[0027] (5) ;

[0028] where, and represent the latitude and longitude, respectively, and represent the ascending and descending data, respectively, and represent the maximum and minimum value operations, respectively, represents the intersection; thus, the rectangular overlapping area with latitude and longitude ranges of , is obtained, and other areas are non-overlapping areas;

[0029] Step 3.2, determine the semi-empirical SINC model for inverting non-overlapping areas as:

[0030] (6) ;

[0031] where, is the coherence amplitude, is the forest height, is the corrected effective vertical wavenumber, and is the semi-empirical parameter to be determined;

[0032] Step 3.3, the forest height of the spaceborne sparse LiDAR control point is controlled, and the root mean square error RMSE is the criterion for estimating the semi-empirical parameter of the ascending and descending non-overlapping areas respectively:

[0033] (7);

[0034] In the formula, indicates the forest height of the spaceborne sparse LiDAR control point , indicates the forest height of the semi-empirical SINC model inversion control point , and n is the number of LiDAR control points. Specifically, by using formula (6) to traverse the forest height under different value conditions in the corresponding value range of and

[0035] , the values of and corresponding to the minimum RMSE of the spaceborne LiDAR forest height are taken as the final semi-empirical parameters of the semi-empirical SINC model; a corresponding set of semi-empirical parameters is obtained for the ascending and descending non-overlapping areas respectively. Step 3.4, based on the formula (6) of the determined semi-empirical parameter, the forest height of the non-overlapping area is inverted.

[0036] Further, the specific process of step 4 includes:

[0037] Step 4.1, calculate the height difference using the ascending and descending InSAR digital elevation models:

[0038] (8);

[0039] In the formula, respectively indicate the digital elevation models obtained by the ascending and descending short-wave double-station InSAR data,

[0040] indicates the height difference between the ascending and descending tracks. Step 4.2, construct a double-baseline forest height inversion model considering the coherence phase and amplitude information at the same time;

[0041] First, the InSAR complex coherence can be expressed as a function of the forest height:

[0042]

[0043] ​​ (9);

[0044] In the formula, Indicating complex coherence, For surface phase, Extinction coefficient, The ratio of the earth's amplitude;

[0045] Subsequently, assuming a positive range slope in the ascending InSAR data, the dual-baseline forest height inversion model is as follows:

[0046] (10);

[0047] When the range slope of the ascending InSAR data is negative, the dual-baseline forest height inversion model is as follows:

[0048] (11);

[0049] In the formula, and These represent the complex coherence of observations during ascent and descent, respectively. and These represent the complex coherence calculated theoretically for ascending and descending orbits, respectively. and These represent the effective vertical wavenumbers after orbital ascent and descent corrections, respectively. and The extinction coefficients for ascending and descending orbits are given. and The ratio of the elevation and descent of the ground body;

[0050] Step 4.3, with the range slope of the ascending InSAR data positive, using spaceborne sparse LiDAR control points forest height To control, and to calculate using formula (10) and Inversion control points under different value conditions within the corresponding value range forest height Calculate the RMSE; when the RMSE is minimized, determine the optimal parameters for the ascending InSAR data with a positive range slope. and :

[0051] (12);

[0052] When the range slope of the ascending InSAR data is negative, use spaceborne sparse LiDAR control points. forest height To control, and to calculate using formula (11) and inversion control points under different value conditions in the corresponding value range forest height , calculate RMSE; when the RMSE is the smallest, determine the optimal parameters of the ascending track InSAR data under the condition that the distance slope is negative and ;

[0053] Step 4.4, under the condition that the distance slope of the ascending track InSAR data is positive, perform pixel-by-pixel forest height inversion on the overlapping area using the formula (10) of the determined optimal parameters and ; under the condition that the distance slope of the ascending track InSAR data is negative, perform pixel-by-pixel forest height inversion on the overlapping area using the formula (11) of the determined optimal parameters and .

