CN116912690A - Forest leaf area index inversion acquisition method and system based on data fusion - Google Patents
Forest leaf area index inversion acquisition method and system based on data fusion Download PDFInfo
- Publication number
- CN116912690A CN116912690A CN202310894150.0A CN202310894150A CN116912690A CN 116912690 A CN116912690 A CN 116912690A CN 202310894150 A CN202310894150 A CN 202310894150A CN 116912690 A CN116912690 A CN 116912690A
- Authority
- CN
- China
- Prior art keywords
- forest
- image
- data
- leaf area
- index
- 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 54
- 230000004927 fusion Effects 0.000 title claims abstract description 16
- 238000011160 research Methods 0.000 claims abstract description 45
- 230000002776 aggregation Effects 0.000 claims abstract description 16
- 238000004220 aggregation Methods 0.000 claims abstract description 16
- 238000012545 processing Methods 0.000 claims abstract description 15
- 238000002310 reflectometry Methods 0.000 claims description 25
- 230000008569 process Effects 0.000 claims description 15
- 238000012216 screening Methods 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 9
- 229930002875 chlorophyll Natural products 0.000 claims description 6
- 235000019804 chlorophyll Nutrition 0.000 claims description 6
- ATNHDLDRLWWWCB-AENOIHSZSA-M chlorophyll a Chemical compound C1([C@@H](C(=O)OC)C(=O)C2=C3C)=C2N2C3=CC(C(CC)=C3C)=[N+]4C3=CC3=C(C=C)C(C)=C5N3[Mg-2]42[N+]2=C1[C@@H](CCC(=O)OC\C=C(/C)CCC[C@H](C)CCC[C@H](C)CCCC(C)C)[C@H](C)C2=C5 ATNHDLDRLWWWCB-AENOIHSZSA-M 0.000 claims description 6
- 238000007637 random forest analysis Methods 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 claims description 4
- WPPDXAHGCGPUPK-UHFFFAOYSA-N red 2 Chemical compound C1=CC=CC=C1C(C1=CC=CC=C11)=C(C=2C=3C4=CC=C5C6=CC=C7C8=C(C=9C=CC=CC=9)C9=CC=CC=C9C(C=9C=CC=CC=9)=C8C8=CC=C(C6=C87)C(C=35)=CC=2)C4=C1C1=CC=CC=C1 WPPDXAHGCGPUPK-UHFFFAOYSA-N 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 2
- 238000006116 polymerization reaction Methods 0.000 claims description 2
- 239000011521 glass Substances 0.000 description 18
- 230000003595 spectral effect Effects 0.000 description 8
- 238000005259 measurement Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 230000003287 optical effect Effects 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 4
- 238000007500 overflow downdraw method Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000012876 topography Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 239000002028 Biomass Substances 0.000 description 2
- 241000218631 Coniferophyta Species 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 241000196324 Embryophyta Species 0.000 description 1
- 208000036855 Left sided atrial isomerism Diseases 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 229920006395 saturated elastomer Polymers 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
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
- 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
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- 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
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
-
- 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
-
- 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
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a forest leaf area index inversion acquisition method and system based on data fusion. The acquisition method comprises the following steps: performing cloud removal processing and cutting on remote sensing image data of a research area to obtain a segmented image; based on land coverage data, forest pixels in each divided image are obtained, and space aggregation matched with resolution is carried out to obtain an aggregated remote sensing image; extracting vegetation indexes from the segmented image map; acquiring forest canopy height data of a research area; and taking reflectance values, vegetation indexes, forest canopy height data and solar sensor related parameters of different wavebands in the aggregated remote sensing image as inputs, and taking forest leaf area indexes as outputs to construct a forest leaf area index fusion prediction model, so as to obtain a prediction result. The LAI distribution result obtained by the method has more space details, has strong applicability on mountain forest ecosystems with large space heterogeneity, and can accurately and rapidly monitor mountain vegetation parameters.
Description
Technical Field
The invention relates to the field of application of remote sensing technology, in particular to the field of obtaining forest leaf area indexes through the remote sensing technology.
Background
Forest canopy height (Forest height), forest Leaf Area Index (LAI) and other Forest parameters are indispensable important characteristic parameters in an ecological process model and a Forest carbon circulation model, can reflect the net primary productivity level of a Forest and the advantages and disadvantages of Forest resources to a certain extent, and are also commonly used for constructing a Forest biomass model. The accurate estimation and acquisition of the forest leaf area index and the like are beneficial to improving the modeling accuracy of forest ground biomass, and have important significance for deeply understanding the regional and even the global climate and environment change law.
At present, a complex inversion model is adopted for acquiring forest parameters, more input parameters and larger field actual measurement working intensity are adopted, the physical model is constructed based on a flat earth surface, the influence of a topography factor is not included, the obtained result deviates from a true value, and the accuracy is not enough. The method aims at the problems of high actual measurement working difficulty, high data acquisition cost, low precision and the like, and an effective technical means is to fully utilize a remote sensing technology, and improve the estimation precision of forest parameters and reduce the difficulty by combining remote sensing data with the existing method. However, currently used LAI remote sensing calculation products such as MODIS LAI and other global surface parameter products are low-resolution products, and are more suitable for large-scale research, while medium-resolution LAI products for small-area research do not consider vegetation diversity and regional differences, and the obtained results are not accurate enough.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide the inversion acquisition method for the forest leaf area index, which fully considers the problems of topography factors and scale matching and fuses multispectral remote sensing data, and can be used in areas such as mountain areas with large topography fluctuation, and the like, so that the accurate and high-resolution forest leaf area index distribution condition can be obtained.
The technical scheme of the invention is as follows:
a forest leaf area index inversion acquisition method based on data fusion comprises the following steps:
s1, preprocessing remote sensing image data of a region to be obtained with forest leaf area indexes, namely a research region, and cutting pixels to obtain a segmented image map which eliminates cloud pixels and is segmented into a plurality of research cells;
s2, extracting and obtaining mask data of a forest area from land coverage data of a research area by a mask extraction method, applying the obtained mask data to the segmented image obtained in the S1 to obtain forest pixels in each segmented image, and performing spatial aggregation on the obtained forest pixels to obtain an aggregate image with improved resolution;
s3, extracting vegetation indexes of a research area from the segmented image according to reflection values of different wave bands in the segmented image;
s4, acquiring forest canopy height data through a geographic information system of a research area, and screening out heterogeneity data in the forest canopy height data through standard deviation and variation coefficients of the forest canopy height data to acquire screened canopy height data;
s5, obtaining a forest leaf area index of the research area through a remote sensing data inversion tool;
s6, constructing and training a forest leaf area index fusion prediction model, and obtaining a forest leaf area index inversion estimated value and a forest leaf area index distribution prediction graph through the trained or trained and tested model; the fusion prediction model is a random forest regression model, the reflectance values of different wave bands in the aggregate image or the optimized aggregate image are used as inputs, the vegetation index values, and the screened canopy height data and the solar zenith angle, the observed zenith angle and the relative azimuth angle obtained according to the solar sensor are used as outputs.
According to some preferred embodiments of the present invention, the remote sensing image data in S1 is remote sensing image data subjected to radiometric calibration and atmospheric correction.
According to some preferred embodiments of the invention, the obtaining of the segmented image map comprises:
s11, performing cloud removal processing on remote sensing image data subjected to radiometric calibration and atmospheric correction in a research area through a remote sensing image processing product to obtain image data subjected to cloud removal;
s12, setting reflectivity thresholds of visible light and near infrared light bands in the image data after cloud removal, and filtering out residual cloud pixels according to the reflectivity thresholds to obtain a preprocessed cloud removal image;
s13, cutting the preprocessed cloud removal image according to boundary division conditions of different research cells to obtain a segmented image corresponding to each research cell.
According to some preferred embodiments of the invention, the reflectivity threshold is 0.3.
According to some preferred embodiments of the invention, the S2 comprises:
s21, extracting a forest region mask through a land coverage data set with the resolution of 10m, and applying the extracted forest region mask to the segmented image obtained in the step S1 to obtain an extracted image of each segmented image, namely a segmented extracted image, wherein only forest region pixels are reserved;
s22, carrying out space aggregation on the segmentation extraction images until the resolution reaches 250m, and obtaining the aggregation image with the improved resolution.
According to some preferred embodiments of the invention, the S2 further comprises:
and matching the pixels in the obtained aggregate image with the forest leaf area index of the research area through a simple average method, calculating standard deviation and variation coefficient in the matching, and screening heterogeneous pixels in the aggregate image according to the variation coefficient to obtain an optimized aggregate image.
According to some preferred embodiments of the invention, the S2 further comprises:
space polymerization is carried out on the forest region mask with the resolution of 10m obtained in the step S21 until the resolution reaches 250m, so that a forest coverage image is obtained;
and removing the misclassified pixels in the aggregate image according to the comparison between the obtained forest coverage image and the aggregate image, so as to obtain an optimized aggregate image.
According to some preferred embodiments of the invention, the vegetation index comprises: normalized vegetation index NDVI, building index IBI, differential vegetation index DVI, ratio vegetation index RVI, and green chlorophyll index CIg.
According to some preferred embodiments of the present invention, the aggregate image is obtained according to a Sentinel-2 remote sensing image, and the reflectance values of different wavebands in the aggregate image or the optimized aggregate image include: the image has a reflectance of a green band, a red band, a near infrared band, a red 1 band, a red 2 band, a red 3 band, a far infrared 1 band, and a far infrared 2 band at a resolution of 10m and a resolution of 20 m.
The invention further provides a forest leaf area index inversion system, which is applied to any of the above forest leaf area index inversion acquisition methods, and specifically comprises the following steps: the data acquisition module is used for acquiring and/or storing the remote sensing image data, the land coverage data, the forest canopy data and the forest leaf area index; the data processing module is used for carrying out the preprocessing and pixel clipping in the step S1, and the screening processes of the step S2, the step S3 and the step S4; and a model processing module for performing step S6.
The invention has the following beneficial effects:
the invention can integrate the advantages of each remote sensing data, fuse multispectral data, invert to obtain the forest leaf area index by using a machine learning algorithm, introduce the terrain factors into inversion, eliminate the influence of the terrain on the inversion of the forest parameters, and solve the problem that the forest parameters are difficult to estimate accurately due to the terrain reasons.
The inversion method can obtain the applicability problem of the medium resolution remote sensing parameter product compared with the prior art, so that the obtained forest parameter data has more regional specificity.
The invention can be used as a research area in areas rich in forest resources and diversified ecosystems, effectively inverts to obtain a leaf area index with high resolution on the premise of not needing ground samples, and has higher precision compared with an inversion model only with two-dimensional optical image data by adding the structural parameters of a forest canopy, thereby helping a decision maker to formulate an effective environmental management policy of the forest ecosystems.
The method can combine optical data with radar-derived canopy parameters, and combine the optical data with the existing LAI product to establish a machine learning model, and overcomes the problem of data saturation of a single data source in the existing LAI estimation by utilizing the complementary advantages of a plurality of different data sources, and the method based on data fusion does not need a complex physical model, does not need too many actually measured vegetation parameters, overcomes the defect that a large number of field sample points are needed when estimating the LAI in a large range, and effectively reduces the operation cost; the invention can be directly applied to a Google Earth Engine platform with PB-level geospatial analysis capability, and can rapidly and efficiently process remote sensing images to generate LAI distribution diagrams of research areas.
The method for estimating leaf area index based on the data fusion method combines spectrum information of multispectral remote sensing data, geometric structure parameters of a solar sensor, vegetation index and canopy structure parameters with medium resolution LAI products, and uses a GEE remote sensing big data platform to process the remote sensing data, and comprises the following steps: the method comprises the steps of pixel screening, spatial heterogeneity screening, scale matching, sample point selection and final model establishment, and inversion to obtain a leaf area index product with 20m resolution of a research area.
The invention contains forest canopy structure parameters based on SAR data inversion, obtains the LAI value relative to a direct LAI inversion result such as GLASS LAI product, has higher inversion precision and can obtain higher spatial resolution.
Drawings
FIG. 1 is a flow chart of the LAI inversion acquisition method of the present invention in a specific embodiment.
FIG. 2 is a schematic diagram of the manner in which the pixel aggregation is performed according to the reduction function in example 1.
Fig. 3 is a graph showing vegetation index extraction obtained in example 1.
Fig. 4 shows the reflectance distribution and the total LAI frequency distribution of the different bands in the LAI interval in example 1.
FIG. 5 is a comparison of the GLASS LAI accuracy assessment results for the different data sets of example 1.
Fig. 6 is a histogram comparison of the frequency distribution of LAI prediction data obtained from different data sets in example 1.
FIG. 7 is a comparison of measured LAI with predicted LAI (left panel) or GLASS LAI (right panel) in example 1.
FIG. 8 is a comparison of the GLASS LAI profile (left panel) of the study area of example 1 with the LAI profile (right panel) obtained by inversion.
Detailed Description
The present invention will be described in detail with reference to the following examples and drawings, but it should be understood that the examples and drawings are only for illustrative purposes and are not intended to limit the scope of the present invention in any way. All reasonable variations and combinations that are included within the scope of the inventive concept fall within the scope of the present invention.
Referring to fig. 1, in some embodiments, the method for obtaining the forest leaf area index inversion based on data fusion according to the present invention includes the following steps:
s1: preprocessing remote sensing image data of a region to be obtained with forest leaf area indexes, namely a research region, and cutting pixels to obtain a segmented image map which eliminates cloud pixels and is segmented into a plurality of research cells.
In some preferred embodiments, the remote sensing image data uses remote sensing image data that has been radiometrically calibrated and atmospheric corrected, such as the remote sensing image data in Google Earth Engine, to eliminate atmospheric effects.
In some specific embodiments, the pre-treatment comprises:
s11, performing cloud removal processing on remote sensing image data of a research area through a remote sensing image processing product such as Sentinel-2L2A to obtain cloud-removed image data;
the remote sensing image data used is preferably image data subjected to radiometric calibration and atmospheric correction in Google Earth Engine (GEE), and when cloud removal processing is carried out on the image data through Sentinel-2L2A, a cloud mask file can be directly obtained to generate cloud-removed Sentinel-2 image data, wherein the resolution is 10m.
S12, classifying the reflectivity of the obtained cloud-removed image data in different spectral bands according to a threshold method in remote sensing image processing, and screening out pixels with reflectivity values larger than 0.3 in visible light and near infrared (VNIR) bands to obtain a preprocessed image without residual cloud pixels.
In some specific embodiments, the pel clipping comprises:
and S13, in the cloud removal image obtained after preprocessing, cutting the cloud removal image according to the boundary division conditions of different research cells to obtain a segmented image corresponding to each research cell.
The research cells can be divided according to different villages and towns in the research area, and the vector boundary files of each village and towns can be directly utilized to conduct batch cutting on the preprocessed cloud image removing images when the research cells are cut, so that images of each research cell are obtained.
S2: extracting and obtaining mask data of a forest area from land coverage data of a research area by a mask extraction method, applying the obtained mask data to the segmented image obtained in the step S1 to obtain forest pixels in each segmented image, and performing spatial aggregation on the obtained forest pixels to obtain an aggregate image with improved resolution.
In a specific embodiment, step S2 is implemented by:
(1) Extracting a forest region mask from a 10m resolution land coverage data set of ESA (European space and space agency) through GEE, wherein the number of the forest region in the data set is 10, applying the extracted forest region mask to the segmented image obtained in the step S1, and obtaining an extracted image of each segmented image, namely a segmented extracted image, which only retains forest region pixels;
(2) Space aggregation is carried out on the segmentation extracted images until the resolution reaches 250m, so that an aggregation image with improved resolution is obtained; the spatial aggregation may be implemented by a reduce resolution function as in GEE, and the pixel weights in the aggregation process may be set by the degree of overlap between small pixels in the aggregated segmented extraction image and large pixels in the output aggregated image.
In some more preferred embodiments, step S2 further comprises:
(3) And matching the pixels in the obtained aggregate image with LAI data of the research area by a simple average method, calculating standard deviation and variation coefficient in the matching, and screening heterogeneous pixels in the aggregate image according to the variation coefficient to obtain an optimized aggregate image.
In a specific embodiment, the step (3) includes: and matching the pixels in the obtained aggregate image with GLASS LAI data of the research area by a simple average method, and calculating standard deviation and variation coefficient of each aggregate pixel with the resolution of 250m in the matching, wherein the variation coefficient is obtained by standard variation normalized by an average value, and is the ratio of the standard deviation to the average value, and the average value represents the spatial reflectivity heterogeneity in one aggregate pixel. If the pixel has a spectral band variation coefficient of greater than 0.1 in any of green (560 nm), red (665 nm) or near infrared (842 nm), indicating that it has high spatial heterogeneity, which may result in a large uncertainty in the relationship of reflectivity to LAI at different spatial scales, the pixel is discarded.
In some more preferred embodiments, step S2 further comprises:
(4) Carrying out space aggregation on the forest region mask with the resolution of 10m extracted in the step (1) until the resolution reaches 250m, so as to obtain a forest coverage image;
further, in step (3), of the 250m resolution pixels, only if at least 90% of the sub-pixels (i.e., at least 563 of 625 sub-pixels of 10m resolution) are classified as forest categories, the pixel is considered as a forest pixel;
(5) And removing the wrongly classified forest areas in the aggregate image according to the comparison between the obtained forest coverage image and the obtained aggregate image.
S3: and (3) extracting a vegetation index from the segmentation image obtained in the step (S1) to obtain a vegetation index value.
In some specific embodiments, the vegetation indices include a normalized vegetation index NDVI, a building index IBI, a differential vegetation index DVI, a ratio vegetation index RVI, and a green chlorophyll index CIg, which are calculated as follows:
DVI=NIR-R
wherein NIR is the reflection value of the near infrared band in the image, R is the reflection value of the red light band, and in the Sentinel-2 remote sensing image, the NIR corresponds to the pixel values of the band 5 and the band 4 respectively; SWIR is the reflection value of the short wave infrared band, G is the reflection value of the green light band, and the values correspond to the pixel values of the band 9 and the band 3 in the Sentinel-2 remote sensing image map respectively.
S4, acquiring forest canopy height data through a geographic information system of the research area, and screening out heterogeneity data in the forest canopy height data through standard deviation and variation coefficient of the forest canopy height data to obtain screened canopy height data.
The forest canopy height data can be obtained through arcGIS extraction.
S5, obtaining LAI inversion data of the research area through a remote sensing data inversion tool.
The remote sensing data inversion tool is GLASS LAI product.
S6, constructing a random forest regression model according to the obtained aggregate image, the screened canopy height data, the vegetation index, the LAI inversion data and the solar parameters obtained by the solar sensor, training to obtain a fusion prediction model of the LAI, and further obtaining an LAI inversion estimated value and an LAI distribution prediction graph with 20m resolution through the model; the input of the random forest regression model is reflectivity data of an aggregate image, normalized solar zenith angle SZA, observed zenith angle VZA, relative azimuth angle RAA, normalized vegetation index NDVI, ratio vegetation index RVI, difference vegetation index DVI, urban building index IBI, green chlorophyll index CIg and screened canopy height data, and the data are output as LAI inversion data.
In a preferred embodiment, the input of the random forest regression model comprises:
reflectivity data for the resulting aggregate image at 10m resolution and 20m resolution, comprising: the reflectance of the green band (560 nm), the red band (665 nm), the near infrared band (842 nm), the red 1 band (705 nm), the red 2 band (740 nm), the red 3 band (783 nm), the far infrared 1 band (1610 nm), and the far infrared 2 band (2190 nm); normalized solar zenith angle SZA, observed zenith angle VZA, relative azimuth angle RAA, normalized vegetation index NDVI, ratio vegetation index RVI, difference vegetation index DVI, urban building index IBI, green chlorophyll index CIg, and canopy height data after screening.
Considering that the blue band is most sensitive to atmospheric scattering effects, it is easy to cause a large uncertainty in the results, and the reflectance data of the blue band is not used in the above embodiment.
In some embodiments, the random forest regression model is trained with an RF tree value set to 100 and a minimum She Qun set to 5; 80% of the samples were used as training data sets and 20% as test data sets; the model may be normalized to fix its value range within the range of 0, 1.
In some preferred embodiments, in the training process, an LAI frequency distribution map may be drawn according to the LAI inversion data obtained in step S5, and the obtained LAI inversion data is matched with the reflectivities of different spectral bands to obtain a distribution map of LAI values varying with the spectral bands, and the difference between the model predicted value and the obtained value is visually observed through the distribution map of LAI values varying with the spectral bands and the LAI frequency distribution map.
As in some embodiments, it comprises:
(1) Grouping the LAI inversion data obtained in the step S5, taking 0.5 as an interval, dividing the LAI inversion data into 8 groups, and removing samples with reflectivity lower than 1.5 quartile range (IQR) in green (560 nm), red (665 nm) and near infrared (842 nm) bands in each group, namely considering samples with reflectivity lower than Q1-1.5IQR and reflectivity higher than Q3+1.5IQR in each group as outliers, wherein Q1 is the first quartile in each group of samples, and Q3 is the third quartile in each group of samples, so as to obtain screened LAI samples;
(2) According to the correspondence of the LAI value and the reflectivities of different spectral bands, drawing a distribution diagram of the screened LAI sample changing along with the spectral bands so as to observe the saturation of the LAI value in a certain band, and in the prediction process, improving the prediction accuracy by dividing saturated and unsaturated data. The corresponding process is as follows: in the screened LAI samples, the LAI value corresponding to the 0-1 group is 0-2, and the corresponding green light wave band reflectivity interval is 0.10-0.12, so that a distribution diagram of the LAI value along with the change of a spectrum wave band can be drawn
(3) And obtaining an LAI frequency distribution map through the GLASS LAI to intuitively display the maximum value, the minimum value and the data distribution trend of the LAI of the area, so as to be convenient for comparison with predicted LAI data.
According to the above method for obtaining the forest leaf area index inversion, the invention can further obtain a forest leaf area index inversion obtaining system, which comprises:
the data acquisition module is used for acquiring and/or storing the remote sensing image data, the land cover data, the forest canopy data and the LAI inversion data; the data processing module is used for carrying out the preprocessing and pixel clipping in the step S1, and the screening processes of the step S2, the step S3 and the step S4; and a model processing module for performing step S6.
Through the system, the invention can overcome the problem of data saturation of a single data source in forest leaf area data estimation by utilizing the advantage complementation of a plurality of different data sources by means of remote sensing data fusion technology; the method solves the defect that a large number of field sample points are needed when estimating the LAI in a large range, and the method based on data fusion does not need to consider a complex physical model, does not need too many actually measured vegetation parameters, and effectively reduces the operation cost; and the remote sensing images can be processed rapidly and efficiently by means of the Google Earth Engine platform, so that a forest leaf area index map of the research area is generated. The system can combine optical data and radar data derived canopy structure parameters, cloud and snow removing treatment is carried out on multispectral data in GEE to obtain surface reflectivity data, solar sensor geometric structure parameters corresponding to the optical data and vegetation indexes are extracted, a forest mask of a research area is extracted by using a land cover product, a machine learning model is built on the data and medium resolution LAI products, and leaf area indexes of the research area are inverted by using a data fusion method.
Example 1
The forest leaf area index inversion is carried out on a certain area through the following process to obtain:
(1) Performing pixel screening on the Sening-2 remote sensing data of the area by utilizing data preprocessing of GEE to obtain a remote sensing image after cloud removal;
(2) Extracting forest canopy data of the region through ArcGIS, and placing the forest canopy data into an inversion model after space aggregation;
(3) And (3) extracting a forest region mask through a 10m resolution land cover product of the ESA, wherein the number of the forest region in the land cover product is 10, the operation is carried out in the GEE, and the extracted forest region mask is applied to the remote sensing image after cloud removal obtained in the step (1) to obtain the remote sensing image of Sentinel-2 pixels only retaining the forest region. In the training sample generation process, the Sentinel-2 image of the forest mask is spatially aggregated to 250m resolution, using the reduced resolution function in the GEE, the pixel weights in the aggregation process are the overlap between the aggregated small pixels and the output projection-specified large pixels.
The principle of the reduction function is shown in fig. 2.
(4) A simple averaging method was used for matching the Sentinle-2 pels to the gloss LAI data. In addition to the mean, the standard deviation and coefficient of variation for each aggregated 250m pel were also calculated. The coefficient of variation is derived from the standard variation normalized by the mean and is the ratio of the standard deviation to the mean. The average value represents the spatial reflectance heterogeneity within one 250m aggregate Sentinel-2 pel. If the coefficient of variation of any one of the spectral bands green (560 nm), red (665 nm) or near infrared (842 nm) is greater than 0.1, the 250m picture element is discarded because its high spatial heterogeneity may lead to a large uncertainty in the relationship of reflectivity to LAI at different spatial scales.
