A multi-baseline forest height inversion method, system and storage medium for P-band heavy-track PolInSAR

Abstract

The present invention discloses a multi-baseline forest height inversion method and system for P-band heavy track PolInSAR, and a storage medium. First, SAR image data is acquired; SKP algorithm is used to decompose different scattering mechanisms of forests, and the influence of ground time incoherence and residual ground scattering is removed; then, the fuzzy interval of pure body coherence coefficient is calculated in combination with the set multi-baseline, and solution space optimization is carried out to remove impossible solution combinations; finally, multi-baseline forest height is inverted, and the coherence strength of pure body coherence coefficient is reduced by assuming body time incoherence, and forest height inversion is realized by using cost function. The method provided by the present invention cooperates with SKP algebraic synthesis method and vegetation coherent scattering model to form a multi-baseline PolInSAR forest height inversion method, focusing on removing the joint influence of the above three aspects of interference, realizing accurate inversion of forest height, and ensuring that the method has high robustness under different regions and time baselines, thereby providing technical support for large-scale forest height and AGB inversion mapping.

Description

Technical Field

[0001] The invention relates to the field of remote sensing inversion technology, and in particular to a multi-baseline forest height inversion method and system and a storage medium for P-band heavy-track PolInSAR. Background Art

[0002] Forest aboveground biomass is an important parameter for characterizing forest carbon storage. Accurate estimation of forest aboveground biomass is of great significance for in-depth research on terrestrial carbon storage, carbon cycle, climate change, etc. Forest height, as an important indicator of forest biomass, can be directly converted into biomass through the allometric equation. The upcoming P-band SAR satellite BIOMASS makes it possible to map forest height over a large area. It is urgent to develop a robust forest height inversion method for the P-band heavy-orbit PolInSAR.

[0003] PolInSAR technology introduces SAR polarization information (sensitive to the shape and orientation of the scatterer) into interferometric technology (sensitive to the vertical height position of the scatterer). By combining the vegetation coherent scattering model and the corresponding inversion algorithm, the inversion of forest height can be achieved. However, most of the current models and inversion algorithms are developed for the L-band. Directly applying them to the P-band heavy-track PolInSAR will face three problems:

[0004] First, the strong penetration of the P band will make the pure body coherence coefficient in the forest area unobservable (that is, there is residual coherence contribution in all polarization channels), which will bring a one-dimensional fuzzy interval (m) along the coherence line direction to the solution of the pure body coherence coefficient. min );

[0005] Second, the canopy temporal decoherence (γ VT ) will bring additional fuzzy intervals to the solution of the pure body coherence coefficient;

[0006] The third is ground time decoherence (γ GT ) will bring errors to the solution of the ground phase.

[0007] The above three problems will lead to the under-determination of the vegetation coherent scattering model, which will seriously affect the accuracy of forest height inversion. Although some methods have been developed to solve some of the above problems, for example, for L-band data, Lavalle et al. (2014) assumed that m min The influence of can be ignored, indicating that the three-baseline observation data can make the RMoG model solvable. Liao et al. (2019) used the difference in fuzzy intervals between multiple baselines to alleviate the γ VT and m minHowever, this method uses the fixed extinction coefficient method as an intermediate step, which limits its universality. Existing methods are limited by their own specific assumptions and it is difficult to completely eliminate all the above problems.

[0008] Therefore, how to develop a multi-baseline forest height inversion method without relying on prior knowledge or assumptions and effectively remove the combined influence of residual ground scattering and temporal incoherence is the key to improving the accuracy and universality of P-band heavy-track PolInSAR forest height inversion. Summary of the invention

[0009] In view of this, the object of the present invention is to provide a multi-baseline forest height inversion method for P-band heavy-track PolInSAR, which realizes a multi-baseline inversion method for forest height of P-band heavy-track PolInSAR data.

[0010] In order to achieve the above object, the present invention provides the following technical solutions:

[0011] The multi-baseline forest height inversion method for P-band heavy-track PolInSAR provided by the present invention comprises the following steps:

[0012] Step S1: Acquire SAR image data of vegetation coverage area;

[0013] Step S2: Use the SKP algorithm to decompose different scattering mechanisms of the forest and obtain the ground coherence matrix R g and the pure body coherence matrix R v , remove the ground time decoherence γ GT and residual ground scatter m min The impact of

[0014] Step S3: Calculate the coherence coefficient γ by combining the set multi-baseline G and the pure body coherence coefficient γ V ;

[0015] Step S4: Using the independent pure body coherence coefficients γ of multiple baselines V Fuzzy interval, optimize the solution space and remove impossible solution combinations;

[0016] Step S5: Multi-baseline forest height inversion, reducing the pure body coherence coefficient γ by assuming body temporal decoherence V The coherence strength of the forest is then inverted using the cost function.

