CN104809471B - A kind of high spectrum image residual error integrated classification method based on spatial spectral information - Google Patents
A kind of high spectrum image residual error integrated classification method based on spatial spectral information Download PDFInfo
- Publication number
- CN104809471B CN104809471B CN201510205088.5A CN201510205088A CN104809471B CN 104809471 B CN104809471 B CN 104809471B CN 201510205088 A CN201510205088 A CN 201510205088A CN 104809471 B CN104809471 B CN 104809471B
- Authority
- CN
- China
- Prior art keywords
- classification
- dictionary
- spectral information
- information
- matrix
- 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.)
- Active
Links
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The high spectrum image residual error integrated classification method based on spatial spectral information that the invention discloses a kind of.Include the following steps, read in hyperspectral image data, determines that sample class number is L；The spatial texture information F of high spectrum image is extracted respectivelyGWith spectral information FN；According to spatial texture information FGWith spectral information FNConstruct correlation matrix；Solve test sample x in high spectrum imagepCoefficient matrix corresponding with correlation matrix；Reconstructed sample calculates test sample xpEach classification corresponding to reconstructed residual；Calculate test sample xpEach classification fusion residual error；According to test sample xpThe fusion residual error of each classification determine test sample xpClassification.The present invention has the advantages of nicety of grading is high, obtained classification chart good visual effect.
Description
Technical field
The invention belongs to remote sensing information process technical field more particularly to classification hyperspectral imagery, one kind being based on space
The high spectrum image residual error integrated classification method of spectral information.
Background technique
High spectrum image is all widely used in many fields, including agricultural production, Minerals identification and detection target area
Domain, disaster alarm, military surveillance and urban planning etc..Classification hyperspectral imagery technology is occupied non-in Hyperspectral imagery processing
Normal consequence, the purpose is to be divided into the pixel in image in its respective classes according to representative sample.It passes
The information being typically used in system EO-1 hyperion processing is original spectrum (Original Spectral, OS) information and principal component point
(Principal Components Analysis, PCA) characteristic information is analysed, data information abundant how is effectively utilized, and
Ensure that processing accuracy obtains more and more extensive concern simultaneously.In numerous classification methods, support vector machines (Support
Vector Machine, SVM) it is a kind of generally acknowledged outstanding sorting algorithm, 2006 on its basis, Camps-Valls was utilized
The property of Mercer kernel function constructs compound kernel function (Composite Kernels, CK), to open easy combination
Spatial information and spectral information wide field, resulting classifier is referred to as SVM-CK.
There are the following problems in traditional Hyperspectral data classification method: 1, high-spectral data cannot obtain well
Expression causes nicety of grading not high.2, the neighborhood information in high spectrum image is not made full use of.3, have in assorting process
Effect information cannot be utilized adequately.
Summary of the invention
The object of the present invention is to provide a kind of classification good visual effect, nicety of grading are high, one kind is believed based on spatial spectral
The high spectrum image residual error integrated classification method of breath.
The present invention is achieved by the following technical solutions:
A kind of high spectrum image residual error integrated classification method based on spatial spectral information, including following steps,
Step 1: reading in hyperspectral image data, carries out dimension conversion to high spectrum image, three-dimensional data is converted into two
Dimension data makees normalized to resulting 2-D data, determines that sample class number is L；
Step 2: the spatial texture information F of high spectrum image is extracted respectivelyGWith spectral information FN；
Step 3: according to spatial texture information FGWith spectral information FNCorrelation matrix is constructed, correlation matrix includes according to space
Texture information FGThe spatial texture information allied signal matrix XJ of buildingG, spatial texture information dictionary AGWith local adaptive space
Texture information dictionary AGL, correlation matrix further includes according to spectral information FNThe spectral information allied signal matrix XJ of buildingN, spectrum
Dictionary of information ANWith local adaptive optical spectrum information dictionary ANL；
Step 4: test sample x in high spectrum image is solvedpCoefficient matrix corresponding with correlation matrix；
Step 5: by spatial texture information dictionary AG, local auto-adaptive spatial texture information dictionary AGL, spectral information dictionary
ANWith local adaptive optical spectrum information dictionary ANLCoefficient matrix corresponding with its is multiplied to obtain reconstructed sample respectively, calculates test specimens
This xpEach classification corresponding to reconstructed residual；
Step 6: test sample x is calculatedpEach classification fusion residual error；
Step 7: according to test sample xpThe fusion residual error of each classification determine test sample xpClassification.
