CN-121978711-A - Interference chromatography SAR forest canopy height estimation optimization method and system

CN121978711ACN 121978711 ACN121978711 ACN 121978711ACN-121978711-A

Abstract

The invention provides an interference chromatography SAR forest canopy height estimation optimization method and system, which belong to the technical field of forest canopy height estimation, wherein the first stage carries out envelope curve fitting on relative reflectivity signals of a forest vertical structure to obtain an initial envelope curve of forest canopy and the ground surface, and determines a corresponding relative reflectivity loss threshold K by referring to LiDAR height, the second stage screens noise interference at the top and the bottom of a forest section canopy by using a signal continuity threshold to obtain a more reasonable envelope curve, and the third stage carries out classification on forest canopy density based on K value optimization of canopy density segmentation, different levels are given to K values of different forest canopy and the ground surface to carry out relative reflectivity envelope curve extraction, and finally more accurate forest height is obtained according to subtraction of height indexes corresponding to envelope curves of the canopy and the ground surface. According to the method, the three-stage error optimization method is adopted, so that the forest canopy height estimation accuracy is effectively improved.

Inventors

LUO HONGBIN
ZHAO XUN
Luo Guangfei
YUE CAIRONG
DUAN YUNFANG
YU QIONGFEN
LI CHURUI

Assignees

西南林业大学

Dates

Publication Date: 20260505
Application Date: 20260408

Claims (10)

1. An interference chromatography SAR forest canopy height estimation optimization method is characterized by comprising the following steps: Reading L-band UAVSAR multi-baseline data, constructing a multi-baseline InSAR covariance matrix based on a ground phase compensation result, and estimating and reconstructing the relative reflectivity of a forest vertical section by adopting a Capon frequency spectrum; performing envelope curve fitting on the relative reflectivity signals of the forest vertical section, and determining an initial envelope curve of a forest canopy, an initial envelope curve of the earth surface and a signal loss threshold K by combining LiDAR height reference data; Noise screening based on vertical signal continuity is carried out on the relative reflectivity signals of the forest vertical section, noise interference at the top of a canopy and noise interference at the bottom of the earth surface are removed, and the denoised forest vertical structure reflectivity section and a more reasonable envelope curve are obtained; Pauli decomposition is carried out on the full-polarization SAR data to obtain a canopy density index, and different forest canopy K values and ground surface K values are given to different levels of pixels according to canopy density grading results; and re-executing envelope extraction on the relative reflectivity signals, and subtracting the height indexes corresponding to the canopy envelope and the surface envelope to obtain the forest canopy height.
2. The method for optimizing the height estimation of a forest canopy of an interferometric SAR of claim 1, wherein reconstructing the relative reflectivity of the forest vertical section using Capon spectral estimation comprises: performing preprocessing on the L-band UAVSAR multi-baseline data; A multi-baseline RVoG three-stage method is called to estimate ground phases, and residual phase compensation is carried out on multi-baseline observation data according to the ground phases; calculating a multi-baseline InSAR covariance matrix R according to the compensated multi-baseline observation data; Inputting the multi-baseline InSAR covariance matrix R into a Capon beam forming power estimator, and calculating the relative reflectivity in the vertical direction of a vegetation area, wherein the expression is specifically as follows: ; In the formula, The method comprises the steps of obtaining a back scattering power vertical distribution function estimated by a Capon algorithm, wherein z represents a height variable in a vertical direction, and a (z) represents a guide vector with the height of z; r represents a multi-baseline InSAR data covariance matrix; Representing the inverse of covariance matrix R.
3. The interference tomography SAR forest canopy height estimation optimization method according to claim 1, wherein determining the forest canopy initial envelope, the earth surface initial envelope, and the signal loss threshold K comprises: According to the maximum forest canopy height of the research area Performing upper limit constraint on the inversion result; Setting threshold values K with different step sizes, and extracting TomoSAR upper and lower envelopes of the relative reflectivity signals; Obtaining forest canopy height estimated values under different threshold values according to the upper envelope curve difference values and the lower envelope curve difference values, and removing inversion heights larger than Is a sample of (2); selecting an optimal threshold K according to the RMSE between the inversion result and LiDAR-RH100, wherein the expression is specifically as follows: ; wherein H represents a forest canopy height estimated value obtained under the current threshold condition; representing a value-taking operation for minimizing the target amount; Representing the upper envelope curve height or the canopy side height amount extracted from the envelope curve under the current threshold condition; representing the lower envelope curve height or the surface side height amount extracted from the envelope curve under the current threshold value condition; K represents a signal loss critical value; representing the maximum forest canopy height of the investigation region.
4. A method of optimizing forest canopy height estimation of an interferometric SAR according to claim 3, wherein performing noise screening based on vertical signal continuity on the relative reflectivity signal of the forest vertical section comprises: Setting a threshold T, and taking the threshold T as the maximum critical value of the upward transition from the forest canopy to the noise and the downward transition from the earth surface to the noise; Comparing the relative reflectivity function F (H) layer by layer from the top to the bottom of the section according to the height index H, and recording the corresponding height index as a feedback value H k when the F (H) < T is satisfied; After the complete height layer screening, a feedback sequence { H k } arranged according to the height sequence is obtained, and the adjacent feedback height difference delta H k =H k+1 −H k is calculated; And selecting two adjacent feedback heights (H k\* ,H k\*+1 ) which generate the maximum delta H k as main body signal boundaries, and extracting the relative reflectivity corresponding to the height index between the main body signal boundaries as the normal distribution range of the denoised forest.
5. The method for optimizing forest canopy height estimation of interference tomography SAR according to claim 4, wherein the step of setting the threshold T comprises uniformly setting the threshold T and the signal loss critical value K so that an envelope extraction threshold and a noise continuity screening threshold perform noise rejection and signal boundary discrimination under the same data standard.
6. The method of optimizing forest canopy height estimation of interferometric SAR of claim 1, wherein performing Pauli decomposition on the fully polarized SAR data to obtain canopy density index comprises: performing Pauli decomposition on the fully polarized SAR data; extracting Pauli decomposition brightness as a forest canopy density index; Dividing a research area into a sparse coverage grade, a medium coverage grade and a dense coverage grade according to the canopy density index; outputting the canopy density grade label to which each pixel belongs.
7. The method for optimizing forest canopy height estimation of interferometric SAR in claim 6, wherein assigning different forest canopy K values and ground surface K values to different grades of pixels according to canopy density classification results comprises: establishing a canopy envelope extraction threshold set and a ground surface envelope extraction threshold set respectively aiming at a sparse coverage level, a medium coverage level and a dense coverage level; calling a corresponding canopy K value and a corresponding earth surface K value according to the canopy density grade to which each pixel belongs; re-extracting a canopy envelope curve and a ground envelope curve respectively by using the called canopy K value and ground K value; And performing adaptive correction on the envelope position offset, and outputting a corrected canopy envelope height index and a ground envelope height index.
8. The interference tomography SAR forest canopy height estimation optimization method according to claim 1, wherein the corresponding canopy K value and earth surface K value are called according to canopy density grade to which each pixel belongs, and the expression is specifically: ; H represents the height of a forest canopy; P (&) represents the relative reflectivity or an evaluation function corresponding to the relative reflectivity; Representing the height differential associated with the canopy side and the earth side at location (r, x); Representing the height corresponding to the surface envelope curve at the position (r, x), K (C) representing the critical value corresponding to the coverage level C, C representing the forest coverage level, C (max) representing the dense coverage level, C (medium) representing the medium coverage level, C (min) representing the sparse coverage level, r, x representing the pixel space position index, K= 0.1,0.2,0.3, ⋯,0.8 representing the traversing search for different critical values.
9. The interference tomography SAR forest canopy height estimation optimization method according to claim 8, wherein the forest canopy height is obtained by subtracting the height indexes corresponding to the canopy envelope and the surface envelope, specifically: reading the height index corresponding to the corrected canopy envelope curve and the height index corresponding to the earth surface envelope curve; And performing difference value calculation on the height index corresponding to the canopy envelope curve and the height index corresponding to the earth surface envelope curve to obtain the pixel-by-pixel forest canopy height, wherein the expression is specifically as follows: ; In the formula, Representing the height differential associated with the canopy side and the earth side at location (r, x); Representing the amount of height corresponding to the surface envelope at location (r, x); And performing space collection on all pixel-by-pixel forest canopy heights, and outputting a forest canopy height result of the research area.
10. An interferometric SAR forest canopy height estimation optimization system, the system comprising: The relative reflectivity calculation module is used for reading L-band UAVSAR multi-baseline data, constructing a multi-baseline InSAR covariance matrix based on a ground phase compensation result, and estimating the relative reflectivity of the reconstructed forest vertical section by adopting a Capon frequency spectrum; The relative reflectivity loss threshold K determining module is used for performing envelope curve fitting on relative reflectivity signals of the forest vertical section and determining an forest canopy initial envelope curve, an earth surface initial envelope curve and a signal loss threshold K by combining LiDAR height reference data; the noise error elimination module is used for performing noise screening based on signal continuity in the vertical direction on the relative reflectivity signal of the vertical section of the forest, eliminating noise interference at the top of the canopy and at the bottom of the earth surface, and obtaining a denoised vertical structure reflectivity section of the forest and a more reasonable envelope; the forest canopy density grading module is used for performing Pauli decomposition on the full-polarization SAR data to obtain canopy density indexes, and different forest canopy K values and ground surface K values are given to different grades of pixels according to canopy density grading results; and the forest height inversion module is used for re-executing envelope extraction on the relative reflectivity signal and obtaining the forest canopy height according to the subtraction of height indexes corresponding to the canopy envelope and the earth surface envelope.

Description

Interference chromatography SAR forest canopy height estimation optimization method and system Technical Field The invention relates to the technical field of forest canopy height estimation, in particular to an interference chromatography SAR forest canopy height estimation optimization method and system. Background Forest height is an important index for reflecting forest growth and health, and is a key parameter for estimating forest biomass and carbon reserves. However, due to the influence of traffic, topography and forest structure complexity, the ground actual forest height measurement is high in cost and difficult. In recent years, development of unmanned aerial vehicles and airborne Lidar technology provides a good technical means for accurate measurement of forest tree height, but due to high data acquisition cost, forest height measurement with large area scale is difficult to develop. The synthetic aperture radar is active remote sensing, has the characteristics of all-weather, strong penetrating power, large-scale observation range and the like, and can penetrate through branches and leaves of most forests to reach the ground, so that the vertical structural information of the forests can be obtained through SAR signals, and the defects of other remote sensing means are overcome. Forest parameters are extracted through a synthetic aperture radar, and a currently common forest canopy height inversion model mainly comprises a ground random body scattering model (random volume on ground, RVoG), an inversion method utilizing the phase difference between a canopy and the ground and TomoSAR. The ground random body scattering model mainly represents models such as RVoG three-stage inversion method, maximum likelihood estimation algorithm and the like, and the model principle is based on an interference complex coherence model, so that the requirement on the setting of an initial value is too high, and the practical effect is poor. The classical representation method in the inversion method using the phase difference between the canopy and the ground is an ESPRIT method, the canopy height inversion is performed by using the difference value of the phase centers by acquiring the phase centers of the canopy and the ground, and the phase centers of the canopy and the ground have errors due to the complex forest structure. TomoSAR is widely applied to forest vertical structure acquisition, and TomoSAR technology has a good advantage in acquiring the height direction of a forest canopy by separating the reflected signal of a target ground object in the height direction to acquire a three-dimensional structure. In Reigber et al in 2000, an L-band airborne dataset is firstly used, a three-dimensional structure of a forest is obtained through a Fast Fourier Transform (FFT) technology, experiments in forest application are successful for the first time, and a foundation is laid for a plurality of scientific researchers to apply the chromatographic SAR to the forest. At present, after the information of the forest vertical structure is obtained through TomoSAR, an envelope is commonly used for extracting the envelope of the relative reflectivity signal of the forest vertical structure, so that the height of the forest canopy is obtained. However, to implement TomoSAR techniques, the SAR data needs to be preprocessed, temporally decohered, reconstructed, and analyzed as a result. The data preprocessing includes performing land leveling phase removal, terrain phase correction and phase calibration on the SAR data, and even if the data is subjected to relevant preprocessing, part of residual phase influence exists due to differences of terrain, target ground objects, sensors and the like. Time decoherence requires consideration of the time interval of the baseline. The beam forming method includes an nonparametric method (Beamforming method, capon method, etc.) and a parametric method (Music method, WSF method, etc.), and the nonparametric method is different from the parametric method in that the nonparametric method does not need to acquire prior knowledge of the number of scatterers, scattering mechanism, etc., and can directly perform correlation processing on signals. In practical application, due to various influencing factors such as terrain, baseline time interval, vegetation complexity, uncertainty of a beam forming method and the like, errors are generated in phase focusing of SAR data, and the influence of partial residual phase exists, so that an envelope line falls to a position beyond forest vertical structure information. Along with the penetration of microwaves to the ground surface when the microwaves pass through different forest covers, the information range in the vertical direction of the forest is enlarged, and the phenomenon that the height of the canopy of the forest is overestimated occurs. In summary, after performing terrain phase correction and phase compensation on data, due to various