CN104656098B - A kind of method of remote sensing forest biomass inverting - Google Patents
A kind of method of remote sensing forest biomass inverting Download PDFInfo
- Publication number
- CN104656098B CN104656098B CN201510056042.1A CN201510056042A CN104656098B CN 104656098 B CN104656098 B CN 104656098B CN 201510056042 A CN201510056042 A CN 201510056042A CN 104656098 B CN104656098 B CN 104656098B
- Authority
- CN
- China
- Prior art keywords
- forest
- variable
- biomass
- variables
- value
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
- G01S17/89—Lidar systems specially adapted for specific applications for mapping or imaging
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Electromagnetism (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of method of remote sensing forest biomass inverting, this method is on the basis of remotely-sensed data is pre-processed, respectively from LiDAR point cloud（Include canopy three-dimensional spatial information）It is and multispectral（Include the spectral information of canopy upper surface）The characteristic variable of extracting data Vegetation canopy；Above LiDAR point cloud and multispectral characteristic variable are screened by correlation analysis, and combined ground actual measurement biomass information passes through Gradual regression analysis model inverting biomass above and below the ground.The Optimization inversion model of the north subtropical forest biomass built by the present invention can be by model " coefficient of determination " R2Improve 3 24%；And forest biomass can be estimated in high precision, 2 10% will be reduced by rRMSE " with respect to root-mean-square error ".It can be applicable to the field such as research of Forestry Investigation, forest resource monitoring, forest carbon storage assessment and forest ecosystem, and the data support of quantification provided for the comprehensive utilization of forest sustainable management and the forest reserves.
Description
Technical Field
The invention relates to the technical field of monitoring and pollution-free control of forest pests, in particular to a method for inverting remote sensing forest biomass.
Background
Accurate biomass estimation has important significance for forest resource monitoring, forest carbon reserve assessment and forest ecosystem research. Meanwhile, the information can also provide quantitative data support for forest sustainable management and comprehensive utilization of forest resources. The traditional biomass investigation method is time-consuming and labor-consuming, can only obtain limited information on points, and is difficult to be practically popularized in a large area; the remote sensing technology can accurately and quickly acquire forest parameters of all scales, and has good practical value and application prospect.
The Landsat series of satellites can acquire multispectral (optical) information of the forest on medium and large scales. Compared with the previous sensors such as TM, the latest Landsat 8OLI sensor (which is mounted on the Landsat 8 satellite launched in 2013, 2, 11 and by the United states space and aviation administration (NASA)) has a great improvement in the arrangement of wave bands and the sensitivity to vegetation. However, optical remote sensing is still difficult to penetrate through forest canopies to obtain vertical structure information of the forest canopies, and forest biomass information is easily saturated when the forest biomass information is obtained in an area with high forest coverage (vegetation is vigorous). LiDAR (Light Detection And Ranging) is an active remote sensing technology which is rapidly developed in recent years, laser pulses emitted by the LiDAR can penetrate through vegetation canopies to obtain three-dimensional structure And energy information of the vegetation canopies, and previous researches show that the LiDAR has great potential in accurately estimating biophysical And structural characteristics of different forest types.
In recent years, the research of inverting forest biomass is as follows: 1) Zheng et al 2004 published "Estimating aboveground biobased using land and sat7ETM + data across a managed land and southern Wisconsin, USA" on Remote Sensing of Environment, volume 35, which inverted the aboveground biomass information of northern coniferous forest, wisconsin, USA by means of a vegetation index such as NDVI extracted from ETM + (Landsat 7) images. 2) The ' correlation analysis between tropical forest vegetation biomass of different age groups and remote sensing geological data ' published in 2004 journal of the plant ecology bulletin the Yangzhou, ming et al ' for estimating the tropical forest vegetation biomass of the West double-Pana of Yunnan by using a TM (Landsat 5) image original wave band method. 3) Ferster et al, 2009, published "above-bound Large Tree Mass Estimation in a Coastal Format in British Columbia Using Plot-Level measurements and Industrial Tree Detection from LiDAR" on Canadian Journal of Remote Sensing, volume 35, by means of a small spot (diameter: 0.1-2 m) and the height of the point cloud and the density of the canopy extracted from the LiDAR data, the aboveground biomass of the temperate forest is inverted. 4) Saatchi et al 2011 published "Benchmark Map of Forest Carbon Stocks in thermoplastic Regions acids Three contacts" at "Proceedings of the National Academy of Sciences of the United States of America" at stage 12 by the method of measuring large spots (diameter: 52-90 m) LiDAR data, extracting characteristic variables related to tree height information, and inverting the biomass of tropical rainforests. However, the above methods only mine traditional "optical" and LiDAR data from a single angle, with low specificity, and shallow mining depths for feature variables (i.e., feature variables are not systematically grouped and extracted and screened from multiple angles), and are not able to accurately invert biomass.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the defects in the prior art, the invention provides a remote sensing forest biomass inversion method which has the characteristics of strong specificity, low cost, easiness in popularization and application and the like.
