CN117077547B - Forest overground biomass estimation method and system - Google Patents
Forest overground biomass estimation method and system Download PDFInfo
- Publication number
- CN117077547B CN117077547B CN202311331655.2A CN202311331655A CN117077547B CN 117077547 B CN117077547 B CN 117077547B CN 202311331655 A CN202311331655 A CN 202311331655A CN 117077547 B CN117077547 B CN 117077547B
- Authority
- CN
- China
- Prior art keywords
- forest
- model
- height
- calculating
- penetration depth
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 136
- 239000002028 Biomass Substances 0.000 title claims abstract description 65
- 230000035515 penetration Effects 0.000 claims abstract description 87
- 230000008030 elimination Effects 0.000 claims abstract description 40
- 238000003379 elimination reaction Methods 0.000 claims abstract description 40
- 238000002790 cross-validation Methods 0.000 claims abstract description 39
- 239000002245 particle Substances 0.000 claims abstract description 38
- 238000012706 support-vector machine Methods 0.000 claims abstract description 37
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 36
- 230000001427 coherent effect Effects 0.000 claims abstract description 30
- 230000008033 biological extinction Effects 0.000 claims abstract description 25
- 238000000926 separation method Methods 0.000 claims abstract description 21
- 238000007637 random forest analysis Methods 0.000 claims description 27
- 238000012795 verification Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 abstract description 11
- 230000008569 process Effects 0.000 description 12
- 230000006870 function Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 8
- 230000010287 polarization Effects 0.000 description 5
- 238000003066 decision tree Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000003384 imaging method Methods 0.000 description 4
- 238000011835 investigation Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 238000012549 training Methods 0.000 description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- 238000005259 measurement Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000002310 reflectometry Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000010521 absorption reaction Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- OYPRJOBELJOOCE-UHFFFAOYSA-N Calcium Chemical compound [Ca] OYPRJOBELJOOCE-UHFFFAOYSA-N 0.000 description 1
- 238000012952 Resampling Methods 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Hardware Design (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- Geometry (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention relates to the technical field of biomass estimation, and discloses a forest overground biomass estimation method and system, wherein the method comprises the following steps: inversion is carried out by using a coherent scattering model RVoG three-stage method to obtain forest canopy height and extinction coefficient, and phase center height and coherent separation degree of complex coherence are calculated; calculating the penetration depth of the bulk scattering complex coherence based on the microwave penetration depth theory, and calculating the forest height and penetration depth ratio of the RVoG model after the penetration compensation; calculating vertical wave numbers, 2 pi fuzzy heights and baseline parameters according to the InSAR geometric relationship; reducing the independent variable dimension by adopting a cross-validation recursive feature elimination method; and optimizing the regression model parameters of the support vector machine by using a global optimal particle swarm algorithm, and selecting independent variable characteristics to estimate forest ground biomass based on a cross-validation recursive characteristic elimination method. According to the method, forest biomass estimation accuracy is effectively improved through integrating the PolInSAR multidimensional parameters and an optimized machine learning method.
Description
Technical Field
The invention relates to the technical field of biomass estimation, in particular to a forest overground biomass estimation method and system.
Background
Forest is an important component of land ecosystem, and monitoring of forest ecosystem is particularly important in the context of global climate change. The accurate and efficient acquisition of the forest overground biomass (AGB) has great significance on the dynamic monitoring of forest resources, the traditional AGB measurement mainly depends on manual field investigation, the method can obtain a relatively accurate result, but has low efficiency, a large amount of manpower and funds are required to be consumed, the method is not applicable to the large-scale investigation, the gap is formed in the forest canopy height monitoring under the regional scale, and the application of the remote sensing technology greatly improves the investigation efficiency of forest parameters and forest resources. At present, the remote sensing technology mainly applied to forest investigation mainly comprises optical remote sensing, laser radar remote sensing, microwave remote sensing and the like, and has the advantages of short reentry cycle, wide coverage area and more available data sources, and is widely applied to the fields of forest biomass, accumulation amount estimation, forest classification, forest change monitoring and the like, but has the limitations that optical data only can acquire canopy reflectivity information, is insensitive to vertical structure information of a forest, and is seriously influenced by cloud and weather; the airborne laser radar (LiDAR) can rapidly acquire the information of the canopy and the internal three-dimensional structure of the stand and is not limited by weather conditions, but the main constraint factors of the LiDAR are the observation scale and the cost of data acquisition, and the airborne platform is flexible and convenient to use, but has limited coverage area of data, no continuity and higher cost; although the coverage of the satellite-borne platforms such as ICEsat-1, ICEsat-2 and GEDI is larger, only the data of the strip scale is obtained due to the limitation of the sensor platform and the data type, the surface coverage cannot be achieved, and the microwave remote sensing is used as the three, and has the characteristics of all weather, initiative and large-scale surface coverage.
The polarization interference synthetic aperture radar (PolInSAR) has the common characteristics of PolSAR and InSAR, can reflect the height and scattering characteristics of the surface vegetation scatterer, and has great potential in regional scale forest biomass inversion. The method has been widely studied in the past twenty years, and currently common inversion models generally comprise a polarized water cloud model, an interference water cloud model, a linear model and a machine learning regression model. The model forms fixed in the polarized water cloud model, the interference water cloud model and the linear model are difficult to describe the real forest scene and are influenced by the saturation of SAR signals. In the research of a machine learning method, the traditional research mostly adopts coherence, geometric parameters, backscattering characteristics and even coherence shape parameters as independent variables and combines a small amount of ground measurement data to invert forest biomass, and the variables do not completely reflect the visual effect of SAR on the vertical structure parameters of the forest, so that the estimation precision is low, and the problem of SAR backscattering signal saturation is also faced. Because the forest biomass can be directly related to the forest height characteristics through the different-speed growth function, the vertical structure information obtained through the mechanism model is used as an independent variable of a machine learning method, and the inversion model is more reasonable to construct by combining ground measurement data. The existing forest canopy height inversion method based on PolInSAR is mainly based on the scattering theory of electromagnetic waves, and the characteristic that microwaves are sensitive to dielectric properties and geometric shapes of ground objects is utilized to obtain scattering information of the ground objects which cannot be obtained by optical remote sensing. The microwaves have certain penetrability, the internal structural information of the forest can be obtained through the canopy of the forest, and the backscattering signals obtained by the microwaves are mainly from the crowns, branches and trunks of the forest and the ground surface, so that the method has unique advantages for obtaining the vertical information of the forest structure. And (3) extracting forest canopy height information based on InSAR/PolInSAR. The most widely used is the RVoG model method which has been successfully applied to InSAR/PolInSAR data of different frequencies, including C, L, P and even the X band, and different forest types are included in the researches. RVoG describes the forest scattering process as a forest bulk scattering layer and an impenetrable ground layer, the bulk scattering layer being regarded as an isotropic homogeneous medium of thickness hv, and the scattering and absorption losses of electromagnetic waves therein being described by an extinction coefficient σ independent of polarization. If assumptions are made about the ground volume amplitude ratio and extinction coefficient σ, or where the terrain is known, the forest canopy height can be inverted from using the single polarized InSAR data based on the RVoG model. RVoG three-stage inversion method estimates and estimates the earth surface phase by fitting the real part and the imaginary part of the complex coherence through a least square method based on the distribution rule of the interference complex coherence, and sets a reasonable forest canopy height and extinction coefficient construction lookup table to perform forest canopy height inversion. The height estimation error caused by the penetration of microwaves can be corrected by adopting the microwave penetration theory.
Therefore, how to provide a model and a method for estimating forest biomass with high accuracy for small samples is a technical problem to be solved.
Disclosure of Invention
The invention mainly aims to provide a forest overground biomass estimation method and system, and aims to improve AGB estimation accuracy by excavating PolInSAR multidimensional forest structural features and adopting an optimized machine learning method.
To achieve the above object, the present invention provides a method for estimating forest floor biomass, the method comprising the steps of:
s1: inversion is carried out by using a coherent scattering model RVoG three-stage method to obtain forest canopy height and extinction coefficient, and phase center height and coherent separation degree of complex coherence are calculated;
s2: calculating the penetration depth of the bulk scattering complex coherence based on the microwave penetration depth theory, and calculating the forest height and penetration depth ratio of the RVoG model after penetration compensation;
s3: calculating vertical wave numbers and 2 pi fuzzy heights and baseline parameters according to InSAR geometric relations of the sensor observation platform;
s4: based on the variable information extracted in the steps S1-S3, a random forest model-based cross verification recursive feature elimination method is adopted to select variable parameters, and the independent variable dimension is reduced; the variable information comprises a phase center height and a coherence separation degree of complex coherence, a RVoG model forest height and a penetration depth ratio after penetration compensation, a vertical wave number, a 2 pi fuzzy height and a baseline parameter;
s5: and optimizing the regression model parameters of the support vector machine by using a global optimal particle swarm algorithm, and estimating the forest aboveground biomass based on the independent variable characteristics selected by the cross-validation recursive characteristic elimination method.
Optionally, in the step S1, a step of inverting to obtain the height and the extinction coefficient of the canopy of the forest by using a coherent scattering model RVoG three-stage method specifically includes: solving ground phase by fitting intersection point of straight line where PolInSAR complex coherence point is located and unit circle by RVoG three-stage method, determining volume scattering complex coherence by referencing the ground phase, and finally obtaining forest canopy height Hv and extinction coefficient by two-dimensional lookup table methodThe method comprises the steps of carrying out a first treatment on the surface of the And phase center height and coherence separation of complex coherence are calculated with reference to ground phase.
Optionally, in the step S2, the step of calculating the penetration depth of the bulk scattering complex coherence based on the microwave penetration depth theory and calculating the forest height and penetration depth ratio of the RVoG model after the penetration compensation specifically includes: and calculating the penetration depth Hd of the bulk scattering based on the microwave penetration theory, and calculating the penetration compensated forest canopy height Hc and the microwave penetration ratio Hrate by combining the forest canopy height Hv in the step S1.
Optionally, in the step S3, a step of calculating the vertical wave number, the 2 pi ambiguity height and the baseline parameter according to the InSAR geometric relationship of the sensor observation platform specifically includes: and determining a baseline selection parameter by adopting a Line algorithm, and calculating vertical wave numbers kz and 2 pi fuzzy high HoA corresponding to the InSAR geometric relationship.
Optionally, in the step S4, a step of selecting variable parameters and reducing independent variable dimensions by adopting a cross-validation recursive feature elimination method based on a random forest model specifically includes: the method comprises the steps of selecting variable parameters by adopting a cross-validation recursive feature elimination method based on a random forest model, evaluating model scores by gradually eliminating features and setting a cross-validation algorithm on the basis of random forest regression, and determining the number of the independent variables by using the combination with the highest score in all combinations so as to reduce the dimension of the independent variables.
Optionally, in the step S5, a step of optimizing the regression model parameters of the support vector machine by using a global optimal particle swarm algorithm specifically includes: and taking the support vector machine regression model as a biomass inversion model, and optimizing C, epsilon and gamma parameters in the support vector machine regression model by adopting a global optimal particle swarm algorithm.
Optionally, in the step S5, the step of estimating the forest above-ground biomass based on the independent variable feature selected by the cross-validation recursive feature elimination method specifically includes: inputting the independent variable characteristics selected by the cross-validation recursive characteristic elimination method in the step S4 into an optimized support vector machine regression model, and obtaining a forest aboveground biomass estimation result from the output of the support vector machine regression model.
In order to achieve the above object, the present invention also provides a forest aboveground biomass estimation system, the system comprising:
the system comprises a height information inversion module, a phase center height and a coherence separation degree, wherein the inversion module is configured to invert by using a coherent scattering model RVoG three-stage method to obtain a forest canopy height and an extinction coefficient, and calculate a phase center height and a coherence separation degree of complex coherence;
the system comprises a height compensation module, a volume scattering complex coherence module and a volume scattering complex coherence module, wherein the compensation module is configured to calculate the penetration depth of volume scattering complex coherence based on a microwave penetration depth theory and calculate the forest height and penetration depth ratio of an RVoG model after penetration compensation;
a geometric parameter calculation module configured to calculate vertical wave numbers and 2pi ambiguity heights and baseline parameters from an InSAR geometric relationship of the sensor observation platform;
a variable selection module configured to select variable parameters using a random forest model-based cross-validation recursive feature elimination method, reducing an independent variable dimension;
an AGB estimation module configured to optimize support vector machine regression model parameters using a global optimal particle swarm algorithm and estimate forest aboveground biomass based on the independent variable features selected by the cross-validation recursive feature elimination method.
The invention has the beneficial effects that: the method comprises the steps of inversion by using a coherent scattering model RVoG three-stage method to obtain forest canopy height and extinction coefficient, and calculating the phase center height and coherent separation degree of complex coherence; calculating the penetration depth of the bulk scattering complex coherence based on the microwave penetration depth theory, and calculating the forest height and penetration depth ratio of the RVoG model after penetration compensation; calculating vertical wave numbers and 2 pi fuzzy heights and baseline parameters according to InSAR geometric relations of the sensor observation platform; selecting variable parameters by adopting a cross verification recursive feature elimination method based on a random forest model, and reducing independent variable dimensions; and optimizing the regression model parameters of the support vector machine by using a global optimal particle swarm algorithm, and estimating the forest aboveground biomass based on the independent variable characteristics selected by the cross-validation recursive characteristic elimination method. According to the method, the forest biomass estimation precision is effectively improved through integrating the PolInSAR multidimensional parameters and the optimized machine learning method, the problem of backward scattering signal saturation in SAR data biomass estimation can be avoided, and the biomass value underestimation after forest height information saturation can be effectively relieved.
Drawings
FIG. 1 is a schematic flow chart of a forest land biomass estimation method of the invention;
FIG. 2 is a schematic diagram of the forest floor biomass estimation of the present invention;
figure 3 is a schematic diagram of an RVoG model in an embodiment of the invention;
figure 4 is a schematic diagram of a RVoG three stage process of the present invention;
FIG. 5 is a schematic view of the penetration of microwaves in a forest according to the present invention;
FIG. 6 is a schematic diagram of a baseline selection according to the present invention;
FIG. 7 is a diagram of InSAR observation geometry according to the present invention;
fig. 8 is a block diagram of the forest floor biomass estimation system of the invention.
The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The embodiment of the invention provides a forest aboveground biomass estimation method, and referring to fig. 1, fig. 1 is a flow chart of an embodiment of the forest aboveground biomass estimation method.
In this embodiment, the method for estimating forest aboveground biomass includes the following steps:
s1: calculating the volume scattering complex coherence gamma using the coherent scattering model RVoG three-stage method H And ground phase phi 0 And inverting to obtain the height of the forest canopy and the extinction coefficient, and simultaneously calculating the complex coherence (gamma H And gamma L ) Phase center height and coherence separation of (a);
s2: calculating the volume scattering complex coherence (gamma) based on the microwave penetration depth theory H ) Calculating the forest height and penetration depth ratio of the RVoG model after the penetration compensation;
s3: calculating vertical wave numbers and 2 pi fuzzy heights and baseline parameters according to InSAR geometric relations of the sensor observation platform;
s4: variable parameter information is obtained based on the S1-S3 stage, a random forest model-based cross verification recursive feature elimination method is adopted to select variable parameters, and independent variable dimensions are reduced; the variable information comprises a phase center height and a coherence separation degree of complex coherence, a RVoG model forest height and a penetration depth ratio after penetration compensation, a vertical wave number, a 2 pi fuzzy height and a baseline parameter;
s5: based on the independent variable information selected in the S4 stage, estimating forest overground biomass by using a support vector machine regression model, and optimizing parameters of the support vector machine regression model by adopting a global optimal particle swarm algorithm.
In a preferred embodiment, in the step S1, the step of inverting to obtain the forest canopy height and the extinction coefficient by using the coherent scattering model RVoG three-stage method specifically includes: solving ground phase by fitting intersection point of straight line where PolInSAR complex coherence point is located and unit circle by RVoG three-stage method, determining volume scattering complex coherence by referencing the ground phase, and finally obtaining forest canopy height Hv and extinction coefficient by two-dimensional lookup table method。
In a preferred embodiment, in the step S2, the step of calculating the penetration depth of the bulk scattering complex coherence based on the microwave penetration depth theory and calculating the forest height and penetration depth ratio of the RVoG model after the penetration compensation specifically includes: and calculating the penetration depth Hd of the bulk scattering based on the microwave penetration theory, and calculating the penetration compensated forest canopy height Hc and the microwave penetration ratio Hrate by combining the forest canopy height Hv in the step S1.
In a preferred embodiment, in the step S3, the step of calculating the vertical wave number, the 2 pi ambiguity height and the baseline parameter according to the InSAR geometric relationship of the sensor observation platform specifically includes: and determining a baseline selection parameter by adopting a Line algorithm, and calculating vertical wave numbers kz and 2 pi fuzzy high HoA corresponding to the InSAR geometric relationship.
In a preferred embodiment, in the step S4, a step of selecting variable parameters and reducing the dimensions of the independent variables by using a cross-validation recursive feature elimination method based on a random forest model specifically includes: the method comprises the steps of selecting variable parameters by adopting a cross-validation recursive feature elimination method based on a random forest model, evaluating model scores by gradually eliminating features and setting a cross-validation algorithm on the basis of random forest regression, and determining the number of the independent variables by using the combination with the highest score in all combinations so as to reduce the dimension of the independent variables.
In a preferred embodiment, in the step S5, the step of optimizing the support vector machine regression model parameters by using a global optimal particle swarm algorithm specifically includes: and taking the support vector machine regression model as a biomass inversion model, and optimizing C, epsilon and gamma parameters in the support vector machine regression model by adopting a global optimal particle swarm algorithm.
In a preferred embodiment, in the step S5, the step of estimating the forest land biomass based on the argument feature selected by the random forest cross-validation recursive feature elimination method specifically includes: inputting the independent variable characteristics selected by the cross-validation recursive characteristic elimination method in the step S4 into an optimized support vector machine regression model, and obtaining a forest aboveground biomass estimation result from the output of the support vector machine regression model.
Therefore, the embodiment provides a forest aboveground biomass estimation method, which effectively improves forest biomass estimation precision by integrating the PolInSAR multidimensional parameter and the optimized machine learning method, can not only avoid the problem of backward scattering signal saturation in SAR data biomass estimation, but also effectively relieve biomass value underestimation after forest height information saturation
For a clearer explanation of the present application, specific examples of the forest above-ground biomass estimation method of the present application are provided below. As shown in fig. 2, the embodiment adopts a forest ground biomass estimation method integrating the PolInSAR multidimensional forest structure parameters and optimizing machine learning. In the prior art, a simple polarization interference characteristic and a common machine learning method are adopted to estimate forest biomass, biomass is estimated by comprehensively excavating PolInSAR multidimensional forest structure parameters (including height, penetration depth, coherent separation degree and the like), a random forest cross-validation recursive characteristic elimination method (RFECV-RF) is used for automatically selecting characteristic combinations in an inversion process, a final biomass estimation link is improved, and a global optimal particle swarm (GLB-PSO) algorithm is adopted to optimize SVR model parameters. Compared with the conventional method, the method can not only avoid the problem of backward scattering signal saturation in SAR data biomass estimation, but also effectively relieve biomass value underestimation after forest height information saturation, and is an effective forest biomass inversion method.
Regarding step S1: inversion is carried out by using a coherent scattering model RVoG three-stage method to obtain the forest canopy height and the extinction coefficient, and the phase center height and the coherent separation degree of complex coherence are calculated. Specifically: forest height and phase center height extraction, comprising the following implementation steps:
(1) Data processing
The L-band airborne multi-baseline PolInSAR data and the ground biomass verification data are both derived from the published data set of the AfriSAR project. The PolInSAR dataset is subjected to polarization scaling, baseline fine registration and spectral filtering and provided in the form of single vision complex numbers, each track containing SLC data (HH, HV, VH and VV) for four polarized channels. The number of tracks in the test area is 8, the data are subjected to multi-view processing, and the PD coherent optimization algorithm is used for calculating complex coherence under different baseline combinations、/>)。
(2) RVoG model
In RVoG, the forest scattering process is described as a forest volume scattering layer and an impenetrable ground layer, the volume scattering layer being regarded as having a thicknessIs combined with an extinction coefficient independent of polarization>Describing the scattering and absorption losses of electromagnetic waves in the bulk scattering layer (see FIG. 3), the ground elevation is z 0 The scattering product height is hv, F (z) is radar reflectivity at z height, σ is extinction ratio, φ 0 is ground phase, μ is ground-to-body amplitude ratio. The general expression for the RVoG model is as follows:
(1)
In the middle ofFor an effective body amplitude ratio, +.>Expressed as ground phase +.>Representing interference complex coherence->Representing the ground volume amplitude ratio,/->Indicating decoherence caused by vegetation alone, irrespective of the contribution of surface scattering +.>Can be expressed as:
(2)
Is a distance gradient>Is of oblique distance and is->For the vertical base-line length,nn=2 in the single-base operation mode depending on the acquisition mode of the radar image; n=1 in the dual base mode of operation. />For the ground phase +.>In order to take into account the relative reflectivity of the vegetation volume scattering fraction, z is vegetation height, +.>For radar incidence angle, +.>And->Is a process parameter, does not have any and +.>In the form of complex expression->For radar wavelength, +.>For vertical wave number>The difference in incidence angle between the main and auxiliary images is represented.
(3) RVoG three-stage method
The RVoG three-stage method is the most classical method for solving an RVoG model, the three-stage forest canopy height inversion method is proposed by Cloude, the phase solution is carried out by fitting a coherent straight line where complex coherence is located, the forest canopy height inversion is carried out by using a lookup table method, and the process is divided into three stages:
the first stage: and (5) fitting a coherent straight line. Fitting baseline corresponding、/>) Two potential ground phases (phi) are obtained 1 And phi 0 ) (FIG. 4-a).
And a second stage: on the basis of the first phase, the ground phase is determined from two potential ground phasesSelecting complex coherent (gamma) with the farthest phase from ground H ) As a bulk scattering complex coherence (fig. 4-b).
And a third stage: forest canopy height (Hv) and extinction coefficient) Is a function of the estimate of (2). Ground phase->And bulk scattering complex coherence (gamma) H ) After the determination, according to +.in the formula (2-5)>And (Hv and->) Reasonable Hv and +.>Creating a two-dimensional lookup table (LUT), wherein the inversion process is to search and +.>Minimum distance +.>Corresponding forest canopy height (Hv) and extinction coefficient (+.>):
(3)
Representing the complex coherence furthest from the earth phase. />、/>Respectively the ground phase and distance.
With known phase, can be obtainedAnd->Distance from ground phase +.>And->Phi from the earth's surface 0 I.e. the phase center height of the complex coherence.
TABLE 1 phase center height and coherence separation
Note that:and->Respectively representing crown top complex coherence and ground complex coherence, phi 0 representing ground phase, abs () representing modulus of complex numbers, +_>() Representing the phase of the complex number
Regarding step S2: and calculating the penetration depth of the bulk scattering complex coherence based on the microwave penetration depth theory, and calculating the forest height and penetration depth ratio of the RVoG model after penetration compensation. Specifically, the microwave penetration depth compensation specifically comprises the following implementation steps:
the RVoG model is established on the basis of separating the phase center of the canopy of the forest microwave scattering signal from the phase center of the ground, and the estimation of the height of the canopy of the forest is realized by combining imaging geometry. When the phase center is separated, the penetration of microwaves to the forest often causes the phase center of the canopy to move downwards, so that the model is underestimated when the height of the canopy of the forest is estimated (see fig. 5), fig. 5 (a) is a schematic diagram of the scattering phase center for the building, and fig. 5 (b) is the scattering phase center for the forest.
In the estimation study for InSAR penetration depth, dall provides the only theoretical framework for estimating infinite depth volume penetration depth so far, which provides a very advantageous help for directly calculating the penetration depth of SAR data. The theory proposes a height deviationThe interference phase can be normalized by the phase>And vertical wave numberkzCalculated, see equation (4).
(4)
In the middle ofRepresenting the 2 pi blur height in air, hoA taking into account the volume refraction, i.e. +.>Can be expressed as:
(5)
Where n is the refractive index, complex coherence in an infinitely deep volume can be expressed as:
(6)
d 2 Is a measure of the depth of penetration in both directions, σ is the extinction coefficient, according to Dall, study d 2 /HoA Vol Normalized interference phase in relation to coherence amplitudeCan be composed ofThe formula (7) gives:
(7)
Assuming negligible refraction n in the scatterer, the penetration depth can be derived from the normalized interference phaseIt follows that the formula (7) can be written as:
(8)
TABLE 2 penetration depth
Regarding step S3: and calculating the vertical wave number and the 2 pi fuzzy height and the baseline parameter according to the InSAR geometric relation of the sensor observation platform. Specifically, the InSAR geometrical parameters specifically comprise the following implementation steps:
(3) Baseline parameters
Because of the difference of the space baselines, the long baselines are suitable for low forests, the short baselines are suitable for higher forests, the multi-baseline PolInSAR technology is more in line with the real forest scene (see figure 6), and the optimal baseline size is selected from a plurality of baseline combinations to invert the forest canopy height, so that the multi-baseline PolInSAR technology is more in line with the actual situation. However, in practical situations, the distribution range of the complex coherence points is approximately elliptical, the objective of baseline selection is to select a most suitable pair from a plurality of baseline combinations to perform forest canopy height inversion, and the basis of judgment is to select the baseline combination which best meets the assumption of the RVoG model according to the separation degree and the distribution shape of the complex coherence. Currently, the most widely used baseline selection methods are mainly VAR, ECC and PROD methods and LINE methods proposed by lavale (2017), which are used herein.
LINE method: the LINE method considers the product of the long axis distance of the coherent region and the minimum distance between the origin of the complex plane and the coherent straight LINE, and is the same as PROD in that the long axis size of the coherent region, namely the coherent separation degree, is considered; the LINE method considers the distance between the distribution of the coherent points and the origin, so as to avoid the situation that the phase difference is small due to the fact that the coherent straight LINE is close to the origin, and therefore, a baseline combination (formula 9) when the LINE value is maximum is selected.
(9)
Where p and t are intermediate variables of the simplified formula.
TABLE 3 Baseline selection parameters
In the embodiment, multi-baseline PolInSAR data are adopted, baseline selection is needed, a Line combination with the largest Line value is selected from different baseline combinations by adopting a Line party according to RVoG model principle, and a group of complex coherence @ corresponding to the optimal baseline combination is determined、/>)。
(4) InSAR imaging geometry
The imaging geometry of the InSAR altimetry is shown in FIG. 7.
TABLE 4 InSAR imaging geometry parameters
Regarding step S4: forest canopy height, microwave penetration depth, canopy height after penetration compensation, and PolInSAR baseline parameters and geometric parameters are obtained from PolInSAR information in the S1-S3 stage, and are used for AGB estimation, but independent variables are required to be screened before a machine learning regression model is constructed. Therefore, a cross-validation recursive feature elimination method based on a random forest model is adopted to select variable parameters, and the independent variable dimension is reduced. Specifically, the cross-validation recursive feature elimination method (RFECV-RF) variable selection based on the random forest regression model specifically comprises the following implementation steps:
random Forest (RF) is a data mining method found by Cutler Adelee and Leo Breiman, which is a combined self-learning and modern regression and classification technique. The random forest can be used for classification and regression as well as clustering and survival analysis, and has the advantages of strong adaptability to a data set, good noise resistance, extremely strong fitting capacity and no overfitting phenomenon compared with other algorithms. The method comprises the steps of extracting a plurality of samples from an original sample by utilizing a bootstrapping resampling method, modeling a decision tree for each bootstrapping sample, then combining predictions of a plurality of decision trees, and obtaining a final prediction result through voting. The internal node tree structure is built according to Gini standard (Gini criterion) best principle. A original variables are set, ai feature variables are randomly selected to split decision trees and grow freely to generate Ntree decision trees, and the number of trees (Ntree) and the size (Ai) of a randomly selected subset in the regression process need to be optimized to obtain the best fitting result. The random forest regression model has the advantages that: 1) Large-scale data sets can be processed, and the roles of thousands of explanatory variables can be predicted; 2) Insensitive to the polynary linear formula, the prediction results of the missing data and the unbalanced data are relatively robust; 3) An importance estimate of the variable can be given; 4) The training speed is high.
The recursive feature elimination method (RFE, recursive Feature Elimination) is a greedy optimization algorithm aimed at finding feature subsets that perform best. It repeatedly creates a model and either retains the best features or rejects the worst features at each iteration, and at the next iteration it builds the next model using features that were not selected in the last modeling until all features are exhausted. The cross-validation recursive feature elimination method (RFECV, recursive Feature Elimination using Cross Validation) is very simple in logic, i.e., with the model you want to train, training with all features first, then progressively reducing the features, and scoring each training by cross-validation, so that a scheme with the least features and highest scores can be obtained.
Regarding step S5: on the basis of S4 independent variable selection, a global optimal particle swarm algorithm is used for optimizing the regression model parameters of the support vector machine, and forest overground biomass is estimated based on the independent variable characteristics selected by the cross-validation recursive characteristic elimination method. Specifically, the support vector machine regression AGB inversion model (GLB-PSO-SVR) based on global optimal particle swarm optimization comprises the following implementation steps:
(1) Support vector machine (SVM, support Vector Machine) algorithm
The method is proposed by Vapnik et al, and simultaneously, the support vector machine algorithm is based on the principle of multi-dimension and minimum risk in statistics, and has the characteristics of supporting small sample number and low risk, namely, the method is used for searching the best method on the premise of existing samples and established precision and accurately and correctly identifying the samples, so that better expectations and universality are obtained. Therefore, the support vector machine algorithm is also the most commonly used algorithm in the current machine learning, and can be used for regression and classification, and the method can be used for selecting kernel functions according to the use requirements, including linear kernel functions, polynomial kernel functions, radial basis kernel functions and S-shaped kernel functions, and has better performance in small sample regression estimation.
(2) GLB-PSO optimization algorithm
In the SVR model, when a radial basis function is adopted, the effect of the model is required to be achieved finally by adjusting parameters C, epsilon and gamma, wherein C is a punishment coefficient, namely tolerance to errors. The higher the specification, the less tolerant the error, and the easier the overfitting. The smaller C, the easier the under-fit. When C is too large or too small, the generalization ability becomes poor. gamma is a kernel function coefficient after selecting an RBF function as a kernel, and implicitly determines the distribution of data mapped to a new feature space, the larger the gamma is, the smaller the gamma value is, the more support vectors are, and the number of the support vectors influences the training and prediction speed. Epsilon specifies Epsilon-tube in an Epsilon-SVR model that approximates the optimum with a given margin called Epsilon-tube (Epsilon represents the width of the tube) taking into account the complexity and error rate of the model.
In order to determine the optimal parameters in the SVR model, the present embodiment uses a global optimal particle swarm (GLB-PSO) algorithm to optimize the selection of C, epsilon, gamma parameters, the main idea of PSO being to place a cluster S of particles Pn into a search space to find the optimal solution. Particle swarm algorithm simulates birds in a bird swarm by designing a mass-free particle that has only two properties: speed, which represents the speed of movement, and position, which represents the direction of movement. Each particle independently searches for an optimal solution in a search space, marks the optimal solution as a current individual extremum, shares the individual extremum with other particles in the whole particle swarm, finds the optimal individual extremum as a current global optimal solution of the whole particle swarm, and adjusts the speed and the position of each particle in the particle swarm according to the current individual extremum found by each particle and the current global optimal solution shared by the whole particle swarm.
The particles are defined as:,
,
and the particle group is composed of N particles, and has a certain position at the time step t:
,
in order to find a global optimum, the population of particles must be moved. According to the velocity V of the particle swarm, the movement is achieved by updating the current position:
,
the speed is then calculated as follows:
,
wherein,and->Representing interval [0,1 ]]Random value in->Is the best position->Is the current individual location,/>Is the optimal location for all particles. Furthermore, the->Is an inertial weight used to control the "memory" of the previous position of the colony. c1 and c2 represent the individual cognitive coefficients and population coefficients of the particles, respectively.
And finally, inputting the independent variable characteristics selected by the cross validation recursive characteristic elimination method in the step S4 into an optimized support vector machine regression model, and obtaining a forest aboveground biomass estimation result from the output of the support vector machine regression model.
Referring to fig. 8, fig. 8 is a block diagram illustrating a forest floor biomass estimation system according to an embodiment of the invention.
As shown in fig. 8, the forest floor biomass estimation system according to the embodiment of the invention includes:
the inversion module 10 is configured to invert to obtain a forest canopy height and an extinction coefficient by using a coherent scattering model RVoG three-stage method, and calculate a phase center height and a coherent separation degree of complex coherence;
a compensation module 20 configured to calculate a penetration depth of the bulk scattering complex coherence based on a microwave penetration depth theory, and calculate a penetration compensated RVoG model forest height and penetration depth ratio;
a geometric parameter calculation module 30 configured to calculate vertical wavenumbers and 2pi ambiguity heights, as well as baseline parameters, from the InSAR geometric relationship of the sensor observation platform;
a variable selection module 40 configured to select variable parameters using a random forest model-based cross-validation recursive feature elimination method to reduce the argument dimensions;
an AGB estimation module 50 configured to optimize support vector machine regression model parameters using a global optimal particle swarm algorithm and estimate forest aboveground biomass based on the independent variable features selected by the cross-validation recursive feature elimination method.
Other embodiments or specific implementations of the forest land biomass estimation system of the present invention may refer to the above-mentioned method embodiments, and will not be described herein.
It is appreciated that in the description herein, reference to the terms "one embodiment," "another embodiment," "other embodiments," or "first through nth embodiments," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.
The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.
Claims (2)
1. A method for estimating forest land biomass, the method comprising the steps of:
s1: inversion is carried out by using a coherent scattering model RVoG three-stage method to obtain forest canopy height and extinction coefficient, and phase center height and coherent separation degree of complex coherence are calculated; in the step S1, inversion is performed by using a coherent scattering model RVoG three-stage method to obtain a forest canopy height and an extinction coefficient, and the method specifically includes: solving ground phase by fitting intersection point of straight line where PolInSAR complex coherence point is located and unit circle by RVoG three-stage method, determining volume scattering complex coherence by referencing the ground phase, and finally obtaining forest canopy height Hv and extinction coefficient by two-dimensional lookup table methodThe method comprises the steps of carrying out a first treatment on the surface of the Simultaneously, calculating the phase center height and the coherence separation degree of complex coherence by referring to the ground phase;
s2: calculating the penetration depth of the bulk scattering complex coherence based on the microwave penetration depth theory, and calculating the forest height and penetration depth ratio of the RVoG model after penetration compensation; in the step S2, the step of calculating the penetration depth of the bulk scattering complex coherence based on the microwave penetration depth theory and calculating the forest height and penetration depth ratio of the RVoG model after the penetration compensation specifically includes: calculating the penetration depth Hd of the bulk scattering based on the microwave penetration theory, and calculating the penetration compensated forest canopy height Hc and the microwave penetration ratio Hrate by combining the forest canopy height Hv in the step S1;
s3: calculating vertical wave numbers and 2 pi fuzzy heights and baseline parameters according to InSAR geometric relations of the sensor observation platform; in the step S3, a step of calculating a vertical wave number, a 2 pi ambiguity height and a baseline parameter according to an InSAR geometric relationship of the sensor observation platform specifically includes: determining a baseline selection parameter by adopting a Line algorithm, and calculating vertical wave number kz and 2 pi fuzzy high HoA corresponding to the InSAR geometric relationship;
s4: based on the variable information extracted in S1-S3, a random forest model-based cross verification recursive feature elimination method is adopted to select variable parameters, and the independent variable dimension is reduced; the variable information comprises a phase center height and a coherence separation degree of complex coherence, a RVoG model forest height and a penetration depth ratio after penetration compensation, a vertical wave number, a 2 pi fuzzy height and a baseline parameter; in the step S4, a variable parameter is selected by adopting a cross-validation recursive feature elimination method based on a random forest model, and the step of reducing the independent variable dimension specifically includes: selecting variable parameters by adopting a cross-validation recursive feature elimination method based on a random forest model, evaluating model scores by gradually eliminating features and setting a cross-validation algorithm on the basis of random forest regression, and determining the number of the independent variables by using the combination with the highest score in all combinations so as to reduce the dimension of the independent variables;
s5: optimizing the regression model parameters of the support vector machine by using a global optimal particle swarm algorithm, and estimating the forest ground biomass based on the independent variable characteristics selected by the cross-validation recursive characteristic elimination method; in the step S5, a step of optimizing the regression model parameters of the support vector machine by using a global optimal particle swarm algorithm specifically includes: taking the support vector machine regression model as a biomass inversion model, and optimizing parameters C, epsilon and gamma in the support vector machine regression model by adopting a global optimal particle swarm algorithm; the method for estimating forest aboveground biomass based on the independent variable characteristics selected by the cross-validation recursive characteristic elimination method specifically comprises the following steps: inputting the independent variable characteristics selected by the cross-validation recursive characteristic elimination method in the step S4 into an optimized support vector machine regression model, and obtaining a forest aboveground biomass estimation result from the output of the support vector machine regression model.
2. A forest floor biomass estimation system, the system comprising:
the system comprises a height information inversion module, a phase center height and a coherence separation degree, wherein the inversion module is configured to invert by using a coherent scattering model RVoG three-stage method to obtain a forest canopy height and an extinction coefficient, and calculate a phase center height and a coherence separation degree of complex coherence; inversion is carried out by using a coherent scattering model RVoG three-stage method to obtain the height of a forest canopy and an extinction coefficient, and the method specifically comprises the following steps: solving ground phase by fitting intersection point of straight line where PolInSAR complex coherence point is located and unit circle by RVoG three-stage method, determining volume scattering complex coherence by referencing the ground phase, and finally obtaining forest canopy height Hv and extinction coefficient by two-dimensional lookup table methodThe method comprises the steps of carrying out a first treatment on the surface of the Simultaneously, calculating the phase center height and the coherence separation degree of complex coherence by referring to the ground phase;
the system comprises a height compensation module, a volume scattering complex coherence module and a volume scattering complex coherence module, wherein the compensation module is configured to calculate the penetration depth of volume scattering complex coherence based on a microwave penetration depth theory and calculate the forest height and penetration depth ratio of an RVoG model after penetration compensation; the method for calculating the penetration depth of the volume scattering complex coherence based on the microwave penetration depth theory, and calculating the forest height and penetration depth ratio of the RVoG model after the penetration compensation specifically comprises the following steps: calculating the penetration depth Hd of the bulk scattering based on the microwave penetration theory, and calculating the penetration compensated forest canopy height Hc and the microwave penetration ratio Hrate by combining the forest canopy height Hv in the step S1;
a parameter calculation module configured to calculate vertical wave numbers and 2pi ambiguity heights and baseline parameters from an InSAR geometry of a sensor observation platform; the method comprises the steps of calculating a vertical wave number, a 2 pi fuzzy height and a baseline parameter according to the InSAR geometric relation of a sensor observation platform, and specifically comprises the following steps: determining a baseline selection parameter by adopting a Line algorithm, and calculating vertical wave number kz and 2 pi fuzzy high HoA corresponding to the InSAR geometric relationship;
a variable selection module configured to select variable parameters using a random forest model-based cross-validation recursive feature elimination method, reducing an independent variable dimension; the method adopts a random forest model-based cross validation recursive feature elimination method to select variable parameters and reduce independent variable dimensions, and specifically comprises the following steps: selecting variable parameters by adopting a cross-validation recursive feature elimination method based on a random forest model, evaluating model scores by gradually eliminating features and setting a cross-validation algorithm on the basis of random forest regression, and determining the number of the independent variables by using the combination with the highest score in all combinations so as to reduce the dimension of the independent variables;
an AGB estimation module configured to optimize support vector machine regression model parameters using a global optimal particle swarm algorithm and estimate forest ground biomass based on the independent variable features selected by the cross-validation recursive feature elimination method; the method for optimizing the regression model parameters of the support vector machine by using the global optimal particle swarm algorithm specifically comprises the following steps: taking the support vector machine regression model as a biomass inversion model, and optimizing parameters C, epsilon and gamma in the support vector machine regression model by adopting a global optimal particle swarm algorithm; estimating forest aboveground biomass based on the independent variable characteristics selected by the cross-validation recursive characteristic elimination method, and specifically comprises the following steps: inputting the independent variable characteristics selected by the cross-validation recursive characteristic elimination method in the step S4 into an optimized support vector machine regression model, and obtaining a forest aboveground biomass estimation result from the output of the support vector machine regression model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311331655.2A CN117077547B (en) | 2023-10-16 | 2023-10-16 | Forest overground biomass estimation method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311331655.2A CN117077547B (en) | 2023-10-16 | 2023-10-16 | Forest overground biomass estimation method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117077547A CN117077547A (en) | 2023-11-17 |
CN117077547B true CN117077547B (en) | 2023-12-26 |
Family
ID=88717546
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311331655.2A Active CN117077547B (en) | 2023-10-16 | 2023-10-16 | Forest overground biomass estimation method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117077547B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117452432B (en) * | 2023-12-21 | 2024-03-15 | 西南林业大学 | Forest canopy height estimation method based on forest penetration compensation |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2011154804A1 (en) * | 2010-06-07 | 2011-12-15 | Universitat Politècnica De Catalunya | Method for estimating the topography of the earth's surface in areas with plant cover |
EP2784537A1 (en) * | 2013-05-15 | 2014-10-01 | Institute of Electronics, Chinese Academy of Sciences | Inversion method and apparatus based on polarimetric interferometric synthetic aperture radar |
CN104361592A (en) * | 2014-11-14 | 2015-02-18 | 中国林业科学研究院资源信息研究所 | Method and device for estimating forest above-ground biomass |
CN109738895A (en) * | 2019-01-31 | 2019-05-10 | 中南大学 | A kind of building and inversion method based on second-order Fourier gear-Legnedre polynomial vegetation height inverse model |
CN110569624A (en) * | 2019-09-20 | 2019-12-13 | 哈尔滨工业大学 | Forest three-layer scattering model determining and analyzing method suitable for PolInSAR inversion |
CN113945927A (en) * | 2021-09-17 | 2022-01-18 | 西南林业大学 | Forest canopy height inversion method through volume scattering optimization |
CN113945926A (en) * | 2021-09-17 | 2022-01-18 | 西南林业大学 | Forest canopy height inversion method improved through underestimation compensation |
CN115062260A (en) * | 2022-06-16 | 2022-09-16 | 电子科技大学 | Forest biomass PolInSAR estimation method and system suitable for heterogeneous forest and storage medium |
CN115079205A (en) * | 2022-06-16 | 2022-09-20 | 电子科技大学 | Multi-baseline forest height inversion method and system for P-band heavy-orbit PolInSAR and storage medium |
-
2023
- 2023-10-16 CN CN202311331655.2A patent/CN117077547B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2011154804A1 (en) * | 2010-06-07 | 2011-12-15 | Universitat Politècnica De Catalunya | Method for estimating the topography of the earth's surface in areas with plant cover |
EP2784537A1 (en) * | 2013-05-15 | 2014-10-01 | Institute of Electronics, Chinese Academy of Sciences | Inversion method and apparatus based on polarimetric interferometric synthetic aperture radar |
CN104361592A (en) * | 2014-11-14 | 2015-02-18 | 中国林业科学研究院资源信息研究所 | Method and device for estimating forest above-ground biomass |
CN109738895A (en) * | 2019-01-31 | 2019-05-10 | 中南大学 | A kind of building and inversion method based on second-order Fourier gear-Legnedre polynomial vegetation height inverse model |
CN110569624A (en) * | 2019-09-20 | 2019-12-13 | 哈尔滨工业大学 | Forest three-layer scattering model determining and analyzing method suitable for PolInSAR inversion |
CN113945927A (en) * | 2021-09-17 | 2022-01-18 | 西南林业大学 | Forest canopy height inversion method through volume scattering optimization |
CN113945926A (en) * | 2021-09-17 | 2022-01-18 | 西南林业大学 | Forest canopy height inversion method improved through underestimation compensation |
CN115062260A (en) * | 2022-06-16 | 2022-09-16 | 电子科技大学 | Forest biomass PolInSAR estimation method and system suitable for heterogeneous forest and storage medium |
CN115079205A (en) * | 2022-06-16 | 2022-09-20 | 电子科技大学 | Multi-baseline forest height inversion method and system for P-band heavy-orbit PolInSAR and storage medium |
Non-Patent Citations (4)
Title |
---|
"Improving Forest Canopy Height Estimation Using a Semi-Empirical Approach to Overcome TomoSAR Phase Errors";Hongbin Luo等;Forests;全文 * |
"基于干涉相位-相干幅度法和基线选择的森林冠层高度反演";罗洪斌;林业资源管理;119-125 * |
基于S-RVoG模型的PolInSAR森林高度非线性复数最小二乘反演算法;解清华;朱建军;汪长城;付海强;张兵;;测绘学报(第10期);1303-1310 * |
干涉、极化干涉SAR技术森林高度估测算法研究进展;张王菲;陈尔学;李增元;赵磊;姬永杰;;遥感技术与应用(第06期);983-997 * |
Also Published As
Publication number | Publication date |
---|---|
CN117077547A (en) | 2023-11-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Medvedev et al. | Relation of internal gravity wave anisotropy with neutral wind characteristics in the upper atmosphere | |
CN117077547B (en) | Forest overground biomass estimation method and system | |
CN102426700B (en) | Level set SAR image segmentation method based on local and global area information | |
Pawluszek-Filipiak et al. | On the importance of train–test split ratio of datasets in automatic landslide detection by supervised classification | |
CN105572736B (en) | A kind of method using seismic data quantitative forecast Slope break belt | |
Nevalainen et al. | Estimating the rut depth by UAV photogrammetry | |
CN111738332B (en) | Underwater multi-source acoustic image substrate classification method and system based on feature level fusion | |
CN113065467B (en) | Satellite image low coherence region identification method and device based on deep learning | |
Wang et al. | A new likelihood function for consistent phase series estimation in distributed scatterer interferometry | |
CN104239901A (en) | Polarized SAR image classification method based on fuzzy particle swarm and target decomposition | |
CN111738278B (en) | Underwater multi-source acoustic image feature extraction method and system | |
Leng et al. | A novel bathymetry signal photon extraction algorithm for photon-counting LiDAR based on adaptive elliptical neighborhood | |
Ma et al. | On the effects of InSAR temporal decorrelation and its implications for land cover classification: The case of the ocean-reclaimed lands of the Shanghai megacity | |
Xu et al. | Feature curve-based registration for airborne LiDAR bathymetry point clouds | |
CN102426699B (en) | Level set synthetic aperture radar (SAR) image segmentation method based on edge and regional information | |
Wang et al. | Hybrid model for estimating forest canopy heights using fused multimodal spaceborne LiDAR data and optical imagery | |
Chen et al. | A high speed method of SMTS | |
Akcay | Landslide fissure inference assessment by ANFIS and logistic regression using UAS-based photogrammetry | |
Mao et al. | Retrieval of boreal forest heights using an improved random volume over ground (RVoG) model based on repeat-pass spaceborne polarimetric SAR interferometry: the case study of Saihanba, China | |
CN117455970B (en) | Airborne laser sounding and multispectral satellite image registration method based on feature fusion | |
CN116817869B (en) | Submarine photon signal determination method using laser radar data | |
Zhou et al. | Automatic reconstruction of 3-D building structures for tomoSAR using neural networks | |
Peng et al. | Underlying topography inversion using TomoSAR based on non-local means for an L-band airborne dataset | |
Managhebi et al. | Forest Height Retrieval Based on the Dual PolInSAR Images | |
Liu et al. | Statistical terrain model with geometric feature detection based on GPU using LiDAR on vehicles |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |