CN115147721A - Remote sensing inversion method and device for forest accumulation - Google Patents
Remote sensing inversion method and device for forest accumulation Download PDFInfo
- Publication number
- CN115147721A CN115147721A CN202210781345.XA CN202210781345A CN115147721A CN 115147721 A CN115147721 A CN 115147721A CN 202210781345 A CN202210781345 A CN 202210781345A CN 115147721 A CN115147721 A CN 115147721A
- Authority
- CN
- China
- Prior art keywords
- image
- forest
- class
- texture
- 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
- 238000009825 accumulation Methods 0.000 title claims abstract description 107
- 238000000034 method Methods 0.000 title claims abstract description 51
- 239000013598 vector Substances 0.000 claims abstract description 109
- 230000003595 spectral effect Effects 0.000 claims abstract description 85
- 238000012545 processing Methods 0.000 claims abstract description 54
- 230000009467 reduction Effects 0.000 claims description 42
- 238000002310 reflectometry Methods 0.000 claims description 39
- 239000011159 matrix material Substances 0.000 claims description 37
- 238000001228 spectrum Methods 0.000 claims description 17
- 238000000354 decomposition reaction Methods 0.000 claims description 15
- 230000002194 synthesizing effect Effects 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 8
- 238000003860 storage Methods 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 6
- 238000012314 multivariate regression analysis Methods 0.000 claims description 6
- 238000012706 support-vector machine Methods 0.000 claims description 6
- 230000001419 dependent effect Effects 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 5
- 238000012795 verification Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000012937 correction Methods 0.000 description 4
- 230000003287 optical effect Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012732 spatial analysis Methods 0.000 description 3
- 230000015572 biosynthetic process Effects 0.000 description 2
- 238000002790 cross-validation Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011835 investigation Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000001308 synthesis method Methods 0.000 description 2
- 238000003786 synthesis reaction Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 230000009919 sequestration Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
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
- G06V10/40—Extraction of image or video features
- G06V10/54—Extraction of image or video features relating to texture
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Processing (AREA)
Abstract
The application provides a remote sensing inversion method and a remote sensing inversion device for forest accumulation, relates to the technical field of remote sensing image processing, and specifically comprises the following steps: acquiring a multi-phase remote sensing image and a woodland small class vector of a region to be evaluated in a period of exuberant forest growth; processing the multi-period remote sensing image and the forest land small class vector to obtain small class spectral data and small class texture data; acquiring a canopy height image of an area to be evaluated, and performing spatial superposition on the canopy height image and a forest land minor-shift vector to obtain a minor-shift average tree height; and processing the small-class spectral data, the small-class texture data and the small-class average tree height by using a pre-trained forest accumulation inversion model to obtain the forest accumulation of the area to be evaluated. The inversion accuracy of the forest accumulation is improved.
Description
Technical Field
The application relates to the technical field of remote sensing image processing, in particular to a remote sensing inversion method and device for forest accumulation.
Background
The forest is an important component in a land ecological system, has huge carbon sequestration capacity, plays a special role in maintaining ecological safety and coping with climate change, and is an important measure for ensuring the healthy and long-term development of forest resources. The forest accumulation is a key index in forest monitoring, and the abundance, health degree and forest management effect of forest resources can be effectively reflected.
The forest accumulation, namely the total volume of all kinds of standing trees existing in a forest of a certain area, is calculated by taking cubic meters as a unit. The traditional forest accumulation estimation mostly adopts a manual investigation mode, and the monitoring mode is time-consuming, labor-consuming, low in precision and not beneficial to large-scale forest accumulation estimation. The remote sensing technology has the advantages of rapidness, large area and multiple frequencies, is widely applied to the forestry industry, and can provide important technology and data basis for the forestry monitoring field by estimating the accumulation amount through remote sensing.
At present, according to the difference of remote sensing data sources, methods for remotely sensing and inverting forest accumulation amount can be divided into two types, namely inversion based on optical remote sensing images and inversion using radar data (including laser radar and synthetic aperture radar), wherein the inversion based on optical spectrum images mainly establishes a regional inversion model by combining spectrum and texture information of the optical images with accumulation amount and combining occlusion degree, chest diameter and the like, so as to further obtain the accumulation amount of a monitoring area; the tree structure information obtained mainly through radar data is utilized for inversion of the radar data, and then the relation between the tree structure information and the accumulation amount is analyzed, so that high-precision inversion is achieved.
The method has certain limitations in the aspects of precision, speed and engineering in the practical application of remote sensing estimation of the forest accumulation, and a quick and accurate forest accumulation remote sensing inversion method which can reduce remote sensing data redundancy, effectively utilize image spectral information and texture information and is convenient for large-scale engineering application is urgently needed.
Disclosure of Invention
In view of this, the present application provides a remote sensing inversion method and apparatus for forest accumulation amount, so as to solve the above technical problems.
In a first aspect, an embodiment of the present application provides a remote sensing inversion method for forest reserves, including:
acquiring a multi-phase remote sensing image and a forest land class vector of a forest growth flourishing period of an area to be evaluated;
processing the multi-period remote sensing image to obtain an earth surface reflectivity mean value image; processing the earth surface reflectivity mean value image to obtain a spectral index image;
synthesizing the earth surface reflectivity mean image and the spectral index image into a synthetic image, performing spectral dimensionality reduction processing on the synthetic image based on truncated singular value decomposition to obtain a dimensionality reduction spectral image, and performing spatial superposition on the dimensionality reduction spectral image and the woodland small class vector to obtain small class spectral data;
calculating an LBP texture image based on the earth surface reflectivity mean image, and carrying out spatial superposition on the LBP texture image and the forest land minor class vector to obtain a texture forest land minor class vector; selecting the class texture data from the texture woodland class vector according to the preset wave band serial number;
acquiring a canopy height image of an area to be evaluated, and performing spatial superposition on the canopy height image and a forest land minor-shift vector to obtain a minor-shift average tree height;
and processing the minor shift spectral data, the minor shift texture data and the minor shift average tree height by using a pre-trained forest accumulation inversion model to obtain the forest accumulation of the area to be evaluated.