The technical scheme is as follows: in order to achieve the purpose of the invention, the invention adopts the technical scheme that:
a method for inverting remote sensing forest biomass comprises the following steps: on the basis of preprocessing remote sensing data, respectively extracting characteristic variables of vegetation canopies from LiDAR point cloud (containing three-dimensional space information of the canopies) and multispectral (containing spectral information of the upper surfaces of the canopies) data; and (3) screening the LiDAR point cloud and the multispectral characteristic variables through correlation analysis, and inverting the aboveground and underground biomass through a stepwise regression model by combining with the ground actual measurement biomass information.
The remote sensing forest biomass inversion method comprises the following steps:
1) OLI image preprocessing: firstly, carrying out radiometric calibration on an original image by means of radiometric calibration parameters of an OLI sensor; converting the original DN value into a pixel radiation brightness value; then, carrying out atmospheric correction on the image by using a FLAASH model so as to convert the radiation brightness value into the actual reflectivity of the earth surface; then, geometrically and finely correcting the image, selecting the same-name feature points, correcting by adopting a quadratic polynomial, controlling the correction error within 0.1 pixel, and resampling by adopting a nearest pixel method.
2) LiDAR data preprocessing:
a) Noise level estimation and data smoothing: firstly, converting original data into a frequency domain, and then taking a low value part with higher frequency as a judgment standard of a noise level; then, a Gaussian filter is selected for smoothing, because the Gaussian filter can effectively smooth data and keep the trend of the original curve to the maximum extent;
b) Gaussian fitting (decomposition) and waveform data point clouding: fitting the waveform data by adopting a nonlinear least square method on the basis of the assumption that the echo data is the accumulation of a plurality of Gaussian functions; then extracting discrete point clouds from the processed waveform data through a local maximum peak detection filtering algorithm, wherein the energy and amplitude information of a return signal is recorded in each discrete point;
c) Generating a digital terrain: the purpose of LiDAR data height normalization is to obtain a "true" vegetation height from which the terrain effect is removed, typically by subtracting the terrain height from the original LiDAR data height information; therefore, accurate generation of a Digital Terrain Model (DTM) is an important prerequisite for calculating the normalized vegetation height; firstly, extracting discrete point clouds from waveform data to classify the discrete point clouds, then carrying out Kraus filtering processing on a last echo to remove non-ground points, and finally using the filtered last echo data and generating a digital terrain model by means of natural proximity interpolation;
d) Then, normalizing the elevation of the vegetation echo point by utilizing a DEM (digital elevation model), namely converting the height value of the vegetation point into a height value relative to the ground, and cutting each sample land through coordinates of a lower left corner and an upper right corner; finally, extracting normalized point cloud data of the coordinate positions corresponding to the 55 sample plots by a GIS analysis tool;
3) OLI characteristic variable extraction: extracting 5 groups of characteristic variables (detailed in table 1), namely original single-waveband variables, waveband combination variables, information enhancement group variables, vegetation index variables and texture information variables, by performing waveband combination, tassel cap transformation, texture information extraction, principal component analysis, minimum noise separation transformation and various vegetation index transformations on the OLI image; wherein the texture analysis is performed on a first principal component of the principal component analysis;
4) Extracted LiDAR point cloud feature variables: the LiDAR point cloud characteristic variables are 4 groups of characteristic variables (detailed in a table 2) calculated on the basis of three-dimensional normalized LiDAR point cloud values, namely height variables, height percentile variables, canopy density variables and canopy coverage variables;
5) Screening characteristic variables: performing Pearson's correlation analysis on the extracted LiDAR characteristic variables and OLI characteristic variables and parameters to be predicted, and selecting the characteristic variables of which the absolute values of Pearson's correlation coefficients are higher than 0.2 as modeling candidate variables; the calculation method of the Pearson's correlation coefficient comprises the following steps:
in the formula, x i For a certain stand characteristic measured on the ground, y i For a certain variable of the LiDAR characteristics,is x i Is determined by the average value of (a),is y i Average value of (a);
6) Statistical analysis: and (3) establishing a multivariate regression model by taking biomass information obtained by ground actual measurement and summary as a dependent variable and characteristic variables extracted by a remote sensing method as independent variables. Using stepwise entry method (stepwise) and checking the coefficient of determination (R) 2 ) To select appropriate variables to enter the model; if there is an independent variable, the statistic F is too small and T test does not reach significant level (P value)>, 0.1), then removing; f value is large and T test reaches a significant level (P value)< 0.05) then get in; using a coefficient of determination (R) 2 ) Root Mean Square Error (RMSE), and relative Root Mean Square Error (RMSE) to evaluate the accuracy of the regression model;
in the formula, x i Is a certain forest stand characteristic measured on the ground,is x i Is determined by the average value of (a) of (b),a certain forest stand characteristic estimated by the model, wherein n is the number of sample plots;
as shown in the formula (3), rRMSE is the RMSE (root mean square error) and(mean of measured values).
