CN116381700A - Forest canopy height remote sensing estimation method and computer readable medium - Google Patents
Forest canopy height remote sensing estimation method and computer readable medium Download PDFInfo
- Publication number
- CN116381700A CN116381700A CN202310265544.XA CN202310265544A CN116381700A CN 116381700 A CN116381700 A CN 116381700A CN 202310265544 A CN202310265544 A CN 202310265544A CN 116381700 A CN116381700 A CN 116381700A
- Authority
- CN
- China
- Prior art keywords
- data
- constructing
- vegetation
- height
- remote sensing
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 61
- 239000013598 vector Substances 0.000 claims abstract description 47
- 239000011159 matrix material Substances 0.000 claims abstract description 28
- 238000010801 machine learning Methods 0.000 claims abstract description 23
- 238000007637 random forest analysis Methods 0.000 claims abstract description 22
- 230000003287 optical effect Effects 0.000 claims abstract description 19
- 238000001914 filtration Methods 0.000 claims abstract description 7
- 238000012216 screening Methods 0.000 claims description 41
- 238000007781 pre-processing Methods 0.000 claims description 18
- 238000006243 chemical reaction Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 7
- 238000011156 evaluation Methods 0.000 claims description 6
- 239000000463 material Substances 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 238000012163 sequencing technique Methods 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 238000012952 Resampling Methods 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 238000009432 framing Methods 0.000 claims description 3
- 230000005855 radiation Effects 0.000 claims description 3
- 230000007613 environmental effect Effects 0.000 abstract description 5
- 238000004364 calculation method Methods 0.000 abstract description 3
- 230000000694 effects Effects 0.000 description 13
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 8
- 229910052799 carbon Inorganic materials 0.000 description 8
- 230000008859 change Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000012417 linear regression Methods 0.000 description 3
- 230000008030 elimination Effects 0.000 description 2
- 238000003379 elimination reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012847 principal component analysis method Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000010187 selection method Methods 0.000 description 2
- 101001057166 Homo sapiens Protein EVI2A Proteins 0.000 description 1
- 102100027246 Protein EVI2A Human genes 0.000 description 1
- 241000422653 Thysanus Species 0.000 description 1
- 238000003915 air pollution Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
- 238000010792 warming Methods 0.000 description 1
- 239000002023 wood Substances 0.000 description 1
Images
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
- G01S17/02—Systems using the reflection of electromagnetic waves other than radio waves
- G01S17/06—Systems determining position data of a target
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The invention provides a forest canopy height remote sensing estimation method and a computer readable medium. The method uses the near-ground vegetation canopy height data of a forest area at a plurality of historical moments as real tag data, and corrects the relative height parameters of the foot print positions of the satellite-borne laser radar data by using a random forest model; and calculating multiband characteristic indexes by using passive optical remote sensing data, synthetic aperture radar data, interference radar data and topographic data at a plurality of historical moments as non-environmental variables, combining space weight matrix calculation to obtain a space characteristic vector set to construct an environmental variable set, using vegetation canopy height parameters at the corrected foot print positions of the satellite-borne laser radar data as real tag data, constructing an optimized lightweight gradient lifting machine learning model, and realizing forest canopy height remote sensing estimation based on space filtering values.
Description
Technical Field
The invention belongs to the technical field of remote sensing, and particularly relates to a forest canopy height remote sensing estimation method and a computer readable medium.
Background
Global climate change has become a focus of general attention of international society, and forest carbon recycling is one of important contents for studying global climate change, and about 77% of vegetation carbon in land ecosystems is stored in forests. In order to relieve the influence caused by global climate warming and protect a forest ecosystem, china starts a national carbon emission trading system, and the accurate estimation of the spatial distribution pattern of forest carbon reserves and the dynamic change thereof is the basis of carbon balance accounting of a land ecosystem. The vegetation canopy height is an important factor and basic data for estimating forest carbon reserves, and is also a key link for accounting carbon balance of the land ecosystem. The traditional forest canopy height investigation method is time-consuming and labor-consuming, only can acquire small-range information, a laser radar (Light Detection and Ranging, liDAR) can penetrate through a canopy so as to accurately acquire forest canopy vertical structure information, height information of the ground surface and the forest canopy, average canopy area, tree density and other information can be measured, great advantages are presented in estimating the forest canopy height, and particularly, the satellite-borne LiDAR data has huge unique advantages in large-area forest canopy height inversion and drawing. The remote sensing technology can objectively and rapidly acquire forest parameters of all scales, so that the remote sensing technology has great advantages in the aspect of forest canopy height drawing. The current optical remote sensing data is a main data source for estimating the height of the forest canopy, can accurately and rapidly estimate the height of the forest canopy in a large range, fully utilizes reflectivity, spatial characteristics, textures, vegetation index information and spectral information of different wave bands to reflect more detailed forest remote sensing information, and can solve the influence of cloud layers on image quality by using annual synthesis indexes and a method for complementing multiple annual images.