Further, performing spectral dimensionality reduction processing on the synthetic image based on truncated singular value decomposition to obtain a dimensionality reduced spectral image, comprising:
to the mean image M of the surface reflectivity O Sum spectral index image M I Synthesized to obtain a synthetic image M' O ，M′ O The number of the wave bands is N ', the height of the image is H, the width is W, and N' = N +2,N is the number of the wave bands of the remote sensing image;
will synthesize image M' O Performing array size transformation to obtain a matrix D with the size of C multiplied by N', wherein C = H multiplied by W;
according to the truncated singular value decomposition theorem, the matrix D can be expressed as:
D＝U∑V
wherein the size of the matrix U is C × N ', the size of the matrix Σ is N ' × N ', the size of the matrix V is N ' × N ', U and V are unitary matrices:
U＝(u 1 ，u 2 ...u C ) T
V＝(v 1 ，v 2 ...v N ′)
wherein λ is 1 ，λ 2 ...λ N′ Is a matrix D T N' characteristic values of D, v 1 ，v 2 ...v N′ Is a matrix D T N' eigenvectors of D; u. of 1 ，u 2 ...u C Is a matrix DD T C feature vectors of (a);
sorting diagonal elements of the matrix sigma from large to small, adjusting vectors of U and V, and respectively intercepting U and sigmaAnd before V K 1 The vectors form a matrix U ', sigma ' and V ', the size of which is C × K 1 ，K 1 ×K 1 And K 1 ×N′，K 1 ＜N′；
Then dimension reduction matrix D' 1 ：
D′ 1 ＝U′∑′V′
Transforming the dimensionality reduction matrix into a dimensionality reduction spectral image D 1 The number of wave bands is K 1 。
Further, performing spatial superposition on the dimension reduction spectrum image and the forest land minor class vector to obtain minor class spectrum data; the method comprises the following steps:
will reduce the dimension spectrum image D 1 Superposing the woodland minor vector to obtain a dimension-reduced spectrum woodland minor vector;
for K 1 Calculating the average spectrum of all pixels in each woodland class according to each wave band to obtain class spectrum data X 1 Size of (K) 1 And S) is the number of the forest land minor shifts.
Further, performing space superposition on the LBP texture image and the woodland minor vector to obtain a texture woodland minor vector; the method comprises the following steps:
texture image D of LBP 2 Superposing the forest land minor class vectors to obtain texture forest land minor class vectors, wherein LBP texture image D 2 The number of wave bands of (1) is N;
and calculating the average LBP texture value of all the image elements in each forest land class for N wave bands, thereby obtaining a texture forest land class vector.
Further, selecting the class texture data from the texture woodland class vector according to the preset wave band sequence number; the method comprises the following steps:
for texture woodland minor vectors of N wave bands, according to preset K 2 Selecting K from sequence numbers of each wave band 2 Average LBP texture value per band, K 2 ＜N；
K of each small forest area 2 Taking the average LBP texture value of each wave band as the small class texture data X 2 Size of (K) 2 And S) is the number of shifts.
Further, a support vector machine is adopted by the forest accumulation inversion model; the training step of the forest accumulation inversion model comprises the following steps:
acquiring a multi-period historical remote sensing image and a forest land small class vector sample of a dense forest region;
acquiring the average tree height and the actually measured hectare accumulation amount of each forest area class from the forest area class vector sample;
processing the multi-period historical remote sensing image to obtain an earth surface reflectivity mean value image sample; processing the earth surface reflectivity mean value image sample to obtain a spectral index image sample;
synthesizing the earth surface reflectivity mean image sample and the spectral index image sample into a synthesized image sample, performing spectral dimension reduction processing on the synthesized image sample based on truncated singular value decomposition to obtain a dimension reduction spectral image sample, and performing spatial superposition on the dimension reduction spectral image sample and a forest field class vector sample to obtain a class spectral data sample;
calculating an LBP texture image sample based on the earth surface reflectivity mean image sample, and performing space superposition on the LBP texture image sample and the forest land minor vector sample to obtain a texture forest land minor vector sample; selecting a small class texture data sample from the texture woodland small class vector sample by utilizing multivariate regression analysis;
inputting the small class spectral data sample, the small class texture data sample and the average tree height of all the woodland small classes into a forest accumulation quantity inversion model to obtain a predicted forest accumulation quantity; determining a loss function value based on the predicted forest accumulation amount and the actually measured hectare accumulation amount;
and updating model parameters of the forest accumulation quantity inversion model based on the loss function value.
Further, selecting a small class texture data sample from the texture woodland small class vector sample by utilizing multivariate regression analysis; the method comprises the following steps:
for the texture woodland minor vector sample, obtaining an average LBP value of each wave band of each minor, wherein the number of the wave bands is N;
selecting K from N wave bands 2 A combination of bands;
using the actual measured hectare accumulation per shift as a dependent variable, K 2 The average LBP value of each wave band is used as an independent variable to construct
A plurality of multiple regression linear equations;
calculating the correlation coefficient of each multiple regression linear equation to obtain the wave band combination corresponding to the maximum correlation coefficient;
selecting K from average LBP values of N wave bands according to wave band serial numbers 2 Taking the average LBP value of each wave band as a small class texture data sample;
and taking the wave band serial number of the wave band combination corresponding to the maximum correlation number as a preset wave band serial number.