The invention discloses a remote sensing forest biomass inversion method, which extracts characteristic variables of a vegetation canopy from point cloud (comprising three-dimensional space information of the canopy) and multispectral (comprising spectral information of the upper surface of the canopy) data respectively on the basis of preprocessing LiDAR and OLI data, and then inverts aboveground and underground biomass through a stepwise regression model by combining with actually measured biomass information on the ground on the basis of screening the characteristic variables, wherein the innovation points and the characteristics are as follows: 1) Fully mining the biological and physical characteristic information of the vegetation in the canopy three-dimensional space information and the spectral information of the upper surface of the canopy in two dimensions; 2) The extracted characteristic variables are screened by means of correlation analysis and used for a final inversion model, so that mechanism explanation and method transplantation are facilitated; 3) Since the OLI data is free data, the method also saves the cost of increasing the biomass inversion accuracy over multiple scales to a large extent.
Has the advantages that: compared with the prior art, the invention fully excavates vegetation biophysical characteristic information contained in the canopy three-dimensional space information and the spectral information dimension of the upper surface of the canopy three-dimensional space information by integrating the LiDAR point cloud and the OLI multispectral characteristic variables, screens the extracted characteristic variables by means of correlation analysis, and finally determines the inversion model. The method is beneficial to mechanism explanation and method transplantation; and because the OLI data are free data, the method also greatly saves the cost of increasing the biomass inversion accuracy on multiple scales. The early-stage experimental verification result shows that the optimized inversion model of the northern subtropical forest biomass (compared with a method based on a single multispectral data source) constructed by the method can determine the coefficient R of the model 2 The improvement is 3 to 24 percent; and forest biomass can be estimated with high precision (reducing the relative root mean square error rRMSE by 2-10%). The method can be applied to the fields of forestry investigation, forest resource monitoring, forest carbon reserve assessment, forest ecosystem research and the like, and provides quantitative data support for forest sustainable management and comprehensive utilization of forest resources.
Drawings
FIG. 1 is a plot of a true color aerial photograph and 55 sample plots of a study area;
FIG. 2 is a method technical roadmap for remote sensing forest biomass inversion;
FIG. 3 is a plot of Pearson's correlation coefficients for various characteristic variables with aboveground biomass;
FIG. 4 is a plot of Pearson's correlation coefficients for various characteristic variables with subsurface biomass;
FIG. 5 is a comparative scattergram of measured values of an OLI model-above ground biomass plot and predicted values of the model;
FIG. 6 is a scatter plot of OLI model-measured values of subterranean biomass plots versus predicted values of the model;
FIG. 7 is a comparative scatter plot of LiDAR model-aboveground biomass plot actual measurements versus model predicted values;
FIG. 8 is a scatter plot of LiDAR model-measured value of a subterranean biomass plot versus predicted value for the model;
FIG. 9 is a comparative scattergram of measured values and predicted values of a model for an integrated model-aboveground biomass plot;
FIG. 10 is a scatter plot comparing model-subterranean biomass sample plot measured values to model predicted values.
Detailed Description
The present invention will be further described with reference to the following examples.
Example 1
A method for inverting biomass of remote sensing forest is disclosed, the technical route is shown in figure 1, the method is a method for inverting biomass of remote sensing forest integrating LiDAR point cloud and OLI multispectral characteristic variable, on the basis of preprocessing remote sensing data, the characteristic variable of vegetation canopy is extracted from LiDAR point cloud (containing canopy three-dimensional space information) and multispectral (containing spectral information of upper surface of canopy) data respectively; and (3) screening the LiDAR point cloud and the multispectral characteristic variables through correlation analysis, and inverting the aboveground and underground biomass through a stepwise regression model by combining with the actually measured biomass information on the ground. The method comprises the following specific steps:
the research district is located in China-Yu mountain forest farm (120 DEG 42 '9.4' E,31 DEG 40 '4.1' N) of mature city in Jiangsu province, belongs to subtropical monsoon climate, has mild climate, annual average precipitation amount of 1054 mm and area of about 1103hm 2 The altitude is 20-261m. The corn poppy field belongs to northern subtropical secondary commingled forests, and the main forest types are coniferous forests, broad-leaved forests and commingled forests, wherein the main conifer species are Pinus massoniana (Pinus massoniana), cedar wood (Cunninghamia lancelolia), slash pine (Pinus elliottii), black pine (Pinus thunbergii) and the like; the main broad-leaved tree species are Quercus acutissima, sweet maple (Liquidambar formosan) and some evergreen broad-leaved tree species, such as Fagaceae (Fagaceae), lauraceae (Lauraceae) and Theaceae (Theaceae).
