Method for identifying vegetation layer under natural forest and estimating biomass of vegetation layer

Abstract

The invention discloses a natural under-forest vegetation layer identification and biomass estimation method, belongs to the technical field of forest stand investigation, and aims to solve the problem that the accurate identification of under-forest vegetation layers is not facilitated due to the fact that the change of under-forest vegetation height is not considered when vegetation layer division is performed through a simple unified height threshold in the prior art. According to the method, a forest canopy point cloud profile is analyzed, a demarcation point is constructed based on a profile statistical function, and an under-forest vegetation layer is automatically identified; multiple groups of under-forest vegetation layer LiDAR point cloud features are extracted, feature variable optimization and modeling are carried out, and the under-forest vegetation layer biomass is estimated with high precision. According to the method, vertical and horizontal distribution variation of the under-forest vegetation is considered, so that the under-forest vegetation layer identification capability and precision are enhanced. The method can fundamentally improve the depiction capability of the canopy space structure of the point cloud data, and further improves the recognition precision of the under-forest vegetation layer and the estimation precision of the biomass of the under-forest vegetation layer.

Description

Technical Field

[0001] The invention belongs to the technical field of forestry investigation, and in particular relates to a method for identifying a natural understory vegetation layer and estimating its biomass. Background Art

[0002] The understory vegetation layer (such as shrub layer, herb layer, etc.) in the forest is an important part of the forest ecosystem, and the biomass of the understory vegetation layer accounts for about 6.2% of the total forest biomass. The accurate extraction of the understory vegetation layer is of great significance for the precise investigation of forest resources and the measurement of carbon sinks. At the same time, the understory vegetation information is of great significance for analyzing the competition and suitability distribution of vegetation, further understanding the biodiversity and functional structure of forests, and for the sustainable management of forests, community succession analysis, and the study of carbon sink processes in the entire ecosystem. Conventional forest resource surveys are mainly carried out on the tree layer, ignoring the contribution of understory vegetation. At the same time, the extraction of survey parameters mainly relies on field surveys and interpretation of aerial or satellite images, which is inefficient and often has low accuracy, making it difficult to promote practical applications in large areas.

[0003] In recent years, the method of estimating understory vegetation structure parameters based on UAV LiDAR point cloud data has been gradually applied to forest resource surveys and ecological applications. UAV LiDAR data can accurately obtain the three-dimensional structural information inside and outside the forest canopy, which includes the spatial distribution relationship of canopy and understory vegetation materials. At the same time, its high point density characteristics help to accurately identify and separate different vegetation layers under the tree canopy. Identifying and separating the understory vegetation point cloud can effectively eliminate the influence of the tree canopy point cloud and improve the estimation accuracy of understory vegetation structure parameters.

[0004] The research on estimating understory vegetation structural parameters based on point cloud data is as follows: "Under-canopy UAV laser scanning for accurate forest field measurements" published in Volume 164 of ISPRS Journal of Photogrammetry and Remote Sensing in 2020. This existing technology uses drones to fly under the forest to obtain understory vegetation point cloud data, and estimates structural parameters such as single tree diameter at breast height and stem curve by combining single tree detection, point cloud clustering, and principal component analysis. "A method for quantifying understory leaf area index in a temperate forest through combining small footprint full-waveform and point cloud lidar data" published in Volume 13 of Remote Sensing in 2021. This existing technology uses aircraft to obtain LiDAR point clouds, divides the forest canopy into upper canopy and lower canopy based on vertical canopy porosity, and estimates the leaf area index of the lower canopy based on this. “Detection of sub-canopy forest structure using airborne LiDAR” was published in Volume 244 of Remote Sensing of Environment in 2020. This existing technology uses LiDAR point clouds obtained by aircraft and combines the ground-measured Lorey's average tree height to divide the forest canopy into canopy and sub-canopy, and then estimates parameters such as the volume of understory vegetation and the number of trees. However, the above methods are all based on a simple unified height threshold for vegetation layer division, and do not consider the changes in the height of understory vegetation, which is not conducive to the accurate identification of the understory vegetation layer. At the same time, there is no comprehensive and in-depth analysis of the distribution characteristics of the forest canopy in the vertical direction, and no method to accurately identify the understory vegetation layer by combining the vertical profile distribution function. Summary of the invention

[0005] The technical problem solved by the present invention is to provide a method for fundamentally increasing the depiction capability of the canopy spatial structure of point cloud data, thereby improving the accuracy of understory vegetation layer identification and biomass estimation.