Modeling methods commonly used for forest canopy height estimation include traditional statistical regression methods and machine learning methods. The statistical regression method has a simple model and is convenient to calculate, but can only describe the linear relation between the forest canopy height and the multi-source remote sensing index, wherein the most representative model is a multiple linear regression model. The machine learning method can better reflect the complex nonlinear relation between the remote sensing index and the vegetation canopy height, wherein a representative random forest algorithm is adopted, a series of decision trees are constructed by randomly sampling samples, a predicted value is obtained by voting, and the algorithm can evaluate the importance of the variable and is also commonly used for selecting the characteristics. In addition, the support vector regression model has good effect on high-dimensional characteristics, and can explain nonlinear regression problems by setting different kernel functions, so that the support vector regression model has stronger generalization capability, but the calculated amount is large when the sample size is very large. The lightweight gradient lifting machine (Light Gradient Boosting Machine, lightGBM) has the advantages of huge anti-interference, anti-overfitting, high calculation speed, strong stability and the like in learning, and has small calculation amount, small required parameter and data input amount compared with a deep learning algorithm, and can obtain a good effect. Meanwhile, the model mobility is very strong, the weights of different models can be adjusted according to the condition of model input data in the training process, and the model mobility training method has certain application in the fields of ecology, environment and the like.
According to the first law of geography, the vegetation canopy height and the spatial distribution of environmental factors have spatial autocorrelation, but the influence of spatial effect in a model is not considered in the vegetation canopy height remote sensing inversion. The space effect is mapped into a feature vector through the space weight matrix feature decomposition constructed by a geographic unit, the space effect affecting the distribution of geographic variables is filtered out through screening a significant feature vector set, the space distribution mode of the geographic variables and the space effect of the geographic units can be added into a model as independent variables, and the variance expansion effect and the regression coefficient offset effect caused by the space autocorrelation in statistical modeling are considered, so that the influence of the space effect on the model is reduced, and the model precision is improved. The method has the advantages that the space influence is expressed by utilizing the feature vector of the space weight matrix, the method has strong expandability, can be directly applied to linear regression and generalized linear regression, and is applied to the fields of air pollution, vegetation coverage, landslide disasters and the like, and the result shows that the feature vector space filtering method can remarkably improve the accuracy of the model.
In summary, in the inversion remote sensing estimation and prediction of the vegetation canopy height model based on machine learning, the influence of the vegetation canopy height and the environmental factor space effect is not considered, and the model mobility is also poor. Therefore, a space effect lightweight gradient lifting machine learning method is needed to predict vegetation canopy height, and provide important support for forest carbon reserves and global climate change.
Disclosure of Invention
The invention aims to provide a forest canopy height estimation method and a computer-readable medium, which are used for realizing high-efficiency and high-precision establishment of a forest canopy height estimation model based on machine learning and remote sensing data and finishing canopy height drawing.