55 square plots (size: 30 x 30m, setting time: 7-8 months in 2012 and 8 months in 2013) are set in the range of the research area according to indexes such as forest type, age and site index in three types of survey history data (1995, 2012). The sample plot is divided into three types of coniferous forest (n = 13), broad-leaved forest (n = 16) and mixed forest (n = 26) according to the tree species composition ratio. During the investigation of the sample plot, for trees with a breast diameter greater than 5cm, the tree species, breast diameter (measured with girth), tree height and under-branch height (measured with a Vertex IV laser altimeter) and crown width (i.e. the projected distance in both main directions, measured with a tape) of a single tree are measured one by one, and the trees with a breast diameter less than 5cm and dead are counted, but not involved in the calculation of biomass. The southwest angular position is likewise determined using differential GPS (Trimble GeoXH6000Handheld GPS units) with an accuracy better than 0.5 m by receiving JSCORS wide area differential signal positioning (fig. 2).
Collecting relevant forest parameters of the sample plot scale according to the single-tree survey data, wherein the relevant forest parameters comprise the aboveground biomass and underground biomass (t.hm) per unit area on the sample plot scale -2 ). Biomass information Individual biomass was calculated by the equation of growth at different rates (following the near principle) and summarized for eachBiomass (W) per unit area of block plot A ) And underground biomass (W) B )。
And (4) preprocessing the OLI image. The 2-7 wave bands in the Landsat 8OLI image (band number 119/38) of 19/7 in 2013 are adopted for research, the spatial resolution of the image is 30m, the radiation resolution is 12bit, and the spectral range covers 11 wave bands. Firstly, the original image is radiometrically calibrated by means of radiometric calibration parameters of the OLI sensor. And converting the original DN value into a pixel radiation brightness value. And then carrying out atmospheric correction on the image by using a FLAASH model (atmospheric model: tropical; aerosol model: urban; aerosol inversion method: 2-Band (K-T); initial visibility: 40; and spectral response function: ldcm _ oli. Sli), thereby converting the radiation brightness value into the actual surface reflectivity. Then, geometric fine correction is carried out on the image, 40 homonymy ground object points are selected, quadratic polynomial is adopted for correction, the correction error is controlled within 0.1 pixel, and the nearest pixel method is adopted for resampling.
LiDAR data preprocessing. 1) Noise level estimation and data smoothing: the original data is first converted into the frequency domain, and the lower value part with higher frequency is used as the judgment standard of the noise level. And then, a Gaussian filter is selected for smoothing, because the Gaussian filter can keep the trend of the original curve to the maximum extent while effectively smoothing data. 2) Gaussian fitting (decomposition) and waveform data point clouding: the waveform data is fitted using a non-linear least squares method based on the assumption that the echo data is an accumulation of multiple gaussian functions. And then extracting discrete point clouds from the processed waveform data through a local maximum peak detection filtering algorithm, wherein the energy and amplitude information of the return signal is recorded in each discrete point. 3) Generating a digital terrain: the purpose of LiDAR data height normalization is to obtain a "true" vegetation height with terrain effects removed, typically using raw LiDAR data height information minus terrain height. Therefore, accurate generation of a Digital Terrain Model (DTM) is an important prerequisite for calculating the normalized vegetation height. Firstly, discrete point clouds extracted from waveform data are classified, then Kraus filtering processing is carried out on the last echo to remove non-ground points, and finally the filtered last echo data are used and a digital terrain model is generated by means of interpolation of a natural proximity method. And then, normalizing the elevation of the vegetation echo point by utilizing the DEM, namely converting the height value of the vegetation point into a height value relative to the ground, and cutting each sample land through coordinates of a lower left corner and an upper right corner. And finally, extracting the normalized point cloud data of the coordinate positions corresponding to the 55 sample plots by a GIS analysis tool.