[0054] Further, the complete and continuous forest height inversion product generated in step 5 is represented as:

[0055] (13) ;

[0056] wherein, is the forest height of any pixel point obtained by final inversion, is the forest height obtained by inversion based on the semi-empirical SINC model in the non-overlapping area of the ascending track, is the forest height obtained by inversion based on the semi-empirical SINC model in the non-overlapping area of the descending track, is the forest height obtained by inversion based on the double baseline model in the overlapping area of the ascending and descending tracks.

[0057] An apparatus for forest height inversion in steep mountainous areas by fusing ascending and descending track short-wave InSAR data, comprising:

[0058] a data preprocessing module, configured to perform SAR image registration, spectrum filtering, interference, baseline estimation, flat earth phase removal, coherence estimation, phase adaptive filtering, phase unwrapping, phase-to-height conversion and geographic coding on the obtained ascending and descending track InSAR data respectively, to obtain complex coherence, a digital elevation model and an effective vertical wave number;

[0059] a terrain effect correction module, configured to calculate a distance slope based on the digital elevation model, and correct the effective vertical wave number by using the distance slope;

[0060] a region division module, configured to determine an overlapping area of the ascending and descending track InSAR data according to the latitude and longitude range;

[0061] A single-baseline forest height inversion module is configured to: based on a semi-empirical SINC model, determine semi-empirical parameters by taking star-borne sparse LiDAR as a control point and taking the minimum root mean square error as a criterion, and then perform forest height inversion on a non-overlapping area pixel by pixel by using the semi-empirical SINC model;

[0062] A double-baseline forest height inversion module is configured to: first, calculate the height difference by using a digital elevation model corresponding to ascending / descending track InSAR; then, construct a double-baseline forest height inversion model considering both coherent phase and amplitude information; then, determine the optimal model parameters of the double-baseline forest height inversion model by taking the star-borne sparse LiDAR forest height as a control; and finally, perform forest height inversion on an overlapping area pixel by pixel by using the double-baseline forest height inversion model.

[0063] A forest height output module is configured to: mosaic the inversion results of the non-overlapping area and the overlapping area, and output a complete and continuous forest height inversion product.

[0064] The present application has the following advantages:

[0065] The present application solves the problem of performance decline caused by insufficient observation information and only using coherent amplitude in the prior art by considering both coherence and phase based on ascending / descending track InSAR data, and realizes adaptive parameter adjustment under different slope conditions by considering the decorrelation effect caused by slope and the difference in penetration depth. Therefore, the present application can guarantee high stability of forest height inversion in steep and complex mountainous areas, different forest types and terrain conditions, and provides a method and technical support for large-scale forest height inversion and above-ground biomass mapping. BRIEF DESCRIPTION OF DRAWINGS

[0066] Figure 1 The present application is described as a flowchart of the method.

[0067] Figure 2 The present application is described as a flowchart of the method. Figure 2 (a) is the geographical position of the test area, Figure 2 (b) is the terrain elevation and star-borne sparse LiDAR coverage range (black dots) of the test area, Figure 2 (c) is the coverage range of the ascending track (red) and descending track (blue) InSAR data used.

[0068] Figure 3 The present application is described as a flowchart of the method. Figure 3 (a) is the distance slope distribution (see Figure 3 (b) is the elevation difference map (see of the ascending / descending track InSAR data in the double-baseline forest height model modeling of the present application.

[0069] Figure 4 Forest height estimated by different methods and techniques. Figure 4 (a) is the forest height inverted by the traditional simplified model using only the coherence amplitude, Figure 4 (b) is the forest height estimated by the method of the present application, Figure 4 (c) is the reference forest height true value.

[0070] Figure 5 Forest height accuracy estimated by different methods. Figure 5 (a) is the accuracy of the forest height inverted by the traditional simplified model using only the coherence amplitude, Figure 5 (b) is the accuracy of the forest height estimated by the method of the present application. DETAILED DESCRIPTION

[0071] The embodiments of the present application are described in detail below, which are based on the technical solutions of the present application, and give detailed implementation modes and specific operation processes, and further explain the technical solutions of the present application.