In a second aspect, an embodiment of the present application provides a remote sensing inversion apparatus for forest accumulation, including:
the system comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a multi-phase remote sensing image and a woodland minor shift vector of a busy forest growth period of an area to be evaluated;
the first processing unit is used for processing the multi-period remote sensing image to obtain an earth surface reflectivity mean value image; processing the earth surface reflectivity mean value image to obtain a spectral index image;
the second processing unit is used for synthesizing the earth surface reflectivity mean image and the spectral index image into a synthetic image, performing spectral dimensionality reduction processing on the synthetic image based on truncated singular value decomposition to obtain a dimensionality reduction spectral image, and performing spatial superposition on the dimensionality reduction spectral image and a woodland small class vector to obtain small class spectral data;
the third processing unit is used for calculating an LBP texture image based on the earth surface reflectivity mean image, and performing space superposition on the LBP texture image and the woodland minor class vector to obtain a texture woodland minor class vector; selecting the class texture data from the texture woodland class vector according to the preset wave band serial number;
the fourth processing unit is used for acquiring a canopy height image of the area to be evaluated, and performing spatial superposition on the canopy height image and the small class vector of the forest land to obtain the average tree height of the small class;
and the inversion unit is used for processing the minor-shift spectral data, the minor-shift texture data and the minor-shift average tree height by using the pre-trained forest accumulation inversion model to obtain the forest accumulation of the area to be evaluated.
In a third aspect, an embodiment of the present application provides an electronic device, including: the remote sensing inversion method comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein when the processor executes the computer program, the remote sensing inversion method for the forest accumulation amount is realized.
In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, where computer instructions are stored, and when executed by a processor, the computer instructions implement the remote sensing inversion method for forest accumulation amount of the embodiment of the present application.
The inversion accuracy of the forest accumulation is improved.
Drawings
In order to more clearly illustrate the detailed description of the present application or the technical solutions in the prior art, the drawings needed to be used in the detailed description of the present application or the prior art description will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is an overall technical route diagram of a remote sensing inversion method of forest accumulation provided in an embodiment of the present application;
FIG. 2 is a flow chart of a remote sensing inversion method of forest accumulation provided by an embodiment of the application;
FIG. 3 is a forest land class vector diagram provided by an embodiment of the present application;
FIG. 4 is a flowchart of a process for processing a surface reflectance image according to an embodiment of the present disclosure;
FIG. 5 is a flow chart of processing of a small shift of spectral data provided by an embodiment of the present application;
FIG. 6 is a flow chart of a training process of a forest accumulation inversion model according to an embodiment of the present disclosure;
FIG. 7 is a schematic diagram illustrating a verification of a forest accumulation inversion model according to an embodiment of the present disclosure;
FIG. 8 is a functional block diagram of a remote sensing inversion apparatus for forest accumulation provided in an embodiment of the present application;
fig. 9 is a block diagram of an electronic device according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. The components of the embodiments of the present application, as generally described and illustrated in the figures herein, could be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
First, the design idea of the embodiment of the present application is briefly introduced.
The method comprehensively utilizes spectral characteristics and textural characteristics in the remote sensing image, adopts the technologies of data dimension reduction, multiple regression analysis, model optimization and the like, reduces the redundancy of remote sensing data, improves the inversion precision and efficiency, realizes the remote sensing inversion of forest accumulation, the method is suitable for most optical multispectral images, has the advantages of high running speed, high inversion precision, convenience for large-area engineering application and the like, plays a key supporting role in comprehensively mastering forest resources and quickly and accurately monitoring the forest, and the whole technical route is shown in figure 1.
After introducing the application scenario and the design concept of the embodiment of the present application, the following describes a technical solution provided by the embodiment of the present application.
As shown in fig. 2, an embodiment of the present application provides a remote sensing inversion method of forest accumulation, including:
step 101: acquiring a multi-phase remote sensing image and a forest land class vector of a forest growth flourishing period of an area to be evaluated;
the forest field sub-class vector is shown in figure 3, the sub-class is a basic unit for carrying out forestry investigation, production and operation, the forest fields with the same site conditions, forest stand factors, felling modes and operation measures and the same Lin Quan are divided into one sub-class, and the area of the sub-class is generally 5-20 hectares.
Step 102: processing the multi-period remote sensing image to obtain an earth surface reflectivity mean value image; processing the earth surface reflectivity mean value image to obtain a spectral index image;
as shown in fig. 4, the multi-phase remote sensing image needs to be subjected to geometric correction and atmospheric correction respectively, and then a multi-phase earth surface reflectivity image is generated and is marked as (M) 1 ，M 2 …M n ) The system comprises N-stage images, wherein the wave band number of each stage image is N;
synthesizing a surface reflectivity mean value image M' by a mean value synthesis method, wherein the synthesis method comprises the following steps:
in the formula M (i，j) J =1,2.. N, the mean image M' contains N bands.
Cutting M' by using the vector data of the research area to finally obtain a mean value image M of the earth surface reflectivity of the research area O The number of bands is N, and the height and width are H and W.
By means of M O Generating spectral index image M I The spectral indexes are selected from NDVI (Normalized Difference vector Index) and EVI (Enhanced vector Index), M I The NDVI and the EVI are formed by M O And generating a corresponding waveband:
wherein C 1 ＝6.0，C 2 ＝7.5，L＝1.0，G＝2.5，ρ NIR ，ρ red ，ρ blue Is M O And the corresponding ground surface reflectivity of near infrared, red and blue wave bands.
