CN116883834A - Pinus koraiensis overground biomass remote sensing estimation method and system - Google Patents

Pinus koraiensis overground biomass remote sensing estimation method and system Download PDF

Info

Publication number
CN116883834A
CN116883834A CN202310703336.3A CN202310703336A CN116883834A CN 116883834 A CN116883834 A CN 116883834A CN 202310703336 A CN202310703336 A CN 202310703336A CN 116883834 A CN116883834 A CN 116883834A
Authority
CN
China
Prior art keywords
biomass
pinus koraiensis
remote sensing
model
pinus
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.)
Granted
Application number
CN202310703336.3A
Other languages
Chinese (zh)
Other versions
CN116883834B (en
Inventor
高菲
陈江平
饶宸
鄢佳芸
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Wuhan University WHU
Original Assignee
Wuhan University WHU
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Wuhan University WHU filed Critical Wuhan University WHU
Priority to CN202310703336.3A priority Critical patent/CN116883834B/en
Publication of CN116883834A publication Critical patent/CN116883834A/en
Application granted granted Critical
Publication of CN116883834B publication Critical patent/CN116883834B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/188—Vegetation
    • 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
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/86—Combinations of radar systems with non-radar systems, e.g. sonar, direction finder
    • G01S13/867—Combination of radar systems with cameras
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/10—Image acquisition
    • G06V10/12—Details of acquisition arrangements; Constructional details thereof
    • G06V10/14—Optical characteristics of the device performing the acquisition or on the illumination arrangements
    • G06V10/143—Sensing or illuminating at different wavelengths
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/30—Noise filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
    • G06V10/421—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation by analysing segments intersecting the pattern
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/54—Extraction of image or video features relating to texture
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G06V10/765—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects using rules for classification or partitioning the feature space
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/766—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using regression, e.g. by projecting features on hyperplanes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Multimedia (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Remote Sensing (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • Medical Informatics (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Educational Administration (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Game Theory and Decision Science (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Image Processing (AREA)

Abstract

The invention provides a method and a system for estimating biomass on the ground of Pinus koraiensis, which comprise the steps of selecting a sampling area from a target area to be analyzed, acquiring Pinus koraiensis ground investigation data, and acquiring laser point cloud and optical images of an unmanned aerial vehicle; calculating the biomass on the single wood floor by utilizing single wood floor survey data; preprocessing laser point cloud and optical images and identifying single wood respectively, extracting remote sensing characteristic variables, and sequencing importance of the characteristic variables by combining biomass on the pinus koraiensis ground to obtain a prediction factor; taking the optimal remote sensing characteristic combination as a prediction variable, taking the biomass on the pinus koraiensis land as a response variable, training a random forest model, correcting and selecting an optimal estimation model; and carrying out Pinus koraiensis single wood extraction and biomass quantitative estimation on the whole region to be estimated. The invention can provide effective technical support for forest environment protection and ecological construction by combining the geospatial information technology and the mathematical analysis model.

Description

Pinus koraiensis overground biomass remote sensing estimation method and system
Technical Field
The invention relates to the field of application of remote sensing technology, in particular to a remote sensing estimation scheme of the biomass on the pinus koraiensis land based on a geospatial information technology and a mathematical analysis model.
Background
Aboveground biomass (AGB) is a fundamental parameter in assessing forest productivity. In the past, the remote sensing research of the aboveground biomass is mostly in a scale of a sample, and a biomass estimation model aiming at a specific tree species is very limited, so that the fine control requirement of a forest can not be completely met. The traditional monitoring method of the biomass on the single wood floor mainly comprises a simvastatin method, an average biomass method, a regression model estimation method, a timber volume source estimation method and the like, and although the estimation result is high in accuracy, a large amount of field investigation is required, time and labor are wasted, certain damage to ecology can be caused, and the method is more suitable for investigation of a small-scale typical forest area. With the development of scientific technology, unmanned Aerial Vehicle (UAV) remote sensing is increasingly applied to ecological mapping research as a novel mapping means, and compared with the traditional method, the UAV remote sensing has the characteristics of low cost, short period, strong maneuverability and rich data details, and provides a powerful means for automatically monitoring forest resources. The passive optical remote sensing of the unmanned aerial vehicle is the earliest and most commonly used UAV mapping method, and the method has wide data coverage and is easy to access. At the same time, laser radar (LiDAR) technology is emerging, also in organic combination with unmanned aerial vehicle platforms. The unmanned aerial vehicle laser radar is an advanced active remote sensing technology, and is characterized in that a distance measuring sensor is carried, the distance and the angle between the sensor and surrounding targets are directly obtained based on the distance measuring principle of a time-of-flight measurement method, the three-dimensional coordinates of a ground object are obtained, meanwhile, the point cloud intensity and echo frequency information are stored, and the data have extremely high angular resolution and distance resolution; in forestry research, three-dimensional parameters of single woods can be obtained through the UAV LiDAR, vertical and horizontal structural information of the forests is accurately captured, and a technical basis is provided for forest biomass estimation at the single wood level.
In summary, the invention improves the existing method for estimating the biomass of the Pinus koraiensis, and provides a method and a system for remotely sensing and estimating the biomass of the Pinus koraiensis on the ground. According to the method for refining the forest AGB estimation research scale to the single wood scale, the Pinus koraiensis AGB estimation based on the unmanned aerial vehicle image and the LiDAR point cloud is effectively realized; meanwhile, the effects of three deviation correction methods of the random forest model are compared, and the improved method can evaluate the carbon sequestration potential and carbon sink quality of the pinus koraiensis in more detail, so that the method is a key basis for related departments to monitor forest ecological level and formulate and adjust forestry protection policies.
Disclosure of Invention
The invention aims to overcome the technical defects and provide an accurate remote sensing estimation scheme for the biomass on the pinus koraiensis land based on a geospatial information technology and a mathematical analysis model.
In order to achieve the aim, the technical scheme adopted by the invention provides a remote sensing estimation method for the biomass on the pinus koraiensis land, which comprises the following steps of,
step 1, based on a target area to be analyzed, selecting a plurality of small blocks as sampling areas, acquiring pinus single wood ground investigation data of the sampling areas, and acquiring unmanned aerial vehicle laser point clouds and optical images of the sampling areas in an acquisition time period according to the acquisition time period of the ground investigation data;
step 2, calculating the overground biomass of the single wood by using the single wood ground survey data obtained in the step 1;
step 3, preprocessing the laser point cloud and the optical image and identifying the single wood respectively, extracting remote sensing characteristic variables, and sequencing the importance of the characteristic variables by combining the biomass on the pinus koraiensis ground to obtain a prediction factor for finally estimating the biomass;
step 4, combining the optimal remote sensing characteristics obtained in the step 3 to serve as a prediction variable of a biomass regression model, training a random forest model by taking the Pinus koraiensis overground biomass calculated in the step 2 as a response variable, correcting a model prediction result through a deviation correction algorithm, evaluating a model effect, and selecting an optimal estimation model;
and 5, carrying out Pinus koraiensis single wood extraction and biomass quantitative estimation on the whole region to be estimated by utilizing the optimal estimation model obtained in the step 4.
In a preferred embodiment of the present invention, in the step 1, the single wood floor survey data includes a position, a breast diameter and a tree height.
In the step 2, the differential speed growth equation is used as a calculation method of the biomass on the pinus koraiensis floor.
As a preferred solution of the present invention, in the step 3, the preprocessing of the laser point cloud includes denoising, separating ground points, and normalizing the heights of tree points, and the preprocessing of the optical image includes geographic registration with the point cloud; the laser point Yun Shanmu identification method is a point cloud distance-based segmentation method, and the optical image single wood identification method is object-oriented multi-scale segmentation.
In the step 4, the importance of the remote sensing feature to the biomass of the Pinus koraiensis is calculated by using a random forest model, and a remote sensing feature variable with high importance is selected as a feature factor; and (2) calculating the correlation relationship in the remote sensing characteristic variables by using Pearson correlation analysis, wherein the corresponding characteristic is considered as a redundant characteristic when the Pearson correlation coefficient is more than 0.8, and the characteristic with the strongest correlation with the Pinus koraiensis biomass calculated in the step (2) in the redundant characteristic of the same category is reserved, so that the biomass optimal estimation characteristic combination is obtained.
As a preferable embodiment of the invention, in the step 4, the random forest bias correction method includes simple linear regression correction, built-in bias correction, and residual error rotation bias correction.
In the step 4, 80% of the biomass on the pinus koraiensis land calculated in the step 2 is selected as a training set, the remaining 20% is used as a verification set, and a leave-one-out cross verification method is adopted to train the random forest model. The model evaluation index comprises a determination coefficient R 2 The interpretable variance/interpretation degree EVS, the mean square error MSE, the mean absolute deviation MAE, the following formula:
wherein m refers to the number of estimated data and represents the total number of the pinus koraiensis units in the verification set, y i Representing the i-th measured value of biomass,represents the mean value of the measured values,/>Representing the ith biomass estimate, var is the data set variance.
The invention provides a remote sensing estimation system for biomass on the pinus koraiensis land, which is used for realizing the remote sensing estimation method for biomass on the pinus koraiensis land.
And, the system comprises a processor and a memory, wherein the memory is used for storing program instructions, and the processor is used for calling the stored instructions in the memory to execute the remote sensing estimation method of the biomass on the pinus koraiensis floor.