A kind of high spectrum image residual error integrated classification method based on spatial spectral information of the present invention can also include:
1, test sample x in high spectrum imagepCoefficient matrix corresponding with correlation matrix are as follows:
Wherein ψGSFor spatial texture information dictionary AGCoefficient matrix, ψGLCFor local auto-adaptive spatial texture information dictionary
AGLCoefficient matrix, ψNSFor spectral information dictionary ANCoefficient matrix, ψNLCFor local auto-adaptive spectral information dictionary ANLBe
Matrix number, λGSFor spatial texture information dictionary AGCoefficient of balance, λGLCFor local auto-adaptive spatial texture information dictionary AGLIt is flat
Weigh coefficient, λNSFor spectral information dictionary ANCoefficient of balance, λNLCFor local auto-adaptive spectral information dictionary ANLCoefficient of balance.
2, test sample xpI-th of classification reconstructed residual are as follows:
3, test sample xpI-th of classification fusion residual error are as follows:
Wherein μ is method coefficient of balance, meets 0≤μ≤1, and η is information balance coefficient, meets 0≤η≤1.
4, sample xpClassification be determined to be in L classification the classification with minimum fusion residual error
The utility model has the advantages that
The advantage of the invention is that its application space-spectral information states that image well, it can be to image
In neighborhood information adequately utilize, while making full use of its effective information with fusion residual error, optimize classification chart
Visual effect, the advantages that improving the precision of classification.
Detailed description of the invention
Fig. 1 is flow chart of the invention.
Fig. 2 is the true picture of the Indian Pines high-spectral data in experiment.
Fig. 3 is original all kinds of atural object distribution map selected in Indian Pines high-spectral data in experiment.
Fig. 4 is sample names and sample number scale selected in Indian Pines high-spectral data in experiment.
Fig. 5 is that the three kinds of distinct methods used in experiment divide the classification of corresponding all kinds of atural objects after data set classification
Butut, Fig. 5-a are the OS+SVM classification distribution map of data set, and Fig. 5-b is SVM-CK classification distribution map, and Fig. 5-c is present invention side
Method classification distribution map.
Fig. 6 is the evaluation index table that the three kinds of distinct methods used in experiment classify to data set.
Specific embodiment
The present invention is described in further details below in conjunction with attached drawing.
The high spectrum image residual error integrated classification method based on space-optical spectrum information that the object of the present invention is to provide a kind of,
With more preferably classifying quality, the advantages that good visual effect of classifying, nicety of grading is high.
Specific step is as follows:
1, hyperspectral image data is read in.EO-1 hyperion high dimensional data is read in, carries out dimension conversion, and make at normalization to it
Reason, wherein being L containing sample class number.
2, high-spectral data information is extracted.Extract its spatial texture information F respectively for imageGWith spectral information FN。
3, correlation matrix is constructed according to high-spectral data information.Respectively according to spatial texture information FGWith spectral information FN
To construct corresponding some correlation matrixes.
4, corresponding coefficient matrix is solved.According to correlation matrix constructed in step 3, to solve in high spectrum image
Corresponding coefficient matrix in test sample.
5, reconstructed sample calculates corresponding residual error.Dictionary is multiplied with its corresponding coefficient matrix first and is reconstructed
Sample.For the sample of reconstruct, the corresponding corresponding reconstructed residual of every class is calculated.
6, residual error merges.Reconstructed residual required in step 5 is merged, the fusion residual error of test sample is obtained.
7, sample class is determined.The classification of test sample is determined according to the residual error of fusion.Under the final output present invention
Classification hyperspectral imagery result Y.
Solution finishes.Y is the classification knot of the high spectrum image residual error integrated classification method based on space-optical spectrum information
Fruit.
As shown in Figure 1, detailed process of the invention are as follows:
1, hyperspectral image data is read in.
Three-dimensional EO-1 hyperion high dimensional data is read in, dimension conversion is carried out to it makes it be converted to 2-D data from three-dimensional with side
Just subsequent processing, and normalized is made to resulting 2-D data, determine that sample class number to be processed is L, here often
The corresponding mathematical form of a sample is a column data of its column in 2-D data.
2, high-spectral data information is extracted.
Extract its spatial texture information and spectral information respectively for image, wherein step 2.1 is and advances with step 2.2
Capable.
2.1 extract image space texture information.
(1), PCA transformation is carried out to image.It is unfolded on the basis of PCA transformed obtained first principal component grayscale image
Subsequent step.
(2), Gabor filtering is carried out.
For Gabor filter, in spatial domain, following real responses can be obtained:
In formula (1), by co-ordinate position information x, y and relational expression x'=xcos θ+ysin θ, y'=-xcos θ+ysin
θ, can obtain the value of x ' He y ', and remaining parameter meaning are as follows: ρ indicates the scale of Gabor function, θ andThen indicate the function
Direction and phase angle, σ and γ are Gauss radius and deflection.
(3), nonlinear transformation.
The filtering image extracted is converted using nonlinear algorithm.Here nonlinear algorithm are as follows:
In formula (2), t and α respectively indicate discrete wavelet coefficient and a certain constant.
(4), Gassian low-pass filter is carried out.Mean absolute deviation is calculated, calculating process are as follows: apply Gaussian low pass
Wave device is handled, and the low-pass filter g (x, y) of selection is expressed as follows:
G (x, y)=exp {-(x2+y2)/2σ2} (3)
The standard deviation of Gauss window function is indicated with σ.
(5), spatial texture information F is generatedG.High-spectrum may finally be obtained by the operation of (1)-(4) in step 2.1
The textural characteristics F of pictureG。
2.2 extract image spectrum information.
The spectral information of high spectrum image is extracted by NWFE algorithm in this algorithm.
(1), sample is calculatedTo the weighted cluster center of classification j
Wherein,For the weight matrix of distribution, For classification i's (1≤i≤L)
K-th of sample, njFor the number of samples of classification j, j (1≤j≤L) is category label.
(2), divergence distribution matrix of k-th of sample to classification j in calculating classification i
Wherein, Euclidean distance of the dist (a, b) between vector a and vector b, niFor the number of samples of classification i.
(3), inter _ class relationship matrix S is calculatedbWith within class scatter matrix Sw。
(4), transformation matrix is calculated.
Transformation matrix J needed for being generated according to optimal characteristics Optimality Criteria relational expression (8).
J=tr (Sw -1Sb) (8)
(5), spectral information F is generatedN。
Spectral information F can be obtained by being multiplied by transformation matrix J with high-spectral dataN。
3, correlation matrix is constructed according to high-spectral data information.
3.1 are based on spatial texture information FGConstruct correlation matrix.
(1), spatial texture information allied signal matrix XJ is constructedG。
Assuming that single pixel xpOn the p of position, the neighborhood that a rectangular size is T × T, neighborhood window size are constructed
For T, the spatial texture information of the pixel in high spectrum image neighborhood is used to construct united matrix XJG, first row is to be in
The corresponding spatial texture information of the test sample of center, remaining is classified as the corresponding sky of neighborhood sample around central sample
Between texture information.
(2), entire spatial texture information dictionary A is constructedG。
Spatial information corresponding to a part of sample, which is respectively chosen, from each classification carrys out composing training collection AG。
(3), local auto-adaptive spatial texture information dictionary A is constructedGL。
Calculate dictionary AGIn each atom and XJGCorrelation, and arranged according to descending, from AGMiddle selection is related
Property maximum first K corresponding atom constitutes test sample xpThe sub- dictionary A of corresponding local auto-adaptiveGL。
3.2 are based on spectral information FNConstruct correlation matrix.
It is similar with step 3.1, construct spectral information allied signal matrix XJN, construct entire spectral information dictionary AN, building
Local auto-adaptive spectral information dictionary ANL, with step 3.1 the difference is that the process object of each step is by spatial information FG
It is changed to spectral information FN。
4, corresponding coefficient matrix is solved.
According to correlation matrix constructed in step 3, surveyed by formula (9)-formula (12) to solve in high spectrum image
This x of samplepIn corresponding coefficient matrix:
ψ thereinGS,ψGLC,ψNS,ψNLCRespectively AG,AGL,AN,ANLCorresponding solves the coefficient matrix come.λGS,
λGLC,λNS,λNLCRespectively AG,AGL,AN,ANLCorresponding coefficient of balance is the number greater than zero, is used for degree of rarefication and again
Structure precision is balanced.|| ||1Indicate 1 norm, | | | |FIndicate Frobenious norm.
