Two-stage improved forest height estimation method

Abstract

The invention relates to the technical field of forest height estimation, and discloses a two-stage improved forest height estimation method, which calculates the interference coherence amplitude γ and the vertical wave number k based on InSAR data. z ; Based on the InSAR forest height inversion model SINC model, analyze the source of the systematic error of the SINC model and conduct simulation tests; On the basis of the error analysis of the SINC model, solve the systematic error correction coefficient of the SINC model to improve the SINC model, complete the first stage improvement and verify the effectiveness; Use the improved Kalman filter algorithm to assimilate the SINC model to reduce the non-systematic error of the SINC model, and realize the second stage improvement; Use the two-stage improved SINC model to estimate the forest height. Therefore, the present invention is based on InSAR data and the SINC model, and realizes the accurate estimation of the forest height of InSAR data based on the two-stage improved SINC model of systematic error correction and Kalman filtering. This method can effectively improve the estimation accuracy of the forest height of InSAR data.

Description

Technical Field

[0001] The present invention relates to the technical field of forest height estimation, and particularly to a two-stage improved forest height estimation method. Background Art

[0002] Forest height is an important basis for forest carbon accounting, forest quality and quantity evaluation, and forest succession monitoring. At present, globally, due to the influence of human activities and natural conditions such as fires, forests are changing every year. Therefore, it is crucial to quickly and efficiently monitor the status of forest resources. Synthetic Aperture Radar (SAR) technology has long been proven to be an effective method for estimating forest height. The biggest advantage of SAR is that it is not affected by weather, has full coverage, and has high spatial resolution, and has broad application prospects in large-scale and even global canopy height mapping.

[0003] Microwaves can penetrate the forest canopy to obtain information on the internal structure of the forest. The backscattering signals obtained by microwaves mainly come from the tree crowns, branches and trunks of trees, and the ground surface. Therefore, the forest height can be obtained through the backscattering model. Currently, the two-layer RVoG coherent scattering model is mainly used as the backscattering model. In RVoG, the forest scattering process is described as a forest volume scattering layer and an impenetrable ground layer. The volume scatterer layer is regarded as an isotropic homogeneous medium with a thickness of, and the extinction coefficient independent of polarization is used to describe the scattering and absorption losses of electromagnetic waves in the volume scattering layer. When the forest conditions meet the assumption that the extinction coefficient is 0 and the surface scattering is not considered, the relationship between the forest height hv and the "pure" volume coherence γv can be expressed by the SINC function, which is called the SINC model.

[0004] Due to temporal decorrelation and inappropriate interferometric baseline lengths, existing repeat-pass InSAR systems are difficult to obtain the observations required for retrieving the canopy height. The existence of this decorrelation will cause the forest canopy height to be overestimated. Research shows that short baselines will reduce the sensitivity to forest height, and long baselines will reduce coherence, thus limiting forest height inversion. Secondly, when the product of the forest height and the vertical wavenumber is less than the ambiguity height (k z ×hv < Hoa), the inversion result is more accurate. Too large or too small vertical beams will increase the interference of decorrelation. When Hoa is 2-4 times the forest height, the estimation result of the forest height is the best.

[0005] Therefore, a large number of studies have been conducted to improve the coherence model. The first one proposed a linear coherence model and used a linear function to replace the SINC function. Another strategy was to add a temporal decoherence compensation coefficient and a forest height compensation coefficient to improve the estimation accuracy. The second one improved the SINC model by adding a systematic effect to compensate for the volume coherence amplitude value. The third one improved the coherence model by adding horizontal motion and dielectric constant factors to the coherence method. The fourth one expanded the semi-empirical SINC model into a multi-layer model and used classification problems to invert the forest height. With the application of GEDI, some studies have used GEDI data to semi-empirically model the SINC model. However, the essence of these studies is to eliminate the systematic errors caused by decoherence and baseline discomfort to improve accuracy. In addition, non-systematic errors in the model will still affect the estimation results, and the causes of these non-systematic errors are often difficult to calculate.

[0006] In summary, for InSAR methods, the SINC model is the most convenient forest inversion method, which is suitable for large-scale, high-efficiency global forest height estimation. At present, different suggestions have been put forward for the improvement of the SINC model. The basic principle is to increase the error correction factor or change the SINC function structure. These methods can effectively reduce the systematic error and improve the estimation accuracy, but they do not eliminate the non-systematic error, and no unified conclusion has been reached. Summary of the invention

[0007] The main purpose of the present invention is to provide a two-stage improved forest height estimation method to solve the above technical problems.

[0008] To achieve the above object, the present invention provides a two-stage improved forest height estimation method, comprising the following steps:

[0009] S1: Obtain InSAR data and calculate the interferometric coherence amplitude γ and vertical wave number k based on the InSAR data z ;

[0010] S2: Based on the InSAR forest height inversion model SINC model, analyze the systematic error sources of the SINC model and conduct simulation tests;

[0011] S3: Based on the error analysis of the SINC model, solve the system error correction coefficient of the SINC model;

[0012] S4: Improve the SINC model using the solved system error correction coefficient, complete the first stage of improvement, and verify the effectiveness;

[0013] S5: Based on the systematic error correction of the SINC model, the SINC model is assimilated using the improved Kalman filter algorithm to reduce the non-systematic error of the SINC model and achieve the second stage improvement;

[0014] S6: Estimation of forest height using the two-stage improved SINC model.

[0015] Optionally, calculating the interference coherence amplitude γ specifically includes: calculating the interference coherence amplitude γ using primary and secondary images in the acquired InSAR data.

[0016] Optionally, calculate the vertical wave number k z Specifically, it includes: using the range slope, incident angle, slant range, vertical baseline and data acquisition mode information in the acquired InSAR data to calculate the vertical wave number k z .

[0017] Optional, expression for the error analysis result of the SINC model, specifically:

[0018] ;

[0019] in, represents the incoherence caused only by vegetation, where kz_c represents k z The correction coefficient of gamma_c represents | γ v |Correction factor.

[0020] Optionally, S3: based on the error analysis of the SINC model, solving the system error correction coefficient of the SINC model, specifically including: based on the error analysis of the SINC model, using part of the LiDAR height as the true value, and using the nonlinear least squares method to solve the system error correction coefficient of the SINC model.

[0021] Optionally, use the solved systematic number error correction coefficient to improve the SINC model, complete the first stage improvement, and verify the effectiveness steps, specifically including: use the solved systematic number error correction coefficient to improve the SINC model, complete the first stage improvement, verify the effectiveness in independent samples, and calculate the determination coefficient R 2 and root mean square error RMSE.

[0022] Optional, S5: Based on the systematic error correction of the SINC model, use the improved Kalman filter algorithm to assimilate the SINC model to reduce the non-systematic error of SINC, and realize the second stage improvement, specifically including: based on the systematic error correction of the SINC model, use the predicted value and observed value generated by the SINC model based on the improved Kalman filter algorithm to update the state estimate value to reduce the non-systematic error of SINC, and realize the second stage improvement.

[0023] Optionally, the improved Kalman filter algorithm adopts an extended Kalman filter.

[0024] Optionally, the extended Kalman filter is configured to linearize the nonlinear problem by performing a first-order Taylor expansion to update the state estimate using the predicted values ​​and observed values ​​generated by the SINC model.

[0025] Optionally, the improved Kalman filter algorithm uses a Bayesian optimization algorithm to optimize the hyperparameters of the noise covariance matrices Q and R.

[0026] The beneficial effect of the present invention is that a two-stage improved forest height estimation method is proposed, by acquiring InSAR data, calculating the interference coherence amplitude γ and the vertical wave number k based on the InSAR data z ; Based on the InSAR forest height inversion model SINC model, analyze the source of the systematic error of the SINC model and conduct simulation tests; Based on the error analysis of the SINC model, solve the systematic error correction coefficient of the SINC model; Use the solved systematic error correction coefficient to improve the SINC model, complete the first stage of improvement, and verify the effectiveness; Based on the systematic error correction of the SINC model, use the improved Kalman filter algorithm to assimilate the SINC model to reduce the non-systematic error of SINC, and achieve the second stage of improvement; Use the two-stage improved SINC model to estimate the forest height. Therefore, the present invention is based on InSAR data and the SINC model, and based on the two-stage improvement of the SINC model of systematic error correction and Kalman filtering to achieve accurate estimation of the forest height of InSAR data. This method can effectively improve the estimation accuracy of the forest height of InSAR data. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0028] Figure 1 It is a flowchart of an embodiment of the two-stage improved forest height estimation method of the present invention.

[0029] Figure 2 This is a geometric diagram of InSAR interferometric imaging in the present invention.

[0030] Figure 3 Schematic diagram of geometric changes of penetration and random motion in the present invention.

[0031] Figure 4 Schematic diagram of geometric changes of penetration and random motion in the present invention.

[0032] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] The embodiment of the present invention provides a two-stage improved forest height estimation method, referring to Figure 1 , Figure 1 It is a flowchart of an embodiment of the two-stage improved forest height estimation method of the present invention.

[0035] In this embodiment, the two-stage improved forest height estimation method includes the following steps:

[0036] S1: Obtain InSAR data and calculate the interferometric coherence amplitude γ and vertical wave number k based on the InSAR data z ;

[0037] S2: Based on the InSAR forest height inversion model SINC model, analyze the systematic error sources of the SINC model and conduct simulation tests;

[0038] S3: Based on the error analysis of the SINC model, solve the system error correction coefficient of the SINC model;

[0039] S4: Improve the SINC model using the solved system error correction coefficient, complete the first stage of improvement, and verify the effectiveness;

[0040] S5: Based on the systematic error correction of the SINC model, the SINC model is assimilated using the improved Kalman filter algorithm to reduce the non-systematic error of the SINC model and achieve the second stage improvement;

[0041] S6: Estimation of forest height using the two-stage improved SINC model.

[0042] In a preferred embodiment, calculating the interference coherence amplitude γ specifically includes: calculating the interference coherence amplitude γ using the primary and secondary images in the acquired InSAR data.

[0043] In a preferred embodiment, the vertical wave number k is calculated z Specifically, it includes: using the range slope, incident angle, slant range, vertical baseline and data acquisition mode information in the acquired InSAR data to calculate the vertical wave number k z.

[0044] In a preferred embodiment, the expression of the SINC model error analysis result is specifically:

[0045] ;

[0046] in, represents the incoherence caused only by vegetation, where kz_c represents k z The correction coefficient of gamma_c represents | γ v |Correction factor.

[0047] In a preferred embodiment, S3: based on the error analysis of the SINC model, solving the system error correction coefficient of the SINC model, specifically including: based on the error analysis of the SINC model, using part of the LiDAR height as the true value, and using the nonlinear least squares method to solve the system error correction coefficient of the SINC model.

[0048] In a preferred embodiment, the steps of improving the SINC model using the solved systematic number error correction coefficient, completing the first stage improvement, and verifying the effectiveness include: improving the SINC model using the solved systematic number error correction coefficient, completing the first stage improvement, verifying the effectiveness in independent samples, and calculating the determination coefficient R 2 and root mean square error RMSE.

[0049] In a preferred embodiment, S5: Based on the systematic error correction of the SINC model, an improved Kalman filter algorithm is used to assimilate the SINC model to reduce the non-systematic error of the SINC model, thereby realizing the second stage improvement, specifically including: Based on the systematic error correction of the SINC model, the predicted values ​​and observed values ​​generated by the SINC model are used based on the improved Kalman filter algorithm to update the state estimate value to reduce the non-systematic error of the SINC, thereby realizing the second stage improvement.

[0050] In a preferred embodiment, the improved Kalman filter algorithm adopts an extended Kalman filter.

[0051] In a preferred embodiment, the extended Kalman filter is configured to linearize the nonlinear problem by performing a first-order Taylor expansion to update the state estimate using the predicted values ​​and observed values ​​generated by the SINC model.

[0052] In a preferred embodiment, the improved Kalman filter algorithm uses a Bayesian optimization algorithm to optimize the hyperparameters of the noise covariance matrices Q and R.

[0053] In this embodiment, based on the basic form of the SINC model, a semi-empirical method is used to correct the systematic errors caused by factors such as decoherence, microwave penetration, and temporal decoherence, and a Kalman filter method is used to further assimilate the model and eliminate non-systematic noise, providing an effective reference method for the accurate estimation of global forest height. Specifically, based on InSAR data and the SINC model, the SINC model is improved in two stages based on systematic error correction and Kalman filtering to achieve accurate estimation of forest height of InSAR data. This method can effectively improve the accuracy of forest height estimation of InSAR data.

[0054] In order to explain the present application more clearly, a specific example of the two-stage improved forest height estimation method of the present invention is provided below, which specifically includes the following steps:

[0055] Step 1: Interference coherence amplitude γ and vertical wave number k z Calculation

[0056] InSAR interferometry coherence is as follows:

[0057] (1)

[0058] S1 and S2 represent the main and auxiliary images respectively.

[0059] In areas covered with vegetation, the interferometric coherence γ is affected by several factors: the volume decorrelation of the vegetation , time decorrelation , signal-to-noise ratio decorrelation , baseline decorrelation and other decoherence Composition, as shown in formula (2):

[0060] (2)

[0061] Vertical wave number k z The calculation of is as follows:

[0062] (3)

[0063] In the formula is the range slope, R is the slope distance, θ is the incident angle, is the vertical baseline, and n depends on the data acquisition mode.

[0064] Step 2: Error analysis of the InSAR forest height inversion model SINC model.

[0065] (1) SINC model:

[0066] The SINC model is simplified from the RVoG model. The general expression of the RVoG model can be expressed as formula (4):

[0067] (4)

[0068] In the formula is the effective volume amplitude ratio, e jφ0 =e jkzz0 Indicates the ground phase, represents the incoherence caused only by vegetation, ignoring the contribution of surface scattering It can be expressed as;

[0069] (5)

[0070] If the extinction coefficient is zero, then the "pure" volume scattering in the RVoG model is coherent It becomes the form of SINC function, which is the coherent amplitude inversion method (SINC model) (Equation 6)

[0071] (6)

[0072] The forest height can be estimated by formula (7):

[0073] (7)

[0074] (2) SINC model error analysis:

[0075] However, when measuring forest height using InSAR / PolInSAR, in addition to the components mentioned in equation (2), two additional phase contributions can be observed at the top of the tree canopy ( Figure 3 ). The first contribution is the phase center shift caused by the penetration of the forest canopy by the low-frequency synthetic aperture radar. The second factor is the phase shift caused by the random movement of the tree canopy between the acquisition of the primary and auxiliary images. These factors not only cause phase shifts, but also reduce coherence, which seriously affects the inversion of forest height.

[0076] In order to correct the error, based on formula (5), add Used to correct for coherence; increase and It is used to correct the phase shift caused by the random movement of the tree canopy and microwave penetration, so the corrected It can be expressed as:

[0077] (8)

[0078] In addition to the errors caused by temporal decoherence and penetration, the spatial baseline has a greater impact, which can be expressed as follows:

[0079] (9)

[0080] In the formula, Represents the vertical baseline correction factor.

[0081] Here, we first conduct simulation experiments to compare the effects of baseline, random motion, and penetration on k z The impact of Figure 4 It can be clearly seen that the change of the spatial baseline has an impact on k z The impact is far greater than random motion and penetration.

[0082] Therefore, formula (8) should be expressed as:

[0083] (10)

[0084] when When , formula (6) becomes:

[0085] (11)

[0086] The magnitude can be expressed as:

[0087] (12)

[0088] The forest height can be calculated by formula (13):

[0089] (13)

[0090] Compared with the ordinary SINC model, two unknown parameters are added: and , in the invention Denoted as gamma_c, it means | γ v The correction factor of | ranges between 0 and 1 and Expressed as kz_c, indicating k z The correction coefficient of is obtained to obtain the SINC model to be solved:

[0091] (14)

[0092] Step 3: Solve the systematic error coefficient of the SINC model.

[0093] A small amount of lidar crown heights are used as references and the unknown parameters gamma_c and kz_c are solved using nonlinear least squares method and incorporated into the model.

[0094] (15)

[0095] Step 4: First-stage system error improvement and verification.

[0096] Use the systematic error correction coefficient obtained in step 3 to improve the SINC model, complete the first stage of improvement, verify the effectiveness in independent samples, and calculate the determination coefficient R 2 and root mean square error RMSE.

[0097] Step 5: Two-stage non-systematic error correction based on improved Kalman filter

[0098] (1) Kalman filter algorithm and SINC model integration:

[0099] In order to further eliminate non-system noise, Kalman filtering is used to assimilate the SINC model. Kalman filtering is a recursive algorithm used to estimate the state of a dynamic system from noisy measurements. Its core steps include prediction and update. The implementation process of the SINC model based on Kalman filtering is as follows:

[0100] 1) Define the state and observation vectors:

[0101] The state vector (x k ): In the location prediction scenario, the state vector usually represents the hv value at different locations or time points. Assuming there are n locations, the state vector can be expressed as:

[0102] (16)

[0103] The observation vector ( ): The observation vector is obtained by using the SINC function according to the observed data γ and k z The calculated hv value is composed of. At each time step, the forest height hv is estimated using this function:

[0104] (17)

[0105] Where H is the observation matrix, v k is the observation noise.

[0106] 2) State transition model:

[0107] The state transition model describes how a system transitions from one state to the next:

[0108] (18)

[0109] Where A is the state transfer matrix, which can be the identity matrix or can be adjusted according to the specific situation, wk is the process noise.

[0110] 3) Prediction and update in Kalman filter:

[0111] First, you need to initialize the state estimate and covariance matrix of the Kalman filter and define the process noise and observation noise covariance matrices. At each time step, the following operations must be performed: predict the state estimate and state covariance.

[0112] (19)

[0113] In the formula, is the state covariance matrix at time step k, and Q is the process noise covariance matrix. Update the state estimate and state covariance using the new observations.

[0114] (20)

[0115] Where R is the observation noise covariance matrix, K k is the Kalman gain, y k is the measurement residual, S k is the residual covariance.

[0116] (2) Improved Kalman filter and SINC model integration:

[0117] In the assimilation phase, the Kalman filter updates its state estimates using the predicted values ​​and observed values ​​generated by the SINC model, gradually improving the accuracy of the estimates. The state estimates and covariances are updated using the Kalman gain. In the improved algorithm, the invention uses an extended Kalman filter (EKF) instead of a standard Kalman filter. This is a significant improvement because the EKF is applicable to nonlinear systems, while the standard Kalman filter is only applicable to linear systems.

[0118] 1) Extended Kalman filter:

[0119] The original Kalman filter assumes that the observations are linear functions of the state. The EKF linearizes the nonlinear problem by performing a first-order Taylor expansion, making it possible to apply the standard Kalman filter algorithm.

[0120] 2) Bayesian Optimization:

[0121] In the improved method, Bayesian optimization is also introduced to optimize the hyperparameters of the noise covariance matrices Q and R.

[0122] Therefore, this specific example proposes a two-stage improved forest height estimation method, which calculates the interference coherence amplitude γ and vertical wave number k based on InSAR data. z; Based on the InSAR forest height inversion model SINC model, analyze the source of the systematic error of the SINC model and conduct simulation tests; Based on the error analysis of the SINC model, solve the systematic error correction coefficient of the SINC model to improve the SINC model, complete the first stage improvement and verify the effectiveness; Based on the improved Kalman filter algorithm, assimilate the SINC model to reduce the non-systematic error of SINC, and realize the second stage improvement; Use the two-stage improved SINC model to estimate the forest height. Therefore, the present invention is based on InSAR data and the SINC model, and realizes the accurate estimation of the forest height of InSAR data based on the two-stage improved SINC model of systematic error correction and Kalman filtering. This method can effectively improve the estimation accuracy of the forest height of InSAR data.

[0123] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate rather than limit the technical solutions of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the present invention can still be modified or replaced by equivalents. Any modification or partial replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A two-stage improved forest height estimation method, characterized in that: The following steps are involved: S1: Obtain InSAR data and calculate the interferometric coherence amplitude γ and vertical wave number k based on the InSAR data z ; S2: Based on the InSAR forest height inversion model SINC model, analyze the source of the systematic error of the SINC model and conduct simulation tests; specifically include: S2.1 is based on the InSAR forest height inversion model SINC model, taking into account the phase center shift caused by the low-frequency synthetic aperture radar penetrating the forest canopy and the phase shift caused by the random movement of the canopy between the acquisition of the main image and the auxiliary image; on the basis of the InSAR forest height inversion model SINC model, the addition Used to correct for coherence; S2.2 Increase and It is used to correct the phase shift caused by random movement of the tree crown and microwave penetration to obtain the corrected The expression is: ； S3: Based on the error analysis of the SINC model, solve the system error correction coefficient of the SINC model; specifically include: S3.1 Based on the error analysis of the SINC model, a simulation experiment was conducted to compare the effects of baseline, random motion and penetration on k z The impact of The expression is: ； S3.2 When , becomes: ； S3.3 The magnitude is expressed as: ； S3.4 Based on the above optimization process, the optimized SINC model expression is obtained, which is: ； in, represents the incoherence caused only by vegetation, where kz_c represents k z The correction coefficient of gamma_c represents | γ v |Correction factor; S3.5 solves the system error correction coefficient of the SINC model, specifically including: using a small amount of lidar canopy height as a reference, using the nonlinear least squares method to solve the unknown parameters gamma_c and kz_c; S4: Use the solved system error correction coefficient to improve the SINC model, complete the first stage of improvement, and verify the effectiveness; S5: Based on the systematic error correction of the SINC model, the SINC model is assimilated using the improved Kalman filter algorithm to reduce the non-systematic error of the SINC model, thereby realizing the second stage improvement; specifically, based on the systematic error correction of the SINC model, the state estimation value is updated using the predicted value and observed value generated by the SINC model based on the improved Kalman filter algorithm to reduce the non-systematic error of the SINC model, thereby realizing the second stage improvement; Among them, the state estimation value is updated based on the prediction value and observation value generated by the SINC model based on the improved Kalman filter algorithm, which specifically includes: S5.1 Define the state and observation vectors: The state vector (x k ): In the location prediction scenario, the state vector represents the hv value at different locations or time points; assuming there are n locations, the state vector is expressed as: ； The observation vector ( ): The observation vector is obtained by using the SINC function according to the observed data γ and k z The calculated hv value consists of: At each time step, the forest height hv is estimated using this function: ； Where H is the observation matrix, v k is the observation noise; S5.2 State Transition Model: The state transition model describes how a system transitions from one state to the next: ； Where A is the state transfer matrix, w k is the process noise; S5.3 Prediction and Update in Kalman Filter: First, you need to initialize the state estimate and covariance matrix of the Kalman filter and define the process noise and observation noise covariance matrices; at each time step, the following operations must be performed: Predict the state estimate and state covariance: ； in, is the state covariance matrix at time step k, Q is the process noise covariance matrix, and the state estimate and state covariance are updated using new observations; ； Where R is the observation noise covariance matrix, K k is the Kalman gain, y k is the measurement residual, S k is the residual covariance; S6: Estimation of forest height using the two-stage improved SINC model.

2. The two-stage improved forest height estimation method according to claim 1, characterized in that: Calculating the interference coherence amplitude γ specifically includes: calculating the interference coherence amplitude γ using the main and auxiliary images in the acquired InSAR data.

3. The two-stage improved forest height estimation method according to claim 1, characterized in that: Calculate the vertical wave number k z Specifically, it includes: using the range slope, incident angle, slant range, vertical baseline and data acquisition mode information in the acquired InSAR data to calculate the vertical wave number k z .

4. The two-stage improved forest height estimation method according to claim 1, characterized in that: S3: Based on the error analysis of the SINC model, the system error correction coefficient of the SINC model is solved, specifically including: based on the error analysis of the SINC model, partial LiDAR height is used as the true value, and the nonlinear least squares method is used to solve the system error correction coefficient of the SINC model.

5. The two-stage improved forest height estimation method according to claim 4, characterized in that: Improve the SINC model using the solved systematic error correction coefficient, complete the first stage of improvement, and verify the effectiveness steps, including: Improve the SINC model using the solved systematic error correction coefficient, complete the first stage of improvement, verify the effectiveness in independent samples, and calculate the determination coefficient R 2 and root mean square error RMSE.

6. The two-stage improved forest height estimation method according to claim 1, characterized in that: The improved Kalman filter algorithm adopts the extended Kalman filter.

7. The two-stage improved forest height estimation method according to claim 6, characterized in that: The extended Kalman filter is configured to linearize a nonlinear problem by performing a first-order Taylor expansion to update a state estimate using predicted values ​​and observed values ​​generated by a SINC model.

8. The two-stage improved forest height estimation method according to claim 7, characterized in that: The improved Kalman filter algorithm uses the Bayesian optimization algorithm to optimize the hyperparameters of the noise covariance matrices Q and R.

Images