(5) The GLASS LAI and Sentinel-2 data and the canopy height data training model are fused by using the RF algorithm of GEE;
(6) Constructing a model at the GEE, wherein the input parameters comprise vegetation indexes, surface multiband reflectivity, corresponding solar sensor geometric structures and canopy heights;
(7) Extracting vegetation indexes of the Sentinel-2 remote sensing image of the Wolong protection area, wherein 5 vegetation indexes are selected, namely a normalized vegetation Index NDVI (Normalized Difference Vegetation Index), a building Index IBI (Index-based Builtup Index), a difference vegetation Index DVI (Difference Vegetation Index), a ratio vegetation Index RVI (Ratio Vegetation Index) and a green chlorophyll Index, and the extraction results are shown in figure 3;
(8) Preprocessing, scale conversion and index extraction are carried out on remote sensing data by using a GEE data analysis function (figure 4);
(9) Regression analysis was performed on the measured effective leaf area index and the Sentinel-2LAI, GLASS LAI in the field, respectively, and the fitting thereof on the data analysis level was observed (FIGS. 5, 6, 7).
(10) And constructing a prediction model by using GEE machine learning modeling capability, and carrying out forest LAI space drawing (figure 8).
The LAI products and the remote sensing data used in the process are the same as the time for collecting the data in the field, and the main distribution tree species in the research area are conifer forests.
Wherein, fig. 4 (a-c) shows the distribution of reflectivity of the green light band, the red light band and the near infrared band of Sentinel-2 in different LAI grouping intervals after removing the abnormal values, as can be seen from fig. 4, the reflectivity of the green light band and the red light band generally show a decreasing trend along with the increase of LAI, while the near infrared reflectivity is relatively gentle and has a small increase. It can be seen from fig. 4 (a) and 4 (b) that when the LAI group number is greater than 5, the LAI value is greater than 2.5, the reflectance of green and red bands of sentinel-2 gradually tends to saturate, whereas the reflectance change of near infrared band in fig. 4 (c) is not significant, which is related to the leaf structure and tissue of the plant, these factors tend not to play a significant role in red and green light of the short band, and furthermore, the LAI value of conifer is generally lower and tends not to saturate relative to the vegetation in the growing season, and thus the change in near infrared band is not significant. Fig. 4 (d) is a histogram of the corresponding GLASS LAI values of all the pixel points, and it can be seen that LAI is normally distributed, its maximum value is 4.2, its minimum value is 0, its mean value is 1.68, and its standard deviation is 0.48. 15635 sample points are left after abnormal values and pixel screening are removed, the screened sample points are regarded as pure pixels, and each pure pixel sample contains a GLASS LAI value, an average reflectivity value of a visible light band, a near infrared band and a short wave infrared band of Sentinel-2 data, a geometric structure parameter of a solar sensor, a forest canopy height value and an extracted vegetation index value.
FIG. 5 is a graph showing the results of accuracy evaluation using training data, test data, and all data, respectively, R 2 The MSE of the training data is 0.13, the RMSE is 0.23, and the fitting degree R is evaluated and verified to be above 0.83 2 Is 0.83 and the pearson correlation coefficient is 0.91; MSE estimated using test data was 0.14, RMSE was 0.22, R 2 A Pearon correlation coefficient of 0.85, and a Pearon correlation coefficient of 0.92; MSE was 0.15, RMSE 0.24, R using all data for evaluation 2 Is 0.83 and the pearson correlation coefficient is 0.91. The model fitting degree of the method is better, and a higher precision evaluation result is obtained by using a data fusion-based method, which shows that GLASS LAI, sentinel-2 and canopy are combinedThe method for fusing the height data can effectively estimate the leaf area index of the forest area, can rapidly process the data and model the data by using the GEE platform, and does not need to resort to field actual measurement data.
FIG. 6 is a histogram of a 20m leaf area index distribution of a forest area predicted using a training dataset, a test dataset, and all data, with three sets of predicted data all subject to normal distribution. Wherein the predicted LAI distribution using training data ranges from 0.47 to 2.89, with an average value of 1.71 and a standard deviation of 0.31; the predicted LAI distribution using training data ranged from 0.5 to 2.84, with an average of 1.53 and a standard deviation of 0.41; the predicted LAI distribution using training data ranged from 0.5 to 2.81 with an average of 1.61 and a standard deviation of 0.36. The histogram distribution of fig. 6 remains the same as that of fig. 4 (d), also indicating that the model has a good predictive effect on leaf area index.
Fig. 7 (a) shows the accuracy evaluation results of the actual values and the predicted values, and fig. 7 (b) shows the fitting accuracy results of the actual values and the gloss LAI products. As can be seen from FIG. 7, the leaf area index of the ground measurement using the canopy analyzer LAI2000 and the estimated leaf area index of the Sentinel data have good agreement, R 2 At 0.78, MSE of 0.16, the accuracy of the fit of LAI2000 measurements to GLASS LAI is substantially no different from the accuracy of the GLASS LAI product itself, R 2 0.74 and MSE 0.17. Under the premise of not considering errors of field actual measurement instruments, the verification accuracy of the leaf area index estimated by using the data fusion method and actual measurement LAI data is slightly higher than that of the GLASS LAI product.
FIG. 8 is a GLASS LAI profile of the study area and a comparison of 20m LAI profiles obtained by inversion in this example. It can be seen that the 20mLAI distribution map predicted and drawn on the GEE platform has a consistent spatial distribution trend with the GLASS LAI products of the same period using the data fusion method, but the Sentinel-2 predicted LAI provides more spatial detail than the GLASS LAI. In both figures, the LAI values are distributed from 0 to 4, and it can be clearly observed on the 20m LAI distribution diagram that in the middle region of the forest coverage, there are points where part of the LAI values are close to 0, and in the eastern region, there are also some smaller values. By observing the detailed enlarged views of the two views, it can be seen that the LAIs in the same area in the two views have the same distribution characteristics, but the 20m LAI view shows more space details for the distribution of leaf area indexes, and the forest LAI view with higher resolution has stronger regional applicability than a medium resolution LAI product, thereby providing assistance for relieving uncertainty caused by space heterogeneity under complex topography.
The above examples are only preferred embodiments of the present invention, and the scope of the present invention is not limited to the above examples. All technical schemes belonging to the concept of the invention belong to the protection scope of the invention. It should be noted that modifications and adaptations to the present invention may occur to one skilled in the art without departing from the principles of the present invention and are intended to be within the scope of the present invention.
Claims (10)
1. A forest leaf area index inversion acquisition method based on data fusion comprises the following steps:
s1, preprocessing remote sensing image data of a region to be obtained with forest leaf area indexes, namely a research region, and cutting pixels to obtain a segmented image map which eliminates cloud pixels and is segmented into a plurality of research cells;
s2, extracting and obtaining mask data of a forest area from land coverage data of a research area by a mask extraction method, applying the obtained mask data to the segmented image obtained in the S1 to obtain forest pixels in each segmented image, and performing spatial aggregation on the obtained forest pixels to obtain an aggregate image with improved resolution;
s3, extracting vegetation indexes of a research area from the segmented image according to reflection values of different wave bands in the segmented image;
s4, acquiring forest canopy height data through a geographic information system of a research area, and screening out heterogeneity data in the forest canopy height data through standard deviation and variation coefficients of the forest canopy height data to acquire screened canopy height data;
s5, obtaining a forest leaf area index of the research area through a remote sensing data inversion tool;
s6, constructing and training a forest leaf area index fusion prediction model, and obtaining a forest leaf area index inversion estimated value and a forest leaf area index distribution prediction graph through the trained or trained and tested model; the fusion prediction model is a random forest regression model, the reflectance values of different wave bands in the aggregate image or the optimized aggregate image are used as inputs, the vegetation index values, and the screened canopy height data and the solar zenith angle, the observed zenith angle and the relative azimuth angle obtained according to the solar sensor are used as outputs.
2. The method according to claim 1, wherein the remote sensing image data in S1 is remote sensing image data subjected to radiometric calibration and atmospheric correction.
3. The method for obtaining the forest leaf area index inversion of claim 1, wherein the obtaining the segmented image map comprises:
s11, performing cloud removal processing on remote sensing image data subjected to radiometric calibration and atmospheric correction in a research area through a remote sensing image processing product to obtain image data subjected to cloud removal;
s12, setting reflectivity thresholds of visible light and near infrared light bands in the image data after cloud removal, and filtering out residual cloud pixels according to the reflectivity thresholds to obtain a preprocessed cloud removal image;
s13, cutting the preprocessed cloud removal image according to boundary division conditions of different research cells to obtain a segmented image corresponding to each research cell.
4. A forest leaf area index inversion acquisition method according to claim 3 wherein the reflectivity threshold is 0.3.
5. The forest leaf area index inversion acquisition method according to claim 1, wherein the S2 includes:
s21, extracting a forest region mask through a land coverage data set with the resolution of 10m, and applying the extracted forest region mask to the segmented image obtained in the step S1 to obtain an extracted image of each segmented image, namely a segmented extracted image, wherein only forest region pixels are reserved;
s22, carrying out space aggregation on the segmentation extraction images until the resolution reaches 250m, and obtaining the aggregation image with the improved resolution.
6. The forest leaf area index inversion acquisition method of claim 5, wherein S2 further comprises:
and matching the pixels in the obtained aggregate image with the forest leaf area index of the research area through a simple average method, calculating standard deviation and variation coefficient in the matching, and screening heterogeneous pixels in the aggregate image according to the variation coefficient to obtain an optimized aggregate image.
7. The forest leaf area index inversion acquisition method of claim 5, wherein S2 further comprises:
space polymerization is carried out on the forest region mask with the resolution of 10m obtained in the step S21 until the resolution reaches 250m, so that a forest coverage image is obtained;
and removing the misclassified pixels in the aggregate image according to the comparison between the obtained forest coverage image and the aggregate image, so as to obtain an optimized aggregate image.
8. The method of claim 1, wherein the vegetation index comprises: normalized vegetation index NDVI, building index IBI, differential vegetation index DVI, ratio vegetation index RVI, and green chlorophyll index CIg.
9. The forest leaf area index inversion acquisition method according to claim 1, wherein the aggregate image is obtained according to a Sentinel-2 remote sensing image, and reflectance values of different wavebands in the aggregate image or the optimized aggregate image include: the image has a reflectance of a green band, a red band, a near infrared band, a red 1 band, a red 2 band, a red 3 band, a far infrared 1 band, and a far infrared 2 band at a resolution of 10m and a resolution of 20 m.
10. A forest leaf area index inversion system applying the forest leaf area index inversion acquisition method of any one of claims 1 to 9, comprising a data acquisition module for acquiring and/or storing the remote sensing image data, the land cover data, the forest canopy data, the forest leaf area index; the data processing module is used for carrying out the preprocessing and pixel clipping in the step S1, and the screening processes of the step S2, the step S3 and the step S4; and a model processing module for performing step S6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310894150.0A CN116912690A (en) | 2023-07-20 | 2023-07-20 | Forest leaf area index inversion acquisition method and system based on data fusion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310894150.0A CN116912690A (en) | 2023-07-20 | 2023-07-20 | Forest leaf area index inversion acquisition method and system based on data fusion |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116912690A true CN116912690A (en) | 2023-10-20 |
Family
ID=88356241
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310894150.0A Pending CN116912690A (en) | 2023-07-20 | 2023-07-20 | Forest leaf area index inversion acquisition method and system based on data fusion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116912690A (en) |
Cited By (3)
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 |
CN117671539A (en) * | 2023-12-14 | 2024-03-08 | 桂林理工大学 | Unmanned aerial vehicle monitoring sample plot vegetation coverage space representative optimization method |
CN117739871A (en) * | 2024-02-20 | 2024-03-22 | 中国科学院空天信息创新研究院 | Leaf area index measurement method, device, system, electronic equipment and storage medium |
-
2023
- 2023-07-20 CN CN202310894150.0A patent/CN116912690A/en active Pending
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117671539A (en) * | 2023-12-14 | 2024-03-08 | 桂林理工大学 | Unmanned aerial vehicle monitoring sample plot vegetation coverage space representative optimization method |
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 |
CN117739871A (en) * | 2024-02-20 | 2024-03-22 | 中国科学院空天信息创新研究院 | Leaf area index measurement method, device, system, electronic equipment and storage medium |
CN117739871B (en) * | 2024-02-20 | 2024-05-03 | 中国科学院空天信息创新研究院 | Leaf area index measurement method, device, system, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Li et al. | Spatio-temporal fusion for remote sensing data: An overview and new benchmark | |
Houborg et al. | Joint leaf chlorophyll content and leaf area index retrieval from Landsat data using a regularized model inversion system (REGFLEC) | |
CN106780091B (en) | Agricultural disaster information remote sensing extraction method based on vegetation index time-space statistical characteristics | |
CN116912690A (en) | Forest leaf area index inversion acquisition method and system based on data fusion | |
Fang | Canopy clumping index (CI): A review of methods, characteristics, and applications | |
Sun et al. | Retrieval of rapeseed leaf area index using the PROSAIL model with canopy coverage derived from UAV images as a correction parameter | |
CN113505635B (en) | Winter wheat and garlic mixed seed area identification method and device based on optics and radar | |
CN104266982A (en) | Large-area insect pest quantization monitoring system | |
Weyermann et al. | Correction of reflectance anisotropy effects of vegetation on airborne spectroscopy data and derived products | |
CN110414738B (en) | Crop yield prediction method and system | |
CN113850139B (en) | Multi-source remote sensing-based forest annual phenological monitoring method | |
Liu et al. | Estimating potato above-ground biomass by using integrated unmanned aerial system-based optical, structural, and textural canopy measurements | |
US6990410B2 (en) | Cloud cover assessment: VNIR-SWIR | |
CN113466143B (en) | Soil nutrient inversion method, device, equipment and medium | |
CN113553697B (en) | Long-time-sequence multi-source data-based vegetation disturbance analysis method for coal mining | |
CN115372282B (en) | Farmland soil water content monitoring method based on hyperspectral image of unmanned aerial vehicle | |
CN107688003A (en) | A kind of leaf reflectance satellite remote sensing extracting method for eliminating Vegetation canopy structure and earth's surface background influence | |
CN108898070A (en) | A kind of high-spectrum remote-sensing extraction Mikania micrantha device and method based on unmanned aerial vehicle platform | |
CN110836870A (en) | GEE-based large-area lake transparency rapid drawing method | |
CN117075138A (en) | Remote sensing measurement and calculation method, system and medium for canopy height of 30-meter forest in area | |
Lou et al. | An effective method for canopy chlorophyll content estimation of marsh vegetation based on multiscale remote sensing data | |
Xu et al. | Vegetation information extraction in karst area based on UAV remote sensing in visible light band | |
CN111175231B (en) | Inversion method and device of canopy vegetation index and server | |
CN112052720B (en) | High-space-time normalization vegetation index NDVI fusion model based on histogram clustering | |
CN115496999A (en) | Method and device for estimating field straw yield |
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 |