5, reconstructed sample calculates corresponding residual error.
First dictionary is multiplied to obtain reconstructed sample with its corresponding coefficient matrix.For the sample of reconstruct, calculate every
The corresponding corresponding reconstructed residual of class.By the residual computations of the i-th class such as formula (13)-formula (16):
WhereinIt Wei not be via ψGS,ψGLC,ψNS,ψNLCAnd AG,AGL,AN,ANLInto
Corresponding reconstructed residual after the reconstruct of row sample.
6, residual error merges.
4 kinds of residual errors required in step 5 are merged according to formula (17), obtain sample xpThe i-th class fusion
Residual error
Wherein μ is method coefficient of balance, meets 0≤μ≤1, and η is information balance coefficient, meets 0≤η≤1.
7, sample class is determined.
According to formula (18), sample x is determined according to the residual error of fusionpClassification.Sample xpClassification be determined to be in L
There is the classification of minimum fusion residual error in class.
It is worth noting that: step 3 to step 7 will be run one time for each test sample in high spectrum image.
Classification hyperspectral imagery result Y by step 1 to step 7, under the final output present invention.
In order to illustrate effectiveness of the invention, spy carries out following experimental demonstration.
The applicability of the method for the present invention is wherein verified using Indian Pines high-spectral data collection.
The test block Indiana, USA Indian Pine image, is acquired in June, 1992 with AVIRIS sensor
It arrives, spatial resolution 20m.Original image shares 220 wave bands, and size is 144 × 144, shares 16 kinds of atural object distributions, will be former
200 wave bands are chosen as simulation object after biggish some wave bands removal affected by noise in 220 wave bands to begin.Sample
This classification L is set to 16, is tested with 16 kinds of atural objects, for convenience to it respectively marked as 1-16 class.Experimental data such as Fig. 2
Shown, the original all kinds of atural object of selection is distributed as shown in figure 3, selected sample names and sample size are as shown in Figure 4.From
This 16 kinds practical atural objects are distributed the data that 10% is uniformly extracted in samples as training sample.
When classifying to high spectrum image, the method for the present invention and two kinds of classical ways i.e.: then passed through using OS information
Svm classifier method (is labeled as OS+SVM)；Using OS information and Gabor texture information, then through SVM-CK classification method (label
It is compared for SVM-CK).
With the classification distribution map of corresponding all kinds of atural objects after three kinds of classifications as shown in figure 5, can be very intuitive
See that classification effect picture of the invention is good relative to OS+SVM and SVM-CK classification.
Three evaluation of classification index, that is, overall classification accuracies of three kinds of classification methods, classification are averaged nicety of grading and Kappa
Coefficient, as shown in fig. 6, be wherein known that overall classification accuracy by the definition of these indexs, classification be averaged nicety of grading and
When Kappa coefficient is higher, the classifying quality of image is better.Compared with OS+SVM and SVM-CK, the totality point of method of the invention
Class precision will be higher by 16% and 4% respectively, and the classification of the method for the invention nicety of grading that is averaged will be higher by 17% and 7% respectively,
The Kappa coefficient of method of the invention will be higher by 18% and 4% respectively.
It can be it is further seen where the method for the present invention advantage by the comparative analysis of experiment.