The technical scheme of the method is a forest canopy height estimation method of spatial filtering values, which comprises the following specific steps:
step 1: acquiring coordinate positions and relative height parameters of foot print points of the satellite-borne laser radar data in a forest area at a plurality of historical moments, and respectively acquiring the coordinate positions of the foot print points of the preprocessed satellite-borne laser radar data at the plurality of historical moments and the relative height parameters of the foot print points of the preprocessed satellite-borne laser radar data at the plurality of historical moments through preprocessing;
step 2: acquiring vegetation canopy height data, land utilization classification data, vegetation coverage data and terrain data acquired near the ground, and obtaining the vegetation canopy height data, the land utilization classification data, the vegetation coverage data and the terrain data which are the same as the coordinate positions of the foot print points of the preprocessed satellite-borne laser radar data at a plurality of historical moments through projection conversion matching;
step 3: the method comprises the steps of constructing land utilization classification data, vegetation coverage data and terrain data, which are the same as the coordinate positions of satellite-borne laser radar data footprint points after preprocessing at a plurality of historical moments, into candidate variable sets, and performing variable screening processing on the candidate variable sets and vegetation canopy height data acquired on the near ground by using a variable screening method to obtain characteristic variable sets which are obviously related to the vegetation canopy height data acquired on the near ground;
step 4, constructing a random forest learning model with optimal model parameters, calculating the relative height parameters of the preprocessed satellite-borne laser radar data at a plurality of historical moments through the random forest learning model with the optimal model parameters, and obtaining the vegetation canopy height after calibration of each footprint point of the satellite-borne laser radar data at each historical moment;
step 5: acquiring passive optical remote sensing data, synthetic aperture radar data, interference radar data and topographic data at a plurality of historical moments, and acquiring the preprocessed passive optical remote sensing data, the synthetic aperture radar data, the interference radar data and the topographic data at the plurality of historical moments by a radar remote sensing data preprocessing method;
step 6: respectively calculating a remote sensing vegetation index set and an annual material weather index set through the preprocessed passive optical remote sensing data at a plurality of historical moments, calculating an annual image statistic value set by combining the preprocessed passive optical remote sensing data, synthetic aperture radar data and interferometric radar data at a plurality of historical moments, constructing a topographic feature index set through topographic data at a plurality of historical moments, and further constructing a non-spatial feature vector set;
step 7: constructing a space weight matrix through the space distance relation between the coordinate positions of the foot print points of the preprocessed satellite-borne laser radar data at a plurality of historical moments, carrying out centering conversion on the space weight matrix, calculating to obtain characteristic values and characteristic vectors of the space weight matrix, and arranging the characteristic vectors of the space weight matrix from large to small according to the corresponding characteristic values to obtain the characteristic vectors of the space weight matrix after sequencing;
step 8: screening the feature vectors of the ordered space weight matrixes with feature values larger than a threshold value corresponding to the feature vectors of the ordered space weight matrixes, and constructing a space feature vector primary screening set; constructing a non-space feature vector set and a space feature vector preliminary screening set into an environment variable set;
step 9: and constructing an optimized lightweight gradient lifting machine learning model, and calculating the preprocessed passive optical remote sensing data, the synthetic aperture radar data and the interference radar data at a plurality of historical moments through the optimized lightweight gradient lifting machine learning model to obtain a forest canopy height predicted value.
Preferably, the pretreatment in step 1 comprises the following steps:
respectively and sequentially carrying out projection conversion, outlier processing and noise point removal processing on the coordinate positions and the relative height parameters of the satellite-borne laser radar data footprint points at a plurality of historical moments;
preferably, the projective transformation matching in step 2 is specifically as follows:
converting vegetation canopy height data, land utilization classification data, vegetation coverage data and terrain data into a coordinate system corresponding to the coordinate position of the satellite-borne laser radar data footprint point through projection, and matching the coordinate positions of the satellite-borne laser radar data footprint point after preprocessing at a plurality of historical moments in the step 1 to obtain vegetation canopy height data, land utilization classification data, vegetation coverage data and terrain data which are the same as the coordinate positions of the satellite-borne laser radar data footprint point after preprocessing at a plurality of historical moments;
preferably, the random forest learning model with optimal model parameters is constructed in the step 4, and the method specifically comprises the following steps:
constructing a random forest learning model, sequentially taking each element in a characteristic variable set which is obviously related to vegetation canopy height data collected near the ground as a sample to input, predicting to obtain a vegetation canopy height result, constructing a loss function model by combining the vegetation canopy height data collected near the ground as a real label, utilizing root mean square error and a decision coefficient as evaluation indexes, obtaining the random forest learning model of the optimal model parameters through optimization training,
preferably, the preprocessing method of the radar remote sensing data in the step 5 is as follows:
carrying out radiation correction, missing value processing, resampling, splicing, framing and projection conversion in sequence;
preferably, the remote sensing vegetation index set in the step 6 is composed of a plurality of types of remote sensing vegetation index indexes;
the annual weatherable index set in the step 6 is composed of various types of annual weatherable indexes;
step 6, the annual image statistic value set consists of various types of annual image statistic value indexes;
step 6, the topographic feature index set is composed of various types of topographic feature indexes;
step 6, constructing a non-spatial feature vector set, which is specifically as follows:
sequentially using a variable screening method to carry out screening treatment on variables which are obviously related to the vegetation canopy height after calibration of each footprint point of the satellite-borne laser radar data at each historical moment in the step 5 to obtain a remote sensing vegetation index set after screening treatment, an annual material weather index set after screening treatment, an annual image statistic value set after screening treatment and a topographic feature index set after screening treatment;
preferably, the optimized lightweight gradient lifting machine learning model in step 9 is specifically as follows:
the method comprises the steps of constructing a lightweight gradient lifting machine learning model, sequentially inputting an environment variable set as a sample to predict to obtain vegetation canopy height, constructing a loss function model by combining the vegetation canopy height calibrated by each footprint point of satellite-borne laser radar data at each historical moment as a real label, and obtaining the optimized lightweight gradient lifting machine learning model through optimization training by using root mean square error and a decision coefficient as evaluation indexes.