[0017] Furthermore, the use of the cost function in step S5 to achieve forest height inversion is specifically performed according to the following formula:

[0018]

[0019] Among them, N sIndicates the number of baselines used in the multi-baseline data. is the ground phase, γ V0 is the pure body coherence coefficient simulated by RVoG model, h v is the forest height, σ is the extinction coefficient, k z is the vertical beam;

[0020] F represents the cost function; N s represents the number of baselines in the multi-baseline data; j represents the selected baseline number; γ V-j represents the pure body coherent scattering coefficient of the jth baseline; i represents the imaginary unit.

[0021] Further, the SKP algorithm in step S2 is performed according to the following formula:

[0022]

[0023] Where E represents the covariance matrix solution; H represents the conjugate solution; y represents the SAR complex observation data, K is the total number of scattering mechanisms, C k and R k denote the polarization matrix and coherence matrix of the kth scattering mechanism, respectively.

[0024] Furthermore, the ground coherence coefficient γ in step S3 is G and the pure body coherence coefficient γ V Calculate according to the following formulas:

[0025] γ G =R g (B1,B2);

[0026] γ V =R v (B1,B2);

[0027] Among them, B1 represents the first baseline of the interference pair; B2 represents the second baseline of the interference pair;

[0028] Furthermore, each spatial baseline γ in step S4 V The fuzzy interval γ Amb The calculation is carried out according to the following formula:

[0029]

[0030] Where abs represents the absolute value, imag is the imaginary part, real is the real part, angle is the phase (-π-π), a' and b' are the parameters of the coherence line equation (y = a'x + b');

[0031] γ Amb represents the fuzzy interval; γ m-mirepresents the observable complex coherence coefficient with the minimum ground amplitude ratio;

[0032] Further, the method further comprises the following steps:

[0033] For solutions outside the fuzzy interval, a maximum value will be added to the cost function to exclude them from participating in the forest height inversion.

[0034] Furthermore, in the multi-baseline forest height inversion calculation in step S5, the forest height inversion is achieved using the following cost function:

[0035]

[0036] Among them, N s Indicates the number of baselines used in the multi-baseline data. is the ground phase, γ V0 is the pure body coherence coefficient simulated by RVoG model, h v is the forest height, σ is the extinction coefficient, k z is the vertical beam;

[0037] F represents the cost function; N s represents the number of baselines in the multi-baseline data; j represents the selected baseline number; γ V-j represents the pure body coherent scattering coefficient of the jth baseline; i represents the imaginary unit.

[0038] The multi-baseline forest height inversion system for P-band heavy-track PolInSAR provided by the present invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any one of claims 1 to 7 when executing the program.

[0039] The storage medium provided by the present invention stores a computer program on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the external storage medium detection method according to any one of claims 1 to 7 are implemented.

[0040] The beneficial effects of the present invention are:

[0041] The present invention provides a multi-baseline forest height inversion method for P-band heavy-track PolInSAR.

[0042] This method cooperates with the SKP algebraic synthesis method and the vegetation coherent scattering model to form a multi-baseline PolInSAR forest height inversion method, which focuses on removing the combined influence of the above three interferences, realizes the accurate inversion of forest height, and ensures that the method has high robustness under different regional and time baselines, thereby providing technical support for large-scale forest height and AGB inversion mapping.

[0043] This method makes full use of the response differences of multi-baseline observations to fuzzy intervals and develops a multidimensional fuzzy interval removal method and a solution space optimization method, so as to effectively reduce the impact of the three major interference factors of ground residual scattering, volume time incoherence, and ground time incoherence on the P-band re-orbit PolInSAR, thereby achieving the purpose of improving the accuracy and robustness of forest height inversion.

[0044] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to make the purpose, technical solution and beneficial effects of the present invention clearer, the present invention provides the following drawings for illustration:

[0046] Figure 1 The accuracy verification results of the single baseline inversion algorithms (FHIM1 and FHIM2) based on TropiSAR 2009 data are presented.

[0047] Figure 2 The accuracy verification results of the multi-baseline inversion algorithm (FHIM3 and the proposed method FHIM4) based on TropiSAR 2009 data are shown in Figure 2.

[0048] Figure 3 The spatial distribution results of the multi-baseline inversion algorithm (FHIM3 and the method proposed in this invention FHIM4) based on BioSAR 2007 data. DETAILED DESCRIPTION

[0049] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0050] Example 1

[0051] like Figure 1 As shown in the figure, the multi-baseline forest height inversion method for P-band heavy-track PolInSAR provided in this embodiment is intended to solve the three major interferences (γ VT , γ GT and m min ) problem, namely, the canopy temporal decoherence (γ VT ), ground time decoherence (γ GT ) and residual ground scatter (m min) interference, the specific steps of the multi-baseline forest height inversion method provided in this embodiment are as follows:

[0052] Step 1: SKP decomposition. The SKP algorithm is used to decompose different scattering mechanisms of the forest and obtain the pure body coherence matrix (R v ) and the ground coherence matrix (R g ), removing the ground time decoherence γ GT and residual ground scatter m min impact.

[0053]

[0054] Where E represents the covariance matrix solution; H represents the conjugate solution; y represents the SAR complex observation data, K is the total number of scattering mechanisms, C k and R k denote the polarization matrix and coherence matrix of the kth scattering mechanism, respectively.

[0055] Given the coherent baselines B1 and B2, the ground coherence coefficient (γ G ) and the pure body coherence coefficient (γ V ):

[0056] γ G =R g (B1,B2) (2)

[0057] γ V =R v (B1,B2) (3)

[0058] Step 2: Based on the pure body coherence coefficient γ V The solution space optimization method of the fuzzy interval, due to the pure body coherence coefficient γ V The solution of will be affected by SKP decomposition error and other errors, thus reducing the accuracy of forest height inversion. To this end, the independent pure body coherence coefficient γ of multiple baselines is used. V The fuzzy interval is used to optimize the solution space and remove impossible solution combinations to improve the accuracy and robustness of forest height inversion. Each spatial baseline γ is calculated independently. V The fuzzy interval (γ Amb ), which is expressed as follows:

[0059]

[0060] Where abs represents the absolute value, imag is the imaginary part, real is the real part, angle is the phase (-π-π), a' and b' are the parameters of the coherence line equation (y = a'x + b');

[0061] γ Amb represents the fuzzy interval; γm-m represents the observable complex coherence coefficient with the minimum ground amplitude ratio;

[0062] For solutions outside the fuzzy interval, a maximum value (such as 100) will be added to the cost function to exclude them from participating in the forest height inversion.

[0063] Step 3: Multi-baseline forest height inversion, by assuming that volume temporal decoherence only reduces the pure volume coherence coefficient (γ V ) and the forest height inversion is achieved using the following cost function:

[0064]

[0065] Among them, N s Indicates the number of baselines used in the multi-baseline data. is the ground phase, γ V0 is the pure body coherence coefficient simulated by RVoG model, h v is the forest height, σ is the extinction coefficient, k z is the vertical beam;

[0066] F represents the cost function; N s represents the number of baselines in the multi-baseline data; j represents the selected baseline number; γ V-j represents the pure body coherent scattering coefficient of the jth baseline; i represents the imaginary unit.

[0067] This equation is used to find γ V0 Make it consistent with γ V The cost function does not need to introduce volume temporal incoherence into the lookup table simulation, which can significantly improve the inversion efficiency.

[0068] Example 2

[0069] The accuracy and robustness of the multi-baseline forest height inversion method provided in this embodiment are verified below in combination with specific embodiments and drawings in the specification, and compared and analyzed with the traditional three-stage inversion algorithm, the fixed extinction coefficient method, and the published multi-baseline inversion algorithm.

[0070] (1) Dataset and Method Introduction

[0071] Dataset: Two datasets, TropiSAR 2009 and BioSAR 2007, are used, including airborne P-band re-orbit multi-baseline PolInSAR data and LiDAR CHM data that can be used to verify the accuracy of forest height inversion. For TropiSAR 2009, three spatial baselines are used to carry out forest height inversion, and their vertical beams are approximately 0.05, 0.10 and 0.15 rad / m. All data are acquired on the same day, that is, the time baseline is 0 day. The detailed information of the data is shown in Table 1. For BioSAR 2007, three spatial baselines are also used to carry out multi-baseline forest height inversion, and their vertical beams are approximately 0.07, 0.13 and 0.20 rad / m. In order to further verify the robustness of the method provided in this embodiment to temporal decoherence, observation data of three different time baselines in BioSAR 2007 are selected to carry out multi-baseline forest height inversion and verification, and the time baselines are 0 day, 30 day and 56 day respectively. The detailed information of the data is shown in Table 2.

[0072] Table 1. Multi-baseline PolInSAR data information selected based on TropiSAR 2009

[0073]

[0074] Table 2. Multi-baseline PolInSAR data information selected based on BioSAR 2007

[0075]

[0076] In the table, MB1 indicates the number of the multi-baseline dataset in the first study area; Tropi0206, 07biosar0406, etc. indicate the numbers of the adopted SLC images; site2-MB1 indicates the number of the multi-baseline dataset in the second study area;

[0077] In addition to the method provided in this embodiment, three published forest height inversion methods are selected for comparative analysis.

[0078] FHIM1: The traditional three-stage inversion algorithm proposed by Cloude et al. (2003). This method assumes that the complex coherence coefficient with the minimum ground amplitude ratio is the pure body coherence coefficient, and assumes that the ground phase is the intersection of the coherence line and the unit circle. This method is also affected by the residual ground scattering (m min ), ground time decoherence (γ GT ) and canopy temporal decoherence (γ VT ) interference.

[0079] FHIM2: Fixed extinction coefficient method. This method assumes that the intersection of the coherence line and the RVoG solution space after the fixed extinction coefficient is the pure body coherence coefficient, and assumes that the ground phase is the intersection of the coherence line and the unit circle. This method can remove m min , but will be affected by γ GT and γ VT impact.

[0080] FHIM3: A multi-baseline inversion method proposed by Liao et al. (2019). This method uses the overlapping area between the multi-baseline fuzzy intervals to solve and combines the fixed extinction coefficient method to optimize the forest inversion results, which can alleviate the m min and γ VT However, this method will be affected by γ GT At the same time, the fixed extinction coefficient process it adopts will also affect the robustness of forest height inversion.

[0081] FHIM4: The multi-baseline method provided in this embodiment can theoretically completely remove all ground time-decoherent γ GT , canopy temporal decoherence γ VT and residual ground scatter m min The impact of three major interferences.

[0082] (2) Verification of forest height inversion accuracy

[0083] Figure 1 The accuracy verification of two single baseline inversion methods, FHIM1 and FHIM2, is demonstrated. The traditional three-stage inversion algorithm FHIM1 has serious overestimation errors in all three spatial baselines, and the overestimation error shows a trend of gradually decreasing with the increase of spatial baselines. When FHIM2 is used, the overestimation error in the forest area with higher altitude is well reduced, but the overestimation error in the shorter forest area is aggravated. Overall, FHIM2 has improved accuracy compared with FHIM1, but both single baseline inversion methods have large inversion errors. The best accuracy is the FHIM2 inversion result under the BL3 baseline, R 2 =0.45, RMSE=4.7m, ubRMSE=4.7m.

[0084] Figure 2 The accuracy verification of the two multi-baseline inversion methods, FHIM3 and FHIM4, is demonstrated. Compared with FHIM2, FHIM3 further removes the influence of the temporal decoherence of the forest volume, and its inversion accuracy is further improved. 2 =0.69, RMSE = 4.8m, ubRMSE = 3.7m. However, this method is still affected by ground time decoherence, and its robustness will be reduced due to the fixed extinction coefficient. When the FHIM4 inversion method provided in this embodiment is adopted, R 2The inversion results are more consistent with LiDAR CHM in terms of spatial distribution, especially for the shorter forest areas.

[0085] (3) Verification of the robustness of forest height inversion

[0086] Figure 3 The robustness comparison and validation of forest height inversion using the BioSAR 2007 dataset is presented. Figure 3 (a), (b) and (c) respectively represent the comparison of inversion results under different time baseline lengths. The results show that a longer time baseline will bring greater temporal incoherence, resulting in more serious overestimation errors in forest height inversion, and this effect is particularly evident for shorter forests. Compared with the FHIM3 method, which also uses multi-baseline data inversion, the method provided in this embodiment is obviously less affected by the time baseline. The average height difference between the 30-day inversion result and the 0-day result of this method is 1.2m, while for FHIM3, the average height difference increases to 2.8m. For 0 days and 56 days, the height difference of the inversion results of this method is 3.1m, while that of FHIM3 is 6.4m. This shows that even when the time baseline grows to nearly 2 months, this method can still effectively alleviate the impact of temporal incoherence and obtain relatively stable forest height inversion results. The revisit period of the future BIOMASS satellite design is less than 45 days, so the high robustness of this method has the potential to realize large-scale forest height mapping based on BIOMASS satellite data.

[0087] This method comprehensively removes the multi-dimensional fuzzy interval problem caused by the above errors, effectively improves the accuracy and universality of forest height inversion, and uses airborne P-band re-orbit PolInSAR in different study areas as an example to illustrate the method. Among them, the SKP decomposition method is introduced into the RVoG model to remove the γ GT and m min The fuzzy interval caused by the multi-baseline inversion method is further developed to remove the γ VT The fuzzy interval caused by this can achieve accurate inversion of forest height and provide technical support for large-scale forest AGB estimation.

[0088] The above-described embodiments are only preferred embodiments for fully illustrating the present invention, and the protection scope of the present invention is not limited thereto. Equivalent substitutions or changes made by those skilled in the art based on the present invention are within the protection scope of the present invention. The protection scope of the present invention shall be subject to the claims.

Claims

1. A multi-baseline forest height inversion method for P-band heavy-track PolInSAR, characterized by: The following steps are involved: Step S1: Acquire SAR image data of vegetation coverage area; Step S2: Use the SKP algorithm to decompose different scattering mechanisms of the forest and obtain the ground coherence matrix R g and the pure body coherence matrix R v ; Step S3: Calculate the coherence coefficient γ by combining the set multi-baseline G and the pure body coherence coefficient γ V ; Step S4: Using the independent pure body coherence coefficients γ of multiple baselines V Fuzzy intervals, find their overlapping intervals to optimize the solution space and remove impossible solution combinations; Step S5: Multi-baseline forest height inversion, reducing the pure body coherence coefficient γ by assuming body temporal decoherence V The coherence strength of the forest is inverted using the cost function; The use of the cost function in step S5 to invert the forest height is specifically performed according to the following formula: Among them, N s Indicates the number of baselines used in the multi-baseline data. is the ground phase, γ V0 is the pure body coherence coefficient simulated by RVoG model, h v is the forest height, σ is the extinction coefficient, k z is the vertical beam; F represents the cost function; j represents the selected baseline number; γV ―j represents the pure body coherent scattering coefficient of the jth baseline; i represents the imaginary unit; angle is the phase (-π-π).

2. The multi-baseline forest height inversion method for P-band heavy-track PolInSAR as claimed in claim 1, characterized in that: The SKP algorithm in step S2 is performed according to the following formula: Where E represents the covariance matrix solution; H represents the conjugate solution; y represents the SAR complex observation data, K is the total number of scattering mechanisms, C k and R k denote the polarization matrix and coherence matrix of the kth scattering mechanism respectively.

3. The multi-baseline forest height inversion method for P-band heavy-track PolInSAR as claimed in claim 1, characterized in that: The ground coherence coefficient γ in step S3 G and the pure body coherence coefficient γ V Calculate according to the following formulas: γ G =R g (B1,B2); γ V =R v (B1,B2); Among them, B1 represents the first baseline of the interference pair; B2 represents the second baseline of the interference pair.

4. The multi-baseline forest height inversion method for P-band heavy-track PolInSAR as claimed in claim 1, characterized in that: Each spatial baseline γ in step S4 V The fuzzy interval γ Amb The calculation is carried out according to the following formula: Among them, abs represents the absolute value, imag is the imaginary part, and real is the real part. a' and b' are parameters of the expression equation of the coherence line, and the equation is y=a'x+b'; γ Amb represents the fuzzy interval; γ m―min represents the complex coherence coefficient with the minimum observable ground amplitude ratio.

5. The multi-baseline forest height inversion method for P-band heavy-track PolInSAR as claimed in claim 4, characterized in that: The following steps are also included: The solutions outside the fuzzy interval are removed so that they do not participate in the forest height inversion.

6. A multi-baseline forest height inversion system for P-band heavy track PolInSAR, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 5 is implemented.

7. A storage medium, characterized in that The storage medium stores a computer program, which, when executed by a processor, implements the steps of the multi-baseline forest height inversion method for P-band heavy-track PolInSAR as described in any one of claims 1 to 5.

Images