Claims (1)
1. a kind of high spectrum image residual error integrated classification method based on spatial spectral information, it is characterised in that: including following several
A step,
Step 1: reading in hyperspectral image data, carries out dimension conversion to high spectrum image, three-dimensional data is converted into two-dimemsional number
According to, normalized is made to resulting 2-D data, determine sample class number be L；
Step 2: the spatial texture information F of high spectrum image is extracted respectivelyGWith spectral information FN；
Step 3: according to spatial texture information FGWith spectral information FNCorrelation matrix is constructed, correlation matrix includes according to spatial texture
Information FGThe spatial texture information allied signal matrix XJ of buildingG, spatial texture information dictionary AGWith local adaptive space texture
Dictionary of information AGL, correlation matrix further includes according to spectral information FNThe spectral information allied signal matrix XJ of buildingN, spectral information
Dictionary ANWith local adaptive optical spectrum information dictionary ANL；
Step 4: test sample x in high spectrum image is solvedpCoefficient matrix corresponding with correlation matrix；
Step 5: by spatial texture information dictionary AG, local auto-adaptive spatial texture information dictionary AGL, spectral information dictionary ANWith
Local auto-adaptive spectral information dictionary ANLCoefficient matrix corresponding with its is multiplied to obtain reconstructed sample respectively, calculates test sample xp
Each classification corresponding to reconstructed residual；
Step 6: test sample x is calculatedpEach classification fusion residual error；
Step 7: according to test sample xpThe fusion residual error of each classification determine test sample xpClassification；
If single pixel xpOn the p of position, the neighborhood that a rectangular size is T × T is constructed, neighborhood window size is T, high
The spatial texture information of pixel in spectrum picture neighborhood is used to construct united matrix XJG；Calculate dictionary AGIn each original
Son and XJGCorrelation, and arranged according to descending, from AGFirst K corresponding atom of middle selection correlation maximum is tested to constitute
Sample xpThe sub- dictionary A of corresponding local auto-adaptiveGL, construct spectral information allied signal matrix XJN, construct entire spectral information
Dictionary AN, construct local auto-adaptive spectral information dictionary ANL；
Test sample x in the high spectrum imagepCoefficient matrix corresponding with correlation matrix are as follows:
Wherein ψGSFor spatial texture information dictionary AGCoefficient matrix, ψGLCFor local auto-adaptive spatial texture information dictionary AGL's
Coefficient matrix, ψNSFor spectral information dictionary ANCoefficient matrix, ψNLCFor local auto-adaptive spectral information dictionary ANLCoefficient square
Battle array, λGSFor spatial texture information dictionary AGCoefficient of balance, λGLCFor local auto-adaptive spatial texture information dictionary AGLBalance system
Number, λNSFor spectral information dictionary ANCoefficient of balance, λNLCFor local auto-adaptive spectral information dictionary ANLCoefficient of balance；It is described
Test sample xpI-th of classification reconstructed residual are as follows:
The test sample xpI-th of classification fusion residual error are as follows:
Wherein μ is method coefficient of balance, meets 0≤μ≤1, and η is information balance coefficient, meets 0≤η≤1；
The sample xpClassification be determined to be in L classification the classification with minimum fusion residual error
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510205088.5A CN104809471B (en) | 2015-04-27 | 2015-04-27 | A kind of high spectrum image residual error integrated classification method based on spatial spectral information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510205088.5A CN104809471B (en) | 2015-04-27 | 2015-04-27 | A kind of high spectrum image residual error integrated classification method based on spatial spectral information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104809471A CN104809471A (en) | 2015-07-29 |
| CN104809471B true CN104809471B (en) | 2019-01-15 |
Family
ID=53694282
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510205088.5A Active CN104809471B (en) | 2015-04-27 | 2015-04-27 | A kind of high spectrum image residual error integrated classification method based on spatial spectral information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104809471B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106469316B (en) * | 2016-09-07 | 2020-02-21 | 深圳大学 | Hyperspectral image classification method and system based on superpixel-level information fusion |
| CN107944474B (en) * | 2017-11-06 | 2021-04-09 | 中国地质大学（北京） | Multi-scale collaborative expression hyperspectral classification method based on local adaptive dictionary |
| CN108388917A (en) * | 2018-02-26 | 2018-08-10 | 东北大学 | A kind of hyperspectral image classification method based on improvement deep learning model |
| CN108764357A (en) * | 2018-05-31 | 2018-11-06 | 西安电子科技大学 | Polymerization residual error network hyperspectral image classification method based on compression-excitation |
| CN109785302B (en) * | 2018-12-27 | 2021-03-19 | 中国科学院西安光学精密机械研究所 | Space-spectrum combined feature learning network and multispectral change detection method |
| CN111046844B (en) * | 2019-12-27 | 2020-11-27 | 中国地质大学（北京） | Hyperspectral image classification method based on neighborhood selection constraint |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102129676A (en) * | 2010-01-19 | 2011-07-20 | 中国科学院空间科学与应用研究中心 | Microscopic image fusing method based on two-dimensional empirical mode decomposition |
| CN103559500A (en) * | 2013-10-15 | 2014-02-05 | 北京航空航天大学 | Multispectral remote sensing image land feature classification method based on spectrum and textural features |
| CN103903007A (en) * | 2014-03-10 | 2014-07-02 | 哈尔滨工程大学 | Hyperspectral semi-supervised classification method based on space-spectral information |
| CN103914705A (en) * | 2014-03-20 | 2014-07-09 | 西安电子科技大学 | Hyperspectral image classification and wave band selection method based on multi-target immune cloning |
-
2015
- 2015-04-27 CN CN201510205088.5A patent/CN104809471B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102129676A (en) * | 2010-01-19 | 2011-07-20 | 中国科学院空间科学与应用研究中心 | Microscopic image fusing method based on two-dimensional empirical mode decomposition |
| CN103559500A (en) * | 2013-10-15 | 2014-02-05 | 北京航空航天大学 | Multispectral remote sensing image land feature classification method based on spectrum and textural features |
| CN103903007A (en) * | 2014-03-10 | 2014-07-02 | 哈尔滨工程大学 | Hyperspectral semi-supervised classification method based on space-spectral information |
| CN103914705A (en) * | 2014-03-20 | 2014-07-09 | 西安电子科技大学 | Hyperspectral image classification and wave band selection method based on multi-target immune cloning |
Non-Patent Citations (3)
| Title |
|---|
| Hyperspectral image classification by nonlocal joint collabortive representation with a locally adaptive dictionary;J Li等;《IEEE》;20141231;第3707-3719页 |
| Spectral-Spatial Hyperspectral Image Classification via Mulitiscale Adaptive Sparse Representation;Fang L等;《IEEE》;20141231;第7738-7749页 |
| 基于空间-光谱特征和稀疏表达的高光谱图像分类算法;杨京辉等;《应用地球物理》;20141231;第11卷(第4期);第489-499页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104809471A (en) | 2015-07-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104809471B (en) | A kind of high spectrum image residual error integrated classification method based on spatial spectral information | |
| CN103971123B (en) | Hyperspectral image classification method based on linear regression Fisher discrimination dictionary learning (LRFDDL) | |
| Charles et al. | Learning sparse codes for hyperspectral imagery | |
| Al Bashish et al. | A framework for detection and classification of plant leaf and stem diseases | |
| CN107451614B (en) | Hyperspectral classification method based on fusion of space coordinates and space spectrum features | |
| CN103886336B (en) | Polarized SAR image classifying method based on sparse automatic encoder | |
| CN106503739A (en) | The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics | |
| CN109359623A (en) | High spectrum image based on depth Joint Distribution adaptation network migrates classification method | |
| CN106886760B (en) | A kind of EO-1 hyperion Ship Detection combined based on empty spectrum information | |
| CN105718942B (en) | High spectrum image imbalance classification method based on average drifting and over-sampling | |
| CN108182449A (en) | A kind of hyperspectral image classification method | |
| CN104036289A (en) | Hyperspectral image classification method based on spatial and spectral features and sparse representation | |
| CN110533077A (en) | Form adaptive convolution deep neural network method for classification hyperspectral imagery | |
| CN108446582A (en) | Hyperspectral image classification method based on textural characteristics and affine propagation clustering algorithm | |
| CN106096506A (en) | Based on the SAR target identification method differentiating doubledictionary between subclass class | |
| CN102938072A (en) | Dimension reducing and sorting method of hyperspectral imagery based on blocking low rank tensor analysis | |
| CN108229551A (en) | A kind of Classification of hyperspectral remote sensing image method based on compact dictionary rarefaction representation | |
| Gu et al. | Multiple kernel sparse representation for airborne LiDAR data classification | |
| CN105160351B (en) | Semi-supervised hyperspectral classification method based on anchor point sparse graph | |
| CN109034213B (en) | Hyperspectral image classification method and system based on correlation entropy principle | |
| CN115205590A (en) | Hyperspectral image classification method based on complementary integration Transformer network | |
| Shihavuddin et al. | Automated classification and thematic mapping of bacterial mats in the north sea | |
| CN105138966A (en) | Quick density peak value clustering based polarimetric SAR image classification method | |
| Tombe et al. | Adaptive deep co-occurrence feature learning based on classifier-fusion for remote sensing scene classification | |
| Tao et al. | Deep Learning for Hyperspectral Image Analysis and Classification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |