Accurate inversion method and system for aboveground biomass of urban vegetations considering vegetation type
The present disclosure provides an accurate inversion method and system for aboveground biomass of urban vegetations considering vegetation types. The method includes: based on different research regions, performing corresponding biomass calculation after plot sampling; preprocessing high-resolution optical data and space-borne photon counting LiDAR data, performing fine extraction on urban vegetation information, extracting spectral features of high-resolution remote-sensing image data and constructing an optical-biomass inversion sub-model, extracting vegetation height data of the space-borne photon counting LiDAR data and constructing a space-borne LiDAR-biomass inversion sub-model; with the support of the information on fine classification on the urban vegetation types, performing integration on the biomass inversion sub-models; based on the sample plot biomass data of the urban vegetations, obtaining an integrated inversion model fusing the horizontal information and the three-dimensional structure information of the urban land surface vegetations by training to perform accurate mapping of the aboveground biomass of the urban vegetations.
1 . An inversion method of aboveground biomass of urban vegetations considering vegetation types, the method comprising the following steps:
at step a, calculating sample plot biomass data within a research region, comprising achieving corresponding biomass calculation after plot sampling based on major urban vegetation types within the research region;
at step b, preprocessing high-resolution optical data and space-borne photon counting Light Detection and Ranging (LiDAR) data, comprising performing geometric and radiometric correction for the high-resolution optical data and performing data cleaning and data extraction on the space-borne photon counting LiDAR data;
at step c, performing fine extraction on urban vegetation information, comprising obtaining dominant vegetations growing within cities and traits thereof based on field sampling and various statistic data, defining urban surface land utilization types respectively, extracting features of different urban vegetation types, and performing fine classification on the urban vegetation types to provide basic data for an estimation of the aboveground biomass of the urban vegetations;
at step d, extracting spectral features of high-resolution remote-sensing image data and constructing an optical-biomass inversion sub-model, comprising with the spectral features as independent variable and with biomass AGB as dependent variable, by using sampling points, constructing a feature-biomass relationship; by using the sample plot biomass data obtain in the step a and information on fine classification on the urban vegetation types obtained in the step c, based on the obtained spectral features, selecting a first regression model based on an estimation effect to construct the optical-biomass sub-model to provide a sub-model for integrated inversion;
at step e, extracting vegetation height data of the space-borne photon counting LiDAR data and constructing a space-borne LiDAR-biomass inversion sub-model, comprising after performing filtering and denoising on urban adaptive photon point cloud data, based on a resolution of the high-resolution optical data, determining a size of a window of the vegetation height data and performing extraction on vegetation height features; by using the sample plot biomass data obtained in the step a and the information on fine classification on the urban vegetation types obtained in the step c, based on the obtained vegetation height features, selecting a second regression model with good based on the estimation effect to construct the space-borne LiDAR-biomass inversion sub-model for integrated inversion;
at step f, performing inversion on aboveground biomass of large-scope urban land surface vegetations, comprising, with the support of the information on the fine classification on the urban vegetation types obtained in the step c, based on an integration model constructed by decision tree, performing integration on the optical-biomass inversion sub-model obtained in step d and the space-borne LiDAR-biomass inversion sub-model obtained in step e; based on the sample plot biomass data of the urban vegetations obtained in the step a, obtaining an integrated inversion model fusing horizontal information and three-dimensional structure information of the urban land surface vegetations by training to perform a mapping of the aboveground biomass of the urban vegetations.