The invention also provides a computer readable medium storing a computer program for execution by an electronic device, which when run on the electronic device causes the electronic device to perform the steps of the method for estimating the forest canopy height of the spatial filter.
The beneficial effects of the invention are as follows: the invention provides a forest canopy height remote sensing prediction method of a lightweight gradient hoisting machine based on a spatial filtering value, which considers complex nonlinear relation between the vegetation canopy height and remote sensing multiband and environmental factors in the vegetation canopy height estimation, considers the influence of a spatial effect and adds the space effect into a lightweight gradient hoisting machine model in the form of a spatial characteristic vector, can construct a vegetation canopy height model more accurately, and can improve the robustness of the model, thereby improving the accuracy of vegetation canopy height prediction.
Drawings
Fig. 1: the method of the embodiment of the invention is a flow chart.
Fig. 2: a highly calibrated model schematic of an embodiment of the present invention.
Fig. 3: the embodiment of the invention provides a lightweight gradient lifting machine model schematic diagram.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
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.
The following describes a method for estimating the height of a forest canopy by using a spatial filter according to the technical scheme of the embodiment of the invention with reference to fig. 1 to 3, which is specifically as follows:
a flow chart of the method of the present invention is shown in fig. 1.
Step 1: the method comprises the steps of obtaining coordinate positions and relative height parameters of foot print points of the satellite-borne laser radar data in a forest area at a plurality of historical moments, and obtaining the coordinate positions and the relative height parameters of the foot print points of the preprocessed satellite-borne laser radar data at the plurality of historical moments through preprocessing. The acquired metadata of the satellite-borne laser radar comprises, but is not limited to, parameters affecting the acquisition quality of the satellite-borne laser radar data, such as footprint coordinates, percentage-related height parameters, data acquisition time, laser model, acquisition gesture, footprint topography and the like;
the pretreatment process in the step 1 is as follows:
respectively and sequentially carrying out projection conversion, outlier processing and noise point removal processing on the coordinate positions and the relative height parameters of the satellite-borne laser radar data footprint points at a plurality of historical moments;
step 2: the method comprises the steps of obtaining vegetation canopy height data, land utilization classification data, vegetation coverage data and terrain data acquired near the ground, and obtaining the vegetation canopy height data, the land utilization classification data, the vegetation coverage data and the terrain data which are the same as the coordinate positions of the foot print points of the preprocessed satellite-borne laser radar data at a plurality of historical moments through projection conversion matching. Vegetation canopy height data collected on the ground are divided into four categories: the method comprises the steps of a vegetation canopy height model collected by a laser radar carried by an aviation aircraft, a vegetation canopy height model collected by a laser radar carried by an unmanned aerial vehicle, a vegetation canopy height model collected by a foundation laser radar and a manually measured single wood height. The method comprises the steps of carrying out a first treatment on the surface of the
The projective transformation matching in the step 2 is specifically as follows:
converting vegetation canopy height data, land utilization classification data, vegetation coverage data and terrain data into a coordinate system corresponding to the coordinate position of the satellite-borne laser radar data footprint point through projection, and matching the coordinate positions of the satellite-borne laser radar data footprint point after preprocessing at a plurality of historical moments in the step 1 to obtain vegetation canopy height data, land utilization classification data, vegetation coverage data and terrain data which are the same as the coordinate positions of the satellite-borne laser radar data footprint point after preprocessing at a plurality of historical moments;
step 3: the method comprises the steps of constructing land utilization classification data, vegetation coverage data and terrain data which are the same as the coordinate positions of foot print points of satellite-borne laser radar data after preprocessing at a plurality of historical moments into candidate variable sets, and carrying out variable screening processing on the candidate variable sets and vegetation canopy height data acquired near the ground by using a variable screening method, wherein the common variable screening method comprises random forest, minimum shrinkage and selection operator, recursive feature elimination, a selection model based on LightGBM, a stepwise regression model, a principal component analysis method, a subset selection method and the like, so as to obtain a feature variable set which is obviously related to the vegetation canopy height data acquired near the ground;
step 4, constructing a random forest learning model with optimal model parameters, calculating the relative height parameters of the preprocessed satellite-borne laser radar data at a plurality of historical moments through the random forest learning model with the optimal model parameters, and obtaining the vegetation canopy height after calibration of each footprint point of the satellite-borne laser radar data at each historical moment;
and 4, constructing a random forest learning model with optimal model parameters, wherein the random forest learning model is specifically as follows:
and (3) constructing a random forest learning model, sequentially taking each element in a characteristic variable set which is obviously related to the vegetation canopy height data collected on the near ground as a sample, inputting the sample, predicting to obtain a vegetation canopy height result, constructing a loss function model by combining the vegetation canopy height data collected on the near ground as a real label, and obtaining the random forest learning model with optimal model parameters through optimization training by taking root mean square error and a decision coefficient as evaluation indexes, wherein the random forest learning model is shown in figure 2.