[0072] In order to better illustrate the method and steps of the present application, TanDEM-X bistatic InSAR data and spaceborne ICESat-2 ATL08 data are used in a test area located in northern Spain to further illustrate the present application in detail. It should be noted that the specific implementation described here is only used to explain the present application and is not used to limit the present application.

[0073] The present embodiment provides a steep mountainous forest height inversion method fusing ascending and descending track shortwave InSAR data, as shown in Figure 1 , which comprises the following steps:

[0074] Step 1, acquiring ascending and descending track shortwave bistatic InSAR data and sparse LiDAR data in a forest coverage area, respectively pre-processing the ascending and descending track InSAR data to obtain complex coherence, digital elevation model and effective vertical wave number;

[0075] The present embodiment selects TanDEM-X bistatic InSAR data in northern Spain Figure 2 (a), the ascending track and descending track data are shown in the red and blue rectangular boxes of Figure 2 (c) respectively; the spaceborne LiDAR data selects ICESat-2 ATL08 data, and the data coverage range is shown by the black dot circle in Figure 2 (b) ; the entire test area is covered by airborne LiDAR forest height products, which are used for subsequent verification and analysis.

[0076] In the pre-processing step of SAR data, SAR image registration, spectral filtering, interferometry, baseline estimation, flat earth phase removal, coherence estimation, phase adaptive filtering, phase unwrapping, phase-to-height conversion and geocoding are all mature techniques and methods, and this embodiment will not be described in detail. For the specific implementation process, please refer to the book: Jin Guowang, Xu Qing, Zhang Hongmin. Synthetic aperture radar interferometry [M]. National Defense Industry Press, 2014.

[0077] Then, in order to invert the forest height, an important parameter is to calculate the effective vertical wavenumber:

[0078] (1);

[0079] In the formula, k is the effective vertical wavenumber, B is the vertical baseline, λ is the radar wavelength, R is the slant range (the distance from the satellite antenna center to the ground surface), θ is the incidence angle.

[0080] Step 2, calculate the range slope based on the digital elevation model, and use the range slope to correct the terrain effect of the effective vertical wavenumber.

[0081] First, calculate the range slope in the radar coordinate system according to the digital elevation model generated in step 1:

[0082] (2);

[0083] In the formula, S is the range slope, DEM is the digital elevation model, and respectively represent the range and azimuth coordinates, R represents the ground surface distance corresponding to a single pixel;

[0084] Then, calculate the local incidence angle :

[0085] (3);

[0086] Finally, use the local incidence angle to correct the effective vertical wavenumber calculated by pre-processing:

[0087] (4);

[0088] In the formula, k is the corrected effective vertical wavenumber.

[0089] Step 3, determining the overlapping area of the ascending and descending InSAR data according to the latitude and longitude range; for the non-overlapping area, based on the semi-empirical SINC model, taking the spaceborne sparse LiDAR as the control point and the minimum root mean square error as the criterion to determine the semi-empirical parameters, and then using the semi-empirical SINC model to perform forest height inversion on the non-overlapping area pixel by pixel.

[0090] Step 3.1, determining the spatial overlapping area of the ascending and descending InSAR data according to the latitude and longitude range. Let the latitude and longitude range of the ascending data be: 、 and the latitude and longitude range of the descending data be: 、 then the overlapping area of the ascending and descending InSAR data is:

[0091] (5);

[0092] wherein, and represent the latitude and longitude respectively, and represent the ascending and descending data respectively, and represent the maximum and minimum value operations respectively, represents the intersection; thus the rectangular overlapping area with the latitude and longitude range of 、 is obtained, and other areas are non-overlapping areas. As shown in Figure 2 (c), the yellow range is the identified overlapping area, and other InSAR data coverage areas are non-overlapping areas.

[0093] Step 3.2, determining the semi-empirical SINC model for the non-overlapping area as:

[0094] (6);

[0095] wherein, is the coherence amplitude, is the forest height, and are semi-empirical parameters to be determined.