2 . The inversion method of claim 1 , wherein in the step c, classifying urban regional land surface to define the urban land surface types respectively as impervious surface LU i , water body LU w , bare land LU s , and vegetation LU v ; performing superpixel segmentation on high-resolution multispectral data of the research region obtained in the step b to extract spectral features, texture features and other features for the defined urban surface land utilization type to enable an inter-class difference of different land utilization types to be maximal and an intra-class difference to be minimal; by combining different influential factors, determining the land utilization types, with the expression shown below:
LU
=
a
×
f
(
Feature
S
)
+
b
×
+
(
1
-
a
-
b
)
×
f
(
Feature
O
)
wherein f(Feature S ) is a discriminative model of the spectral features, f(Feature T ) is a discriminative model of the texture features, f(Feature O ) is a discriminative model of other features, and a, b and c are respective weights of the discriminative models of the spectral features, the texture features, and other features;
performing further division based on a scope of LU v , and based on dominant trees, performing fine extraction on the urban vegetation types within the research region to respectively extract spectra features, texture feature and other features of different typical vegetations, where the vegetation type expression during the fine extraction of the urban vegetation types is
L
U
v
=
a
v
×
f
(
Feature
S
v
)
+
b
v
×
f
(
Feature
T
v
)
+
(
1
-
a
v
-
b
v
)
×
f
(
Feature
O
v
)
.
3 . The inversion method of claim 1 , wherein the implementation manner of constructing the optical-biomass inversion sub-model in the step d is that, with spectral features Feature S as independent variable, the biomass AGB as dependent variable, simulating a relationship between the spectral features Feature S and the biomass AGB, and with the sample points as true values, solution is made with the spectral feature Feature S and the biomass AGB to obtain the optical-biomass sub-model AGB S =f(Feature S ); during a model selection process, different regression analysis methods are compared to select a good model.
4 . The inversion method of claim 3 , wherein the implementation manner of constructing the space-borne LiDAR-biomass inversion sub-model in the step e is that, with space-borne LiDAR vegetation height features Feature as independent variable, and with the biomass AGB as dependent variable, a relationship between the space-borne LiDAR vegetation height features Feature L and the biomass AGB is simulated, and with the biomass AGB of the sample points as true values, solution is made with the space-borne LiDAR vegetation height features Feature L and the biomass AGB to obtain the space-borne LiDAR-biomass inversion sub-model AGB L =f(Feature L ); during a model selection process, different regression analysis methods are compared to select a good model.
5 . The inversion method of claim 4 , wherein the implementation manner of constructing a space-borne LiDAR-optical integrated model in the step f is that, with well-performing models obtained in the steps d and e as independent variable, and with the biomass AGB as dependent variable, by using the urban vegetation types obtained in the step c as constraint conditions, several relations between models and biomass are simulated by decision tree; with the sample points as true values, solution is made with the simulation biomass of several models and true biomass to obtain the space-borne LiDAR-optical integrated model.
6 . An inversion system for aboveground biomass of urban vegetations considering vegetation types, applied to perform the inversion method of the aboveground biomass of the urban vegetations considering vegetation types according to claim 1 ;
wherein the inversion system comprises a processor and a memory, wherein the memory is configured to store program instructions and the processor is configured to invoke the program instructions in the memory to perform the inversion method of the aboveground biomass of the urban vegetations considering vegetation types.
7 . The inversion system of claim 6 , wherein the processor is further configured to:
calculate the sample plot biomass data within the research region, comprising achieving corresponding biomass calculation after plot sampling based on the major urban vegetation types within the research region;
preprocess the high-resolution optical data and the space-borne photon counting LiDAR data, comprising performing the geometric and radiometric correction for the high-resolution optical data and performing data cleaning and data extraction on the space-borne photon counting LiDAR data;
perform fine extraction on the urban vegetation information within the research region, comprising performing feature extraction on different urban vegetation types, and performing fine classification on the urban vegetation types to provide basic data for the estimation of the aboveground biomass of the urban vegetations;
extract the spectral features of the high-resolution remote-sensing image data and construct the optical-biomass inversion sub-model, comprising by using the calculated sample plot biomass data and the information on fine classification on the urban vegetation types, based on the obtained spectral features, selecting a first regression model based on the estimation effect to construct the optical-biomass sub-model to provide a sub-model for integrated inversion;
extract the vegetation height data of the space-borne photon counting LiDAR data and construct the space-borne LiDAR-biomass inversion sub-model, comprising after performing filtering and denoising on the urban adaptive photon point cloud data, based on the resolution of the high-resolution optical data, determining the size of the window for statistics of the vegetation height data and performing extraction on the vegetation height features; by using the calculated sample plot biomass data and the information on fine classification on the urban vegetation types, based on the obtained vegetation height features, selecting a second regression model based on the estimation effect to construct the space-borne LiDAR-biomass inversion sub-model to provide a sub-model for integrated inversion;
perform inversion on aboveground biomass of large-scope urban land surface vegetations, comprising, with the support of the information on fine classification on the urban vegetation types, based on the integration model constructed by the decision tree, performing integration on the optical-biomass inversion sub-model and the space-borne LiDAR-biomass inversion sub-model; based on the sample plot biomass data of the urban vegetations, obtaining an integrated inversion model fusing the horizontal information and the three-dimensional structure information of the urban land surface vegetations by training to perform the mapping of the aboveground biomass of the urban vegetations.
8 . The inversion system of claim 6 , comprising a readable storage medium storing computer programs, wherein the computer programs are executed to perform the inversion method of the aboveground biomass of the urban vegetations considering vegetation types.
9 . The inversion method of claim 2 , wherein the implementation manner of constructing the optical-biomass inversion sub-model in the step d is that, with spectral features Feature S as independent variable, biomass AGB as dependent variable, simulating a relationship between the spectral features Feature S and the biomass AGB, and with the sample points as true values, solution is made with the spectral feature Feature S and the biomass AGB to obtain the optical-biomass sub-model AGB S =f(Feature S ); during a model selection process, different regression analysis methods are compared to select a good model.
10 . The inversion method of claim 9 , wherein the implementation manner of constructing the space-borne LiDAR-biomass inversion sub-model in the step e is that, with space-borne LiDAR vegetation height features Feature L as independent variable, and with the biomass AGB as dependent variable, a relationship between the space-borne LiDAR vegetation height features Feature L and the biomass AGB is simulated, and with the biomass AGB of the sample points as true values, solution is made with the space-borne LiDAR vegetation height features Feature L and the biomass AGB to obtain the space-borne LiDAR-biomass inversion sub-model AGB L =f(Feature L ); during a model selection process, different regression analysis methods are compared to select a good model.
11 . The inversion method of claim 10 , wherein the implementation manner of constructing a space-borne LiDAR-optical integrated model in the step f is that, with well-performing models obtained in the steps d and e as independent variable, and with the biomass AGB as dependent variable, by using the urban vegetation types obtained in the step c as constraint conditions, several relations between models and biomass are simulated by decision tree; with the sample points as true values, solution is made with the simulation biomass of several models and true biomass to obtain the space-borne LiDAR-optical integrated model.
12 . An inversion system for aboveground biomass of urban vegetations considering vegetation types, applied to perform the inversion method of the aboveground biomass of the urban vegetations considering vegetation types according to claim 2 ;
wherein the inversion system comprises a processor and a memory, wherein the memory is configured to store program instructions and the processor is configured to invoke the program instructions in the memory to perform the inversion method of the aboveground biomass of the urban vegetations considering vegetation types.
13 . The accurate-inversion system of claim 12 , wherein the processor is further configured to:
calculate the sample plot biomass data within the research region, comprising achieving corresponding biomass calculation after plot sampling based on the major urban vegetation types within the research region;
preprocess the high-resolution optical data and the space-borne photon counting LiDAR data, comprising performing the geometric and radiometric correction for the high-resolution optical data and performing data cleaning and data extraction on the space-borne photon counting LiDAR data;
perform fine extraction on the urban vegetation information within the research region, comprising performing feature extraction on different urban vegetation types, and performing fine classification on the urban vegetation types to provide basic data for the estimation of the aboveground biomass of the urban vegetations;
extract the spectral features of the high-resolution remote-sensing image data and construct the optical-biomass inversion sub-model, comprising by using the calculated sample plot biomass data and the information on fine classification on the urban vegetation types, based on the obtained spectral features, selecting a first regression model based on the estimation effect to construct the optical-biomass sub-model to provide a sub-model for integrated inversion;
extract the vegetation height data of the space-borne photon counting LiDAR data and construct the space-borne LiDAR-biomass inversion sub-model, comprising after performing filtering and denoising on the urban adaptive photon point cloud data, based on the resolution of the high-resolution optical data, determining the size of the window for statistics of the vegetation height data and performing extraction on the vegetation height features; by using the calculated sample plot biomass data and the information on fine classification on the urban vegetation types, based on the obtained vegetation height features, selecting a second regression model based on the estimation effect to construct the space-borne LiDAR-biomass inversion sub-model to provide a sub-model for integrated inversion;
perform inversion on aboveground biomass of large-scope urban land surface vegetations, comprising, with the support of the information on fine classification on the urban vegetation types, based on the integration model constructed by the decision tree, performing integration on the optical-biomass inversion sub-model and the space-borne LiDAR-biomass inversion sub-model; based on the sample plot biomass data of the urban vegetations obtained by the processor, obtaining an integrated inversion model fusing the horizontal information and the three-dimensional structure information of the urban land surface vegetations by training to perform the mapping of the aboveground biomass of the urban vegetations.
14 . The accurate-inversion system of claim 12 , comprising a readable storage medium storing computer programs, wherein the computer programs are executed to perform the inversion method of the aboveground biomass of the urban vegetations considering vegetation types.
15 . An inversion system for aboveground biomass of urban vegetations considering vegetation types, applied to perform the inversion method of the aboveground biomass of the urban vegetations considering vegetation types according to claim 3 ;
wherein the inversion system comprises a processor and a memory, wherein the memory is configured to store program instructions and the processor is configured to invoke the program instructions in the memory to perform the inversion method of the aboveground biomass of the urban vegetations considering vegetation types.
16 . The inversion system of claim 15 , wherein the processor is further configured to:
calculate the sample plot biomass data within the research region, comprising achieving corresponding biomass calculation after plot sampling based on the major urban vegetation types within the research region;
preprocess the high-resolution optical data and the space-borne photon counting LiDAR data, comprising performing the geometric and radiometric correction for the high-resolution optical data and performing data cleaning and data extraction on the space-borne photon counting LiDAR data;
perform fine extraction on the urban vegetation information within the research region, comprising performing feature extraction on different urban vegetation types, and performing fine classification on the urban vegetation types to provide basic data for the estimation of the aboveground biomass of the urban vegetations;
extract the spectral features of the high-resolution remote-sensing image data and construct the optical-biomass inversion sub-model, comprising by using the calculated sample plot biomass data and the information on fine classification on the urban vegetation types, based on the obtained spectral features, selecting a first regression model based on the estimation effect to construct the optical-biomass sub-model to provide a sub-model for integrated inversion;
extract the vegetation height data of the space-borne photon counting LiDAR data and construct the space-borne LiDAR-biomass inversion sub-model, comprising after performing filtering and denoising on the urban adaptive photon point cloud data, based on the resolution of the high-resolution optical data, determining the size of the window for statistics of the vegetation height data and performing extraction on the vegetation height features; by using the calculated sample plot biomass data and the information on fine classification on the urban vegetation types, based on the obtained vegetation height features, selecting a second regression model based on the estimation effect to construct the space-borne LiDAR-biomass inversion sub-model to provide a sub-model for integrated inversion;
perform inversion on aboveground biomass of large-scope urban land surface vegetations, comprising, with the support of the information on fine classification on the urban vegetation types, based on the integration model constructed by the decision tree, performing integration on the optical-biomass inversion sub-model and the space-borne LiDAR-biomass inversion sub-model; based on the sample plot biomass data of the urban vegetations, obtaining an integrated inversion model fusing the horizontal information and the three-dimensional structure information of the urban land surface vegetations by training to perform the mapping of the aboveground biomass of the urban vegetations.
17 . The inversion system of claim 15 , comprising a readable storage medium storing computer programs, wherein the computer programs are executed to perform the inversion method of the aboveground biomass of the urban vegetations considering vegetation types.