Compared with the prior art, the invention has the following beneficial effects:
1. in the past, the remote sensing research of forest biomass is mostly in a sample scale, and a biomass estimation model aiming at specific tree species is very limited, so that the fine control requirement of the forest can not be completely met; whereas biomass monitoring at the single wood level is relatively labor dependent. According to the method, unmanned aerial vehicle images and LiDAR point cloud data are combined, a method for refining forest AGB estimation research scale to a single wood scale is provided, and the Pink AGB estimation is effectively achieved.
2. Aiming at the problems of overestimation and underestimation of the predicted value of the random forest model, the invention introduces three deviation correction methods of the random forest model, compares the effects of the three deviation correction methods and reveals the optimal feature combination of the Pinus koraiensis AGB prediction.
Drawings
FIG. 1 is a schematic diagram of a remote sensing estimation flow of biomass on the surface of Pinus koraiensis.
FIG. 2 is a scatter plot of predicted and measured biomass models of Pinus koraiensis according to an embodiment of the present invention, wherein part a RF is a random forest model, part b RF-SLR is an SLR bias corrected random forest model, part c RF-BC is a BC bias corrected random forest model, and part d RF-RR is an RR bias corrected random forest model.
Detailed Description
The following describes the remote sensing estimation scheme of the biomass on the pinus koraiensis ground in detail by referring to the drawings and the examples.
The invention provides a pinus korotus overground biomass remote sensing estimation scheme, which comprises the steps of selecting a plurality of small blocks from a target area to be analyzed as a sampling area, acquiring pinus korotus terrotus ground investigation data of the sampling area, and acquiring unmanned plane laser point cloud and optical images of the sampling area in an acquisition time period according to the acquisition time period of the ground investigation data; calculating the biomass on the single wood ground by using the acquired single wood ground survey data; preprocessing laser point cloud and optical images and identifying single wood respectively, extracting remote sensing characteristic variables, and sequencing importance of the characteristic variables by combining the biomass on the pinus koraiensis ground to obtain a prediction factor for finally estimating the biomass; the obtained optimal remote sensing feature combination is used as a prediction variable of a biomass regression model, the calculated biomass on the pinus koraiensis land is used as a response variable, a random forest model is trained, a model prediction result is corrected through a deviation correction algorithm, a model effect is evaluated, and an optimal estimation model is selected; and (3) extracting the pinus koraiensis and quantitatively estimating the biomass by utilizing an optimal estimation model. The invention can provide effective technical support for forest environment protection and ecological construction by combining the geospatial information technology and the mathematical analysis model.
Referring to fig. 1, when a service provider needs to perform red pine log biomass estimation, the method for remote sensing estimation of red pine log ground biomass provided by the embodiment of the invention can be used for performing the following steps to realize reasonable analysis,
step S01, selecting a plurality of small blocks from a target area to be analyzed as sampling areas, and acquiring Pinus koraiensis single wood ground investigation data of the sampling areas;
further, the single wood floor survey data includes location, thoracorn diameter, and tree height.
The embodiment investigates single wood data in two 25m x 25m size plots of the Pinus koraiensis of 2022, obtains the space coordinates of the forest from the fixed position measurement in the plots by using a thousands of search RTK device, obtains the single wood (x, y) coordinates as single wood actual measurement data, and the distance resolution of the device is 2.5mm; trunk breast diameter measurement is carried out on the position, which is 1.3m away from the ground, of each tree by using a girth ruler (breast diameter ruler), and the precision is 0.1mm; the height of the tree tops of a plurality of trees is measured by using a tower ruler instrument, and the resolution of the height of the equipment is 0.01m. And 81 pieces of red pine single wood data are obtained, and the space coordinates, tree species, breast diameter and tree height information of the single wood are recorded.
Then, according to the acquisition time period of the ground survey data, acquiring unmanned aerial vehicle laser point cloud and optical images of a sampling area in the acquisition time period;
in the embodiment of the invention, unmanned plane laser radar point cloud numbers of two pinus subsamplesThe single point cloud coverage is 300m by 300m and comprises 25m by 25m ground actual measurement ranges, wherein the single point cloud coverage is obtained by a laser scanning sensor of the Haida cloud ARS200 unmanned aerial vehicle; the original laser spot density is about 100/m 2 The data record information such as (x, y) coordinates, height, laser intensity, echo times and the like of the laser point. The size of the single unmanned aerial vehicle image aerial photo of the pinus two-block sample plot is 300m, the single unmanned aerial vehicle image aerial photo is basically the same as LiDAR data, the single unmanned aerial vehicle image aerial photo comprises landmark reflectivity values of R, G, B wave bands, and the spatial resolution is about 0.25m.
TABLE 1 different growth equation Table 1.Allometric equation
Description: d is the breast diameter of the tree, H is the height of the tree, and the total biomass of overground parts=trunk biomass+branch biomass+leaf biomass step S02 is carried out, and the overground biomass of the single tree is calculated by using the single-tree ground survey data obtained in the step 1;
the embodiment of the invention uses an abnormal speed growth equation as a calculation method of the biomass on the pinus koraiensis land, and the specific abnormal speed growth equation is shown in table 1.
Step S03, respectively preprocessing the laser point cloud and the optical image and identifying single wood, and extracting remote sensing characteristic variables;
(1) The embodiment carries out preprocessing on the laser point cloud, including denoising, separating ground points and tree point height normalization, and carries out preprocessing on the optical image, including geographic registration with the point cloud; the laser point Yun Shanmu identification method is a point cloud distance-based segmentation method, and the optical image single wood identification method is object-oriented multi-scale segmentation.
(2) In order to eliminate the interference of ground features such as understory bushes, the invention selects normalized point clouds with the height of more than 0.5 meter for research, and extracts 33 LiDAR characteristic variables including height variable, density variable, intensity variable and leaf area index from the normalized point clouds, wherein the detailed description is shown in table 2.
(1) Height variable. The height variables extracted by the invention comprise the maximum height and the average height in the single wood point cloud set, and the height percentile value with the adjacent interval of 10 percent and from 10 percent to 90 percent.
(2) Density variable. Refers to dividing the Shan Mudian cloud set from low to high into 10 slices of equal height, and counting the proportion of the number of loops per layer.
(3) Intensity variable. The intensity variables extracted by the method comprise intensity maximum values, intensity average values and intensity percentile values, wherein the intensity maximum values and the intensity average values in a single wood point cloud set are 10% and the adjacent intervals are from 10% to 90%.
(4) Leaf area index. I.e. half of the surface area of all blades on the unit area of the ground surface, and the blade area index is calculated by normalized Shan Mudian cloud data.