[0096] Step 3.3, taking the spaceborne sparse LiDAR forest height as the control and the minimum root mean square error RMSE as the criterion, respectively estimating the semi-empirical parameters of the non-overlapping area of the ascending and descending data.

[0097] (7);

[0098] wherein, represents the spaceborne sparse LiDAR control point forest height, forest height, forest height, the number of LiDAR control points.

[0099] Specifically, the value range of is set to [0.7, 1], the value range of is set to [0.7, 2]; by using formula (6) to traverse and calculate the forest height under different and corresponding to the minimum RMSE of the spaceborne LiDAR forest height, and and are the final non-overlapping area semi-empirical parameters.

[0100] Step 3.4, based on the semi-empirical parameters obtained in step 3.3, the forest height of the non-overlapping area is inverted by using formula (6).

[0101] wherein, step 3.3 needs to estimate the corresponding semi-empirical parameters according to ascending and descending InSAR data, and then apply to forest height inversion of ascending / descending non-overlapping area.

[0102] Step 4, for the overlapping area, first, the elevation difference is calculated by using the digital elevation model obtained by ascending / descending InSAR; then, a dual baseline forest height inversion model considering both coherent phase and amplitude information is constructed; then, the optimal model parameters of the dual baseline forest height inversion model are determined by taking the spaceborne sparse LiDAR forest height as control; finally, the dual baseline forest height inversion model is used to perform forest height inversion pixel by pixel for the overlapping area.

[0103] Step 4.1, the elevation difference is calculated by using the digital elevation model of ascending / descending InSAR:

[0104] (8);

[0105] wherein, respectively represent the digital elevation model obtained by ascending and descending short-wave dual station InSAR data, represents the elevation difference between ascending and descending tracks. Figure 3 (a) shows the distance slope statistical chart of the overlapping area, it can be found that the area has significant terrain undulation, and the slope of most areas is greater than 10°. Figure 3 (b) shows the relationship between the calculated elevation difference and forest height, it can be found that due to the influence of the penetration depth difference caused by the terrain slope, there is a significant elevation difference between ascending and descending tracks, which provides a basis for subsequent construction of the dual baseline model considering phase.

[0106] Step 4.2. Building the dual-baseline forest height inversion model considering both the coherent phase and the amplitude information.

[0107] Firstly, the complex coherence can be expressed as a function of forest height:

[0108] (9) ;

[0109] where, is the complex coherence, is the surface phase, is the extinction coefficient, is the terrain amplitude ratio.

[0110] Secondly, when the range slope of the ascending InSAR data is positive, the dual-baseline forest height inversion model is:

[0111] (10) ;

[0112] When the range slope of the ascending InSAR data is negative, the dual-baseline forest height inversion model is:

[0113] (11) ;

[0114] where, and are the observed complex coherence of the ascending and descending tracks, respectively, and are the calculated theoretical complex coherence of the ascending and descending tracks, respectively, and are the corrected effective vertical wave numbers of the ascending and descending tracks, respectively, and are the extinction coefficients of the ascending and descending tracks, and are the terrain amplitude ratios of the ascending and descending tracks.

[0115] Step 4.3. From equations (10) and (11), it can be seen that, due to the lack of observed information, the values of and need to be determined in advance when solving the model. It should be noted that when the range slope of the ascending InSAR data is positive, the range slope of the descending InSAR data at this geographic location is negative. Due to the positive slope penetration depth enhancement, the SAR signal is relatively insensitive to the change of the extinction coefficient, so for the data with positive slope, the dual-baseline forest height inversion model adopts the fixed way. Due to the weak negative slope penetration depth, the short-wave InSAR signal hardly penetrates the forest layer, so it is relatively insensitive to the change of the terrain amplitude ratio, so for the negative slope, the fixed way can be adopted.

[0116] Therefore, for the geographical locations with positive range slope of spaceborne sparse LiDAR coverage and ascending InSAR data (i.e. the yellow region in (c) and the locations with ICESat-2 black dot coverage), the value range of is set to [0, 1], Figure 2 the value range of is set to [0, 2], and the different values of are calculated at intervals of 0.01. and the double-baseline forest height inverted under ; then, taking the spaceborne sparse LiDAR forest height as control, the RMSE of the forest height inverted under different and is calculated, and when the RMSE is the smallest, the optimal parameters and (i.e. formula (10)) are determined.