By performing band combination, tassel cap transformation, texture information extraction, principal component analysis, minimum noise separation transformation and multiple vegetation index transformation on an OLI image, 5 groups (53 in total) of characteristic variables, 6 original single-band variables, 10 band combination variables, 10 information enhancement group variables, 18 vegetation index variables and 9 texture information variables are extracted, wherein the texture analysis is performed on a first principal component of the principal component analysis, the window size is 3 x 3, and the lag distance is 1 pixel. The meanings of the 53 variables and the calculation formula (see table 1). The LiDAR feature variables were calculated 4 sets (34) of feature variables based on three-dimensional normalized LiDAR point cloud values: 10 height variables, 13 height percentile variables, 10 canopy density variables, 1 canopy coverage variable. The meanings of 34 variables and the calculation formula (see table 2).
TABLE 1 OLI characteristic variables and descriptions
And carrying out Pearson correlation analysis on the extracted 87 (34 LiDAR characteristic variables and 53 OLI characteristic variables) variables and the parameters needing to be predicted, and selecting the characteristic variables of which the absolute values of Pearson's correlation coefficients are higher than 0.2 as modeling candidate variables. The calculation method of the Pearson's correlation coefficient comprises the following steps:
in the formula, x i For a certain stand characteristic measured on the ground, y i For a certain variable of the LiDAR characteristics,is x i Is determined by the average value of (a) of (b),is y i Average value of (d);
the study selected 34 candidate variables whose absolute values of their Pearson's correlation coefficients with the parameter to be predicted are shown in fig. 3 and 4. The 34 variables have high correlation coefficients and significant relation with the biomass Pearson's, which shows that the 34 variables have good linear relation with the biomass. These 34 feature variables are therefore used as final modeling candidate variables.
TABLE 2 LiDAR feature variables and descriptions
And (3) establishing a multivariate regression model by taking biomass information obtained by ground actual measurement and summary as a dependent variable and characteristic variables extracted by a remote sensing method as independent variables. Using a stepwise entry method (stepwise) and checking the coefficient of determination (R) 2 ) To select the appropriate variables to enter the model. If there is an independent variable, the statistic F is too small and T test does not reach significant level (P value)>, 0.1), then removing; f value is large and T test reaches a significant level (P value)<, 0.05) is entered. Adopt and decideConstant coefficient (R) 2 ) Root Mean Square Error (RMSE), and relative Root Mean Square Error (RMSE) to evaluate the accuracy of the regression model;
in the formula, x i Is a certain forest stand characteristic measured on the ground,is x i Is determined by the average value of (a),a certain forest stand characteristic estimated by the model, wherein n is the number of sample plots;
as shown in the formula (3), rRMSE is the ratio of the RMSE (root mean square error) to the RMSE(mean of measured values) percentage.
Firstly, an OLI biomass estimation model (OLI model for short) and a LiDAR biomass estimation model (LiDAR model for short) are respectively constructed by utilizing OLI characteristic variables and LiDAR characteristic variables, and then a comprehensive biomass estimation model (comprehensive model for short) is constructed based on the two characteristic variables. When the three models are constructed, the two conditions are divided equally for analysis, firstly, statistical analysis is carried out on all sample plots without distinction, and then the sample plots are divided into coniferous forests, broad-leaved forests and mixed forests according to tree species composition for analysis respectively. The accuracy of each model for the different forest types was evaluated as in table 3. Scatter based on measured values of aboveground and underground biomass plots for different forest types versus predicted values for the models (OLI model, liDAR model, and integrated model) are shown in fig. 5-10.
Table 3 model accuracy evaluation table for different forest types
Wherein R is 2 Determining coefficients for the model; RMSE is root mean square error; rmse is relative root mean square error.
The above results show that the inversion result of the integrated model is better than that of a single characteristic variable model (i.e., an OLI model and a LiDAR model), and the model fitting effect is improved as follows: r 2 The improvement is 3 to 24 percent; the estimation accuracy is improved as follows: rRMSE decreases by 2-10%.
Claims (4)
1. A method for inverting remote sensing forest biomass is characterized by comprising the following steps:
1) OLI image preprocessing: firstly, carrying out radiometric calibration on an original image by means of radiometric calibration parameters of an OLI sensor, and converting an original DN value into a pixel radiometric brightness value; then, carrying out atmospheric correction on the image subjected to radiometric calibration by using a FLAASH model so as to convert the radiance value into the actual reflectivity of the earth surface; then, geometrically and finely correcting the image after atmospheric correction, selecting the same-name feature points, correcting by using a quadratic polynomial, controlling the correction error within 0.1 pixel, and resampling by using a nearest pixel method;
2) LiDAR data preprocessing: firstly, noise level estimation and data smoothing are carried out, then Gaussian fitting and waveform data point clouding are carried out to generate a digital terrain, and then the elevation of a vegetation echo point is normalized by utilizing a DEM;
3) OLI characteristic variable extraction: extracting 5 groups of characteristic variables which are respectively an original single-waveband variable, a waveband combination variable, an information enhancement group variable, a vegetation index variable and a texture information variable by carrying out waveband combination, tassel cap transformation, texture information extraction, principal component analysis, minimum noise separation transformation and various vegetation index transformations on an OLI image; wherein the texture analysis is performed on a first principal component of the principal component analysis;
4) Extracted LiDAR point cloud feature variables: the LiDAR point cloud characteristic variables are 4 groups of characteristic variables calculated based on a three-dimensional normalized LiDAR point cloud value and are respectively a height variable, a height percentile variable, a canopy density variable and a canopy coverage variable;
5) Screening characteristic variables: performing Pearson's correlation analysis on the extracted LiDAR characteristic variables and OLI characteristic variables and parameters to be predicted, and selecting the characteristic variables of which the absolute values of Pearson's correlation coefficients are higher than 0.2 as modeling candidate variables; the calculation method of the Pearson's correlation coefficient comprises the following steps:
in the formula, x i For a certain stand characteristic measured on the ground, y i For a certain variable of a LiDAR feature,is x i Is determined by the average value of (a),is y i Average value of (a);
6) Statistical analysis: establishing a multivariate regression model by taking biomass information obtained by ground actual measurement and summary as a dependent variable and taking a characteristic variable extracted by a remote sensing method as an independent variable; using stepwise entry method stepwise and checking the coefficient of determination R 2 To select appropriate variables to enter the model; if the independent variable makes the F value of the statistic too small and the T test does not reach the significant level, rejecting the statistic; the F value is larger and the T test reaches a significant level, then the method is entered; using a determining coefficient R 2 The root mean square error RMSE and the relative root mean square error rRMSE evaluate the accuracy of the regression model;
in the formula, x i A certain forest stand characteristic measured for the ground,is x i Is determined by the average value of (a),a certain forest stand characteristic estimated by the model, wherein n is the number of sample plots;
in step 2), gaussian fitting and waveform data point cloud are carried out: fitting the waveform data by adopting a nonlinear least square method; then extracting discrete point clouds from the processed waveform data through a local maximum peak detection filtering algorithm, wherein the energy and amplitude information of a return signal is recorded in each discrete point;
in step 2), the noise level estimation and data smoothing are: firstly, converting original data into a frequency domain, and then taking a low value part with higher frequency as a judgment standard of a noise level; then, a gaussian filter is selected for smoothing.
2. The method of remote sensing forest biomass inversion of claim 1, wherein: in step 2), the digital terrain is generated as follows: firstly, discrete point clouds extracted from waveform data are classified, then Kraus filtering processing is carried out on the last echo to remove non-ground points, and finally the filtered last echo data are used and a digital terrain model is generated by means of interpolation of a natural proximity method.
3. The method of remote sensing forest biomass inversion of claim 1, wherein: in the step 2), the elevation of the vegetation echo point is normalized by using the DEM, and the normalization processing is as follows: converting the height value of the vegetation points into a height value relative to the ground, and cutting each sample land through coordinates of a lower left corner and an upper right corner; and finally, extracting the normalized point cloud data of the coordinate positions corresponding to the 55 sample plots by a GIS analysis tool.
4. The method of remote sensing forest biomass inversion of claim 1, wherein: in the step 6), the T test can not reach the significance level, namely the P value is >0.1, and the T test can reach the significance level, namely the P value is less than 0.05.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510056042.1A CN104656098B (en) | 2015-02-03 | 2015-02-03 | A kind of method of remote sensing forest biomass inverting |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510056042.1A CN104656098B (en) | 2015-02-03 | 2015-02-03 | A kind of method of remote sensing forest biomass inverting |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104656098A CN104656098A (en) | 2015-05-27 |
| CN104656098B true CN104656098B (en) | 2018-04-13 |
Family
ID=53247452
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510056042.1A Active CN104656098B (en) | 2015-02-03 | 2015-02-03 | A kind of method of remote sensing forest biomass inverting |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104656098B (en) |
Families Citing this family (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105354534B (en) * | 2015-09-29 | 2018-11-20 | 南京林业大学 | A kind of tree species classification method based on multi-source same period high-definition remote sensing data |
| CN105608293B (en) * | 2016-01-28 | 2019-10-11 | 武汉大学 | Merge Biomass retrieval method and system on the woodland of spectrum and textural characteristics |
| CN105913016A (en) * | 2016-04-08 | 2016-08-31 | 南京林业大学 | Strip LiDAR data upscaling-based forest biomass estimating method |
| CN105913017A (en) * | 2016-04-08 | 2016-08-31 | 南京林业大学 | Corresponding period double high resolution remote sensing image-based forest biomass estimation method |
| CN106199627B (en) * | 2016-09-14 | 2018-07-10 | 中国农业科学院农业资源与农业区划研究所 | A kind of accuracy improvements method of unmanned plane LIDAR invertings grassland vegetation parameter |
| CN107247809B (en) * | 2017-07-19 | 2020-05-26 | 南京林业大学 | New method for forest age space mapping of artificial forest |
| CN108020211B (en) * | 2017-12-01 | 2020-07-07 | 云南大学 | Method for estimating biomass of invasive plants through aerial photography by unmanned aerial vehicle |
| CN108007438B (en) * | 2017-12-01 | 2020-07-24 | 云南大学 | Estimation method for plant biomass of unmanned aerial vehicle aerial photography remote sensing wetland |
| CN108876917A (en) * | 2018-06-25 | 2018-11-23 | 西南林业大学 | A kind of forest ground biomass remote sensing estimation universal model construction method |
| CN109164459A (en) * | 2018-08-01 | 2019-01-08 | 南京林业大学 | A kind of method that combination laser radar and high-spectral data classify to forest species |
| CN109031344B (en) * | 2018-08-01 | 2020-11-10 | 南京林业大学 | Method for jointly inverting forest structure parameters by full-waveform laser radar and hyperspectral data |
| CN108921885B (en) * | 2018-08-03 | 2020-05-12 | 南京林业大学 | Method for jointly inverting forest aboveground biomass by integrating three types of data sources |
| CN109063657B (en) * | 2018-08-08 | 2021-10-15 | 武汉大学 | Aboveground biomass estimation and scale conversion method facing homogeneous region spectrum unit |
| CN109061601A (en) * | 2018-08-09 | 2018-12-21 | 南京林业大学 | A method of based on unmanned plane laser radar inverting artificial forest forest structural variable |
| CN109118484A (en) * | 2018-08-10 | 2019-01-01 | 中国气象局气象探测中心 | A method of acquisition vegetation coverage and leaf area index based on machine vision |
| CN108981616B (en) * | 2018-08-15 | 2020-06-30 | 南京林业大学 | Method for inverting effective leaf area index of artificial forest by unmanned aerial vehicle laser radar |
| CN109325433A (en) * | 2018-09-14 | 2019-02-12 | 东北农业大学 | Introduce the black soil region soybean biomass multi-temporal remote sensing inversion method of terrain factor |
| CN109344550A (en) * | 2018-11-26 | 2019-02-15 | 国智恒北斗科技集团股份有限公司 | A kind of forest reserves inversion method and system based on domestic high score satellite remote sensing date |
| CN109884664B (en) * | 2019-01-14 | 2022-12-02 | 武汉大学 | Optical microwave collaborative inversion method and system for urban overground biomass |
| CN109946714A (en) * | 2019-04-03 | 2019-06-28 | 海南省林业科学研究所 | A kind of method for building up of the forest biomass model based on LiDAR and ALOS PALSAR multivariate data |
| CN110222656B (en) * | 2019-06-11 | 2020-05-05 | 成都理工大学 | Ground vegetation ecological water quantitative inversion method based on remote sensing technology |
| CN110287457B (en) * | 2019-07-02 | 2023-02-17 | 吉林大学 | Corn biomass inversion measurement method based on satellite radar remote sensing data |
| CN111489388A (en) * | 2020-04-20 | 2020-08-04 | 黑龙江工程学院 | Biomass inversion method based on effective crown information |
| CN111860328B (en) * | 2020-07-21 | 2021-04-06 | 杭州时光坐标影视传媒股份有限公司 | Biomass estimation method based on bidirectional reflection function and forest scene illumination effect modeling |
| CN111860359B (en) * | 2020-07-23 | 2021-08-17 | 江苏食品药品职业技术学院 | Point cloud classification method based on improved random forest algorithm |
| CN112434617B (en) * | 2020-11-26 | 2021-08-13 | 南京观微空间科技有限公司 | Forest biomass change monitoring method and system based on multi-source remote sensing data |
| CN114118835B (en) * | 2021-12-01 | 2022-06-21 | 中南大学 | Quantitative remote sensing inversion prediction result evaluation method and system |
| CN114924034B (en) * | 2022-05-06 | 2024-05-14 | 贵州师范大学 | Forestry carbon metering system based on ecological process model |
| CN115062260B (en) * | 2022-06-16 | 2024-06-14 | 电子科技大学 | Forest biomass PolInSAR estimation method and system suitable for heterogeneous forests and storage medium |
| CN115561773B (en) * | 2022-12-02 | 2023-03-10 | 武汉大学 | Forest carbon reserve inversion method based on ICESat-2 satellite-borne LiDAR data and multispectral data |
| CN117368118B (en) * | 2023-10-09 | 2024-10-18 | 太原理工大学 | Mining area biomass monitoring method based on multispectral and point cloud data processing |
| CN117313959A (en) * | 2023-11-28 | 2023-12-29 | 吉林省林业科学研究院(吉林省林业生物防治中心站) | Forestry carbon sink monitoring method and system based on big data |
-
2015
- 2015-02-03 CN CN201510056042.1A patent/CN104656098B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN104656098A (en) | 2015-05-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104656098B (en) | A kind of method of remote sensing forest biomass inverting | |
| CN108921885B (en) | Method for jointly inverting forest aboveground biomass by integrating three types of data sources | |
| Miraki et al. | Individual tree crown delineation from high-resolution UAV images in broadleaf forest | |
| Yan et al. | Urban land cover classification using airborne LiDAR data: A review | |
| CN109031344B (en) | Method for jointly inverting forest structure parameters by full-waveform laser radar and hyperspectral data | |
| Chen et al. | Lidar remote sensing of vegetation biomass | |
| Omasa et al. | 3D lidar imaging for detecting and understanding plant responses and canopy structure | |
| Chen et al. | Isolating individual trees in a savanna woodland using small footprint lidar data | |
| CN113204998B (en) | Airborne point cloud forest ecological estimation method and system based on single wood scale | |
| Li et al. | Combined use of airborne LiDAR and satellite GF-1 data to estimate leaf area index, height, and aboveground biomass of maize during peak growing season | |
| CN111950336B (en) | Vegetation canopy ecological water estimation method based on backpack type laser radar | |
| CN111091079B (en) | TLS-based method for measuring vegetation advantage single plant structural parameters in friable region | |
| CN112729130A (en) | Method for measuring height of tree canopy by satellite remote sensing | |
| CN104502919A (en) | Method for utilizing airborne laser radar point cloud to extract urban vegetation three-dimensional coverage map | |
| CN108959705B (en) | Method for predicting subtropical forest biomass | |
| CN110988909A (en) | TLS-based vegetation coverage determination method for sandy land vegetation in alpine and fragile areas | |
| Chen et al. | Site quality assessment of a Pinus radiata plantation in Victoria, Australia, using LiDAR technology | |
| CN108038433A (en) | Urban trees carbon content method of estimation based on more echo airborne laser scanning datas | |
| CN109146951A (en) | A method of ginkgo artificial forest leaf area index is estimated based on unmanned plane laser radar porosity model | |
| Wieser et al. | ULS LiDAR supported analyses of laser beam penetration from different ALS systems into vegetation | |
| Hollaus et al. | Full-waveform airborne laser scanning systems and their possibilities in forest applications | |
| CN113534083B (en) | SAR-based corn stubble mode identification method, device and medium | |
| Huang et al. | Information fusion approach for biomass estimation in a plateau mountainous forest using a synergistic system comprising UAS-based digital camera and LiDAR | |
| Yusup et al. | Trunk volume estimation of irregular shaped Populus euphratica riparian forest using TLS point cloud data and multivariate prediction models | |
| Zhao et al. | Evaluation of the soil profile quality of subsided land in a coal mining area backfilled with river sediment based on monitoring wheat growth biomass with UAV systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20150527 Assignee: Beijing Huamei Wanxiang Technology Co., Ltd. Assignor: Nanjing Forestry University Contract record no.: 2018320000235 Denomination of invention: Method for inverting remote sensing forest biomass Granted publication date: 20180413 License type: Common License Record date: 20181024 Application publication date: 20150527 Assignee: Nanjing city Pukou District moon farm Assignor: Nanjing Forestry University Contract record no.: 2018320000234 Denomination of invention: Method for inverting remote sensing forest biomass Granted publication date: 20180413 License type: Common License Record date: 20181024 |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20150527 Assignee: Huangdun Nursery Farm, Ganyu District, Lianyungang City Assignor: Nanjing Forestry University Contract record no.: 2018320000376 Denomination of invention: Method for inverting remote sensing forest biomass Granted publication date: 20180413 License type: Common License Record date: 20181212 |
|
| EE01 | Entry into force of recordation of patent licensing contract |