CN106228186A - Classification hyperspectral imagery apparatus and method - Google Patents
Classification hyperspectral imagery apparatus and method Download PDFInfo
- Publication number
- CN106228186A CN106228186A CN201610571286.8A CN201610571286A CN106228186A CN 106228186 A CN106228186 A CN 106228186A CN 201610571286 A CN201610571286 A CN 201610571286A CN 106228186 A CN106228186 A CN 106228186A
- Authority
- CN
- China
- Prior art keywords
- pixel
- classification
- probability
- image
- level
- 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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a kind of classification hyperspectral imagery apparatus and method, relate to image processing field.Described hyperspectral classification image method is by extracting the Pixel-level structural eigenvector of a high spectrum image, sub-pixel spectral mixing characteristic vector and super-pixel level sky spectrum similarity feature vector；Utilize support vector machine that high spectrum image carries out classification respectively pixel-by-pixel according to Pixel-level structural eigenvector, sub-pixel spectral mixing characteristic vector and super-pixel level sky spectrum similarity feature vector to estimate, thus obtain three class probability matrixes；The joint classification probability matrix structure energy function obtained according to class label, default regulatory factor and three class probability matrixes, asks for the minima of energy function, thus obtains classification results image array.The classification hyperspectral imagery apparatus and method that the present invention provides, can promote the nicety of grading of the smooth region to high spectrum image, can reduce again the misclassification situation of the structural texture close quarters to high spectrum image.
Description
Technical field
The present invention relates to image processing field, in particular to a kind of classification hyperspectral imagery apparatus and method.
Background technology
High spectrum image has multiple spectral band, for identification (such as, lawn, the road of the scene in high spectrum image
And river etc.) provide abundant diversity spectral information.But, done by picture noise and spectral mixing characteristic
Disturb, rely solely on spectral information and be difficult to obtain high-precision classification results.Therefore, the empty spectrum signature of more separability is effectively extracted
Become the study hotspot in classification hyperspectral imagery field in recent years.
In prior art, generally use empty profile classification method based on super-pixel segmentation that high spectrum image is split,
Empty profile classification method based on super-pixel segmentation can make full use of the local similarity spatially and spectrally information of high spectrum image
Promoting the classification results of high spectrum image smooth region, can effectively solve to classify in the case of picture noise exists difficult problem, but
It is to use empty profile classification method based on super-pixel segmentation that high spectrum image is split, there is the limit to high spectrum image
The problem of the mistake segmentation in the region that edge, structure, texture are intensive, splits very inaccuracy.
Summary of the invention
In view of this, the purpose of the embodiment of the present invention is to provide a kind of classification hyperspectral imagery apparatus and method.
First aspect, embodiments provides a kind of classification hyperspectral imagery device, described classification hyperspectral imagery
Device includes:
Characteristic vector pickup unit, for extracting the Pixel-level structural eigenvector of a high spectrum image, sub-pixel light
Spectrum composite character vector and super-pixel level sky spectrum similarity feature vector；
Class probability matrix obtains unit, for according to described Pixel-level structural eigenvector, described sub-pixel spectrum
Composite character is vectorial and described super-pixel level sky spectrum similarity feature vector utilizes support vector machine to described high spectrum image
Carry out classification pixel-by-pixel respectively to estimate, thus obtain three class probability matrixes；
Class probability sequence obtains unit, for obtaining each described class probability matrix sort according to default rule
Three class probability sequences；
Probability Estimation confidence level obtains unit, estimates for obtaining three probability respectively according to three described class probability sequences
Meter confidence level；
Joint classification probability matrix obtains unit, melts for three described probability Estimation confidence levels are carried out adaptive line
Close, thus obtain joint classification probability matrix；
Classification results image array obtains unit, for according to described joint classification probability matrix, described high spectrum image
Each neighborhood in class label corresponding to two pixels and default regulatory factor structure energy function, and according to based on
Figure hugger opinion-expansion algorithm asks for the minima of described energy function, and obtain point according to the minima of described energy function
Class result images matrix.