[0006] Technical solution: In order to solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0007] The present invention relates to a method for natural forest understory vegetation layer identification and biomass estimation based on UAV LiDAR point cloud, specifically, it refers to a method for generating a digital terrain model by performing flight strip splicing, denoising, ground point identification, and interpolation on UAV LiDAR point cloud; then, segmenting the LiDAR point cloud based on a three-dimensional voxel framework, combining Gaussian density distribution, multi-peak Weibull distribution and Gamma mixed distribution function, and performing multi-scale profile function fitting on the point cloud in each voxel; finally, constructing a demarcation point automatic identification (DPI) method, combining profile function and curve shape index to accurately identify the upper tree and lower vegetation point cloud in natural forests, and constructing an understory vegetation biomass estimation model through the understory vegetation point cloud features. It includes the following steps:

[0008] S1: Collect lidar data with the help of drone platform;

[0009] S2: remote sensing data preprocessing;

[0010] S3: constructing a vertical profile function model of forest point cloud;

[0011] S4: vertical stratification of forest;

[0012] S5: Extracting the underlying vegetation LiDAR point cloud feature variables;

[0013] S6: Screening characteristic variables through correlation and collinearity analysis;

[0014] S7: The measured understory vegetation biomass was used as the dependent variable, and the screened characteristic variables were used as independent variables to establish a multivariate linear regression model.

[0015] Furthermore, in step S1, a survey plot is set up on the ground, and the tree species, height, breast diameter, and ground diameter of the lower vegetation in the plot are recorded, and the aboveground biomass of the lower vegetation is calculated according to the allometric growth equation of the corresponding tree species.

[0016] Furthermore, in step S2, the single-strip LiDAR data of the drone is firstly strip-matched and spliced, and the noise points in the point cloud are identified based on the neighborhood search algorithm to remove the noise. At the same time, the improved progressive triangulation filtering algorithm is used to identify the ground points in the point cloud, and the digital elevation model is generated by calculating the average value of the laser point height in each pixel.

[0017] Furthermore, the implementation of step S3 includes the following steps:

[0018] S31: Using a voxel framework to divide the forest point cloud into several vertical voxels, and using DEM to perform height normalization on the point cloud in each voxel;

[0019] S32: using a Gaussian probability density function to fit a vertical profile of the point cloud in each voxel, and constructing a Weibull distribution function of the corresponding peak number according to the number of peaks in the vertical profile;

[0020] S33: For the point cloud within the first wavelength range of the Weibull distribution function, a detailed point cloud vertical profile function model in each voxel is constructed through the Gamma mixed distribution function.

[0021] Furthermore, in step S4, the first local minimum point and the last local maximum point in the point cloud vertical profile function model within each voxel are used as the boundary points of the understory vegetation layer-ground and the upper trees-understory vegetation layer, respectively, to perform stratification of the upper trees-lower vegetation-ground layers.

[0022] Furthermore, in step S5, for the identified and separated lower vegetation point cloud, n point cloud structure characteristic variables are extracted, mainly including: canopy height distribution percentiles; the percentage of point clouds with the number of points above each percent height in all point clouds; the average and maximum heights; the coefficient of variation of the canopy point cloud distribution; the volume proportion of each canopy structure category, including the four canopy structure categories of open layer, light-transmitting layer, low-light layer and closed layer, and the volume percentage of each canopy structure category.

[0023] Furthermore, in step S5, the correlation matrix is ​​first calculated by principal component analysis to obtain the control factor k as the collinearity index to screen the characteristic variables with collinearity between characteristic variables lower than 30, and then the characteristic variables with correlation with the biomass of the lower vegetation higher than 0.8 are screened by correlation.

[0024] Furthermore, in step S6, the correlation matrix is ​​first calculated by principal component analysis to obtain the control factor k as the collinearity index to screen the characteristic variables with collinearity between characteristic variables less than 30, and then the characteristic variables with correlation with the biomass of the lower vegetation higher than 0.8 are screened by correlation.

[0025] Furthermore, in step S7, the stepwise regression method is used to select the variables to be included in the model, and the best model is selected.

[0026] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0027] 1) Previous methods all divide the understory vegetation layer based on a single fixed height threshold, while the present invention analyzes the forest canopy point cloud profile in depth, and then constructs a demarcation point recognition method based on the profile statistical function to automatically identify the understory vegetation layer. Since the present invention takes into account the vertical and horizontal distribution variation of understory vegetation, the present invention enhances the ability and accuracy of understory vegetation layer recognition.

[0028] 2) Previous methods did not use UAV LiDAR technology that can record the spatial distribution of materials inside the forest canopy in detail. Therefore, the present invention can fundamentally increase the ability to depict the canopy spatial structure of point cloud data, thereby improving the accuracy of identifying the understory vegetation layer and estimating its biomass.

[0029] 3) Based on the identification and separation of the understory vegetation layer, the present invention comprehensively and deeply extracts multiple groups of LiDAR point cloud features of the understory vegetation layer, and performs feature variable optimization and modeling, thereby estimating the biomass of the understory vegetation layer with high precision. At the same time, the invention is not only conducive to the identification and separation of the understory vegetation layer, but also eliminates the interference of the upper canopy trees from a mechanism perspective, which is conducive to the analysis and modeling of the understory vegetation in the forest, and is also easy to transplant the method (that is, it can also be applied to different forest types in different regions).

[0030] 4) The verification results show that the relative root mean square error of the aboveground biomass of vegetation in the natural forest understory estimated by the present invention is improved by more than 10% compared with other similar remote sensing methods for estimating the aboveground biomass of the forest understory. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is the regional lidar point cloud data map after being decomposed by the voxel framework;

[0032] Figure 2 It is the point cloud map after DEM height normalization in each voxel;

[0033] Figure 3 It is a schematic diagram of the vertical distribution profile function model of the point cloud within the voxel;

[0034] Figure 4 It is a schematic diagram of the vertical stratification of the forest;

[0035] Figure 5 It is a point cloud of trees, bamboos and ground identified by vertical stratification within the voxel;

[0036] Figure 6 It is a rendering of the tree layer, bamboo layer and ground layer of the regional forest identified by the present invention;

[0037] Figure 7 It is a correlation diagram between each characteristic variable and structural parameter of the understory bamboo point cloud of the present invention. DETAILED DESCRIPTION

[0038] The present invention is further illustrated below in conjunction with specific examples. The examples are implemented based on the technical solutions of the present invention. It should be understood that these examples are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0039] Example 1

[0040] The method for identifying the natural understory vegetation layer and estimating its biomass of the natural understory vegetation layer of the present invention performs flight strip stitching, denoising, ground point identification, and interpolation on the UAV LiDAR point cloud to generate a digital terrain model, segments the LiDAR point cloud based on a three-dimensional voxel framework, and combines Gaussian density distribution, multi-peak Weibull distribution and Gamma mixed distribution functions to construct a multi-scale profile function model of the point cloud within the voxel.

[0041] A demarcation point identification (DPI) method was constructed, and the profile model and curve shape index were combined to accurately identify the point clouds of upper-layer trees and underlayer vegetation in natural forests. A biomass estimation model of underlayer vegetation was constructed based on the point cloud features of underlayer vegetation.

[0042] Specifically including steps S1 to S7:

[0043] S1: Collect lidar data with the help of drone platform.

[0044] Survey plots were set up on the ground to record the tree species, height, breast diameter and ground diameter of the lower vegetation in the plots, and the aboveground biomass of the lower vegetation was calculated based on the allometric growth equation of the corresponding tree species.

[0045] S2: Remote sensing data preprocessing.

[0046] The existing technology is used for preprocessing remote sensing data. First, the single-strip LiDAR data of the UAV is strip-matched and spliced, and the noise points in the point cloud are identified based on the neighborhood search algorithm to remove the noise (Liu et al., 2018). At the same time, the improved progressive triangulation filtering algorithm is used to identify the ground points in the point cloud, and the digital elevation model (DEM) (spatial resolution is 0.1m) is generated by calculating the average height of the laser point in each pixel.

[0047] S3: Construct a vertical profile function model of the forest point cloud. The specific contents are as follows:

[0048] S31: First, the forest point cloud is divided into several vertical voxels using the voxel framework, and the point cloud in each voxel is highly normalized using the DEM with a spatial resolution of 0.1 m;

[0049] S32: Then, a Gaussian probability density function is used to fit the vertical profile of the point cloud in each voxel, and a Weibull distribution function of the corresponding peak number is constructed according to the number of peaks in the vertical profile;

[0050] The Gaussian probability density function is:

[0051]

[0052] Where h is a height random variable, α and β are the mean and standard deviation respectively.

[0053] The multimodal Weibull distribution function is:

[0054]

[0055] Where h is a height random variable, α and β are shape and scale parameters, respectively.

