CN108280486A - A kind of high spectrum image solution mixing method based on end member cluster - Google Patents
A kind of high spectrum image solution mixing method based on end member cluster Download PDFInfo
- Publication number
- CN108280486A CN108280486A CN201810104197.1A CN201810104197A CN108280486A CN 108280486 A CN108280486 A CN 108280486A CN 201810104197 A CN201810104197 A CN 201810104197A CN 108280486 A CN108280486 A CN 108280486A
- Authority
- CN
- China
- Prior art keywords
- end member
- spectrum
- cluster
- image
- pixel
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
- G06F18/2193—Validation; Performance evaluation; Active pattern learning techniques based on specific statistical tests
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
A kind of novel high spectrum image solution mixing method based on end member cluster, the adaptive accurate solution that high spectrum image is reached by extraction and two steps of abundance inverting based on end member cluster of end member cluster are mixed.Steps are as follows for this method：(1) rarefaction representation based on global image；(2) the alternative end member screening based on ballot；(3) end member cluster structure is carried out by extracting spectral shape feature；(4) the excessively complete dictionary of piecemeal is constructed；(5) whole image is traversed, for each pixel select based on the sparse best end member of block；(6) abundance estimation, output end member cluster spectrum and abundance result are carried out using staff cultivation least square method.This method can effectively reduce the error brought due to spectrum variation mainly for the variation problem of spectrum present in high spectrum image, improve solution and mix precision, while the extraction of the end member cluster based on image itself makes algorithm adaptivity strong.
Description
Technical field
The present invention relates to a kind of novel high spectrum image solution mixing methods based on end member cluster, can have by technological means
Effect reduces the error brought due to spectrum variation when solution is mixed, and belongs to field of remote sensing image processing.
Background technology
High-spectrum remote-sensing refers to the remote sensing technology with high spectral resolution, and the wave band of detection is with the spectrum of Nano grade point
Resolution cover including ultraviolet, visible light, near-infrared and in infrared and thermal infrared SPECTRAL REGION (0.4 μm -2.5 μm).Bloom
Spectrum remote sensing can continuously and subtly describe object spectrum, have outstanding advantage in terms of exploration, detection, identification.EO-1 hyperion is distant
It is huge to feel development potentiality, is considered together most important two in remote sensing technology with imaging radar after the advent of the 1980s
Item technological break-through.Since the 1990s, high-spectrum remote-sensing be increasingly becoming in the world the main flow direction of photoelectric remote-sensing and
The heat subject of remote sensing technology.China is a few state for possessing independent high spectrum resolution remote sensing technique intellectual property in the world
One of family.In recent years, country energetically support high spectrum resolution remote sensing technique development, many researchers 863 projects, country from
Under the supports of projects such as right science fund, in researchs such as resource investigation, environmental monitoring, engineering construction, agriculture identification, medical diagnosis
Field achieves notable achievement.
By spectrometer spatial resolution limit and atural object distribution it is multifarious influence, in high spectrum image a pixel is past
Toward comprising a variety of atural objects, such pixel is called mixed pixel by we.Corresponding with mixed pixel, we are containing only one
The pixel of kind atural object is known as Pure pixel or end member, while in mixed pixel, the percentage shared by each substance is referred to as rich
Degree.The mixed purpose of EO-1 hyperion solution is exactly to obtain the characteristic spectrum of end member in image, and find out whereby in mixed pixel eachly
Ratio shared by object.EO-1 hyperion solution is mixed can to solve being stranded for the atural object brought because of mixed pixel distribution detection and Object Classification
Difficulty, therefore all occupy critically important status in the theoretical research and application of high-spectrum remote sensing.
Existing EO-1 hyperion solution mixes algorithm, generally represents an end member by calculating or selecting a curve of spectrum.So
And only an end member is represented with a curve of spectrum and had some limitations.In piece image, due to uneven illumination, mine
Organic matter these factors different from impurity that object size distribution is different, is included, it is easy to appear spectrum for the spectrum of same substance
Variation phenomenon.Spectrum, which makes a variation, causes the variation of spectral value, then causes to solve mixed error.In order to solve EO-1 hyperion solution under complex environment
The mixed spectrum variation problem occurred in the process, the present invention utilize the concept of end member cluster, i.e., indicate a kind of substance with one group of spectrum,
In conjunction with sparse representation theory model, propose that a set of completely new efficient, accurate EO-1 hyperion solution mixes model.
Invention content
For the spectrum variation phenomenon that image occurs, the present invention designs a kind of completely new EO-1 hyperion solution mixing method, method
Core is that the adaptive structure of end member cluster is mixed with the solution based on end member cluster, is subtracted to the comprehensive expression of end member by end member cluster
Solution caused by dim light composes variation phenomenon mixes error.The present invention can directly extract end member cluster from image and carry out solution and mix, and be not required to
Want library of spectra as prior information, the adaptivity of method is strong, and suitable environment is wide, can effectively improve solution and mix precision.
To achieve the goals above, the technical solution adopted by the present invention is：A kind of completely new EO-1 hyperion solution mixing method, mainly
Including：Alternative Endmember extraction, end member cluster structure and abundance estimate three steps.The purpose of alternative Endmember extraction is looked for from image
The point of end member property is provided, the inside includes the variation spectrum of each end member.The structure of end member cluster is needed in unsupervised situation
Lower that similar spectrum condenses together, the present invention completes accurate spectral classification by extracting the shape feature of spectrum.Finally
On the basis of the end member cluster extracted, best match spectra combination is found out for each pixel to carry out abundance estimation.
Method flow according to the present invention includes the following steps：(1) rarefaction representation based on global image；(2) it is based on
The alternative end member screening of ballot；(3) end member cluster structure is carried out by extracting spectral shape feature；(4) the excessively complete word of piecemeal is constructed
Allusion quotation；(5) whole image is traversed, for each pixel select based on the sparse best end member of block；(6) utilize staff cultivation minimum
Square law carries out abundance estimation, output end member cluster spectrum and abundance result.
Each step of this method flow is described in detail below.
(1) rarefaction representation based on global image
A given width size is m rows n row, the high spectrum image X={ x of l wave band1,x2,...,xi,...,xm×n, wherein
xi={ xi1,xi2,...,xil}.Then rarefaction representation is carried out to each pixel using match tracing method, for pixel xi's
Complete dictionary is crossed to be configured toSubstance classes number in image, i.e. end member number
Mesh is k, and the degree of rarefication in rarefaction representation is likewise provided as k.It recycles match tracing method to carry out sparse expression, obtains sparse
Coefficient yi, is as follows：
1) initialization residual error r0=xi, the set of indexesAtom collectionAnd iterations t=1.
2) selection is closest with residual error, i.e. the maximum atom of inner product, records its index λt.Then update set of indexes Λt=
Λt-1∪{λtAnd atom collection Atomt=[Atomt-1；dλt]。
λt=argmax | ＜ rt-1, D ＞ | (1)
3) coefficient of estimation is obtained by least square problem, and updates residual error rt
rt=xi-εtAtomt (3)
4) t=t+1 is updated, and returns to step 2), until residual error gradually restrains or when t=K stops iteration.
5) sparse coefficient y is finally obtainedi=εt。
(2) the alternative end member screening based on ballot
For pixel xiSparse coefficient y corresponding with itsi, we find out sparse coefficient yiMiddle maximum absolute value it is sparse right
The atom answered, and vote it.Image includes m × n pixel, i.e., launches m × n tickets altogether, adds up the poll that each atom obtains,
Then all atoms are ranked up according to poll.The number cn of alternative end member is calculated by end member number k, here cn=k ×
5, and the atom of cn before number of votes obtained alternately end member is selected, obtain alternative end member set Xcand。
(3) end member cluster structure is carried out by extracting spectral shape feature
This step carries out feature extraction to the alternative end member obtained in step 2, to achieve the purpose that accurately to build end member cluster.
For including the spectrum of l wave bandIt is l that spectrum, which is cut into N number of length, first0Spectrum segment, wherein
N=[l/l0].If l cannot be by l0Divide exactly, the length of spectrum final stage is remaining wave band number.
Each section of spectrum is fitted followed by straight line, the slope for extracting straight line represents this spectrum segment shape, then obtains
Entire slope of a curve vectorP=1,2...N.
Fitting a straight line uses least-square fitting approach
It can obtain alternative end member XcandFeature Fcand=[f1；f2；...；fc...；fcn], then feature is inputted
Into unsupervised segmentation device, k classification F is obtainedbundle=[Fb1；Fb2...；Fbi；...；Fbk].Corresponding alternative end member is pressed again
End member cluster X is formed according to classification resultsbundle=[Xb1；Xb2...；Xbi；...Xbk],Xbi=[xh|fh∈Fbi], and export result.
(4) the excessively complete dictionary of piecemeal is constructed
For acquired end member cluster Xbundle=[Xb1；Xb2...；Xbi；...Xbk], each piece of excessively complete dictionary by
One end member cluster is constituted, i.e. D=[Db1；Db2；...；Dbi；...；DbK],Dbi=Xbi。
(5) whole image is traversed, for each pixel select based on the sparse best end member of block
For arbitrary pixel xi, specifically calculating step is：
1) initialization residual error r0=xi, the set of indexesEnd member collectionDictionary D0=D, setting iterations t=
1, iteration ends number K.
2) index λ is found in entire dictionarytIt is set to meet (6)
λt=arg max | ＜ rt-1,Dt-1＞ | (6)
Update set of indexes Λt=Λt-1∪{λt, end member collectionDictionary
3) least square method, the optimization problem in solution (7) are utilized, and updates residual error rt。
4) if residual error is less than the threshold value of setting, stop iteration, otherwise increases t=t+1, and return 2), until t=
K。
(6) abundance estimation, output end member cluster spectrum and abundance result are carried out using staff cultivation least square method.
Description of the drawings
Fig. 1 is the flow chart of the high spectrum image solution mixing method based on end member cluster.
Fig. 2 is the schematic diagram of curve shape feature extraction：A) it is primitive curve, b) it is the feature extracted.
Fig. 3 be menology high-spectral data and solution it is mixed after obtain as a result, a) being the position handled on image and its menology
Schematic diagram, b) the end member cluster spectrum that is, c) be the mixed obtained each substance of solution abundance distribution figure.
Specific implementation mode
The application process of the present invention is described further with reference to example.
The interference imaging spectral that this example is carried using the lunar probe Chang'E-1 that China was succeeded in sending up October 24 in 2007
The hyperspectral image data that instrument takes carries out solving mixed processing.The size of the high-spectral data of this processing is 300 × 128, wave
Long ranging from 480-960nm, including 32 wave bands, wherein 20 wave bands are for calculating for selection, spatial resolution 200m/
pixel。
(1) rarefaction representation based on global image
The excessively complete dictionary that size is 38400 × 20 is built, and rarefaction representation is carried out to each pixel.It is dilute in this example
It dredges the degree of rarefication indicated and is set as 3, the number of alternative end member is set as 15.
(2) the alternative end member screening based on ballot
For the sparse coefficient of each pixel and its acquisition, ballot gives the maximum item of sparse coefficient corresponding atom.Statistics
The number of votes obtained of all atoms chooses preceding 15 pixels alternately end member in this example.
(3) end member cluster structure is carried out by extracting spectral shape feature
Alternative endmember spectra is segmented, every segment length is 2 wave bands in this example, and then carrying out shape feature to curve carries
It takes, obtains feature vectorP=1,2...N.Curve is based on least square fitting, wherein slope system
Several method for solving are
After obtaining slope characteristics, the characteristic use k-means methods extracted are polymerized to 3 classes, complete the structure of end member cluster
It builds.
(4) the excessively complete dictionary of piecemeal is constructed
Using the 3 class end member cluster spectrum extracted, the block dictionary D=[Db that size is 15 × 20 are constructed1；Db2；Db3]。
(5) whole image is traversed, for each pixel select based on the sparse best end member of block
For arbitrary pixel xi,
1) initialization residual error r0=xi, the set of indexesEnd member collectionDictionary D0=D, setting iterations t=
1, iteration ends number K.
2) index λ is found in entire dictionarytIt is set to meet (8)
λt=argmax | ＜ rt-1,Dt-1＞ | (8)
Update set of indexes Λt=Λt-1∪{λt, end member collectionDictionary
3) least square method, the optimization problem in solution (9) are utilized, and updates residual error rt。
4) if residual error is less than the threshold value of setting, stop iteration, otherwise increases t=t+1, and return 2), until t=
K。
(6) abundance estimation, output end member cluster spectrum and abundance result are carried out using staff cultivation least square method.
Claims (2)
1. a kind of end member cluster extracting method, it is characterised in that：Based on relationship between rarefaction representation analysis pixel, sparse coefficient pair is utilized
Alternative end member is screened in pixel ballot, and extraction spectral shape feature carries out end member cluster structure, and this method can be realized based on image
End member cluster automatically extracts, and its step are as follows：
Step 1, the rarefaction representation based on global image
A given width size is m rows n row, the high spectrum image X={ x of l wave band1,x2,...,xi,...,xm×n, wherein xi=
{xi1,xi2,...,xil}；Rarefaction representation is carried out to each pixel first with match tracing method, for pixel xiIt is excessively complete
Standby dictionary is configured toSubstance classes number in image, i.e. end member number are
Degree of rarefication in rarefaction representation is likewise provided as k, match tracing method is recycled to carry out sparse expression, obtains sparse coefficient by k
yi, sparse coefficient solution be as follows：
1) initialization residual error r0=xi, the set of indexesAtom collectionAnd iterations t=1；
2) selection is closest with residual error, i.e. the maximum atom of inner product, records its index λt.Then update set of indexes Λt=Λt-1∪
{λtAnd atom collection Atomt=[Atomt-1；dλt]；
λt=argmax | ＜ rt-1, D ＞ | (1)
3) the sparse of estimation is obtained by least square problem, and updates residual error rt；
rt=xi-εtAtomt (3)
4) t=t+1 is updated, and returns to step 2), until residual error gradually restrains or when t=K stops iteration；
5) sparse coefficient y is finally obtainedi=εt；
Step 2, the alternative end member screening based on ballot
For pixel xiSparse coefficient y corresponding with itsi, we find out sparse coefficient yiMiddle maximum absolute value it is sparse corresponding
Atom, and vote it；Image includes m × n pixel, i.e., launches m × n tickets altogether, adds up the poll that each atom obtains, then
All atoms are ranked up according to poll；The number cn of alternative end member, cn=k × 5 are calculated by end member number k, and are selected
The atom of cn alternately end member, obtains alternative end member set X before number of votes obtainedcand；
Step 3, end member cluster structure is carried out by extracting spectral shape feature
This step carries out feature extraction to the alternative end member obtained in step 2, to achieve the purpose that accurately to build end member cluster；For
Include the spectrum of l wave bandIt is l that spectrum, which is cut into N number of length, first0Spectrum segment, wherein N=
[l/l0]；If l cannot be by l0Divide exactly, the length of spectrum final stage is remaining wave band number；
Each section of spectrum is fitted followed by straight line, the slope for extracting straight line represents this spectrum segment shape, then obtains entire
Slope of a curve vectorP=1,2...N；
Fitting a straight line uses least-square fitting approach
It can obtain alternative end member XcandFeature Fcand=[f1；f2；...；fc...；fcn]；Then no prison is input the feature into
It superintends and directs in grader, obtains k classification Fbundle=[Fb1；Fb2...；Fbi；...；Fbk], then by corresponding alternative end member according to classification
As a result end member cluster X is formedbundle=[Xb1；Xb2...；Xbi；...Xbk],Xbi=[xh|fh∈Fbi], and export result.
2. a kind of abundance estimation method based on end member cluster, it is characterised in that：Using known end member cluster library of spectra, combined block is sparse
Thought selects the combination of most mating end member for each pixel, using staff cultivation Least-squares inversion, obtains abundance distribution,
It is as follows：
Step 1, the excessively complete dictionary of piecemeal is constructed
For acquired end member cluster Xbundle=[Xb1；Xb2...；Xbi；...Xbk], each piece of excessively complete dictionary is held by one
First cluster is constituted, i.e. D=[Db1；Db2；...；Dbi；...；DbK],Dbi=Xbi；
Step 2, whole image is traversed, for each pixel select based on the sparse best end member of block
For arbitrary pixel xi, specifically calculating step is：
1) residual error r0=x, the set of indexes are initializedEnd member collectionDictionary D0Iterations t=1 is arranged, repeatedly in=D
In generation, terminates number K；
2) index λ is found in entire dictionarytIt is set to meet (6)
λt=argmax | ＜ rt-1,Dt-1＞ | (6)
Update set of indexes Λt=Λt-1∪{λt, end member collectionDictionary
3) least square method, the optimization problem in solution (7) are utilized, and updates residual error rt；
4) if residual error is less than the threshold value of setting, stop iteration, otherwise increases t=t+1, and return 2), until t=K；
Step 3, abundance estimation, output end member cluster spectrum and abundance result are carried out using staff cultivation least square method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810104197.1A CN108280486B (en) | 2018-02-02 | 2018-02-02 | Hyperspectral image unmixing method based on end member cluster |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810104197.1A CN108280486B (en) | 2018-02-02 | 2018-02-02 | Hyperspectral image unmixing method based on end member cluster |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108280486A true CN108280486A (en) | 2018-07-13 |
| CN108280486B CN108280486B (en) | 2020-06-23 |
Family
ID=62807325
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810104197.1A Active CN108280486B (en) | 2018-02-02 | 2018-02-02 | Hyperspectral image unmixing method based on end member cluster |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108280486B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109696406A (en) * | 2018-11-29 | 2019-04-30 | 北京航空航天大学 | A kind of menology high spectrum image shadow region solution mixing method based on compound end member |
| CN109840544A (en) * | 2018-12-19 | 2019-06-04 | 嘉兴市恒创电力设备有限公司 | A kind of high spectrum image multiterminal member spectral mixture analysis method and device |
| CN112733867A (en) * | 2021-02-04 | 2021-04-30 | 大连民族大学 | End member extraction method of hyperspectral image |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060233421A1 (en) * | 2005-04-15 | 2006-10-19 | The Boeing Company | Methods and apparatus for adaptive foreground background analysis |
| CN102314685A (en) * | 2011-07-23 | 2012-01-11 | 北京航空航天大学 | Hyperspectral image sparse unmixing method based on random projection |
| CN103020912A (en) * | 2012-12-18 | 2013-04-03 | 武汉大学 | Remote sensing image restoration method combining wave-band clustering with sparse representation |
| CN104036509A (en) * | 2014-06-19 | 2014-09-10 | 哈尔滨工业大学 | Method for unmixing hyperspectral mixed pixel based on compressed sensing |
| CN105320959A (en) * | 2015-09-30 | 2016-02-10 | 西安电子科技大学 | End member learning based hyperspectral image sparse unmixing method |
-
2018
- 2018-02-02 CN CN201810104197.1A patent/CN108280486B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060233421A1 (en) * | 2005-04-15 | 2006-10-19 | The Boeing Company | Methods and apparatus for adaptive foreground background analysis |
| CN102314685A (en) * | 2011-07-23 | 2012-01-11 | 北京航空航天大学 | Hyperspectral image sparse unmixing method based on random projection |
| CN103020912A (en) * | 2012-12-18 | 2013-04-03 | 武汉大学 | Remote sensing image restoration method combining wave-band clustering with sparse representation |
| CN104036509A (en) * | 2014-06-19 | 2014-09-10 | 哈尔滨工业大学 | Method for unmixing hyperspectral mixed pixel based on compressed sensing |
| CN105320959A (en) * | 2015-09-30 | 2016-02-10 | 西安电子科技大学 | End member learning based hyperspectral image sparse unmixing method |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109696406A (en) * | 2018-11-29 | 2019-04-30 | 北京航空航天大学 | A kind of menology high spectrum image shadow region solution mixing method based on compound end member |
| CN109696406B (en) * | 2018-11-29 | 2020-10-02 | 北京航空航天大学 | Moon table hyperspectral image shadow region unmixing method based on composite end member |
| CN109840544A (en) * | 2018-12-19 | 2019-06-04 | 嘉兴市恒创电力设备有限公司 | A kind of high spectrum image multiterminal member spectral mixture analysis method and device |
| CN112733867A (en) * | 2021-02-04 | 2021-04-30 | 大连民族大学 | End member extraction method of hyperspectral image |
| CN112733867B (en) * | 2021-02-04 | 2023-08-08 | 大连民族大学 | End member extraction method of hyperspectral image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108280486B (en) | 2020-06-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Three-dimensional convolutional neural network model for tree species classification using airborne hyperspectral images | |
| CN104952050B (en) | High spectrum image adaptive de mixing method based on region segmentation | |
| CN108460391B (en) | Hyperspectral image unsupervised feature extraction method based on generation countermeasure network | |
| CN107292317B (en) | Polarization SAR classification method based on shallow feature and T matrix deep learning | |
| CN104978573B (en) | A kind of non-negative matrix factorization method applied to Hyperspectral imagery processing | |
| CN107145836B (en) | Hyperspectral image classification method based on stacked boundary identification self-encoder | |
| CN103208011B (en) | Based on average drifting and the hyperspectral image space-spectral domain classification method organizing sparse coding | |
| Jia et al. | Vegetation classification method with biochemical composition estimated from remote sensing data | |
| CN108171122A (en) | The sorting technique of high-spectrum remote sensing based on full convolutional network | |
| CN108229551B (en) | Hyperspectral remote sensing image classification method based on compact dictionary sparse representation | |
| CN109543723B (en) | Robust image clustering method | |
| Lin et al. | Integrating model-and data-driven methods for synchronous adaptive multi-band image fusion | |
| CN108280486A (en) | A kind of high spectrum image solution mixing method based on end member cluster | |
| Chen et al. | Agricultural remote sensing image cultivated land extraction technology based on deep learning | |
| CN108470192B (en) | Hyperspectral classification method and device | |
| Feng et al. | Analyzing the role of spatial features when cooperating hyperspectral and LiDAR data for the tree species classification in a subtropical plantation forest area | |
| Zhang et al. | Local information interaction transformer for hyperspectral and LiDAR data classification | |
| Bao et al. | Method for wheat ear counting based on frequency domain decomposition of MSVF-ISCT | |
| Huang et al. | Background purification framework with extended morphological attribute profile for hyperspectral anomaly detection | |
| CN108734672B (en) | Hyperspectral data unmixing method based on spectral library cutting and collaborative sparse regression | |
| Gupta et al. | Fusion of texture and wavelet features of PALSAR image using LDA and PCA for land cover classification | |
| CN109447009B (en) | Hyperspectral image classification method based on subspace nuclear norm regularization regression model | |
| CN111460943A (en) | Remote sensing image ground object classification method and system | |
| CN111985501B (en) | Hyperspectral image feature extraction method based on self-adaptive high-order tensor decomposition | |
| Ye et al. | A novel semi-supervised learning framework for hyperspectral image classification |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |