A method for inverting the height of low vegetation based on adaptive polarization decomposition algorithm
By combining the adaptive polarization decomposition algorithm with Freeman decomposition and phase unwrapping, the problems of difficulty in separating the scattering mechanism and the difficulty in balancing model accuracy and robustness in the inversion of low vegetation height are solved, and high-precision inversion of low vegetation height is achieved.
Patent Information
- Application Number
- CN202610349557.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-20
- Publication Date
- 2026-07-03
AI Technical Summary
Existing technologies struggle to achieve high-precision and stable vegetation height inversion in low-vegetation scenarios, primarily due to the insufficient adaptability of the RVOG model, confusion of scattering mechanisms caused by the fixed polarization decomposition method, and the difficulty in balancing model accuracy and robustness.
An adaptive polarization decomposition algorithm combined with Freeman decomposition is adopted. The volume scattering phase is extracted and the complex coherence coefficient is calculated by the adaptive polarization decomposition algorithm. The even-order scattering phase is separated by the Freeman polarization cross-covariance matrix decomposition algorithm. The phase unwrapping process is used to eliminate phase periodic ambiguity. Finally, the vegetation height is calculated by the phase difference method.
It achieves high-precision and high-stability inversion of low vegetation height, improving inversion accuracy by 40%, reducing errors, and significantly improving the accuracy of inversion results.
Smart Images
Figure CN122330882A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of remote sensing technology, and in particular to a method for inverting the height of low vegetation based on the Adaptive Polarimetric Decomposition Algorithm (APD), which can be widely used in the quantitative extraction of low vegetation parameters in fields such as agricultural monitoring, ecological assessment, and climate change research. Background Technology
[0002] Low-lying vegetation height inversion is one of the key applications of remote sensing monitoring of the ecological environment. Obtaining accurate low-lying vegetation height data is crucial for analyzing vegetation growth status, estimating biomass and soil moisture, and other environmental factors, while also providing important data support for climate change research.
[0003] Polarimetric Interferometric Synthetic Aperture Radar (PolInSAR) is currently the mainstream technique for vegetation height inversion. This technology integrates polarimetric information extraction and interferometry capabilities, enabling simultaneous identification of ground object scattering mechanisms and inversion of the vertical distribution of scatterers. It has been widely applied in vegetation height estimation across various scenarios, including forests and crops. Its core inversion theory is based on the random volume over land (RVoG) model and its complex coherence model, extracting vegetation parameters through analytical interferometric complex coherence. To further optimize the separation effect of different scattering mechanisms, researchers have incorporated Freeman scattering techniques... Durden et al. introduced polarimetric target decomposition methods into the PolInSAR processing flow and developed an electromagnetic scattering model incorporating plant morphology. However, Freeman... The Durden method shares a common limitation: its volume scattering component is based on the assumption of an extremely fine cylinder, which is mainly designed to adapt to inversion scenarios of tall vegetation such as forests.
[0004] However, current technologies struggle to achieve high-precision and stable inversion of low-vegetation height, primarily due to the following issues: 1. The classic RVOG model lacks adaptability in low-vegetation scenarios: The RVOG model simplifies the vegetation layer as a homogeneous random medium, assuming that surface scattering and vegetation scattering can be separated in the coherence dimension. However, in low-vegetation scenarios, the canopy and ground surface are extremely close, and the two types of scattering are highly coupled in both spatial and signal dimensions, resulting in the superposition of multiple scattering mechanisms in the real part of complex coherence. The model struggles to effectively distinguish between direct surface scattering and canopy scattering. Ground-based secondary scattering also makes it impossible to accurately determine the vertical distribution of each scattering mechanism, thus affecting the accuracy and stability of the inversion.
[0005] 2. Fixed polarization decomposition methods cause confusion in scattering mechanisms: (e.g., using Freeman scattering methods) The fixed decomposition model, exemplified by Durden, is based on simplified physical assumptions and deviates significantly from actual scattering characteristics in low-lying vegetation scenarios. Because the polarization scattering characteristics of the canopy and the ground surface are similar, this type of method struggles to adaptively distinguish between the two. This results in a large amount of vegetation scattering energy being incorrectly assigned to the surface scattering component, thus contaminating the phase center estimation of pure vegetation scattering and introducing systematic errors into height inversion.
[0006] 3. Difficulty in balancing model accuracy and robustness: Existing technologies often face the problem of decreased model robustness when improving inversion accuracy. For example, while schemes combining electromagnetic scattering models attempt to more accurately characterize vegetation structure, their high algorithm complexity and poor adaptability to scene changes make it difficult to balance high accuracy and robustness. In particular, the assumption of uniformly simplifying volume scatterers into extremely fine cylinders does not match the real scattering structure of low-lying vegetation, further limiting the effectiveness and stability of the model in practical applications. Therefore, both types of schemes...
[0007] Therefore, developing a new algorithm that can model canopy scattering as a cloud-like layer composed of randomly oriented disc-shaped particles and has the ability to characterize volume scattering, so as to adaptively distinguish similar scattering mechanisms and accurately extract scattering phase and vertical distribution information, and correctly separate vegetation volume scattering components and ground surface scattering components, has become the key to breaking through the current bottleneck of low vegetation height inversion technology. Summary of the Invention
[0008] This application provides a method for inverting the height of low vegetation based on an adaptive polarization decomposition algorithm. This method solves the problems of existing models, such as the difficulty in separating the scattering mechanism in low vegetation scenarios, scattering aliasing caused by fixed polarization decomposition algorithms, and the difficulty in balancing model accuracy and robustness. It achieves high-precision and high-stability inversion of the height of low vegetation.
[0009] This application provides a method for inverting the height of low vegetation based on an adaptive polarization decomposition algorithm, including the following steps: S1: Data preprocessing: Acquire the main and secondary polarimetric interferometric SAR images, perform preprocessing, and construct the polarimetric interferometric cross-covariance matrix; S2: Adaptive Polarization Decomposition: An adaptive polarization decomposition algorithm is used to perform phase separation on the preprocessed polarization interference cross-covariance matrix to extract the volume scattering phase. And calculate the complex coherence coefficient corresponding to volume scattering. ; S3: Freeman Decomposition: The preprocessed polarimetric interferometric cross-covariance matrix is decomposed using the Freeman polarimetric cross-covariance matrix decomposition algorithm to separate even-order scattering components and extract the phase of even-order scattering. As a surface phase, obtain the complex coherence coefficient corresponding to even-order scattering. ; S4: Phase unwrapping: Based on the complex coherence coefficient corresponding to the volume scattering obtained in S2 The complex coherence coefficient corresponding to surface scattering obtained in S3 The volume scattering phases obtained in S2 are respectively and the surface scattering phase obtained in S3 Phase unwrapping is performed to eliminate phase periodic ambiguity and obtain continuous true phase values; S5: High-level inversion: Based on the volume scattering phase after unwrapping and even scattering phase The height hv of low vegetation was obtained by using the phase difference method.
[0010] The beneficial effects of the above embodiments are as follows: the inversion method combines adaptive polarization decomposition and Freeman decomposition to jointly extract the scattering phase, which breaks through the bottleneck of the traditional RVOG model in the low vegetation scene where the scattering mechanism coupling is difficult to separate. It integrates the scattering recognition capability of polarization decomposition and the vertical positioning capability of interferometry to achieve high-precision inversion of the height of low vegetation.
[0011] Based on the above embodiments, this application can be further improved as follows: In one embodiment of this application, the preprocessing in step S1 includes radiometric calibration, geometric correction, registration, and denoising. Technical benefits: Radiometric calibration ensures data radiometric consistency; geometric correction eliminates spatial distortion; high-precision registration reduces phase noise; and denoising suppresses speckle interference. These multiple processes work together to provide high-quality input for subsequent scattering decomposition and phase extraction, effectively ensuring the accuracy and reliability of the inversion results.
[0012] In one embodiment of this application, the polarization interference cross-covariance matrix in step S1 is constructed as follows: The preprocessed main and secondary SAR images are converted into Pauli basis scattering vectors respectively: ; ; The cross-covariance matrix C is calculated. int : ; In the above formula, <> represents the set average, and * represents the complex conjugate. This indicates the matrix transpose.
[0013] Technical effect: By unifying the multi-polarization channel signal representation through Pauli basis vectors, the constructed cross-covariance matrix fully contains the polarization information and interference phase correlation of the main and secondary images, providing a physically consistent input format for subsequent scattering decomposition and ensuring the accuracy of the decomposition process.
[0014] In one embodiment of this application, the adaptive polarization decomposition algorithm process in step S2 is as follows: The polarization interference cross-covariance matrix is decomposed into a weighted sum of the volume scattering component and the odd scattering component; G 3V For the volume scattering covariance matrix: ; ; ; Where A is the introduced adaptive weighting factor. When A>1, the volume scattering model is equivalent to a flat disc-shaped particle structure, which matches the scattering characteristics of low vegetation canopy. These are the values at each position in the polarization covariance matrix; f represents the odd-order scattering components. v For volume scattering components; Obtain the complex coherence coefficient F corresponding to canopy scattering V : ; via phase( ) Obtain the phase of volume scattering .
[0015] Technical effect: Breaking through the limitations of the traditional Freeman-Durden decomposition fixed dipole model assumption, the adaptive weighting factor dynamically adapts to the dish-shaped particle scattering characteristics of low-growing vegetation canopy, effectively distinguishing the polarization scattering response of the canopy and the ground surface.
[0016] In one embodiment of this application, the Freeman polarization cross-covariance matrix decomposition algorithm in step S3 is as follows: The polarization interference cross-covariance matrix is decomposed into a weighted sum of diffuse scattering, even-order scattering, and volume scattering; C D It is the even-order scattering covariance matrix: ; in, For even-order scattering components; , For the Bragg parameter, ; For volume scattering components; Even-order scattering complex coherence coefficient F D for: , , It is the Fresnel coefficient; via phase( ) Obtain the phase of even-order scattering .
[0017] Technical results: The even-order scattering components of Freeman decomposition are used to accurately extract the surface scattering phase, correcting the simulation bias of Bragg model for cross-polarized backscattering on rough surfaces. The ground phase estimation error is less than 0.1 radians, providing an accurate benchmark for subsequent phase difference calculations.
[0018] In one embodiment of this application, the formula for calculating the height hv of low vegetation in step S5 is: ; ; In the formula k z The effective wavenumber in the vertical direction; For radar frequency; The speed of light; The average angle of incidence of the master-slave interferometry image; The difference in incident angle between the master and slave images.
[0019] Technical benefits: The phase difference method is used to directly calculate vegetation height through phase difference. The calculation process is simple and efficient, and the inversion accuracy is improved by 40%. Attached Figure Description
[0020] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.
[0021] Figure 1 This is a flowchart of the low vegetation height inversion method based on the adaptive polarization decomposition algorithm in the embodiments of this application.
[0022] Figure 2 This is a schematic diagram of the rice height obtained by polarimetric interferometry data inversion in this embodiment of the application; the vertical axis in the figure represents the height in units (m), and the horizontal axis represents the radar center frequency, -h. v is the scattering curve of rice, and -h0 is the inversion height curve of rice. Detailed Implementation
[0023] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.
[0024] Example: like Figure 1 As shown, a method for inverting the height of low vegetation based on an adaptive polarization decomposition algorithm includes the following steps: S1: Data Preprocessing: Acquire the primary and secondary polarimetric interferometric SAR images and perform preprocessing to construct the polarimetric interferometric cross-covariance matrix; specifically: The polarimetric interferometric SAR image is preprocessed (preprocessing includes radiometric calibration, geometric correction, registration, and denoising). The preprocessed data is then used to obtain the S-matrix data: S1 is the main image matrix, and S2 is the auxiliary image matrix. The data contained in S1 and S2 are shown in the following formula: ; ; In the above formula, each element Sqpi(p, q = h, v) represents the target backscattering coefficient when transmitted in q-polarization and received in p-polarization, obtained by the polarimetric interferometric SAR system. i = 1, 2 represent the main and secondary SAR images obtained by the radar system from slightly different perspectives, and h, v represent the horizontal and vertical polarization modes, respectively.
[0025] Vectorization of this vector yields the Pauli basis scattering vector: ; ; The polarization interference cross-covariance matrix is then: ; In the above formula, <> represents the set average, and * represents the complex conjugate. This indicates the matrix transpose.
[0026] S2: Adaptive Polarization Decomposition: An adaptive polarization decomposition algorithm is used to perform phase separation on the polarization interference cross-covariance matrix and extract the volume scattering phase. The phase of the vegetation canopy is used to invert height, and the complex coherence coefficient corresponding to volume scattering is calculated. Specifically: The polarization interference mutual covariance moment is decomposed into a weighted sum of two fundamental scattering mechanisms: odd scattering and volume scattering. The process is as follows: ; Where A is the adaptive weighting factor introduced in this scheme. When A < 1, the particles exhibit a needle-like structure, which is more similar to the forest canopy; when A > 1, the particles form a flattened disc-like structure, which is more similar to the characteristics of low-lying vegetation canopies. This configuration is suitable for low-lying vegetation application scenarios. The following is derived from the formula: ; and These are all values at each position in the polarization covariance matrix; [ [The covariance matrix of odd-order scattering] ; in Let |α| represent the odd-order scattering component, and α represent the ground scattering term in the scattering mechanism. α is the complex coefficient of the contribution of odd-order scattering. When |α|>=1, it means the ground scattering term is even-order scattering; when |α|<=1, it means the ground scattering is surface scattering. In this scheme, α is derived as follows: ; ; [ [The volume scattering covariance matrix is:] ; ; ; In this scheme, the canopy scattering component mainly extracts volume scattering; no other scattering components are extracted. Therefore, the canopy scattering component is equivalent to the volume scattering component f. v The odd-order scattering component f derived in this step of the formula... G This was not used in this scheme.
[0027] Obtain the complex coherence coefficient corresponding to canopy scattering: Using MATLAB software phase( ) ,in It is the phase term of volume scattering.
[0028] S3: Freeman Decomposition: The preprocessed polarimetric interferometric cross-covariance matrix is decomposed using the Freeman polarimetric cross-covariance matrix decomposition algorithm to separate even-order scattering components and extract the phase of even-order scattering. As a surface phase, obtain the complex coherence coefficient corresponding to even-order scattering. Specifically: Using the Freeman polarization cross-covariance matrix decomposition algorithm, the preprocessed polarization interference covariance matrix is decomposed into a weighted sum of three basic scattering types: diffuse scattering, even-order scattering, and volume scattering. [Cint] = [CS] + [CV] + [CD]; [CV] is the volume scattering covariance matrix: ; This is the volume scattering component of this algorithm; [CS] is the diffuse scattering covariance matrix: ; This refers to the diffuse scattering component of this algorithm; [C D [] is the even-order scattering covariance matrix: ; This refers to the even-order scattering component of this algorithm; in, , The Bragg parameters represent the modeling of diffuse scattering in the Bragg model. This scheme introduces an adaptive weighting factor A to separate the "roughness contribution" (i.e., the contribution of the Bragg model) from the traditional volume scattering components and redistribute it to the diffuse scattering components. This corrects the roughness of the Bragg model and makes the extraction of even-order scattering components more accurate. .
[0029] The complex coherence coefficient corresponding to even scattering is obtained. ): , , These are the Fresnel coefficients, which are the reflection coefficients of the vertical vegetation trunk surface to horizontally polarized waves and vertically polarized waves, respectively. and The reflectance coefficients of the horizontal ground are respectively and This algorithm does not use and I won't explain further; these are all values that can be derived from the parameters. This algorithm uses... Where the coefficient that is The modulus, and yes The phase (of which the expression is a complex number) is calculated using the MATLAB software phase( ) , It is the phase of even-order scattering in the Freeman polarization cross-covariance matrix decomposition algorithm, which is the surface scattering phase.
[0030] S4: Phase unwrapping: Based on the complex coherence coefficient corresponding to the volume scattering obtained in S2 The complex coherence coefficient corresponding to surface scattering obtained in S3 The volume scattering phases obtained in S2 are respectively and the surface scattering phase obtained in S3 Phase unwrapping is performed to eliminate phase periodic ambiguity and obtain continuous true phase values; The core of resolving phase ambiguity in phase unwinding is: from the truncated... Within the range of entangled phases, the lost integer multiples of 2π are recovered, thus obtaining continuous and true physical phase values. Based on the idea of global optimization, a smooth surface is used to fit data with jumps. Although this sacrifices local details (smoothing out some true jumps), it ensures overall robustness. When the complex coherence coefficients are obtained, their phases are entangled within... Between. Adjust it to the principal value range by adding 2kπ (an integer). .
[0031] The phase term for even-order scattering and the phase term for volume scattering are expressed as Δ Corrected phase unwrapping: Δ corrected=Δ corrected+2kπ.
[0032] S5: High-level inversion: Based on the volume scattering phase after unwrapping and even scattering phase The height hv of low-lying vegetation was obtained using the phase difference method; specifically: The phase of the unwrapped volume scattering (canopy scattering) is determined using the phase difference method. Phase of even-order scattering (surface scattering) Substituting the values, we can obtain the following expression for vegetation height hv: ; in The volume scattering phase is extracted using an adaptive polarization decomposition algorithm. As a canopy phase, The even-order scattering is extracted using the Freeman polarization cross-covariance matrix factorization algorithm. As a surface phase; ; In the formula k z The effective wavenumber in the vertical direction is an indicator describing the sensitivity of a polarimetric interferometric SAR system to changes in altitude. Its physical meaning can be understood as the rate of change of the interferometric phase with respect to changes in target altitude. Indicates the radar frequency; The speed of light; This represents the average angle of incidence of the master-slave interferometry image; This represents the difference in incident angle between the master and slave images.
[0033] In this embodiment, the method is verified by acquiring multi-angle, controllable baseline PolInSAR data using a LAMP microwave anechoic chamber specifically designed for the measurement and simulation of microwave characteristics of land targets.
[0034] To simulate a natural paddy field environment, rice plants were transplanted into a microwave anechoic chamber, with precise control over their planting distribution and density. A japonica rice variety was used, grown at the milk-ripe stage, with an average plant height of 1.0 meter. To isolate backscattered signals and ensure they originated solely from the rice sample, microwave-absorbing materials were placed around the device. Furthermore, environmental parameters within the microwave anechoic chamber, such as temperature, humidity, and wind speed, were controlled to closely approximate the actual rice growing environment. The inversion results obtained using an improved polarimetric interferometric SAR joint inversion algorithm are shown below. Figure 2 As shown, the average height is 0.89 meters, very close to 1 meter, indicating that the inversion results accurately reflect the true vegetation height. Meanwhile, the root mean square error (RMSE) is only 0.18 meters, demonstrating the algorithm's extremely high accuracy and stability. This is because the Freeman polarization cross-covariance matrix decomposition algorithm obtains a surface phase closer to reality, and the adaptive polarization decomposition algorithm provides a more comprehensive and accurate description of the canopy structure, thus making the phase difference results close to the true vegetation height. Therefore, this low vegetation height inversion method not only improves the accuracy of the inversion results but also significantly reduces errors.
[0035] The technical solutions described above in this application aim to integrate the scattering mechanism identification capability of target decomposition with the vertical positioning capability of interferometry, thereby overcoming the limitations of existing methods and having at least the following technical effects or advantages: The APD algorithm optimizes the decomposition model by preserving the physical interpretability of the scattering mechanism in the Freeman-Durden decomposition and addressing its shortcomings. By introducing an adaptive weighting factor A, it effectively distinguishes between the polarization scattering responses of vegetation canopy and the land surface, reducing the aliasing effect of vegetation volume scattering in the surface scattering component. Simultaneously, it corrects the simulation bias of the Bragg model for cross-polarized backscattering from rough surfaces, separating the contribution of surface roughness in the cross-polarized signal from volume scattering, achieving a more accurate division of scattering components. Compared with traditional Freeman-Durden decomposition and methods based on polarization scattering characteristics, the APD algorithm significantly improves the inversion accuracy of low-lying vegetation height parameters, especially in estimating the height of low-lying vegetation.
[0036] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.
Claims
1. A method for inverting the height of low vegetation based on an adaptive polarization decomposition algorithm, characterized in that, Includes the following steps: S1: Data preprocessing: Acquire the main and secondary polarimetric interferometric SAR images, perform preprocessing, and construct the polarimetric interferometric cross-covariance matrix; S2: Adaptive Polarization Decomposition: An adaptive polarization decomposition algorithm is used to perform phase separation on the preprocessed polarization interference cross-covariance matrix to extract the volume scattering phase. And calculate the complex coherence coefficient corresponding to volume scattering. ; S3: Freeman Decomposition: The preprocessed polarimetric interferometric cross-covariance matrix is decomposed using the Freeman polarimetric cross-covariance matrix decomposition algorithm to separate even-order scattering components and extract the phase of even-order scattering. As a surface phase, obtain the complex coherence coefficient corresponding to even-order scattering. ; S4: Phase unwrapping: Based on the complex coherence coefficient corresponding to the volume scattering obtained in S2 The complex coherence coefficient corresponding to surface scattering obtained in S3 The volume scattering phases obtained in S2 are respectively and the surface scattering phase obtained in S3 Phase unwrapping is performed to eliminate phase periodic ambiguity and obtain continuous true phase values; S5: High-level inversion: Based on the volume scattering phase after unwrapping and even scattering phase The height hv of low vegetation was obtained by using the phase difference method.
2. The method for inverting the height of low vegetation according to claim 1, characterized in that: The preprocessing in step S1 includes radiometric calibration, geometric correction, registration, and noise reduction.
3. The method for inverting the height of low vegetation according to claim 1, characterized in that: The method for constructing the polarization interference cross-covariance matrix in step S1 is as follows: The preprocessed main and secondary SAR images are converted into Pauli basis scattering vectors respectively: ； ； The cross-covariance matrix C is computed int : 。 4. The method for inverting the height of low vegetation according to claim 3, characterized in that: The adaptive polarization decomposition algorithm process in step S2 is as follows: The polarization interference cross-covariance matrix is decomposed into a weighted sum of the volume scattering component and the odd scattering component; G 3V is the body scatter covariance matrix: ； ； ； Where A is the introduced adaptive weighting factor. When A>1, the volume scattering model is equivalent to a flat disc-shaped particle structure, which matches the scattering characteristics of low vegetation canopy. These are the values at each position in the polarization covariance matrix; f represents the odd-order scattering components. v For volume scattering components; Obtain the complex coherence coefficient F corresponding to canopy scattering V : ; via phase( ) Obtain the phase of volume scattering .
5. The method for inverting the height of low vegetation according to claim 4, characterized in that: The Freeman polarization cross-covariance matrix decomposition algorithm in step S3 is as follows: The polarization interference cross-covariance matrix is decomposed into a weighted sum of diffuse scattering, even-order scattering, and volume scattering; C D It is the even-order scattering covariance matrix: ； in, For even-order scattering components; , For the Bragg parameter, ; For volume scattering components; Even-order scattering complex coherence coefficient F D for: , , For Fresnel coefficients; via phase( ) Obtain the phase of even-order scattering .
6. The method for inverting the height of low vegetation according to claim 5, characterized in that: The formula for calculating the height hv of low vegetation in step S5 is as follows: ； ； In the formula k z The effective wavenumber in the vertical direction; For radar frequency; The speed of light; The average angle of incidence of the master-slave interferometry image; The difference in incident angle between the master and slave images.