TABLE 2LiDAR feature variable extraction Table 2.LiDAR feature variable extraction
(2) The invention extracts 17 image characteristic variables including area, spectral parameters, vegetation index and texture characteristics from a single wood segmentation object of the UAV aerial photo.
(1) Crown area. Refers to the area of a single crown obtained after the image single tree is extracted and segmented.
(2) Spectral parameters. The invention extracts the average reflectivity of three visible light wave bands of canopy Red (Red, R), green (Green, G) and Blue (Blue, B).
(3) A vegetation index. The Visible light vegetation extracted by the invention comprises 5 kinds of Normalized difference greenness Index (Normalized Difference Greenness Vegetation Index, NDGI), red vegetation Index (Red Index, RI), visible light wave band difference vegetation Index (Visible-band Difference Vegetation Index, VDVI), normalized Green-blue difference Index (Normalized Green-Blue Difference Index, NGBDI) and ultra-Green and ultra-Red Index (ExG-ExR), and the specific calculation formulas are as follows.
NDGI＝(G-R)/(G+R)
RI＝(R-G)/(R+G)
VDVI＝(2G-R-B)/(2G+R+B)
NGBDI＝(G-B)/(G+B)
ExG-ExR＝3g-2.4r-b
Wherein R, G, B represents the average reflectivities of red, green and blue bands of the image, and r, g and b represent the average reflectivities of the red, green and blue bands after normalization.
(4) Texture features. Texture characteristics of each pixel 3*3 of the canopy object are calculated in ENVI5.3 image processing software through Gray-Level Co-occurrence Matrix, GLCM, including 8 texture parameters including Mean (Mean), variance (Var), synergy (Hom), contrast (Con), dissimilarity (Dis), entropy (entopy, end), angular second moment (Angle Second Moment, asm), correlation (correptation, cor), and the calculation formula is as follows.
Mean＝Σ i Σ j p(i,j)*i
Var＝Σ i Σ j p(i,j)*(i-Mean) 2
Con＝∑ i ∑ j p(i,j)*(i-j) 2
Dis＝∑ i ∑ j p(i,j)*|i-j|
Ent＝Σ i ∑ j p(i,j)*logp(i,j)
Asm＝∑ i ∑ j p(i,j) 2
Where i, j represent pixel values of (x, y), (x+d, y+l), respectively, and p (i, j) represents probability of occurrence of (i, j) in the image.
Then, carrying out importance ranking on the characteristic variables by combining the biomass on the pinus koraiensis land to obtain a prediction factor for finally estimating the biomass;
in the embodiment, a random forest model is used for calculating the importance of the remote sensing characteristics on the biomass of the Pinus koraiensis, and the remote sensing characteristic variable with high importance is selected as a characteristic factor; calculating the correlation relationship in the remote sensing characteristic variables by using Pearson correlation analysis, wherein the corresponding characteristic is considered as a redundant characteristic when the Pearson correlation coefficient is more than 0.8, and the characteristic with the strongest correlation with the Pinus biomass in the redundant characteristic of the same category is reserved, so that the optimal biomass estimation characteristic combination is obtained as follows: the method comprises the following steps of 90% of point cloud dividing height H90, 10% of point cloud dividing height H10, average point cloud height H_mean, leaf area index LAI, point cloud dividing densities D0 and D4, 70% of point cloud dividing intensity I70, maximum point cloud intensity I_max, crown area, average green wave band reflectivity G, average blue wave band reflectivity B, vegetation index ExG-ExR, variance texture factor Var and correlation texture factor Cor.
Step S04, combining the optimal remote sensing characteristics obtained in the step S03 to serve as a prediction variable of a biomass regression model, and training a random forest model by taking the Pinus koraiensis overground biomass calculated in the step S02 as a response variable;
according to the method, the importance of the remote sensing features on the biomass of the pinus koraiensis is calculated by using a random forest model, and the remote sensing feature variable with high importance is selected as a feature factor; and (2) calculating the correlation relationship in the remote sensing characteristic variables by using Pearson correlation analysis, wherein the corresponding characteristic is considered as a redundant characteristic when the Pearson correlation coefficient is more than 0.8, and the characteristic with the strongest correlation with the Pinus koraiensis biomass calculated in the step (2) in the redundant characteristic of the same category is reserved, so that the biomass optimal estimation characteristic combination is obtained.
In the embodiment, 80% of the biomass on the pinus koraiensis land is selected as a training set, the other 20% is selected as a verification set, and a leave-one-out cross verification method is adopted to train a random forest model. The optimal RF model parameters were set by python3.9 sklearn library as: the number of decision trees is 18, the maximum depth of each tree is 4, the minimum number of samples required by leaf nodes is 4, and the minimum number of samples of each division when dividing the nodes is 10.
Then, correcting a model prediction result through a random forest deviation correction algorithm, evaluating a model effect, and selecting an optimal estimation model;
the random forest Bias correction method adopted in the embodiment comprises simple linear regression correction (Simple Linear Regression, SLR), built-in Bias Correction (BC) and residual rotation Bias correction (Residual Rotation, RR), and is specifically described below.
(1) SLR offset correction
The SLR deviation correction method considers that a linear relation exists between the predicted value and the true value of the RF model, and the predicted value of the test set data is corrected by establishing a model of the predicted value and the actual measured value in the data in the bag by using the model. The model formula is described as:
y obs ＝a+by pre
in which y obs Y is the true observed value pre For the RF model predictors, a and b are linear regression coefficients.
(2) BC offset correction
The BC bias correction considers that the residual error of the RF model predicted value has a quantitative relation with the sample independent variable and the RF predicted value, the estimation of the verification set residual error is carried out through the RF algorithm, and the correction is carried out by being overlapped on the original predicted result, so that the method is a simple, convenient and efficient method for RF bias correction. The method comprises the following specific steps:
(1) the RF model is trained by randomly selecting partial data, and the residual error r of the training set is calculated by the specific formula:
r＝y obs -y pre
in which y obs Y is the true observed value pre Predicted values for the RF model.
(2) And establishing a residual prediction random forest model by combining the independent variable X of the training set and the predicted value, wherein the residual prediction random forest model is shown in the following formula:
r＝RF res (X,y pre )
(3) calculating a validation set prediction using the RF model of step (1)Validating set argument->Predictive value->Substitution formula to calculate verification set residual +.>
(4) Residual error is setSuperimposed on the validation set RF model predictions to obtain final correction data y BCRF And (5) completing BC deviation correction.
(3) RR bias correction
The RR bias correction principle consists in rotating the residuals of the validation set such that their trend lines are close to y=0, thus making the trend lines of the predicted and real values close to y=x. The method comprises the following specific steps:
(1) the RF model is trained by randomly selecting partial data, and the residual error r of the training set is calculated by the specific formula:
r＝y obs -y pre
in which y obs Y is the true observed value pre Predicted values for the RF model.
(2) And establishing a residual prediction random forest model by combining the independent variable X of the training set, wherein the residual prediction random forest model is shown as the following formula:
r＝RF res (X)
(3) will verify set argumentAnd RF prediction value->Substitution formula to calculate verification set residual +.>
(4) Establishing a validation set predictorResidual->Is corrected by rotating the matrix to make the trend line approach y=0, and the residual error generated by rotation is obtained>
(5) Residual error is setSuperimposed on the validation set RF model predictions to obtain final correction data y RRRF And finishing RR deviation correction.
The embodiment of the invention selects the decision coefficient (R 2 ) And an interpretable variance/interpretation degree (EVS) index reflects the accuracy and reliability of the biomass estimation model, and a Mean Square Error (MSE) and mean absolute deviation (MAE) index is selected to reflect the deviation of the predicted data from the actual data. The calculation formula of each index is as follows:
wherein m refers to the estimated data number, and in the invention, represents the total number of the pinus koraiensis, y i Representing the i-th measured value of biomass,represents the mean value of the measured values,/>Representing the ith biomass estimate, var is the data set variance.
TABLE 3 evaluation results of biomass model Table 3.Evaluation results of biomass model
Table 3 and fig. 2 show the evaluation results of the example model. RF in FIG. 2 represents a random forest model, RF-SLR represents an SLR bias corrected random forest model, RF-BC represents a BC bias corrected random forest model, and RF-RR represents an RR bias corrected random forest model. After the RF model is introduced into the offset correction method, the fitting precision is improved compared with that before the RF model is introduced, and the offset correction method with the highest fitting precision is SLR correction (R 2 =84%) with the lowest MSE (mse= 782.49 kg) 2 ) And MAE (mae=17.62 kg). The BC correction fitting accuracy is slightly inferior to the SLR correction (R 2 =83%) MSE and MAE are slightly higher than SLR correction and trend line coefficients are smaller, but BC correction prediction outliers are less than SLR correction, indicating that BC correction has better stability and adaptability to UAV multisource telemetry data. Although in the inventive embodiment the decision coefficient of the BC correction is smaller than the SLR correction, the difference between the two is small, and the BC correction prediction result shows better stability in the scatter diagram, so in summary, the BC correction of the RF model is considered to be the optimal machine learning model for modeling estimation of red pine single wood biomass in the research area.
And S05, carrying out Pinus koraiensis single wood extraction and biomass quantitative estimation on the whole region to be estimated by utilizing the optimal estimation model obtained in the step S04.
According to the method for estimating the biomass on the red pine single wood ground remotely provided by the flow, based on the target area to be analyzed, a plurality of small blocks are selected from the target area to serve as sampling areas, red pine single wood ground survey data of the sampling areas are obtained, and unmanned plane laser point clouds and optical images of the sampling areas in the collecting time period are obtained according to the collecting time period of the ground survey data; calculating the biomass on the single wood ground by using the acquired single wood ground survey data; preprocessing laser point cloud and optical images and identifying single wood respectively, extracting remote sensing characteristic variables, and sequencing importance of the characteristic variables by combining the biomass on the pinus koraiensis ground to obtain a prediction factor for finally estimating the biomass; the obtained optimal remote sensing feature combination is used as a prediction variable of a biomass regression model, the calculated biomass on the pinus koraiensis land is used as a response variable, a random forest model is trained, a model prediction result is corrected through a deviation correction algorithm, a model effect is evaluated, and an optimal estimation model is selected; and (3) extracting the pinus koraiensis and quantitatively estimating the biomass by utilizing an optimal estimation model. The obtained result is used for analyzing, predicting and automatically alarming the biomass on the area pinus koraiensis land, for example, when the biomass on the pinus koraiensis land is obtained by automatic evaluation, a corresponding alarm is sent to related responsible personnel through a preset communication link, and corresponding loss can be avoided in time.
Those skilled in the art will appreciate that the above-described methods provided by the present invention may be implemented in a modular form by software programming. Suitable programming languages may include, for example, but are not limited to, C language, VB, java, and the like. XML techniques may also be used to build pinus koraiensis biomass estimation models, etc.
The method and the system for estimating the biomass on the red pine single wood ground provided by the embodiment of the invention are designed and realized by fully considering the availability and sharing degree of various scientific data in China at present on the basis of ensuring scientificity according to the actual business demands of national ecological management and carbon sink.
In particular, the method according to the technical solution of the present invention may be implemented by those skilled in the art using computer software technology to implement an automatic operation flow, and a system apparatus for implementing the method, such as a computer readable storage medium storing a corresponding computer program according to the technical solution of the present invention, and a computer device including the operation of the corresponding computer program, should also fall within the protection scope of the present invention.
In some possible embodiments, a method and a system for remotely sensing and estimating the biomass on the ground of Pinus koraiensis are provided, which comprise the following modules,
the system comprises a first module, a second module and a third module, wherein the first module is used for selecting a block from a target area to be analyzed as a sampling area, acquiring the Pinus monowood ground investigation data of the sampling area, and acquiring unmanned plane laser point clouds and optical images of the sampling area in the acquisition time period according to the acquisition time period of the ground investigation data;
the second module is used for calculating the single wood ground biomass by using single wood ground survey data;
the third module is used for respectively preprocessing the laser point cloud and the optical image and identifying the single wood, extracting remote sensing characteristic variables, and carrying out importance sequencing on the characteristic variables by combining the biomass on the pinus koraiensis ground to obtain a prediction factor for finally estimating the biomass;
the fourth module is used for taking the obtained optimal remote sensing characteristic combination as a prediction variable of a biomass regression model, taking the calculated biomass on the pinus koraiensis land as a response variable, training a random forest model, correcting a model prediction result through a deviation correction algorithm, evaluating a model effect and selecting an optimal estimation model;
and a fifth module for extracting Pinus koraiensis and quantitatively estimating the biomass by utilizing the biomass optimal estimation model.
In some possible embodiments, a method and a system for remotely sensing and estimating the biomass on the pinus koraiensis ground are provided, which comprise a processor and a memory, wherein the memory is used for storing program instructions, and the processor is used for calling the stored instructions in the memory to execute the method and the system for remotely sensing and estimating the biomass on the pinus koraiensis ground.
In some possible embodiments, a method and a system for remotely sensing and estimating the biomass on the surface of the pinus koraiensis are provided, which comprise a readable storage medium, wherein the readable storage medium is stored with a computer program, and the computer program realizes the method and the system for remotely sensing and estimating the biomass on the surface of the pinus koraiensis when being executed.
The above description is only an example of the present invention and is not intended to limit the present invention. Any modification, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (9)