[0056] S33: Finally, for the point cloud within the first wavelength range of the Weibull distribution function, a fine point cloud vertical profile function model in each voxel is constructed through the Gamma mixed distribution function:

[0057]

[0058] Where h is a height random variable, and α and β are shape parameters.

[0059] S4: Forest vertical stratification, the details are as follows:

[0060] The first local minimum point and the last local maximum point in the point cloud vertical profile function model in each voxel are used as the boundary points of the understory vegetation layer-ground and the upper tree-understory vegetation layer, respectively, to perform the upper tree-understory vegetation-ground stratification. This can then realize the identification and separation of the upper tree and understory vegetation layers in the entire region.

[0061] S5: Extract the underlying vegetation LiDAR point cloud feature variables.

[0062] For the identified and separated understory vegetation point cloud, 16 point cloud structure characteristic variables were extracted, including: canopy height distribution percentile (25 th ,50 th ,75 th ,95 th ); the number of point clouds at each percent height 10 th , 30 th , 50 th , 70 th , 90 th , that is, the percentage of points above d1, d3, d5, d7, d9) in all point clouds; the average and maximum height (h mean ,h max ); Coefficient of variation of canopy point cloud distribution (H cv ); The volume percentage of each canopy structure category, including the open layer, light-transmitting layer, low-light layer and closed layer, and the volume percentage of each canopy structure category (i.e. O g ,E,C g ,O).

[0063] S6: Screen characteristic variables through correlation and collinearity analysis.

[0064] Firstly, the correlation matrix was calculated by principal component analysis to obtain the control factor k (i.e., the ratio of the square root of the largest characteristic root to the smallest characteristic root) as the collinearity index to screen the characteristic variables with collinearity less than 30 between characteristic variables, and then the characteristic variables with correlation higher than 0.8 with the biomass of understory vegetation were screened by correlation.

[0065] S7: The measured understory vegetation biomass was used as the dependent variable, and the screened characteristic variables were used as independent variables to establish a multivariate linear regression model.

[0066] The stepwise regression method was used to select the variables entering the model, and the Akaike information criterion (AIC) was used to optimize the best model, that is, the model with the lowest AIC was the best model.

[0067] The present invention adopts the coefficient of determination (R 2 ), root mean square error (RMSE) and relative root mean square error (rRMSE) were used to evaluate the effect of regression model fitting and estimation accuracy:

[0068]

[0069]

[0070] In the formula, x i is the measured value of aboveground biomass of understory vegetation; is the measured average aboveground biomass of understory vegetation; is the model estimated value of the aboveground biomass of understory vegetation; n is the number of sample plots; i is a specific sample plot.

[0071] Example 2

[0072] The study area is located in Foping Giant Panda National Nature Reserve, Shaanxi Province (107°47′39.59″E, 33°38′53.55″N), with an area of ​​approximately 290km 2, with an elevation range of 980-2900m. The forest type in the study area belongs to natural forest, which can be subdivided into two types of forest stands dominated by coniferous species and broad-leaved species. The main coniferous and broad-leaved deciduous species include Pinus tabulaeformis, Tsuga chinensis, Oak (Quercus variabilis) and Quercusaliena var.acutiserrata. The understory vegetation is mainly Qinling Arrow Bamboo (Fargesia qinlingensis) and Bashan Wood Bamboo (Arundinaria fargesii). Among them, Qinling Arrow Bamboo is mainly distributed in areas above 1700m above sea level, while Bashan Wood Bamboo is mainly distributed in the forest below 1900m above sea level. These two types of bamboo, as the main food source of giant pandas, together with the upper trees constitute a complex natural forest ecosystem.

[0073] 52 square plots (20×20m 2 ), including 22 Qinling Arrow Bamboo plots and 30 Wood Bamboo plots. The coordinates of the center point of the plot were determined by differential GPS, which was positioned by receiving Qianxun differential signals with an accuracy better than 0.5m. The tree species, height, breast diameter, and ground diameter of the understory vegetation in the plot were recorded, and the aboveground biomass of the understory vegetation was calculated according to the allometric growth equation of the corresponding tree species. The plot-scale understory bamboo structural parameters summarized based on the ground survey are shown in Table 1.

[0074] Table 1 Summary of the characteristics of understory vegetation measured in the sample plots

[0075]

[0076] LiDAR data was collected with the help of UAV platform. Firstly, strip matching and splicing of single strip LiDAR data of UAV were performed, and noise points in point cloud were identified based on neighborhood search algorithm to remove noise. At the same time, the improved progressive triangulation filtering algorithm was used to identify ground points in point cloud, and the digital elevation model (DEM) (spatial resolution of 0.1m) was generated by calculating the average height of laser points in each pixel.

[0077] The vertical profile function model of forest point cloud is constructed based on Gaussian probability density function, multi-peak Weibull distribution function and Gamma mixed distribution function. Firstly, the voxel framework is used to divide the forest point cloud into several vertical voxels, and the point cloud in each voxel is highly normalized using the DEM with a spatial resolution of 0.1m; then, the point cloud in each voxel is fitted with a vertical profile, and the Weibull distribution function with the corresponding peak number is constructed according to the number of vertical profile peaks; finally, for the point cloud within the first wavelength range of the Weibull distribution function, the Gamma mixed distribution function is used to construct a fine point cloud vertical profile function model in each voxel. The first local minimum point and the last local maximum point in the point cloud vertical profile function model in each voxel are used as the boundary points of the understory vegetation layer-ground and the upper tree-understory vegetation layer, respectively, to perform the stratification of the upper tree-understory vegetation-ground layer. In this way, the identification and separation of the upper tree and understory vegetation layers in the entire area can be achieved.

[0078] For the identified and separated lower vegetation point cloud, 16 point cloud structure characteristic variables were extracted, including: canopy height distribution percentile (25 th ,50 th ,75 th ,95 th ); the number of point clouds at each percent height 10 th , 30 th , 50 th , 70 th , 90 th , that is, the percentage of points above d1, d3, d5, d7, d9) in all point clouds; the average and maximum height (h mean ,h max ); Coefficient of variation of canopy point cloud distribution (H cv ); The volume percentage of each canopy structure category, including the open layer, light-transmitting layer, low-light layer and closed layer, and the volume percentage of each canopy structure category (i.e. O g ,E,C g ,O).

[0079] Characteristic variables were screened by correlation and collinearity analysis. First, the correlation matrix was calculated by principal component analysis to obtain the control factor k (i.e., the ratio of the square root of the largest characteristic root to the smallest characteristic root) as the collinearity index to screen characteristic variables with collinearity less than 30 between characteristic variables. Then, characteristic variables with correlation between characteristic variables and the biomass of understory vegetation higher than 0.8 were screened by correlation.

[0080] The measured understory vegetation biomass was used as the dependent variable, and the characteristic variables after screening were used as independent variables to establish a multiple linear regression model. The variables entering the model were selected using the stepwise regression method, and the Akaike information criterion (AIC) was used to select the best model, that is, the model with the lowest AIC was the best model. The present invention uses the coefficient of determination (R 2 ), root mean square error (RMSE) and relative root mean square error (rRMSE) were used to evaluate the effect of regression model fitting and estimation accuracy:

[0081]

[0082]

[0083] In the formula, x i is the measured value of aboveground biomass of understory vegetation; is the measured average aboveground biomass of understory vegetation; is the model estimated value of the aboveground biomass of understory vegetation; n is the number of sample plots; i is a specific sample plot.

[0084] The aboveground biomass estimation model of understory bamboo and its prediction accuracy are shown in Table 2. The correlation between the characteristic variables and structural parameters of the understory bamboo point cloud is shown in Figure 2 .

[0085] Table 2 Estimation model of aboveground biomass of understory vegetation and model prediction accuracy

[0086] Model formula <![CDATA[ a R 2 ]]> RMSE rRMSE(%) k Aboveground biomass <![CDATA[exp(-0.52+3.34lnh mean +0.42lnE)×1.081]]> <![CDATA[0.85 *** ]]> 8.63mg / ha 29.41 3.43

[0087] Note: h mean is the average height of the bamboo point cloud in the sample plot, and E is the volume ratio of the light-transmitting space in the canopy of the bamboo point cloud in the forest.

[0088] Verification results show that the determination coefficient error of the aboveground biomass of vegetation in the natural forest understory estimated by the present invention is improved by 10% or more compared with other similar remote sensing methods for estimating the aboveground biomass of the forest understory.

[0089] Table 3 Comparison of aboveground biomass estimation accuracy of different methods

[0090]

[0091] The present invention can be applied in the fields of fine-scale forest resource survey and forest carbon sink measurement, which is conducive to in-depth understanding of the relationship between vegetation growth and spatial distribution within the forest and the environment, and further grasping the growth and competition laws of different vegetation layers in the vertical direction within the forest ecosystem. It is of great significance for the sustainable management of forests, community succession analysis and research on the carbon sink process of the entire ecosystem.

[0092] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for identifying a natural understory vegetation layer and estimating its biomass, characterized in that: Analyze the forest canopy point cloud profile, construct dividing points based on the profile statistical function, and automatically identify the understory vegetation layer; extract multiple groups of LiDAR point cloud features of the understory vegetation layer, and perform feature variable optimization and modeling to estimate the biomass of the understory vegetation layer with high precision.

2. The method for identifying the natural understory vegetation layer and estimating its biomass according to claim 1, characterized in that: The following steps are involved: S1: Collect lidar data with the help of drone platform; S2: remote sensing data preprocessing; S3: constructing a vertical profile function model of forest point cloud; S4: vertical stratification of forest; S5: Extracting the underlying vegetation LiDAR point cloud feature variables; S6: Screening characteristic variables through correlation and collinearity analysis; S7: The measured understory vegetation biomass was used as the dependent variable, and the screened characteristic variables were used as independent variables to establish a multivariate linear regression model.

3. The method for identifying the natural understory vegetation layer and estimating its biomass according to claim 2, characterized in that: In step S1, a survey plot is set up on the ground, and the tree species, height, breast diameter, and ground diameter of the lower vegetation in the plot are recorded, and the aboveground biomass of the lower vegetation is counted according to the allometric growth equation of the corresponding tree species.

4. The method for identifying the natural understory vegetation layer and estimating its biomass according to claim 2, characterized in that: In step S2, the single-strip LiDAR data of the drone is firstly strip-matched and spliced, and the noise points in the point cloud are identified based on the neighborhood search algorithm to remove the noise. At the same time, the improved progressive triangulation filtering algorithm is used to identify the ground points in the point cloud, and the digital elevation model is generated by calculating the average value of the laser point height in each pixel.

5. The method for identifying the natural understory vegetation layer and estimating its biomass according to claim 2, characterized in that: The implementation in step S3 includes the following steps: S31: Using a voxel framework to divide the forest point cloud into several vertical voxels, and using DEM to perform height normalization on the point cloud in each voxel; S32: using a Gaussian probability density function to fit a vertical profile of the point cloud in each voxel, and constructing a Weibull distribution function of the corresponding peak number according to the number of peaks in the vertical profile; S33: For the point cloud within the first wavelength range of the Weibull distribution function, a detailed point cloud vertical profile function model in each voxel is constructed through the Gamma mixed distribution function.

6. The method for identifying the natural understory vegetation layer and estimating its biomass according to claim 1, characterized in that: In step S4, the first local minimum point and the last local maximum point in the point cloud vertical profile function model within each voxel are used as the boundary points of the understory vegetation layer-ground and the upper tree-understory vegetation layer, respectively, to perform stratification of the upper tree-lower vegetation-ground layers.

7. The method for identifying the natural understory vegetation layer and estimating its biomass according to claim 2, characterized in that: In step S5, for the identified and separated lower vegetation point cloud, n point cloud structure characteristic variables are extracted, mainly including: canopy height distribution percentiles; the percentage of point clouds with the number of points above each percent height in all point clouds; the average and maximum heights; the coefficient of variation of the canopy point cloud distribution; the volume proportion of each canopy structure category, including the four canopy structure categories of open layer, light-transmitting layer, low-light layer and closed layer, and the volume percentage of each canopy structure category.

8. The method for identifying the natural understory vegetation layer and estimating its biomass according to claim 2, characterized in that: In step S5, the correlation matrix is ​​first calculated by principal component analysis to obtain the control factor k as the collinearity index to screen the characteristic variables with collinearity less than 30 between the characteristic variables, and then the characteristic variables with correlation with the biomass of the lower vegetation higher than 0.8 are screened by correlation.

9. The method for identifying the natural understory vegetation layer and estimating its biomass according to claim 2, characterized in that: In step S6, the correlation matrix is ​​first calculated by principal component analysis to obtain the control factor k as the collinearity index to screen the characteristic variables with collinearity less than 30 between the characteristic variables, and then the characteristic variables with correlation with the biomass of the lower vegetation higher than 0.8 are screened by correlation.

10. The method for identifying the natural understory vegetation layer and estimating its biomass according to claim 3, characterized in that: In step S7, the stepwise regression method is used to select the variables to be included in the model, and the best model is selected.

Images