Forest height estimation method combining polarization rotation domain characteristics

Abstract

The invention relates to a forest height estimation method in combination with polarization rotation domain characteristics. The method comprises the following steps: acquiring a polarization scattering matrix under a specific imaging geometric condition by using a polarization radar; performing rotation processing on the polarization scattering matrix around the radar sight line, and expanding the polarization scattering matrix to a polarization rotation domain around the radar sight line; calculating oscillation characteristics and polarization correlation characteristics of the polarization rotation domain; calculating the relative importance of all the polarization rotation domain features to the forest height, sorting the relative importance of the polarization rotation domain features, selecting the polarization rotation domain features in the front of the sorting to perform modeling in sequence, and determining the polarization rotation domain features with higher contribution degree to the model according to error change to perform feature optimization; and constructing a forest height prediction model by taking the polarization rotation domain features obtained through optimization as modeling features, and realizing forest height prediction by using the forest height prediction model.

Description

Technical Field

[0001] The present application relates to the field of radar polarization information processing and application technology, and in particular to a forest height estimation method combining polarization rotation domain features. Background Art

[0002] Synthetic Aperture Radar (SAR) is unaffected by sunlight and weather, enabling all-day, all-weather surface observation. Compared to traditional SAR, polarimetric SAR can obtain complete polarization information of a target. By transmitting and receiving electromagnetic waves in different polarization states, it provides richer surface features, thereby enhancing target identification and classification capabilities. Forest height is a key indicator for assessing forest health. Accurately estimating forest height can improve the efficiency of forest resource monitoring and contribute to sustainable forest management and global climate change research.

[0003] The composition and structure of forests are influenced by forest species, age, and density, resulting in distinct scattering signatures in SAR images. Complex layered structures lead to multiple scattering and attenuation, further enhancing scattering diversity. Traditional polarization interpretation methods are generally not adapted to target scattering diversity, and therefore often suffer from interpretation ambiguity and ambiguity. Expanding the initial polarization information acquired by polarimetric radar to the polarization rotation domain around the radar's line of sight is expected to unlock the rich information contained in target scattering diversity, thereby improving the accuracy of forest height estimation and providing a new reference for forest resource monitoring and sustainable management. Therefore, developing a forest height estimation method that incorporates the polarization rotation domain has significant application value. Summary of the Invention

[0004] Based on this, it is necessary to provide a forest height estimation method that combines polarization rotation domain characteristics and is directly applied to the target polarization scattering matrix data obtained by the polarization radar system, which has application value in the fields of forest resource survey and environmental monitoring.

[0005] A forest height estimation method combining polarization rotation domain features, the method comprising: The polarization scattering matrix under specific imaging geometric conditions is obtained using polarimetric radar. The polarization scattering matrix is ​​rotated around the radar line of sight and expanded to the polarization rotation domain around the radar line of sight. Calculate the oscillation characteristics and polarization-related characteristics of the polarization rotation domain; calculate the relative importance of all polarization rotation domain features to forest height, and rank the relative importance of polarization rotation domain features. Select the polarization rotation domain features with the highest ranking to model in sequence. Determine the polarization rotation domain features that contribute more to the model based on the error change for feature optimization; A forest height prediction model is constructed based on the optimally obtained polarization rotation domain features as modeling features, and forest height prediction is achieved using the forest height prediction model.

[0006] The aforementioned forest height estimation method, which incorporates polarization rotation domain features, first uses polarimetric radar to obtain a polarization scattering matrix under specific imaging geometry. This matrix is ​​then rotated around the radar's line of sight and expanded into the polarization rotation domain. This changes the relative geometric relationship between the target and the radar's line of sight, and a sequence of polarization correlation values ​​is obtained over the rotation angles. By calculating the oscillation and polarization correlation features in the polarization rotation domain and parameterizing the polarization correlation value sequence in the rotation domain, this method fully describes the variation in the target's polarization correlation values ​​in the rotation domain and provides a detailed interpretation of the scattering mechanism underlying forest height variations. By detecting changes in the scattering response at different rotation angles, polarization rotation domain features reveal structural gradients within the vegetation canopy, enabling high-precision forest height estimation. The relative importance of these polarization rotation domain features to forest height is further calculated and ranked, and key features are selected to construct a forest height prediction model. This avoids interference from redundant features, improves model accuracy and efficiency, and enables high-precision forest height estimation. In forest resource surveys, accurate forest height data is crucial for assessing forest health and stock volume, improving monitoring efficiency and facilitating sustainable management. In environmental monitoring, accurate forest height information is crucial for studying global climate change and assessing carbon sequestration capacity. This application visualizes and parameterizes polarization-related features in the polarization rotation domain, extracting polarization rotation domain features. Through sensitivity analysis, the response of polarization rotation domain features to forest height is determined. This data can be directly applied to target polarization scattering matrix data obtained by polarimetric radar systems, demonstrating its application in forest resource surveys, environmental monitoring, and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 1 is a flow chart of a forest height estimation method combining polarization rotation domain features in one embodiment; Figure 2 is a schematic diagram of the relative importance of polarization characteristics in one embodiment; Figure 3 is an error variation diagram in one embodiment; Figure 3 (a) The error variation diagram of different polarization characteristics obtained by modeling typical polarization characteristics, Figure 3 (b) Polarization rotation domain feature modeling to obtain error change diagrams and Figure 3 (c) Error variation diagram of different polarization features obtained by joint feature modeling; Figure 4 A comparison diagram of forest height estimation results in another embodiment; Figure 4 (a) is the forest height prediction result diagram of typical polarization characteristics, Figure 4(b) is the polarization rotation domain feature forest height prediction result map and Figure 4 (c) The result of achieving high-precision forest height estimation by combining the model errors constructed with the two features. DETAILED DESCRIPTION

[0008] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0009] In one embodiment, Figure 1 As shown, a forest height estimation method combining polarization rotation domain features is provided, comprising the following steps: Step 102: Using a polarimetric radar, a polarimetric scattering matrix under specific imaging geometric conditions is obtained; the polarimetric scattering matrix is ​​rotated around the radar line of sight, and the polarimetric scattering matrix is ​​expanded to a polarimetric rotation domain around the radar line of sight.

[0010] In forest monitoring, radar waves in low-lying forests more easily penetrate the tree canopy, typically exhibiting a mixture of even-order and volume scattering. As forest height increases, the vertical complexity of the tree canopy enhances volume scattering, while even-order scattering is reduced due to limited radar penetration. By rotating the polarimetric scattering matrix, acquired under specific imaging geometry, around the radar line of sight, the relative geometric relationship between the target pose and the radar line of sight is altered. A sequence of polarimetric correlation values ​​is obtained by traversing each rotation angle in the rotation domain, effectively capturing the complex and diverse changes in forest scattering characteristics.

[0011] Step 104, calculate the oscillation characteristics and polarization-related characteristics of the polarization rotation domain; calculate the relative importance of all polarization rotation domain features to the forest height, and rank the relative importance of the polarization rotation domain features, select the polarization rotation domain features at the top of the ranking to model in turn, and determine the polarization rotation domain features that contribute more to the model according to the error change for feature optimization.

[0012] By calculating the oscillation characteristics and polarization-correlation characteristics of the polarization rotation domain and parameterizing the polarization-correlation value sequence in the rotation domain, the changing characteristics of the target polarization-correlation value in the rotation domain can be fully described, allowing for a detailed interpretation of the scattering mechanism of forest height variations. This detailed characterization can uncover feature information closely related to forest height, providing a basis for accurate estimation of forest height. The relative importance of all polarization rotation domain features to forest height is calculated and ranked, and the top features are selected for modeling in sequence. The features that contribute most to the model are then selected based on the change in error. This approach avoids interference from redundant features, allowing the constructed forest height prediction model to focus on key features, improve model accuracy and efficiency, and more accurately predict forest height.

[0013] Step 106: construct a forest height prediction model based on the optimally obtained polarization rotation domain features as modeling features, and use the forest height prediction model to achieve forest height prediction.

[0014] In this forest height estimation method, which incorporates polarization rotation domain features, polarimetric radar is first used to obtain a polarization scattering matrix under specific imaging geometry. This matrix is ​​then rotated around the radar's line of sight and expanded into the polarization rotation domain. This changes the relative geometric relationship between the target and the radar's line of sight, and a sequence of polarization correlation values ​​is obtained over the rotation angles. By calculating the oscillation and polarization correlation features in the polarization rotation domain and parameterizing the polarization correlation value sequence in the rotation domain, the polarization rotation domain can be fully described, allowing for a detailed interpretation of the scattering mechanism underlying forest height variations. By detecting changes in the scattering response at different rotation angles, the polarization rotation domain features reveal structural gradients within the vegetation canopy, enabling high-precision forest height estimation. The relative importance of these polarization rotation domain features to forest height is further calculated and ranked, and key features are selected to construct a forest height prediction model. This avoids interference from redundant features, improves model accuracy and efficiency, and enables high-precision forest height estimation. In forest resource surveys, accurate forest height data is crucial for assessing forest health and stock volume, improving monitoring efficiency and facilitating sustainable management. In environmental monitoring, accurate forest height information is crucial for studying global climate change and assessing carbon sequestration capacity. This application visualizes and parameterizes polarization-related features in the polarization rotation domain, extracting polarization rotation domain features. Through sensitivity analysis, the response of polarization rotation domain features to forest height is determined. This data can be directly applied to target polarization scattering matrix data obtained by polarimetric radar systems, demonstrating its application in forest resource surveys, environmental monitoring, and other fields.

[0015] In one embodiment, obtaining a polarization scattering matrix under specific imaging geometric conditions using a polarimetric radar includes: The polarization scattering matrix obtained by polarimetric radar under specific imaging geometric conditions is:

[0016] in, is the complex backscatter coefficient obtained under horizontal polarization transmission and horizontal polarization reception conditions; is the complex backscatter coefficient obtained under the conditions of vertical polarization transmission and horizontal polarization reception; is the complex backscatter coefficient obtained under the conditions of horizontal polarization transmission and vertical polarization reception; is the complex backscatter coefficient obtained under the conditions of vertical polarization transmission and vertical polarization reception.

[0017] In one embodiment, rotating the polarization scattering matrix around the radar line of sight includes: The polarization scattering matrix is ​​rotated around the radar line of sight, and the rotated polarization scattering matrix is ​​obtained as follows:

[0018] Among them, the rotation matrix , superscript is the matrix transpose processing, is the rotation angle in the rotation domain, .

[0019] In one embodiment, calculating the oscillation characteristics and polarization-dependent characteristics of the polarization rotation domain includes: Polarization coherence matrix Each element of can be uniformly represented by a sine function in the polarization rotation domain:

[0020] in, is the polarization oscillation amplitude, is the polarization oscillation center, is the angular frequency, is the initial angle, and these features constitute the polarization rotation domain oscillation feature set ; Therefore, the oscillation feature set includes 4 oscillation amplitude features, 2 oscillation center features and 5 initial angle features; the 4 oscillation amplitude features include , , and ; The two oscillation center features include and ; The 5 initial corner features include , , , and ;in, Represents the elements of the polarization coherence matrix, a is the row element, b is the column element, represents the real part, represents the imaginary part, represents an imaginary number, Indicates conjugation.

[0021] In one embodiment, the correlation between the two polarization channels of the polarization radar contains rich target information. The correlation characteristics of the two polarization channels are extended to the polarization rotation domain to obtain a two-dimensional polarization correlation pattern interpretation tool.

[0022] in, The value range is .

[0023] The polarization correlation characteristics in the rotation domain characterize the scattering characteristics of the polarization radar target in the rotation domain around the radar line of sight. 、 、 and Polarization-related features are extracted and parameterized.

[0024] In one embodiment, the polarization-related features include original polarization-related features, maximum polarization-related features, minimum polarization-related features, polarization-related degree features, polarization-related fluctuation features, polarization-related contrast features, polarization-related anti-entropy features, maximized rotation angle features, minimized rotation angle features, and polarization-related width features.

[0025] In one embodiment, the original polarization-related features are not subjected to any rotation processing. The polarization-related feature value at ; the maximum value of the polarization-related feature is the maximum value of the polarization-related feature in the polarization rotation domain; the minimum value of the polarization-related feature is the minimum value of the polarization-related feature in the polarization rotation domain; the polarization-related degree feature is the mean value of the polarization-related feature values ​​in the polarization rotation domain; the polarization-related fluctuation feature is the standard deviation of the polarization-related feature values ​​in the polarization rotation domain; the polarization-related contrast feature is the difference between the polarization-related maximum feature and the polarization-related minimum feature; the polarization-related anti-entropy feature is the ratio of the difference between the polarization-related maximum feature and the polarization-related minimum feature to the sum of the polarization-related maximum feature and the polarization-related minimum feature; the maximized rotation angle feature is the rotation angle corresponding to the polarization-related maximum feature in the main value interval of the polarization-related directional pattern; the minimized rotation angle feature is the rotation angle corresponding to the polarization-related minimum feature in the main value interval of the polarization-related directional pattern; the polarization-related width feature is the value of the polarization-related feature in the main value interval of the polarization-related directional pattern not less than The rotation angle range.

[0026] In a specific embodiment, the original polarization-related eigenvalue is defined as the value without any rotation processing. The polarization-related characteristic value at is:

[0027] Polarization-dependent characteristic maximum value: defined as the maximum value of the polarization-dependent characteristic in the polarization rotation domain, that is:

[0028] The polarization correlation maximum value feature represents the upper limit of the correlation value between two polarization channels that can be obtained by adjusting the polarization rotation domain angle.

[0029] Minimum value of polarization-related characteristics: defined as the minimum value of the polarization-related characteristics in the polarization rotation domain, that is:

[0030] The polarization correlation minimum feature represents the lower limit of the correlation value between two polarization channels that can be obtained by adjusting the polarization rotation domain angle.

[0031] Polarization correlation feature: It is defined as the mean of the polarization correlation feature values ​​in the polarization rotation domain, that is:

[0032] Polarization correlation is an indicator that measures the average decorrelation effect of the target in the polarization rotation domain. The larger the polarization correlation, the weaker the decorrelation phenomenon.

[0033] Polarization-dependent fluctuation characteristic: It is defined as the standard deviation of the polarization-dependent characteristic values ​​in the polarization rotation domain, that is:

[0034] The polarization-dependent fluctuation feature represents the degree of polarization-dependent fluctuation and can effectively measure the diversity of target scattering in the polarization rotation domain. For scatterers that are polarization-rotation invariant, the feature value will be reduced to zero.

[0035] Polarization-related contrast feature: defined as the difference between the polarization-related maximum feature and the polarization-related minimum feature, that is:

[0036] The polarization-dependent contrast feature reflects the absolute contrast of the polarization-dependent features in the polarization rotation domain.

[0037] Polarization-related anti-entropy characteristics: defined as the ratio of the difference between the polarization coherence maximum value characteristic and the polarization coherence minimum value characteristic to the sum of the polarization coherence maximum value characteristic and the polarization coherence minimum value characteristic, that is,

[0038] The polarization-dependent anti-entropy feature reflects the relative contrast of the polarization-dependent features in the polarization rotation domain. The polarization-dependent anti-entropy feature is a complement to the polarization-dependent contrast feature.

[0039] Maximized rotation angle feature: defined as the rotation angle corresponding to the polarization-related maximum value feature in the main value interval of the polarization-related pattern, that is:

[0040] The maximized rotation angle feature indicates a special state of the polarization rotation domain that can minimize the desired decorrelation effect between two given polarization channels.

[0041] Minimize the rotation angle feature: It is defined as the rotation angle corresponding to the polarization-related minimum value feature in the main value interval of the polarization-related pattern, that is:

[0042] The minimized rotation angle feature also indicates a special state of the polarization rotation domain that can maximize the desired decorrelation effect of two given polarization channels.

[0043] Polarization-related width feature: defined as the polarization-related feature value in the main value interval of the polarization-related direction pattern is not less than The rotation angle range is ,

[0044] The polarization correlation width characteristic reflects the sensitivity of target scattering to azimuth. The smaller its value is, the greater the target decorrelation and azimuth dependence will be.

[0045] In one embodiment, the relative importance of all polarization rotation domain features to forest height is calculated and ranked. The top-ranked features are then selected for modeling. The polarization features that contribute most to the model are then prioritized based on the error change. Finally, the prioritized polarization rotation domain features are used for subsequent forest height modeling.

[0046] The relative importance is characterized by the average gain, and the expression of the average gain is defined as:

[0047] in, represents the polarization rotation domain characteristics, Indicates in Samples in the feature The first derivative (gradient) on , Indicates in Samples in the feature The second derivative (gradient) on ; It is a regularization parameter used to control the size of node weights to prevent overfitting and is a complexity penalty term.

[0048] In one embodiment, a forest height prediction model is constructed based on the optimally obtained polarization rotation domain features as modeling features, including: The forest height prediction model is constructed based on the optimal polarization rotation domain features as the modeling features:

[0049]

[0050]

[0051] in, For the The modeling features of samples, for The true value of for The prediction results, For the The function of a decision tree, is the number of trees to be constructed, is the number of training samples; is the regularization term, is the number of leaf nodes, is the number of leaf node splits; and are the control coefficients to prevent overfitting.

[0052] In one embodiment, the predicted forest height value and the actual measured forest height are determined by the coefficient of determination. , root mean square error and mean absolute error Conduct model accuracy evaluation and use the evaluation results to guide the optimization of the forest height prediction model; The evaluation index is calculated as follows:

[0053]

[0054]

[0055] in, and are the model prediction values, is the mean of the measured values, is the sample size.

[0056] In a specific embodiment, typical polarization features such as amplitude features and polarization decomposition features are extracted using Radarsat-2 full polarization data and compared with the polarization rotation domain features in terms of relative importance. Figure 2 The polarization features ranked by relative importance are shown in Figure 1. Amplitude features include polarization total power and polarization channel backscatter coefficients, along with their texture features. Polarization decomposition features are derived from Cloude Pottier decomposition, Freeman Durden decomposition, and Yamaguchi decomposition. Amplitude features do not contain phase information and are less sensitive to forest height. Polarization decomposition features have clear physical meaning, but due to the diversity of target scattering and the complexity of forest hierarchies, model-based polarization target decomposition results often suffer from interpretation ambiguity and ambiguity, limiting the ability of polarization decomposition features to describe differences in forest height. In contrast, polarization rotation domain features reveal structural gradients in the forest canopy by detecting changes in scattering response at different rotation angles, more flexibly capturing the nonlinear variations in scattering mechanisms with forest height. Forests at different heights exhibit distinct polarization rotation domain characteristics, providing a theoretical basis for forest height estimation.

[0057] Figure 3 Is the error change diagram. Based on Figure 3 The relative importance of (a) typical polarization features, (b) polarization rotation domain features, and (c) combined features were ranked and modeled sequentially to determine the error variation associated with each polarization feature. The overall error of the forest height estimation model based on polarization rotation domain features was significantly lower than that of the typical polarization features, indicating that polarization rotation domain features contribute more significantly to forest height modeling. Combining these two types of features complements each other and further reduces model error.

[0058] Figure 4 This figure compares forest height estimation results. Polarimetric SAR data is Radarsat-2 fully polarimetric data, acquired on October 9, 2023, at an angle of incidence of 23.4°. Measured forest survey data was obtained in July 2023 at the Huangfengqiao Forest Farm in Hunan Province. It includes 93 measured forest height data points, ranging from 7.8 to 22.9 meters. Figure 4 (a) and (b) are the forest height prediction results of typical polarization features and polarization rotation domain features respectively. The model constructed by polarization rotation domain features has better fitting effect and lower prediction error than the former. In addition, Figure 4 (c) shows that the error of the model constructed by combining the two features is further reduced, achieving high-precision forest height estimation. The prediction results of this application are consistent with the actual forest height, verifying that the present invention has high accuracy in forest height estimation.

[0059] It should be understood that although Figure 1The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0060] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0061] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0062] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and such modifications and improvements are intended to fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A forest height estimation method combining polarization rotation domain features, characterized in that: The method comprises: A polarization scattering matrix under specific imaging geometric conditions is obtained using a polarization radar; the polarization scattering matrix is ​​rotated around the radar line of sight to extend the polarization scattering matrix to a polarization rotation domain around the radar line of sight; Calculate the oscillation characteristics and polarization-related characteristics of the polarization rotation domain; calculate the relative importance of all polarization rotation domain features to forest height, and rank the relative importance of polarization rotation domain features. Select the polarization rotation domain features with the highest ranking to model in sequence. Determine the polarization rotation domain features that contribute more to the model based on the error change for feature optimization; A forest height prediction model is constructed based on the optimally obtained polarization rotation domain features as modeling features, and forest height prediction is achieved using the forest height prediction model.

2. The method according to claim 1, characterized in that Polarimetric radar is used to obtain the polarimetric scattering matrix under specific imaging geometric conditions, including: The polarization scattering matrix obtained by polarimetric radar under specific imaging geometric conditions is: in, is the complex backscatter coefficient obtained under horizontal polarization transmission and horizontal polarization reception conditions; is the complex backscatter coefficient obtained under the conditions of vertical polarization transmission and horizontal polarization reception; is the complex backscatter coefficient obtained under the conditions of horizontal polarization transmission and vertical polarization reception; is the complex backscatter coefficient obtained under the conditions of vertical polarization transmission and vertical polarization reception.

3. The method according to claim 1, characterized in that Rotating the polarization scattering matrix around the radar line of sight includes: The polarization scattering matrix is ​​rotated around the radar line of sight to obtain a rotated polarization scattering matrix: Among them, the rotation matrix , superscript is the matrix transpose processing, is the rotation angle in the rotation domain, .

4. The method according to claim 1, wherein Compute oscillatory and polarization-dependent features of polarization-rotating domains, including: Polarization coherence matrix Each element of can be uniformly represented by a sine function in the polarization rotation domain: in, is the polarization oscillation amplitude, is the polarization oscillation center, is the angular frequency, is the initial angle, and these features constitute the polarization rotation domain oscillation feature set ; Therefore, the oscillation feature set includes 4 oscillation amplitude features, 2 oscillation center features and 5 initial angle features; the 4 oscillation amplitude features include , , and ; The two oscillation center features include and ; The five initial angle features include , , , and ;in, Represents the elements of the polarization coherence matrix, a is the row element, b is the column element, represents the real part, represents the imaginary part, represents an imaginary number, Indicates conjugation.

5. The method according to claim 4, characterized in that The method further comprises: Through four independent polarization-dependent patterns 、 、 and Polarization-related features are extracted and parameterized.

6. The method according to claim 5, characterized in that The polarization-related features include original polarization-related features, maximum polarization-related features, minimum polarization-related features, polarization-related degree features, polarization-related fluctuation features, polarization-related contrast features, polarization-related anti-entropy features, maximized rotation angle features, minimized rotation angle features, and polarization-related width features.

7. The method according to claim 6, characterized in that The original polarization-related features are those without any rotation processing. The polarization-related feature value at ; the maximum value of the polarization-related feature is the maximum value of the polarization-related feature in the polarization rotation domain; the minimum value of the polarization-related feature is the minimum value of the polarization-related feature in the polarization rotation domain; the polarization-related degree feature is the mean value of the polarization-related feature values ​​in the polarization rotation domain; the polarization-related fluctuation feature is the standard deviation of the polarization-related feature values ​​in the polarization rotation domain; the polarization-related contrast feature is the difference between the polarization-related maximum feature and the polarization-related minimum feature; the polarization-related anti-entropy feature is the ratio of the difference between the polarization-related maximum feature and the polarization-related minimum feature to the sum of the polarization-related maximum feature and the polarization-related minimum feature; the maximized rotation angle feature is the rotation angle corresponding to the polarization-related maximum feature in the main value interval of the polarization-related directional pattern; the minimized rotation angle feature is the rotation angle corresponding to the polarization-related minimum feature in the main value interval of the polarization-related directional pattern; the polarization-related width feature is the polarization-related feature value in the main value interval of the polarization-related directional pattern that is not less than The rotation angle range.

8. The method according to claim 1, characterized in that The relative importance is characterized by the average gain, and the expression of the average gain is defined as: in, represents the polarization rotation domain characteristics, Indicates in The characteristics of the samples in the polarization rotation domain The first derivative on , Indicates in The characteristics of the samples in the polarization rotation domain The second derivative on ; is the regularization parameter.

9. The method according to claim 1, wherein A forest height prediction model is constructed based on the optimally selected polarization rotation domain features as modeling features, including: The forest height prediction model is constructed based on the optimal polarization rotation domain features as the modeling features: in, For the The modeling features of samples, for The true value of for The prediction results, For the The function of a decision tree, is the number of trees to be constructed, is the number of training samples; is the regularization term, is the number of leaf nodes, is the number of leaf node splits; and are the control coefficients to prevent overfitting.

10. The method according to claim 1, characterized in that The method further comprises: The predicted forest height value and the actual measured forest height are compared by the determination coefficient , root mean square error and mean absolute error Conduct model accuracy evaluation and use the evaluation results to guide the optimization of the forest height prediction model; The evaluation index is calculated as follows: in, and are the model prediction values, is the mean of the measured values, is the sample size.

Images