1.A pinus koraiensis overground biomass remote sensing estimation method is characterized in that: comprises the steps of,
step 1, based on a target area to be analyzed, selecting a plurality of small blocks as sampling areas, acquiring pinus single wood ground investigation data of the sampling areas, and acquiring unmanned aerial vehicle laser point clouds and optical images of the sampling areas in an acquisition time period according to the acquisition time period of the ground investigation data;
step 2, calculating the overground biomass of the single wood by using the single wood ground survey data obtained in the step 1;
step 3, preprocessing the laser point cloud and the optical image and identifying the single wood respectively, extracting remote sensing characteristic variables, and sequencing the importance of the characteristic variables by combining the biomass on the pinus koraiensis ground to obtain a prediction factor for finally estimating the biomass;
step 4, combining the optimal remote sensing characteristics obtained in the step 3 to serve as a prediction variable of a biomass regression model, training a random forest model by taking the Pinus koraiensis overground biomass calculated in the step 2 as a response variable, correcting a model prediction result through a deviation correction algorithm, evaluating a model effect, and selecting an optimal estimation model;
and 5, carrying out Pinus koraiensis single wood extraction and biomass quantitative estimation on the whole region to be estimated by utilizing the optimal estimation model obtained in the step 4.
2. The method for remotely sensing and estimating the biomass on the pinus koraiensis floor according to claim 1, wherein the method comprises the following steps of: in the step 1, the single wood ground investigation data comprise positions, breast diameters and tree heights.
3. The method for estimating the biomass on the surface of the Pinus koraiensis according to claim 2, wherein in the step 2, an abnormal growth equation is used as a calculation method of the biomass on the surface of the Pinus koraiensis.
4. The method for remotely sensing and estimating the biomass on the pinus koraiensis floor according to claim 1, wherein the method comprises the following steps of: in the step 3, the preprocessing of the laser point cloud comprises denoising, separating ground points and normalizing the heights of tree points, and the preprocessing of the optical image comprises geographic registration with the point cloud; the laser point Yun Shanmu identification implementation mode is a point cloud distance-based segmentation method, and the optical image single-wood identification implementation mode is object-oriented multi-scale segmentation.
5. The method for remotely sensing and estimating the biomass on the pinus koraiensis floor according to claim 1, wherein the method comprises the following steps of: in the step 4, the importance of the remote sensing feature on the biomass of the Pinus koraiensis is calculated by using a random forest model, and a remote sensing feature variable with high importance is selected as a feature factor; and (2) calculating the correlation relationship in the remote sensing characteristic variables by using Pearson correlation analysis, wherein the corresponding characteristic is considered as a redundant characteristic when the Pearson correlation coefficient is more than 0.8, and the characteristic with the strongest correlation with the Pinus koraiensis biomass calculated in the step (2) in the redundant characteristic of the same category is reserved, so that the biomass optimal estimation characteristic combination is obtained.
6. The method for remotely sensing and estimating the biomass on the pinus koraiensis floor according to claim 1, wherein the method comprises the following steps of: in the step 4, the random forest deviation correction method comprises simple linear regression correction, built-in deviation correction and residual error rotation deviation correction.
7. The method for remotely sensing and estimating the biomass on the pinus koraiensis floor according to claim 1, wherein the method comprises the following steps of: in the step 4, 80% of the biomass on the Pinus koraiensis wood calculated in the step 2 is selected as trainingThe rest 20% of the set is used as a verification set, and a leave-one-out cross verification method is adopted to train the random forest model. The model evaluation index comprises a determination coefficient R 2 The interpretable variance/interpretation degree EVS, the mean square error MSE, the mean absolute deviation MAE, the following formula:
wherein m refers to the number of estimated data and represents the total number of the pinus koraiensis units in the verification set, y i Representing the i-th measured value of biomass,represents the mean value of the measured values,/>Representing the ith biomass estimate, var is the data set variance.
8. A pinus korotus overground biomass remote sensing estimation system is characterized in that: a method for performing remote sensing estimation of biomass on the pinus koraiensis as defined in any one of claims 1 to 7.
9. The pinus koraiensis overground biomass remote sensing estimation system according to claim 8, wherein: comprising a processor and a memory for storing program instructions, the processor being adapted to invoke the stored instructions in the memory to perform a method for remotely sensing and estimating biomass on red pine, single wood as defined in any one of claims 1-7.
CN202310703336.3A 2023-06-14 2023-06-14 Pinus koraiensis overground biomass remote sensing estimation method and system Active CN116883834B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310703336.3A CN116883834B (en) 2023-06-14 2023-06-14 Pinus koraiensis overground biomass remote sensing estimation method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310703336.3A CN116883834B (en) 2023-06-14 2023-06-14 Pinus koraiensis overground biomass remote sensing estimation method and system