Step 5: acquiring passive optical remote sensing data, synthetic aperture radar data, interference radar data and topographic data at a plurality of historical moments, wherein the passive optical remote sensing data comprises, but is not limited to, active and passive continuous remote sensing data such as Landsat series satellite data, sentinel series satellite data, PALSAR series satellite data and the like, and acquiring the preprocessed passive optical remote sensing data, the synthetic aperture radar data, the interference radar data and the topographic data at the plurality of historical moments by a radar remote sensing data preprocessing method;
the radar remote sensing data preprocessing method in the step 5 is as follows:
carrying out radiation correction, missing value processing, resampling, splicing, framing and projection conversion in sequence;
step 6: respectively calculating a remote sensing vegetation index set and an annual material weather index set through the preprocessed passive optical remote sensing data at a plurality of historical moments, calculating an annual image statistic value set by combining the preprocessed passive optical remote sensing data, synthetic aperture radar data and interferometric radar data at a plurality of historical moments, constructing a topographic feature index set through topographic data at a plurality of historical moments, and further constructing a non-spatial feature vector set;
the remote sensing vegetation index set in the step 6 is composed of various remote sensing vegetation index indexes, including common remote sensing indexes (EVI, EVI2 and RVI), and related indexes of the thysanus cap transformation (brightness, humidity, greenness, humidity greenness difference value) and the like;
the annual weathered index set in the step 6 is composed of various types of annual weathered indexes, including basic weathered parameters such as an NDVI value at the beginning of summer, an NDVI value in the fallen leaves and the like;
the annual image statistic value set in the step 6 is composed of various types of annual image statistic value indexes, including basic statistic indexes such as maximum value, minimum value, average value, standard deviation, median value and the like;
the terrain characteristic index set in the step 6 is composed of various types of terrain characteristic indexes including gradient, slope direction and elevation;
step 6, constructing a non-spatial feature vector set, which is specifically as follows:
sequentially using a variable screening method to a remote sensing vegetation index set, an annual material index set, an annual image statistical value set and a topographic feature index set, and carrying out screening treatment on variables which are obviously related to the vegetation canopy height after calibration of each footprint point of the satellite-borne laser radar data at each historical moment in the step 5, wherein the common variable screening method comprises a random forest, a minimum shrinkage and selection operator, recursive feature elimination, a selection model based on a LightGBM, a stepwise regression model, a principal component analysis method, a subset selection method and the like, so as to obtain the remote sensing vegetation index set after screening treatment, the annual material index set after screening treatment, the annual image statistical value set after screening treatment and the topographic feature index set after screening treatment;
step 7: constructing a space weight matrix through the space distance relation among the coordinate positions of the foot print points of the preprocessed satellite-borne laser radar data at a plurality of historical moments, carrying out centering conversion on the space weight matrix, calculating to obtain characteristic values and characteristic vectors of the space weight matrix, and arranging the characteristic vectors of the space weight matrix from large to small according to the corresponding characteristic values to obtain the characteristic vectors of the space weight matrix after sequencing. The spatial weight matrix is divided into two classes: a distance-based weight matrix and a topology-based weight matrix. The weight matrix based on the distance mainly aims at the foot print point position of the satellite-borne laser radar, and Gaussian, exponential, double square, triple cube and the like can be selected as weight generating functions; the weight matrix based on the topological relation mainly aims at the grid data of the related ground information acquired by the remote sensing sensor, and can be constructed in adjacent modes such as vehicle adjacent (hook) mode, rear adjacent (Queen) mode and the like; .
Step 8: screening the feature vectors of the ordered space weight matrixes with feature values larger than a threshold value corresponding to the feature vectors of the ordered space weight matrixes, and constructing a space feature vector primary screening set; constructing a non-space feature vector set and a space feature vector preliminary screening set into an environment variable set;
step 9: and constructing an optimized lightweight gradient lifting machine learning model, and calculating the preprocessed passive optical remote sensing data, the synthetic aperture radar data and the interference radar data at a plurality of historical moments through the optimized lightweight gradient lifting machine learning model to obtain a forest canopy height predicted value.
And 9, the optimized lightweight gradient lifting machine learning model is specifically as follows:
the method comprises the steps of constructing a lightweight gradient lifting machine learning model, sequentially inputting an environment variable set as a sample to predict to obtain vegetation canopy height, constructing a loss function model by combining the vegetation canopy height calibrated by each footprint point of satellite-borne laser radar data at each historical moment as a real label, and obtaining the optimized lightweight gradient lifting machine learning model through optimization training by using root mean square error and a decision coefficient as evaluation indexes.
Particular embodiments of the present invention also provide a computer readable medium.
The computer readable medium is a server workstation;
the server workstation stores a computer program executed by the electronic device, and when the computer program runs on the electronic device, the electronic device executes the steps of the method for estimating the forest canopy height of the cross-filter value.
It should be understood that parts of the specification not specifically set forth herein are all prior art.
It should be understood that the foregoing description of the preferred embodiments is not intended to limit the scope of the invention, but rather to limit the scope of the claims, and that those skilled in the art can make substitutions or modifications without departing from the scope of the invention as set forth in the appended claims.
Claims (8)
1. A method for estimating the height of a forest canopy of a space filter value is characterized by comprising the following steps:
constructing a candidate variable set, and performing variable screening treatment on the candidate variable set and vegetation canopy height data acquired near the ground by using a variable screening method to obtain a characteristic variable set which is obviously related to the vegetation canopy height data acquired near the ground;
constructing a terrain feature index set through terrain data at a plurality of historical moments, and further constructing a non-space feature vector set;
constructing a space weight matrix, carrying out centering conversion on the space weight matrix, calculating to obtain characteristic values and characteristic vectors of the space weight matrix, and arranging the characteristic vectors of the space weight matrix from large to small according to the corresponding characteristic values to obtain the characteristic vectors of the space weight matrix after sequencing;
screening the feature vectors of the ordered space weight matrixes with feature values larger than a threshold value corresponding to the feature vectors of the ordered space weight matrixes, and constructing a space feature vector primary screening set; constructing a non-space feature vector set and a space feature vector preliminary screening set into an environment variable set;
and constructing an optimized lightweight gradient lifting machine learning model, and calculating the preprocessed passive optical remote sensing data, the synthetic aperture radar data and the interference radar data at a plurality of historical moments through the optimized lightweight gradient lifting machine learning model to obtain a forest canopy height predicted value.
2. A method for estimating the height of a canopy in a forest according to claim 1, comprising the steps of
Step 1: acquiring coordinate positions and relative height parameters of foot print points of the satellite-borne laser radar data in a forest area at a plurality of historical moments, and acquiring the coordinate positions of the foot print points of the preprocessed satellite-borne laser radar data at the plurality of historical moments and the relative height parameters of the foot print points of the preprocessed satellite-borne laser radar data at the plurality of historical moments through projection conversion, outlier processing and noise point removal processing respectively;
step 2: acquiring vegetation canopy height data, land utilization classification data, vegetation coverage data and terrain data acquired near the ground, and obtaining the vegetation canopy height data, the land utilization classification data, the vegetation coverage data and the terrain data which are the same as the coordinate positions of the foot print points of the preprocessed satellite-borne laser radar data at a plurality of historical moments through projection conversion matching;
step 3: the method comprises the steps of constructing land utilization classification data, vegetation coverage data and terrain data, which are the same as the coordinate positions of satellite-borne laser radar data footprint points after preprocessing at a plurality of historical moments, into candidate variable sets, and performing variable screening processing on the candidate variable sets and vegetation canopy height data acquired on the near ground by using a variable screening method to obtain characteristic variable sets which are obviously related to the vegetation canopy height data acquired on the near ground;
step 4, constructing a random forest learning model with optimal model parameters, calculating the relative height parameters of the preprocessed satellite-borne laser radar data at a plurality of historical moments through the random forest learning model with the optimal model parameters, and obtaining the vegetation canopy height after calibration of each footprint point of the satellite-borne laser radar data at each historical moment;
step 5: acquiring passive optical remote sensing data, synthetic aperture radar data, interference radar data and topographic data at a plurality of historical moments, and acquiring the preprocessed passive optical remote sensing data, the synthetic aperture radar data, the interference radar data and the topographic data at the plurality of historical moments by a radar remote sensing data preprocessing method;
step 6: respectively calculating a remote sensing vegetation index set and an annual material weather index set through the preprocessed passive optical remote sensing data at a plurality of historical moments, calculating an annual image statistic value set by combining the preprocessed passive optical remote sensing data, synthetic aperture radar data and interferometric radar data at a plurality of historical moments, constructing a topographic feature index set through topographic data at a plurality of historical moments, and further constructing a non-spatial feature vector set;
step 7: constructing a space weight matrix through the space distance relation between the coordinate positions of the foot print points of the preprocessed satellite-borne laser radar data at a plurality of historical moments, carrying out centering conversion on the space weight matrix, calculating to obtain characteristic values and characteristic vectors of the space weight matrix, and arranging the characteristic vectors of the space weight matrix from large to small according to the corresponding characteristic values to obtain the characteristic vectors of the space weight matrix after sequencing;
step 8: screening the feature vectors of the ordered space weight matrixes with feature values larger than a threshold value corresponding to the feature vectors of the ordered space weight matrixes, and constructing a space feature vector primary screening set; constructing a non-space feature vector set and a space feature vector preliminary screening set into an environment variable set;
step 9: and constructing an optimized lightweight gradient lifting machine learning model, and calculating the preprocessed passive optical remote sensing data, the synthetic aperture radar data and the interference radar data at a plurality of historical moments through the optimized lightweight gradient lifting machine learning model to obtain a forest canopy height predicted value.
3. The method for estimating the height of a canopy in a forest of a spatial filter according to claim 2, wherein:
the projective transformation matching in the step 2 is specifically as follows:
and (2) converting vegetation canopy height data, land utilization classification data, vegetation coverage data and terrain data into a coordinate system corresponding to the coordinate positions of the satellite-borne laser radar data footprint points through projection, and matching the coordinate positions of the satellite-borne laser radar data footprint points after preprocessing at a plurality of historical moments in combination with the coordinate positions of the satellite-borne laser radar data footprint points after preprocessing at the step (1) to obtain vegetation canopy height data, land utilization classification data, vegetation coverage data and terrain data which are the same as the coordinate positions of the satellite-borne laser radar data footprint points after preprocessing at a plurality of historical moments.
4. A method for estimating a height of a canopy in a forest according to claim 3, wherein:
and 4, constructing a random forest learning model with optimal model parameters, wherein the random forest learning model is specifically as follows:
and constructing a random forest learning model, sequentially taking each element in a characteristic variable set which is obviously related to the vegetation canopy height data collected near the ground as a sample, predicting to obtain a vegetation canopy height result, constructing a loss function model by combining the vegetation canopy height data collected near the ground as a real label, and obtaining the random forest learning model with optimal model parameters through optimization training by taking root mean square error and a decision coefficient as evaluation indexes.
5. The method for estimating a height of a canopy in a forest of spatial filtering values according to claim 4, wherein:
the radar remote sensing data preprocessing method in the step 5 is as follows:
and carrying out radiation correction, missing value processing, resampling, splicing, framing and projection conversion in sequence.
6. The method for estimating a height of a canopy in a forest of spatial filtering values according to claim 5, wherein:
step 6, the remote sensing vegetation index set consists of a plurality of types of remote sensing vegetation index indexes;
the annual weatherable index set in the step 6 is composed of various types of annual weatherable indexes;
step 6, the annual image statistic value set consists of various types of annual image statistic value indexes;
step 6, the topographic feature index set is composed of various types of topographic feature indexes;
step 6, constructing a non-spatial feature vector set, which is specifically as follows:
and (3) sequentially using a variable screening method to carry out screening treatment on variables which are obviously related to the vegetation canopy height after calibration of each footprint point of the satellite-borne laser radar data at each historical moment in the step (5) to obtain a remote sensing vegetation index set after screening treatment, an annual weathered index set after screening treatment, an annual image statistic value set after screening treatment and a topographic feature index set after screening treatment.
7. The method for estimating a height of a canopy in a forest of spatial filtering values according to claim 6, wherein:
and 9, the optimized lightweight gradient lifting machine learning model is specifically as follows:
the method comprises the steps of constructing a lightweight gradient lifting machine learning model, sequentially inputting an environment variable set as a sample to predict to obtain vegetation canopy height, constructing a loss function model by combining the vegetation canopy height calibrated by each footprint point of satellite-borne laser radar data at each historical moment as a real label, and obtaining the optimized lightweight gradient lifting machine learning model through optimization training by using root mean square error and a decision coefficient as evaluation indexes.
8. A computer readable medium, characterized in that it stores a computer program for execution by an electronic device, which computer program, when run on the electronic device, causes the electronic device to perform the steps of the method according to any one of claims 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310265544.XA CN116381700A (en) | 2023-03-17 | 2023-03-17 | Forest canopy height remote sensing estimation method and computer readable medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310265544.XA CN116381700A (en) | 2023-03-17 | 2023-03-17 | Forest canopy height remote sensing estimation method and computer readable medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116381700A true CN116381700A (en) | 2023-07-04 |
Family
ID=86974250
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310265544.XA Pending CN116381700A (en) | 2023-03-17 | 2023-03-17 | Forest canopy height remote sensing estimation method and computer readable medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116381700A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117520733A (en) * | 2024-01-05 | 2024-02-06 | 云南师范大学 | Forest canopy height and geographic environment covariate relation determination method and system |
-
2023
- 2023-03-17 CN CN202310265544.XA patent/CN116381700A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117520733A (en) * | 2024-01-05 | 2024-02-06 | 云南师范大学 | Forest canopy height and geographic environment covariate relation determination method and system |
CN117520733B (en) * | 2024-01-05 | 2024-03-19 | 云南师范大学 | Forest canopy height and geographic environment covariate relation determination method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Halme et al. | Utility of hyperspectral compared to multispectral remote sensing data in estimating forest biomass and structure variables in Finnish boreal forest | |
CN104656098B (en) | A kind of method of remote sensing forest biomass inverting | |
CN114037911B (en) | Large-scale forest height remote sensing inversion method considering ecological zoning | |
CN112434617B (en) | Forest biomass change monitoring method and system based on multi-source remote sensing data | |
CN113034689A (en) | Laser point cloud-based terrain three-dimensional model, terrain map construction method and system, and storage medium | |
CN112348812B (en) | Forest stand age information measurement method and device | |
CN114740180A (en) | Soil organic carbon estimation method and device based on multi-source remote sensing data | |
CN112699959B (en) | Multi-source multi-scale precipitation data fusion method and device based on energy functional model | |
Melin et al. | Assessing the performance of aerial image point cloud and spectral metrics in predicting boreal forest canopy cover | |
CN113204998B (en) | Airborne point cloud forest ecological estimation method and system based on single wood scale | |
CN112381013A (en) | Urban vegetation inversion method and system based on high-resolution remote sensing image | |
CN112861435B (en) | Mangrove quality remote sensing inversion method and intelligent terminal | |
CN115372282B (en) | Farmland soil water content monitoring method based on hyperspectral image of unmanned aerial vehicle | |
CN114723149A (en) | Soil moisture content prediction method and device, electronic equipment and storage medium | |
CN108764527B (en) | Screening method for soil organic carbon library time-space dynamic prediction optimal environment variables | |
CN111222539A (en) | Method for optimizing and expanding supervision classification samples based on multi-source multi-temporal remote sensing image | |
CN116381700A (en) | Forest canopy height remote sensing estimation method and computer readable medium | |
CN113870425A (en) | Forest accumulation space mapping method based on random forest and multi-source remote sensing technology | |
CN115761303A (en) | Ground object classification method based on airborne laser radar point cloud and remote sensing image data | |
CN115147721A (en) | Remote sensing inversion method and device for forest accumulation | |
CN114819737A (en) | Carbon reserve estimation method, system and storage medium for road and road area vegetation | |
CN111814666B (en) | Single tree parameter extraction method, system, medium and equipment under complex forest stand | |
CN117035174A (en) | Method and system for estimating biomass on single-woodland of casuarina equisetifolia | |
Lou et al. | Recognition of thaw slumps based on machine learning and UAVs: A case study in the Qilian Mountains, northeastern Qinghai-Tibet Plateau | |
GB2620469A (en) | Spatial prediction and evaluation method of soil organic matter content based on partition algorithm |
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 |