Step 103: synthesizing the earth surface reflectivity mean image and the spectral index image into a synthetic image, performing spectral dimension reduction processing on the synthetic image based on truncated singular value decomposition to obtain a dimension reduction spectral image, and performing spatial superposition on the dimension reduction spectral image and a woodland minor class vector to obtain minor class spectral data;
because multispectral remote sensing images have more wave bands and larger data volume, in order to improve the operation efficiency of the algorithm on the premise of not reducing remote sensing spectral information, a Truncated Singular Value Decomposition (TSVD) -based M pair is utilized O And from M O Generating a spectral index synthesized image for dimension reduction processing, as shown in fig. 5, the specific process is as follows:
from M O And M I Synthesis of New image M' O ，M′ O The number of bands is N +2, the size is (H, W), and is written as N +2=N'.
Will synthesize image M' O Performing array size transformation to obtain a matrix D with the size of C multiplied by N', wherein C = H multiplied by W;
according to the truncated singular value decomposition theorem, the matrix D can be expressed as:
D＝U∑V
wherein the size of the matrix U is C × N ', the size of the matrix Σ is N ' × N ', the size of the matrix V is N ' × N ', U and V are unitary matrices:
U＝(u 1 ，u 2 ...u C ) T
V＝(v 1 ，v 2 ...v N′ )
wherein λ is 1 ，λ 2 ...λ N′ Is a matrix D T N' characteristic values of D, v 1 ，v 2 ...v N′ Is a matrix D T N' feature vectors of D; u. u 1 ，u 2 ...u C Is a matrix DD T C feature vectors of (a);
sorting diagonal elements of the matrix sigma from large to small, adjusting vectors of U and V, and respectively intercepting U, sigma and K before V 1 The vectors form a matrix U ',' sigma 'and V', the size of which is C multiplied by K respectively 1 ，K 1 ×K 1 And K 1 ×N′，K 1 ＜N′；
Then dimension reduction matrix D' 1 ：
D′ 1 ＝U′∑′V′
Transforming the dimensionality reduction matrix into a dimensionality reduction spectral image D 1 Image size is (H, W), and the number of wave bands is K 1 。
D 1 Superposing the forest land sub-class vectors on the images, performing spatial analysis and calculation, acquiring the average spectrum in each forest land sub-class area, and generating sub-class spectrum data X 1 Size of (K) 1 And S) is the number of shifts.
Step 104: calculating an LBP texture image based on the earth surface reflectivity mean image, and performing space superposition on the LBP texture image and the forest land minor class vector to obtain a texture forest land minor class vector; selecting the class texture data from the texture woodland class vector according to the preset wave band serial number;
the traditional natural image texture information mainly comprises information such as homogeneity, difference, contrast and the like derived from a Gray-level co-occurrence matrix (GLCM), the data volume of a remote sensing image is far larger than that of a natural image, and the GLCM cannot meet the requirement of obtaining the texture information of the remote sensing image in speed. In order to effectively increase the texture information of accumulation inversion and improve the efficiency of generating texture data, local Binary Pattern (LBP) is used for generating the texture information of the inversion accumulation, meanwhile, based on correlation analysis, an inversion accumulation texture sensitive wave band is selected preferably, and based on the texture information of the sensitive wave band, the preferable texture data is generated, and the specific process is as follows:
to M 0 LBP texture information calculation is performed for each band, wherein the coordinate of a pixel point of one band is (x) c ，y c ) The LBP texture value is:
wherein, P samples the number of pixel points, R is the search radius, (x) c ，y c ) And (c) represents the gray values of the p-th sampling pixel point and the central point respectively, and s (x) is a threshold function.
The coordinates of the p-th sampling pixel point are calculated by adopting the following formula:
the LBP texture image D can be obtained through the pixel-by-pixel and wave band-by-wave band operation of the LBP algorithm 2 ，D 2 There are N bands with size (H, W).
Texture image D of LBP 2 Superposing the forest land minor class vectors to obtain texture forest land minor class vectors, wherein LBP texture images D 2 The number of the wave bands is N, and the size is (H, W);
for texture woodland minor vectors of N wave bands, according to preset K 2 Selecting K from sequence numbers of each wave band 2 Average LBP texture value per band, K 2 ＜N；
K of each small forest area 2 Taking the average LBP texture value of each wave band as the small class texture data X 2 Size of (K) 2 And S) is the number of shifts.
Step 105: acquiring a canopy height image of an area to be evaluated, and performing spatial superposition on the canopy height image and a forest land minor-shift vector to obtain a minor-shift average tree height;
in this embodiment, the contemporaneous and regional canopy height images are obtained from the global forest canopy height product manufactured by the university of maryland.
Step 106: and processing the small-class spectral data, the small-class texture data and the small-class average tree height by using a pre-trained forest accumulation inversion model to obtain the forest accumulation of the area to be evaluated.
Firstly, a forest accumulation inversion model is constructed, the model is constructed by utilizing a Support Vector Machine (SVM) algorithm framework, and the SVM has the advantages of strong robustness, high precision, high efficiency and the like, and is suitable for accumulation estimation of remote sensing images.
Using the class spectral data X 1 And small class texture data X 2 And generating accumulation amount sample data X by including actually measured tree height and accumulation amount vector small class data 0 Recording and verifying that the average tree height of each class is X 3 The accumulation amount data is V. Sample data X based on accumulation amount 0 The model framework is as follows:
V p ＝f(X 1 ，X 2 ，X 3 )
wherein, V p For the inverted predicted accumulation, f is the accumulation inversion model, X 1 ，X 2 ，X 3 Is an independent variable of the model.
And selecting 75% of sample data for model training, and 25% of sample data for model verification. The model has three important parameters, kernel (SVM kernel type), gamma (kernel coefficient) and C (regularization parameter). The model sets the kernel value range as (linear, poly, rbf, sigmoid, precomputed), the gamma range as (0.00001,0.0001,0.001,0.1,1, 10, 100, 1000) and the C as 2-30, and optimizes the three parameters by using 10-fold cross validation to obtain the optimal kernel, gamma and C.
The training step of the forest accumulation inversion model comprises the following steps:
acquiring multi-period historical remote sensing images and forest land small class vector samples in a dense forest region;
acquiring the average tree height and actually measured hectare accumulation amount of each forest land sub-class from the forest land sub-class vector sample;
it should be noted here that, in the model training, all the sub-shifts in the selected forest land sub-shift vector sample have the average tree height and the actually measured hectare accumulation amount. However, in the model application, step 101, the woodland shift vector does not include average tree height and measured hectare accumulation data.
Processing the multi-period historical remote sensing image to obtain an earth surface reflectivity mean value image sample; processing the earth surface reflectivity mean value image sample to obtain a spectral index image sample;
synthesizing the earth surface reflectivity mean image sample and the spectral index image sample into a synthesized image sample, performing spectral dimensionality reduction on the synthesized image based on truncated singular value decomposition to obtain a dimensionality reduction spectral image sample, and performing spatial superposition on the dimensionality reduction spectral image sample and the forest field class vector sample to obtain a class spectral data sample;
calculating an LBP texture image sample based on the earth surface reflectivity mean image sample, and performing space superposition on the LBP texture image sample and the forest land minor vector sample to obtain a texture forest land minor vector sample; selecting a small class texture data sample from the texture woodland small class vector sample by utilizing multivariate regression analysis;
selecting a small class texture data sample from the texture woodland small class vector sample by utilizing multivariate regression analysis; the method comprises the following steps:
for the texture woodland minor vector sample, obtaining an average LBP value of each wave band of each minor, wherein the number of the wave bands is N;
selecting K from N wave bands 2 A combination of individual bands;
using the actual measured hectare accumulation per shift as a dependent variable, K 2 The average LBP value of each wave band is used as an independent variable to construct
A plurality of multiple regression linear equations;
calculating the correlation coefficient of each multiple regression linear equation to obtain the wave band combination corresponding to the maximum correlation coefficient;
selecting K from average LBP values of N wave bands according to wave band serial numbers 2 The average LBP value of each wave band is used as a small class texture data sample;
and taking the wave band serial number of the wave band combination corresponding to the maximum correlation number as a preset wave band serial number.
Inputting the spectral data samples of the small class, the texture data samples of the small class and the average tree height of all the woodland small classes into a forest accumulation quantity inversion model to obtain predicted forest accumulation quantity; determining a loss function value based on the predicted forest accumulation amount and the actually measured hectare accumulation amount;
and updating model parameters of the forest accumulation quantity inversion model based on the loss function value.
And substituting the optimal model parameters into the model to obtain an inversion model with the optimal training set sample, inputting verification sample data into the inversion model to estimate the predicted accumulation amount, carrying out comparison analysis with the known actually-measured accumulation amount, carrying out model precision evaluation, and generating the inversion model for the forest accumulation amount after the precision is passed. The training process of the whole model is shown in fig. 6.
sentinel2 is used as a typical multispectral remote sensing image, and the sentinel2 multispectral remote sensing image is selected as a data source for accumulation inversion. The data survey of the shifts containing the actually measured accumulation amount usually lasts for a plurality of months, the time is long, and the time of the remote sensing image capable of representing forest characteristics is generally 5-9 months, so that the sentinel2 image of the tree growth flourishing period of 5-9 months is selected as the original data of the algorithm.
After geometric correction, atmospheric correction, multi-period synthesis, cutting and the like are carried out on the sentinel2 of 5-9 months in 2019, M is generated O Image, M O 11 bands, size (1058, 553);
M O calculated NDVI, EVI; d after TSVD spectrum dimensionality reduction 1 The image with the wave band data of 3 and the size of (1058, 553) is superposed with the forest land minor shift vector to carry out spatial analysis and calculation, the average spectrum in each forest land minor shift area is obtained, and the minor shift spectrum data X is generated 1 And the size is (3, 920).
Setting R of LBP algorithm 3,P to 6, LBP texture image can be generated. Through the band-by-band operation of the LBP algorithm, the LBP texture image D can be further obtained 2 ，D 2 There are 11 bands, size (1058, 553); d 2 Superposing the images with the small class vectors, performing spatial analysis and calculation, and obtaining D in each small class 2 The average LBP value of the image is selected from LBP values of 3 wave bands, a ternary regression model is constructed by the LBP values and the accumulation amount corresponding to each class, 165 multivariate models can be constructed according to an exhaustion method, the correlation coefficient of each model is calculated, the group of combinations with the maximum correlation coefficient is selected, and the wave band serial number is recorded as 3,8, 10.3,8, 10 band synthesized image; and superposing the small class vector data according to the extracted sensitive texture waveband data so as to obtain X of the small class texture data sensitive to the accumulation amount 2 And the size is (3, 920).
Generating accumulation sample data X by utilizing the small class spectral data, the small class texture data and the vector small class data containing the actually measured tree height and the accumulation 0 And the size is (8, 920), 690 sample data are selected for model construction, and the remaining 230 samples are verification samples.
Performing parameter tuning by using 1O-fold cross validation to obtain the optimal kernel as rbf; the gamma is 0.1, the C is 15, the model verification precision is 0.748, the root mean square error is 8.99m < 3 >/ha, and therefore the optimal accumulation remote sensing inversion model is constructed and can be applied to remote sensing engineering. The effect of the model inversion is shown in figure 7.
Based on the foregoing embodiments, an embodiment of the present application provides a remote sensing inversion apparatus of forest storage amount, and referring to fig. 8, a remote sensing inversion apparatus 200 of forest storage amount provided in the embodiment of the present application at least includes:
the acquiring unit 201 is used for acquiring a multi-phase remote sensing image and a forest land minor vector of a forest growth exuberant period of an area to be evaluated;
the first processing unit 202 is configured to process the multi-phase remote sensing image to obtain an earth surface reflectance mean image; processing the earth surface reflectivity mean value image to obtain a spectral index image;
the second processing unit 203 is configured to synthesize the earth surface reflectance mean image and the spectral index image into a synthesized image, perform spectral dimension reduction processing on the synthesized image based on truncated singular value decomposition to obtain a dimension-reduced spectral image, and perform spatial superposition on the dimension-reduced spectral image and a woodland minor-shift vector to obtain minor-shift spectral data;
the third processing unit 204 is configured to calculate an LBP texture image based on the earth surface reflectance mean image, and perform spatial superposition on the LBP texture image and the woodland minor class vector to obtain a texture woodland minor class vector; selecting the class texture data from the texture woodland class vector according to the preset wave band serial number;
a fourth processing unit 205, configured to obtain a canopy height image of the area to be evaluated, and perform spatial superposition on the canopy height image and the woodland minor-shift vector to obtain a minor-shift average tree height;
and the inversion unit 206 is configured to process the small class spectral data, the small class texture data, and the small class average tree height by using a pre-trained forest accumulation amount inversion model to obtain the forest accumulation amount of the area to be evaluated.
It should be noted that the principle of the remote sensing inversion method for forest accumulation amount provided by the embodiment of the present application for solving the technical problem by using the remote sensing inversion device 200 for forest accumulation amount provided by the embodiment of the present application is similar to that of the remote sensing inversion method for forest accumulation amount provided by the embodiment of the present application, and therefore, for implementation of the remote sensing inversion method for forest accumulation amount provided by the embodiment of the present application, reference may be made to implementation of the remote sensing inversion method for forest accumulation amount provided by the embodiment of the present application, and repeated details are not repeated.
As shown in fig. 9, an electronic device 300 provided in the embodiment of the present application at least includes: the remote sensing inversion method for the forest accumulation amount comprises a processor 301, a memory 302 and a computer program which is stored on the memory 302 and can run on the processor 301, wherein the remote sensing inversion method for the forest accumulation amount is realized when the processor 301 executes the computer program.
The electronic device 300 provided by the embodiment of the present application may further include a bus 303 connecting different components (including the processor 301 and the memory 302). Bus 303 represents one or more of any of several types of bus structures, including a memory bus, a peripheral bus, a local bus, and so forth.
The Memory 302 may include readable media in the form of volatile Memory, such as Random Access Memory (RAM) 3021 and/or cache Memory 3022, and may further include Read Only Memory (ROM) 3023.
The memory 302 may also include a program tool 3024 having a set (at least one) of program modules 3025, the program modules 3025 including, but not limited to: an operating subsystem, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.
Electronic device 300 may also communicate with one or more external devices 304 (e.g., keyboard, remote control, etc.), with one or more devices that enable a user to interact with electronic device 300 (e.g., cell phone, computer, etc.), and/or with any device that enables electronic device 300 to communicate with one or more other electronic devices 300 (e.g., router, modem, etc.). Such communication may be through an Input/Output (I/O) interface 305. Also, the electronic device 300 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public Network such as the internet) via the Network adapter 306. As shown in FIG. 9, the network adapter 306 communicates with the other modules of the electronic device 300 over the bus 303. It should be understood that although not shown in FIG. 9, other hardware and/or software modules may be used in conjunction with electronic device 300, including but not limited to: microcode, device drivers, redundant processors, external disk drive Arrays, redundant Array of Independent Disks (RAID) subsystems, tape drives, and data backup storage subsystems, to name a few.
It should be noted that the electronic device 300 shown in fig. 9 is only an example, and should not bring any limitation to the functions and the application scope of the embodiments of the present application.
The embodiment of the application also provides a computer readable storage medium, and the computer readable storage medium stores computer instructions, and the computer instructions are executed by a processor to realize the remote sensing inversion method of the forest accumulation amount provided by the embodiment of the application.
Further, while the operations of the methods of the present application are depicted in the drawings in a particular order, this does not require or imply that these operations must be performed in this particular order, or that all of the illustrated operations must be performed, to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step execution, and/or one step broken down into multiple step executions.
While the preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all alterations and modifications as fall within the scope of the application.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present application.
Claims (10)
1. A remote sensing inversion method of forest accumulation is characterized by comprising the following steps:
acquiring a multi-phase remote sensing image and a forest land class vector of a forest growth flourishing period of an area to be evaluated;
processing the multi-period remote sensing image to obtain an earth surface reflectivity mean value image; processing the earth surface reflectivity mean value image to obtain a spectral index image;
synthesizing the earth surface reflectivity mean image and the spectral index image into a synthetic image, performing spectral dimensionality reduction processing on the synthetic image based on truncated singular value decomposition to obtain a dimensionality reduction spectral image, and performing spatial superposition on the dimensionality reduction spectral image and the woodland small class vector to obtain small class spectral data;
calculating an LBP texture image based on the earth surface reflectivity mean image, and performing space superposition on the LBP texture image and the forest land minor class vector to obtain a texture forest land minor class vector; selecting the class texture data from the texture woodland class vector according to the preset wave band serial number;
acquiring a canopy height image of an area to be evaluated, and performing spatial superposition on the canopy height image and a forest land minor-shift vector to obtain a minor-shift average tree height;
and processing the small-class spectral data, the small-class texture data and the small-class average tree height by using a pre-trained forest accumulation inversion model to obtain the forest accumulation of the area to be evaluated.
2. A remote sensing inversion method of forest accumulation according to claim 1, wherein the spectral dimension reduction processing is performed on the synthetic image based on truncated singular value decomposition to obtain a dimension-reduced spectral image, comprising:
to the mean image M of the surface reflectivity 0 Sum spectral index image M I Synthesized to obtain a synthetic image M' 0 ，M′ 0 The number of the wave bands is N ', the height of the image is H, the width is W, and N' = N +2,N is the number of the wave bands of the remote sensing image;
will synthesize image M' 0 Perform array size variationObtaining a matrix D with the size of C multiplied by N', wherein C = H multiplied by W;
according to the truncated singular value decomposition theorem, the matrix D is represented as:
D＝U∑V
wherein the size of the matrix U is C × N ', the size of the matrix Σ is N ' × N ', the size of the matrix V is N ' × N ', U and V are unitary matrices:
U＝(u 1 ，u 2 ...u C ) T
V＝(v 1 ，v 2 ...v N′ )
wherein λ is 1 ，λ 2 ...λ N′ Is a matrix D T N' eigenvalues of D, v 1 ，v 2 ...v N′ Is a matrix D T N' eigenvectors of D; u. of 1 ，u 2 ...u C Is a matrix DD T C feature vectors of (a);
sorting diagonal elements of the matrix sigma from large to small, adjusting vectors of U and V, and respectively intercepting U, sigma and K before V 1 The vectors form a matrix U ', sigma ' and V ', the size of which is C × K 1 ，K 1 ×K 1 And K 1 ×N′，K 1 ＜N′；
Then dimension reduction matrix D' 1 ：
D′ 1 ＝U′∑′V′
Transforming the dimensionality reduction matrix into a dimensionality reduction spectral image D 1 Number of wave bands is K 1 。
3. A remote sensing inversion method of forest accumulation according to claim 2, characterized in that the dimensionality reduction spectral image and the forest land small class vector are spatially superposed to obtain small class spectral data; the method comprises the following steps:
will reduce the dimension spectrum image D 1 Superposing the forest land class vectors to obtain the dimensionality reduction spectrum forest landA small class vector;
for K 1 Calculating the average spectrum of all pixels in each woodland class according to each wave band to obtain class spectrum data X 1 Size of (K) 1 And S) is the number of the forest land minor shifts.
4. A remote sensing inversion method of forest accumulation according to claim 1, characterized in that LBP texture images and forest land minor class vectors are spatially superposed to obtain texture forest land minor class vectors; the method comprises the following steps:
texture image D of LBP 2 Superposing the forest land minor class vectors to obtain texture forest land minor class vectors, wherein LBP texture images D 2 The number of wave bands of (1) is N;
and calculating the average LBP texture value of all the image elements in each forest land class for N wave bands, thereby obtaining a texture forest land class vector.
5. A remote sensing inversion method of forest accumulation according to claim 4, characterized in that the sub-class texture data is selected from texture forest land sub-class vectors according to a preset wave band sequence number; the method comprises the following steps:
for texture woodland minor vectors of N wave bands, according to preset K 2 Selecting K from sequence numbers of each wave band 2 Average LBP texture value per band, K 2 ＜N；
K of each small forest land area 2 Taking the average LBP texture value of each wave band as the small class texture data X 2 Size of (K) 2 And S) is the number of shifts.
6. A remote sensing inversion method of forest accumulation according to claim 1, characterized in that a support vector machine is adopted for the forest accumulation inversion model; the training step of the forest accumulation inversion model comprises the following steps:
acquiring a multi-period historical remote sensing image and a forest land small class vector sample of a dense forest region;
acquiring the average tree height and the actually measured hectare accumulation amount of each forest area class from the forest area class vector sample;
processing the multi-period historical remote sensing image to obtain an earth surface reflectivity mean value image sample; processing the earth surface reflectivity mean value image sample to obtain a spectral index image sample;
synthesizing the earth surface reflectivity mean image sample and the spectral index image sample into a synthesized image sample, performing spectral dimensionality reduction on the synthesized image based on truncated singular value decomposition to obtain a dimensionality reduction spectral image sample, and performing spatial superposition on the dimensionality reduction spectral image sample and the forest field class vector sample to obtain a class spectral data sample;
calculating an LBP texture image sample based on the earth surface reflectivity mean image sample, and performing space superposition on the LBP texture image sample and the forest land minor vector sample to obtain a texture forest land minor vector sample; selecting a small class texture data sample from the texture woodland small class vector sample by utilizing multivariate regression analysis;
inputting the spectral data samples of the small class, the texture data samples of the small class and the average tree height of all the woodland small classes into a forest accumulation quantity inversion model to obtain predicted forest accumulation quantity; determining a loss function value based on the predicted forest accumulation amount and the actually measured hectare accumulation amount;
and updating the model parameters of the forest accumulation quantity inversion model based on the loss function value.
7. A remote sensing inversion method of forest accumulation according to claim 6, characterized in that a small class texture data sample is selected from texture forest land small class vector samples by utilizing multivariate regression analysis; the method comprises the following steps:
for the texture woodland minor vector sample, obtaining an average LBP value of each wave band of each minor, wherein the number of the wave bands is N;
selecting K from N wave bands 2 A combination of bands;
using the actual measured hectare accumulation per shift as a dependent variable, K 2 The average LBP value of each wave band is used as an independent variable to construct
A plurality of multiple regression linear equations;
calculating the correlation coefficient of each multiple regression linear equation to obtain the wave band combination corresponding to the maximum correlation coefficient;
selecting K from average LBP values of N wave bands according to wave band serial numbers 2 The average LBP value of each wave band is used as a small class texture data sample;
and taking the wave band serial number of the wave band combination corresponding to the maximum relational number as a preset wave band serial number.
8. A remote sensing inversion device of forest accumulation is characterized by comprising:
the system comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a multi-phase remote sensing image and a woodland minor shift vector of a busy forest growth period of an area to be evaluated;
the first processing unit is used for processing the multi-period remote sensing image to obtain an earth surface reflectivity mean value image; processing the earth surface reflectivity mean value image to obtain a spectral index image;
the second processing unit is used for synthesizing the earth surface reflectivity mean image and the spectral index image into a synthetic image, performing spectral dimensionality reduction processing on the synthetic image based on truncated singular value decomposition to obtain a dimensionality reduction spectral image, and performing spatial superposition on the dimensionality reduction spectral image and a woodland small class vector to obtain small class spectral data;
the third processing unit is used for calculating an LBP texture image based on the earth surface reflectivity mean image, and performing space superposition on the LBP texture image and the woodland minor class vector to obtain a texture woodland minor class vector; selecting the class texture data from the texture woodland class vector according to the preset wave band serial number;
the fourth processing unit is used for acquiring a canopy height image of the area to be evaluated, and performing spatial superposition on the canopy height image and the woodland minor-shift vector to obtain the minor-shift average tree height;
and the inversion unit is used for processing the small class spectral data, the small class texture data and the small class average tree height by using the forest accumulation amount inversion model trained in advance to obtain the forest accumulation amount of the area to be evaluated.
9. An electronic device, comprising: a memory, a processor and a computer program stored on the memory and executable on the processor, the processor when executing the computer program implementing the method of remote sensing inversion of forest reserves of any of claims 1 to 7.
10. A computer readable storage medium, characterized in that the computer readable storage medium stores computer instructions which, when executed by a processor, implement a method of remote sensing inversion of forest reserves according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210781345.XA CN115147721A (en) | 2022-07-04 | 2022-07-04 | Remote sensing inversion method and device for forest accumulation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210781345.XA CN115147721A (en) | 2022-07-04 | 2022-07-04 | Remote sensing inversion method and device for forest accumulation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115147721A true CN115147721A (en) | 2022-10-04 |
Family
ID=83410495
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210781345.XA Pending CN115147721A (en) | 2022-07-04 | 2022-07-04 | Remote sensing inversion method and device for forest accumulation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115147721A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115860975A (en) * | 2023-02-15 | 2023-03-28 | 南京航天宏图信息技术有限公司 | Salt lake lithium ore project productivity monitoring method and device based on satellite remote sensing |
| CN116297223A (en) * | 2023-03-24 | 2023-06-23 | 南京大学 | Forest deforestation recovery remote sensing monitoring method and system |
-
2022
- 2022-07-04 CN CN202210781345.XA patent/CN115147721A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115860975A (en) * | 2023-02-15 | 2023-03-28 | 南京航天宏图信息技术有限公司 | Salt lake lithium ore project productivity monitoring method and device based on satellite remote sensing |
| CN116297223A (en) * | 2023-03-24 | 2023-06-23 | 南京大学 | Forest deforestation recovery remote sensing monitoring method and system |
| CN116297223B (en) * | 2023-03-24 | 2023-11-21 | 南京大学 | Forest deforestation recovery remote sensing monitoring method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhao et al. | Comparison of machine learning algorithms for forest parameter estimations and application for forest quality assessments | |
| Li et al. | Automatic organ-level point cloud segmentation of maize shoots by integrating high-throughput data acquisition and deep learning | |
| Kershaw Jr et al. | Forest mensuration | |
| Halme et al. | Utility of hyperspectral compared to multispectral remote sensing data in estimating forest biomass and structure variables in Finnish boreal forest | |
| CN115147721A (en) | Remote sensing inversion method and device for forest accumulation | |
| CN114399685B (en) | Remote sensing monitoring and evaluating method and device for forest pest and disease damage | |
| Ali et al. | Machine learning methods’ performance in radiative transfer model inversion to retrieve plant traits from Sentinel-2 data of a mixed mountain forest | |
| Champagne et al. | A bootstrap method for assessing classification accuracy and confidence for agricultural land use mapping in Canada | |
| Bremer et al. | Derivation of tree skeletons and error assessment using LiDAR point cloud data of varying quality | |
| CN104318243B (en) | High-spectral data dimension reduction method based on rarefaction representation and empty spectrum Laplce's figure | |
| CN107895136B (en) | Coal mine area identification method and system | |
| Li et al. | Crop type mapping using time-series Sentinel-2 imagery and U-Net in early growth periods in the Hetao irrigation district in China | |
| CA3241328A1 (en) | Systems and methods for predicting soil carbon content | |
| CN116381700A (en) | Forest canopy height remote sensing estimation method and computer readable medium | |
| US20240337599A1 (en) | Systems and methods for predicting soil properties | |
| CN117036928A (en) | Urban forest carbon sink measuring and calculating method and system based on remote sensing images and machine learning | |
| Patel et al. | Deep Learning-Based Plant Organ Segmentation and Phenotyping of Sorghum Plants Using LiDAR Point Cloud | |
| CN114384023A (en) | Soil arsenic concentration spatial distribution inversion method and device and computer equipment | |
| Da Silva et al. | Netzeroco 2, an ai framework for accelerated nature-based carbon sequestration | |
| US20110054797A1 (en) | Maximum entropy application methods and systems | |
| CN116503681A (en) | Crop identification feature optimization method and system for plain and hilly areas | |
| CN117173565A (en) | Method and system for estimating vegetation coverage of large-area wheat by deep migration learning | |
| Gadepally et al. | Sampling operations on big data | |
| Kalbi et al. | Estimation of forest attributes in the Hyrcanian forests, comparison of advanced space-borne thermal emission and reflection radiometer and satellite poure I’observation de la terre-high resolution grounding data by multiple linear, and classification and regression tree regression models | |
| CN116297281A (en) | System and method for predicting sample characteristics based on spectral measurements |
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 |