Publications (2)

Publication Number Publication Date
CN116883834A true CN116883834A (en) 2023-10-13
CN116883834B CN116883834B (en) 2024-06-11

Family

ID=88253785

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310703336.3A Active CN116883834B (en) 2023-06-14 2023-06-14 Pinus koraiensis overground biomass remote sensing estimation method and system

Country Status (1)

Country Link
CN (1) CN116883834B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117387581A (en) * 2023-12-12 2024-01-12 北京麦麦趣耕科技有限公司 Cinnamomum camphora biomass estimation method, apparatus and storage medium

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
CN107832681A (en) * 2017-10-16 2018-03-23 福州大学 The high evaluation method of forest list ebon of joint LiDAR point cloud and synchronous remote sensing image
US20200225075A1 (en) * 2019-01-14 2020-07-16 Wuhan University Method and system for optical and microwave synergistic retrieval of aboveground biomass
CN113204998A (en) * 2021-04-01 2021-08-03 武汉大学 Airborne point cloud forest ecological estimation method and system based on single wood scale
CN113205565A (en) * 2021-04-13 2021-08-03 武汉大学 Forest biomass remote sensing mapping method based on multi-source data
CN114882361A (en) * 2022-05-13 2022-08-09 东南大学 Deep learning forest overground biological estimation method based on multi-source remote sensing fusion
CN115294147A (en) * 2022-06-17 2022-11-04 南宁师范大学 Method for estimating aboveground biomass of single trees and forests based on unmanned aerial vehicle laser radar
WO2023095039A1 (en) * 2021-11-24 2023-06-01 Dendra Systems Ltd. Generating an above ground biomass density prediction model
CN116224359A (en) * 2022-12-20 2023-06-06 中国地质调查局烟台海岸带地质调查中心 Forest overground biomass estimation method based on airborne LiDAR and small sample data