[0117] (12).

[0118] Finally, the optimal and under the condition that the range slope of ascending InSAR data is negative, i.e. formula (11), are determined by using the same method.

[0119] Step 4.4. Perform pixel-by-pixel forest height inversion for the overlapping area.

[0120] Specifically, when the range slope of the ascending InSAR data of the inverted pixel is positive, formula (10) is used for inversion; otherwise, formula (11) is used for inversion.

[0121] Step 5. Mosaic the forest height of the overlapping area and the forest height of the non-overlapping area to finally generate a complete and continuous forest height inversion product.

[0122] In order to ensure that the forest height at the joint edge of the overlapping area and the non-overlapping area is continuous and does not jump, spatial smoothing processing is performed on the joint edge area by using Gaussian smoothing. The final forest height result is represented as:

[0123] (13).

[0124] In the formula, is the forest height of any pixel point finally inverted, is the forest height of the ascending non-overlapping area inverted based on the semi-empirical SINC model, is the forest height of the descending non-overlapping area inverted based on the semi-empirical SINC model, and ​​The forest height is obtained by inversion based on the double baseline model for the overlapping area of the ascending and descending tracks.

[0125] Figure 4 (b) shows the spatial distribution of the forest height inverted by the embodiment of the present application, which is more consistent with the reference LiDAR true value (see Figure 4 (c)) than the forest height inverted by the conventional simplified model using only the coherent amplitude (see Figure 4 (a)), which shows the advantage of the present application in the inversion of high-precision forest height in complex mountainous areas. Specifically, for the selected area in the embodiment, both the forest height and the terrain show high variability, and the method of the conventional simplified model overestimates the forest height due to the serious slope decorrelation (see Figure 4 (a)), in contrast, the method of the present application effectively alleviates this overestimation phenomenon by fusing the InSAR data of the ascending and descending tracks to provide additional observations and establishing a double baseline forest height model considering the phase difference (see Figure 4 (b)).

[0126] In order to more obviously highlight the advantage of the present application in the inversion of high-precision forest height, Figure 5 scatter plots of the forest height estimated by the conventional method (see Figure 5 (a)) and the method of the present application (see Figure 5 (b)) and the reference true value are drawn. The root mean square error (RMSE) of the forest height estimated by the method of the present application is 4.62 m, which is about 43% higher than that of the conventional method (RMSE = 8.11 m).

[0127] The above embodiments are preferred embodiments of the present application, and those of ordinary skill in the art can also make various transformations or improvements on the basis of the above embodiments, and these transformations or improvements should all belong to the scope of protection of the present application without departing from the general concept of the present application.

Claims

1. A method for forest height inversion in steep mountainous area by fusing long-wavelength InSAR data of ascending and descending tracks, characterized in that, Comprise: Step 1, obtain ascending and descending track shortwave bistatic InSAR data and sparse LiDAR data of forest coverage area, and respectively preprocess the ascending and descending track InSAR data to obtain complex coherence, digital elevation model and effective vertical wave number; Step 2, calculate the range slope based on the digital elevation model, and correct the effective vertical wave number using the range slope; Step 3, determine the overlapping area of the ascending and descending track InSAR data according to the latitude and longitude range; for the non-overlapping area, based on the semi-empirical SINC model, take the spaceborne sparse LiDAR as the control point and the minimum root mean square error as the criterion to determine the semi-empirical parameters, and then use the semi-empirical SINC model to inversely calculate the forest height of the non-overlapping area pixel by pixel; Step 4, for the overlapping area, first, calculate the height difference using the digital elevation model obtained by the ascending and descending track InSAR; then, construct a double-baseline forest height inversion model considering the coherent phase and amplitude information; then, determine the optimal model parameters of the double-baseline forest height inversion model by taking the spaceborne sparse LiDAR forest height as the control; finally, use the double-baseline forest height inversion model to inversely calculate the forest height of the overlapping area pixel by pixel; Step 5, mosaic the forest height of the overlapping area and the forest height of the non-overlapping area to generate a complete and continuous forest height inversion product.

2. The method according to claim 1, wherein The preprocessing of InSAR data includes: SAR image registration, spectrum filtering, interference, baseline estimation, flat phase removal, coherence estimation, phase adaptive filtering, phase unwrapping, phase-height conversion and geographic coding.

3. The steep mountainous forest height inversion method of fusing ascending and descending tracks short-wave InSAR data according to claim 1, characterized in that, The way to calculate the effective vertical wave number in the preprocessing is: (1)； wherein, represents the effective vertical wave number calculated by preprocessing, is the vertical baseline, is the radar wavelength, is the slant range, is the incidence angle.

4. The steep mountainous forest height inversion method of fusing ascending and descending tracks short-wave InSAR data according to claim 1, characterized in that, The process of the terrain effect correction in step 2 is: First, calculate the range slope in the radar coordinate system according to the digital elevation model generated in step 1: (2)； wherein, denotes the distance-wise slope, denotes a digital elevation model, and denotes the distance-wise and azimuth-wise coordinates of a pixel, respectively, denotes the ground surface distance corresponding to a single pixel; Afterwards, the local angle of incidence is calculated : (3)； In the formula, is the angle of incidence; Finally, the local incidence angle The effective vertical wave number correction: (4)； In the formula, represents the effective vertical wave number calculated by preprocessing, is the corrected effective vertical wave number.

5. The steep mountainous forest height inversion method of fusing ascending and descending tracks short-wave InSAR data according to claim 1, characterized in that, The specific process of step 3 includes: Step 3.1, determine the spatial overlap area of ascending and descending InSAR data according to their latitude and longitude ranges; let the latitude and longitude ranges of ascending data be: , , and the latitude and longitude ranges of descending data be: , , then the overlap area of ascending and descending InSAR data is: (5)； wherein, and denote latitude and longitude, respectively, and denote ascending and descending track data, respectively, and denote max and min operations, respectively, denotes intersection; thus, the latitude and longitude ranges are , the rectangular overlap region, and other regions are non-overlapping regions; Step 3.2, the semi-empirical SINC model used to inversely calculate the non-overlapping area is: (6)； wherein is the coherence amplitude, is the forest height, is the corrected effective vertical wavenumber, and is a semi-empirical parameter to be determined. Step 3.3, taking the forest height of the spaceborne sparse LiDAR control point as the control and the minimum root mean square error RMSE as the criterion, estimate the semi-empirical parameters of the ascending track and descending track non-overlapping areas respectively: (7)； wherein represents the forest height at the control points of the spaceborne sparse LiDAR, represents the forest height at the control points of the spaceborne sparse LiDAR, represents the forest height at the control points of the semi-empirical SINC model inversion, represents the forest height at the control points of the semi-empirical SINC model inversion, is the number of LiDAR control points; Specifically, by using formula (6) to traverse and corresponding to the range of different values of the forest height, the value corresponding to the minimum RMSE of the spaceborne LiDAR forest height is taken as the final value of the semi-empirical parameter of the semi-empirical SINC model and the value is the final semi-empirical parameter of the semi-empirical SINC model; a corresponding set of semi-empirical parameters is obtained for the non-overlapping area of the ascending orbit and the descending orbit respectively; Step 3.4, inversely calculate the forest height of the non-overlapping area based on the formula (6) of the determined semi-empirical parameters.

6. The steep mountainous forest height inversion method of fusing ascending and descending tracks short-wave InSAR data according to claim 1, characterized in that, The specific process of step 4 includes: Step 4.1, calculate the height difference using the digital elevation model of the ascending and descending track InSAR: (8)； wherein respectively represent the digital elevation models obtained from ascending and descending short-wavelength bistatic InSAR data, denotes the ascending-descending track elevation difference; Step 4.2, construct a double-baseline forest height inversion model considering the coherent phase and amplitude information; First, the InSAR complex coherence is expressed as a function of forest height: （9）； wherein denotes complex coherence, is the surface phase, is the extinction coefficient, is the surface amplitude ratio; is the incidence angle, is the forest height, is the corrected effective vertical wave number; Then, in the case that the range slope of the ascending track InSAR data is positive, the double-baseline forest height inversion model is: (10)； In the case that the range slope of the ascending track InSAR data is negative, the double-baseline forest height inversion model is: (11)； where and denote the complex coherences of the ascending and descending tracks observations, and denote the complex coherences of the ascending and descending tracks theoretical calculations, and denote the corrected effective vertical wave numbers of the ascending and descending tracks, and are the extinction coefficients of the ascending and descending tracks, and are the crustal amplitude ratios of the ascending and descending tracks. Step 4.3, with the range slope of the ascending InSAR data positive, using spaceborne sparse LiDAR control points forest height To control, and to calculate using formula (10) and Inversion control points under different value conditions within the corresponding value range forest height Calculate the RMSE; when the RMSE is minimized, determine the optimal parameters for the ascending InSAR data with a positive range slope. and : (12)； In the case of negative range slope of ascending InSAR data, the forest height of control point of spaceborne sparse LiDAR is controlled, and the forest height of control point is calculated by formula (11) and The forest height of control point is inverted under different value conditions in the corresponding value range, and the RMSE is calculated; when the RMSE is the smallest, the optimal parameters and in the case of negative range slope of ascending InSAR data are determined and ; Step 4.4, in the case of positive range slope of ascending InSAR data, using the formula (10) of the determined optimal parameters and to perform the forest height inversion pixel by pixel in the overlapping area; in the case of negative range slope of ascending InSAR data, using the formula (11) of the determined optimal parameters and to perform the forest height inversion pixel by pixel in the overlapping area.

7. The steep mountainous forest height inversion method of fusing ascending and descending tracks short-wave InSAR data according to claim 1, characterized in that, The complete and continuous forest height inversion product generated in step 5 is expressed as: (13)； In the formula, is the forest height of any pixel point obtained by final inversion, is the forest height of the ascending track non-overlapping area obtained by inversion based on the semi-empirical SINC model, is the forest height of the descending track non-overlapping area obtained by inversion based on the semi-empirical SINC model, is the forest height of the ascending and descending track overlapping area obtained by inversion based on the double baseline model.

8. A device for fusing forest height inversion of steep mountainous area from long-wave InSAR data of ascending and descending tracks, characterized in that, Comprise: The data preprocessing module is configured to perform SAR image registration, spectrum filtering, interference, baseline estimation, flat earth phase removal, coherence estimation, phase adaptive filtering, phase unwrapping, phase-to-height conversion and geocoding on the obtained ascending and descending track InSAR data respectively, to obtain complex coherence, a digital elevation model and an effective vertical wave number; The terrain effect correction module is configured to calculate a range slope based on the digital elevation model, and correct the effective vertical wave number based on the range slope; The region division module is configured to determine an overlapping region of the ascending and descending track InSAR data according to latitude and longitude ranges; The single-baseline forest height inversion module is configured to determine semi-empirical parameters based on a semi-empirical SINC model, with spaceborne sparse LiDAR as a control point and minimum root mean square error as a criterion, and then perform forest height inversion on non-overlapping regions pixel by pixel using the semi-empirical SINC model; The double-baseline forest height inversion module is configured to first calculate an elevation difference using the digital elevation model obtained by the ascending and descending track InSAR; then construct a double-baseline forest height inversion model considering both coherent phase and amplitude information; then determine optimal model parameters of the double-baseline forest height inversion model with spaceborne sparse LiDAR forest height as a control; and finally perform forest height inversion on overlapping regions pixel by pixel using the double-baseline forest height inversion model; The forest height output module is configured to mosaic the inversion results of the non-overlapping regions and the overlapping regions, and output a complete and continuous forest height inversion product.

Images