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
CN107832681A (en) * 2017-10-16 2018-03-23 福州大学 The high evaluation method of forest list ebon of joint LiDAR point cloud and synchronous remote sensing image
US20200225075A1 (en) * 2019-01-14 2020-07-16 Wuhan University Method and system for optical and microwave synergistic retrieval of aboveground biomass
CN113204998A (en) * 2021-04-01 2021-08-03 武汉大学 Airborne point cloud forest ecological estimation method and system based on single wood scale
CN113205565A (en) * 2021-04-13 2021-08-03 武汉大学 Forest biomass remote sensing mapping method based on multi-source data
WO2023095039A1 (en) * 2021-11-24 2023-06-01 Dendra Systems Ltd. Generating an above ground biomass density prediction model
CN114882361A (en) * 2022-05-13 2022-08-09 东南大学 Deep learning forest overground biological estimation method based on multi-source remote sensing fusion
CN115294147A (en) * 2022-06-17 2022-11-04 南宁师范大学 Method for estimating aboveground biomass of single trees and forests based on unmanned aerial vehicle laser radar
CN116224359A (en) * 2022-12-20 2023-06-06 中国地质调查局烟台海岸带地质调查中心 Forest overground biomass estimation method based on airborne LiDAR and small sample data

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
FAN WANG等: "Development of Estimation Models for Individual Tree Aboveground Biomass Based on TLS-Derived Parameters", 《FORESTS》, vol. 14, no. 2, 9 February 2023 (2023-02-09), pages 8 - 13 *
LINGTING LEI等: "Estimating Individual Tree Above-Ground Biomass of Chinese Fir Plantation: Exploring the Combination of Multi-Dimensional Features from UAV Oblique Photos", 《REMOTE SENSING》, vol. 14, no. 3, 21 January 2022 (2022-01-21), pages 3 - 10 *
赵颖慧等: "基于随机森林偏差校正和多源遥感数据的天然次生林地上生物量估测", 《中南林业科技大学学报》, vol. 41, no. 7, 15 July 2021 (2021-07-15), pages 100 - 102 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117387581A (en) * 2023-12-12 2024-01-12 北京麦麦趣耕科技有限公司 Cinnamomum camphora biomass estimation method, apparatus and storage medium
CN117387581B (en) * 2023-12-12 2024-02-13 北京麦麦趣耕科技有限公司 Cinnamomum camphora biomass estimation method, apparatus and storage medium

Also Published As

Publication number Publication date
CN116883834B (en) 2024-06-11

Similar Documents

Publication Publication Date Title
Hermosilla et al. Regional detection, characterization, and attribution of annual forest change from 1984 to 2012 using Landsat-derived time-series metrics
Moffiet et al. Airborne laser scanning: Exploratory data analysis indicates potential variables for classification of individual trees or forest stands according to species
US8897483B2 (en) System and method for inventorying vegetal substance
Peuhkurinen et al. Preharvest measurement of marked stands using airborne laser scanning
CN111414891A (en) Power transmission line channel tree height inversion method based on laser radar and optical remote sensing
Yu et al. Obtaining plotwise mean height and volume growth in boreal forests using multi‐temporal laser surveys and various change detection techniques
Blanchette et al. Predicting wood fiber attributes using local-scale metrics from terrestrial LiDAR data: A case study of Newfoundland conifer species
CN102959354A (en) Method and apparatus for for analyzing tree canopies with LiDAR data
CN116883834B (en) Pinus koraiensis overground biomass remote sensing estimation method and system
Pippuri et al. Exploring horizontal area-based metrics to discriminate the spatial pattern of trees and need for first thinning using airborne laser scanning
Rowell et al. Development and validation of fuel height models for terrestrial lidar–RxCADRE 2012
CN113156394B (en) Forest resource monitoring method and device based on laser radar and storage medium
Nyström et al. Change detection of mountain birch using multi-temporal ALS point clouds
CN113378785A (en) Forest type identification method and device
Douss et al. Extraction of individual trees based on Canopy Height Model to monitor the state of the forest
Beyene et al. Forest inventory and aboveground biomass estimation with terrestrial LiDAR in the tropical forest of Malaysia
Rouzbeh Kargar et al. Stem and root assessment in mangrove forests using a low-cost, rapid-scan terrestrial laser scanner
Räty et al. Comparing nearest neighbor configurations in the prediction of species-specific diameter distributions
Dalponte et al. Analysis on the use of multiple returns LiDAR data for the estimation of tree stems volume
Maltamo et al. A comparative study of the use of laser scanner data and field measurements in the prediction of crown height in boreal forests
CN114384534A (en) Tree barrier growth prediction analysis method for unmanned aerial vehicle line patrol
Wardius et al. Terrestrial laser scanning vs. manual methods for assessing complex forest stand structure: a comparative analysis on plenter forests
Puletti et al. Enhancing wall-to-wall forest structure mapping through detailed co-registration of airborne and terrestrial laser scanning data in Mediterranean forests
CN113343808A (en) Tropical forest resource measuring method based on satellite remote sensing technology
Popescu et al. Estimation of biomass components by airborne laser scanning

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