Second aspect, the embodiment of the present invention additionally provides a kind of hyperspectral classification image method, described hyperspectral classification figure
Image space method includes:
Extract the Pixel-level structural eigenvector of a high spectrum image, sub-pixel spectral mixing characteristic vector and super picture
Element level sky spectrum similarity feature vector；
According to described Pixel-level structural eigenvector, described sub-pixel spectral mixing characteristic vector and described super-pixel
Level empty spectrum similarity feature vector utilizes support vector machine that described high spectrum image carries out classification estimation respectively pixel-by-pixel, thus
Obtain three class probability matrixes；
According to default rule, each described class probability matrix sort is obtained three class probability sequences；
Three probability Estimation confidence levels are obtained respectively according to three described class probability sequences；
Three described probability Estimation confidence levels are carried out adaptive line fusion, thus obtains joint classification probability matrix；
According to described joint classification probability matrix, described high spectrum image each neighborhood in class corresponding to two pixels
Distinguishing label and default regulatory factor structure energy function, and ask for described energy according to α-expansion algorithm based on figure hugger opinion
The minima of flow function, and obtain classification results image array according to the minima of described energy function.
Compared with prior art, the classification hyperspectral imagery apparatus and method that the present invention provides, can effectively promote Gao Guang
The nicety of grading of the smooth region of spectrogram picture, can effectively reduce again the structural texture close quarters to high spectrum image simultaneously and occur
Misclassification situation, and can also effectively be lifted at the classification performance under noise and spectral mixing serious conditions.
For making the above-mentioned purpose of the present invention, characteristic vector and advantage to become apparent, preferred embodiment cited below particularly, and
Coordinate appended accompanying drawing, be described in detail below.
Accompanying drawing explanation
For making the purpose of the embodiment of the present invention, technical scheme and advantage clearer, below in conjunction with the embodiment of the present invention
In accompanying drawing, the technical scheme in the embodiment of the present invention is clearly and completely described, it is clear that described embodiment is
The a part of embodiment of the present invention rather than whole embodiments.Generally implement with the present invention illustrated described in accompanying drawing herein
The assembly of example can be arranged with various different configurations and design.Therefore, reality to the present invention provided in the accompanying drawings below
The detailed description executing example is not intended to limit the scope of claimed invention, but is merely representative of the selected enforcement of the present invention
Example.Based on the embodiment in the present invention, those of ordinary skill in the art are obtained under not making creative work premise
Every other embodiment, broadly falls into the scope of protection of the invention.
The block diagram of the server that Fig. 1 provides for the embodiment of the present invention；
The functional unit schematic diagram of the classification hyperspectral imagery device that Fig. 2 provides for the embodiment of the present invention；
The subelement schematic diagram of the characteristic vector pickup unit that Fig. 3 provides for the embodiment of the present invention；
The flow chart of the hyperspectral image classification method that Fig. 4 provides for the embodiment of the present invention.
Wherein, the corresponding relation between reference and component names is as follows: classification hyperspectral imagery device 100, service
Device 101, processor 102, memorizer 103, storage control 104, Peripheral Interface 105, characteristic vector pickup unit 201, classification
Probability matrix obtains unit 202, and class probability sequence obtains unit 203, and probability Estimation confidence level obtains unit 204, combines point
Class probability matrix obtains unit 205, and classification results image array obtains unit 206, and morphology sequence obtains subelement 301, as
Element level structure characteristic vector pickup subelement 302, cluster centre obtains subelement 303, sub-pixel characteristic vector pickup list
Unit 304, basic image forms subelement 305, and super-pixel level similarity feature vector extracts subelement 306.
Detailed description of the invention
Below in conjunction with accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Ground describes, it is clear that described embodiment is only a part of embodiment of the present invention rather than whole embodiments.Generally exist
Can arrange and design with various different configurations with the assembly of the embodiment of the present invention that illustrates described in accompanying drawing herein.Cause
This, be not intended to limit claimed invention to the detailed description of the embodiments of the invention provided in the accompanying drawings below
Scope, but it is merely representative of the selected embodiment of the present invention.Based on embodiments of the invention, those skilled in the art are not doing
The every other embodiment obtained on the premise of going out creative work, broadly falls into the scope of protection of the invention.
The classification hyperspectral imagery apparatus and method that the embodiment of the present invention proposes, it is provided that a kind of classification hyperspectral imagery side
Method, this hyperspectral image classification method is applicable to server 101.This server 101 may be, but not limited to, network service
Device, database server, cloud server etc..
As it is shown in figure 1, be the block diagram of described server 101.Described server 101 includes classification hyperspectral imagery
Device 100, processor 102, memorizer 103, storage control 104 and Peripheral Interface 105.
Described memorizer 103, storage control 104 and processor 102, each element is the most electrical
Connect, to realize the transmission of data or mutual.Such as, these elements can pass through one or more communication bus or letter each other
Number line realizes being electrically connected with.Described classification hyperspectral imagery device 100 includes that at least one can be with software or firmware
(firmware) form is stored in described memorizer 103 or is solidificated in the operating system of described server 101
Software function module in (operating system, OS).Described processor 102 is for performing storage in memorizer 103
Executable module, such as, software function module that described classification hyperspectral imagery device 100 includes or computer program.
Wherein, memorizer 103 may be, but not limited to, random access memory (Random Access Memory,
RAM), read only memory Read Only Memory, ROM), programmable read only memory (Programmable Read-Only
Memory, PROM), erasable read-only memory (Erasable Programmable Read-Only Memory, EPROM),
Electricallyerasable ROM (EEROM) (Electric Erasable Programmable Read-Only Memory, EEPROM) etc..
Wherein, memorizer 103 is used for storing program, and described processor 102, after receiving execution instruction, performs described program, aforementioned
The method performed by server 101 flowing through Cheng Dingyi that embodiment of the present invention any embodiment discloses can apply to processor
In 102, or realized by processor 102.
Processor 102 is probably a kind of IC chip, has the disposal ability of signal.Above-mentioned processor 102 can
To be general processor, including central processing unit (Central Processing Unit is called for short CPU), network processing unit
(Network Processor is called for short NP) etc.；Can also is that digital signal processor (DSP), special IC (ASIC),
Ready-made programmable gate array (FPGA) or other PLDs, discrete gate or transistor logic, discrete firmly
Part assembly.Can realize or perform disclosed each method, step and the logic diagram in the embodiment of the present invention.General processor
Can be microprocessor or this processor can also be the processor etc. of any routine.
Various input/output devices are coupled to processor and memorizer 103 by described Peripheral Interface 105.Real at some
Executing in example, Peripheral Interface 105, processor 102 and storage control 104 can realize in one single chip.Some other
In example, they can be realized by independent chip respectively.
Refer to Fig. 2, a kind of classification hyperspectral imagery device 100 that the embodiment of the present invention provides, described high spectrum image
Sorter 100 includes that characteristic vector pickup unit 201, class probability matrix obtain unit 202, class probability sequence obtains single
Unit 203, probability Estimation confidence level obtain unit 204, joint classification probability matrix acquisition unit 205 and classification results image moment
Battle array obtains unit 206.
Described characteristic vector pickup unit 201 is for extracting the Pixel-level structural eigenvector of a high spectrum image, sub-picture
Element level spectral mixing characteristic vector and super-pixel level sky spectrum similarity feature vector.
As it is shown on figure 3, specifically, described characteristic vector pickup unit 201 includes: morphology sequence obtains subelement 301,
For utilizing principal component analysis method to extract the top n main component of described high spectrum image, respectively each main component is entered
Row opening operation, closed operation and calculus of differences, thus obtain difference morphology sequence.Further, described morphology sequence obtains
Take subelement 301 for utilizing principal component analysis method to extract the top n main component of described high spectrum image, foundation formulaEach main component is carried out opening operation and obtains morphology sequence MPYa, according to formulaEach main component is carried out closed operation, thus obtains morphology sequence MPYe, to phase
Adjacent morphology sequence MPYaOr morphology sequence MPYeCarry out calculus of differences thus obtain based on main constituent IjDifference morphology
Sequence.It is used for utilizing principal component analysis method to extract described high-spectrum it is preferred that described morphology sequence obtains subelement 301
Front 3 main components of picture, of course, morphology sequence obtains subelement 301 and not only can be used for utilizing principal component analysis
Method extracts front 3 main components of described high spectrum image, it is also possible to for front 1 main component, 2 main components or 4 masters
Want composition, do not limit at this, be merely illustrative at this.
Pixel-level structural eigenvector extracts subelement 302, for extracting Pixel-level according to described difference morphology sequence
Structural eigenvector.
Specifically, described Pixel-level structural eigenvector extracts subelement 302 for according to formula
Wherein, xiFor pixel,For pixel xiCorresponding dot structure characteristic vector.
Cluster centre obtains subelement 303, is used for using minimal noise separation converter technique that high spectrum image is carried out dimensionality reduction,
And utilize quick K-mean algorithm that the high spectrum image after dimensionality reduction is clustered, thus obtain C cluster centre, wherein, C is
The kind number of thing to be sorted.
Sub-pixel characteristic vector pickup subelement 304, for obtaining C cluster centre according to hybrid modulation filtering method
Abundance, and according to described abundance extract sub-pixel characteristic vector.
Specifically, described sub-pixel characteristic vector pickup subelement 304 is for according to formula
(i.e. hybrid modulation filtering method) is estimated in C cluster
The abundance of the heart, and according to formulaExtraction sub-pixel characteristic vector, wherein, For pixel xiCorresponding sub-pixel characteristic vector, EcFor cluster centre, R is correlation matrix,For abundance.
Basic image forms subelement 305, forms basic image for the top n main component according to described high spectrum image.
Super-pixel level similarity feature vector extracts subelement 306, is used for utilizing entropy rate over-segmentation method to extract described base
Super-pixel in image, and extract super-pixel level similar features vector according to described super-pixel and mean operator.
Using entropy rate over-segmentation method to extract super-pixel in basic image, the number of super-pixel can be predefined forOwing to each super-pixel comprising the spectral pixel that spectrum is similar with spatial information, therefore by profit
Use formula(i.e. mean operator) can obtain super-pixel level similarity feature.Wherein, M × N is EO-1 hyperion
The number of pixels comprised in image, NSUPFor super-pixel number, SiRepresent all collection of pixels that i-th super-pixel comprises, MiTable
Show i-th super-pixel SiThe spectral pixel number comprised.
Described class probability matrix obtains unit 202 for according to described Pixel-level structural eigenvector, described sub-pix
Level spectral mixing characteristic vector and described super-pixel level sky spectrum similarity feature vector utilize support vector machine to described Gao Guang
Spectrogram picture carries out classification pixel-by-pixel respectively and estimates, thus obtains three class probability matrixes.
Described class probability sequence obtains unit 203 for arranging each described class probability matrix according to default rule
Sequence obtains three class probability sequences.
Specifically, can be respectively by the class probability matrix obtained based on sub-pixel feature, based on Pixel-level feature point
The class probability matrix of class probability matrix and super-pixel level characteristic probability sorts from high to low and obtains class probability sequence.
Described probability Estimation confidence level obtains unit 204 for obtaining three respectively according to three described class probability sequences
Probability Estimation confidence level.
Specifically, described probability Estimation confidence level obtains unit 204 for according to described three described class probability sequences
And formula
Obtain three probability Estimation confidence levels, wherein,Respectively
Represent with pixel xiFor the class probability sequence of object of classification,For class probability sequence based on Pixel-level structural eigenvector
Row probability Estimation confidence level,For the probability Estimation confidence level of class probability sequence based on sub-pixel characteristic vector,For based on super-pixel level similarity feature vector class probability sequence probability Estimation confidence level.
Described joint classification probability matrix obtains unit 205 for three described probability Estimation confidence levels are carried out self adaptation
Linear fusion, thus obtain joint classification probability matrix.
Specifically, described joint classification probability matrix obtains unit 205 for according to probability Estimation confidence levelProbability
Estimate confidence levelProbability Estimation confidence levelAnd formula
Enter
Row adaptive line merges, thus obtains joint classification probability matrix, wherein,For joint classification probability matrix,
Described classification results image array obtains unit 206 for according to described joint classification probability matrix, described Gao Guang
Class label that two pixels in each neighborhood of spectrogram picture are corresponding and default regulatory factor structure energy function, and depend on
The minima of described energy function is asked for according to α-expansion algorithm based on figure hugger opinion, and according to the minima of described energy function
Obtain classification results image array.
Specifically, described classification results image array obtains unit 206 for according to described joint classification probability matrix, institute
State class label corresponding to two pixels in each neighborhood of high spectrum image and default regulatory factor structure energy letter
Number, and the minima of described energy function is asked for according to α-expansion algorithm based on figure hugger opinion, and according to described energy function
Minima obtain classification results image array, wherein, the formula of energy function isE
(Si) it is energy function value, Y (Si) it is classification results image array, and have
β is regulatory factor, yi, yjFor pixel xiAnd xjThe most corresponding class label.If in the neighborhood of high spectrum image
Two pixel (xiAnd xj) class label identical, namely yi=yj, δ (y in this casei,yj)=1, vice versa.β
It is balance Di(xi) and V (yi, yj) regulatory factor, such as, β=0.5.
Referring to Fig. 4, the embodiment of the present invention additionally provides a kind of hyperspectral image classification method, it should be noted that this
The hyperspectral image classification method that embodiment is provided, the technique effect of its ultimate principle and generation is identical with above-described embodiment,
For briefly describing, the not mentioned part of the present embodiment part, refer to the corresponding contents in above-described embodiment, described hyperspectral classification
Image method includes:
Step S401: extract the Pixel-level structural eigenvector of a high spectrum image, sub-pixel spectral mixing feature to
Amount and super-pixel level sky spectrum similarity feature vector.
Characteristic vector pickup unit 201 is utilized to extract the Pixel-level structural eigenvector of a high spectrum image, sub-pixel
Spectral mixing characteristic vector and super-pixel level sky spectrum similarity feature vector.Specifically, step S401 can include utilizing master
Become componential analysis to extract the top n main component of described high spectrum image, respectively each main component is carried out opening operation, closes
Computing and calculus of differences, thus obtain difference morphology sequence, extract Pixel-level structure according to described difference morphology sequence
Characteristic vector.Specifically, principal component analysis method is utilized to extract the top n main component of described high spectrum image, according to formulaEach main component is carried out opening operation and obtains morphology sequence MPYa, according to calculating
FormulaEach main component is carried out closed operation, thus obtains morphology sequence MPYe, right
Adjacent morphology sequence MPYaOr morphology sequence MPYeCarry out calculus of differences thus obtain based on main constituent IjDifference form
Learn sequence, according to formulaWherein, xiFor pixel,For pixel xiCorresponding dot structure feature to
Amount.
Use minimal noise separation converter technique that high spectrum image carries out dimensionality reduction, and utilize quick K-mean algorithm to dimensionality reduction
After high spectrum image cluster, thus obtain C cluster centre, wherein, C is the kind number of thing to be sorted, according to mixed
Close modulation filtering method and obtain the abundance of C cluster centre, and extracting sub-pixel characteristic vector according to described abundance.Specifically
Ground, according to formula(i.e. hybrid modulation filtering method) estimates that C is gathered
The abundance at class center, and according to formulaExtraction sub-pixel characteristic vector, wherein, For pixel xiCorresponding sub-pixel characteristic vector, EcFor cluster centre, R is dependency square
Battle array,For abundance.
Form basic image according to the top n main component of described high spectrum image, utilize entropy rate over-segmentation method to extract institute
State the super-pixel in basic image, and extract super-pixel level similar features vector according to described super-pixel and mean operator.
Step S402: according to described Pixel-level structural eigenvector, described sub-pixel spectral mixing characteristic vector and
Described super-pixel level sky spectrum similarity feature vector utilizes support vector machine that described high spectrum image carries out class respectively pixel-by-pixel
Do not estimate, thus obtain three class probability matrixes.
Class probability matrix is utilized to obtain unit 202 according to described Pixel-level structural eigenvector, described sub-pixel light
Spectrum composite character is vectorial and described super-pixel level sky spectrum similarity feature vector utilizes support vector machine to described high-spectrum
Estimate as carrying out classification pixel-by-pixel respectively, thus obtain three class probability matrixes.
Step S403: each described class probability matrix sort is obtained three class probability sequences according to default rule
Row.
Utilize class probability sequence to obtain unit 203 according to default rule, each described class probability matrix sort to be obtained
Obtain three class probability sequences.
Step S404: obtain three probability Estimation confidence levels respectively according to three described class probability sequences.
Utilize probability Estimation confidence level to obtain unit 204 and obtain three probability respectively according to three described class probability sequences
Estimate confidence level.
Specifically, step S404 can include according to described three described class probability sequences and formula
Obtain three probability Estimation confidence levels, wherein,Respectively
Represent with pixel xiFor the class probability sequence of object of classification,For class probability sequence based on Pixel-level structural eigenvector
Row probability Estimation confidence level,For the probability Estimation confidence level of class probability sequence based on sub-pixel characteristic vector,For based on super-pixel level similarity feature vector class probability sequence probability Estimation confidence level.
Step S405: three described probability Estimation confidence levels are carried out adaptive line fusion, thus obtains joint classification
Probability matrix.
Utilize joint classification probability matrix to obtain unit 205 and three described probability Estimation confidence levels are carried out adaptive line
Merge, thus obtain joint classification probability matrix.
Specifically, step S405 can include according to probability Estimation confidence levelProbability Estimation confidence levelGenerally
Rate estimates confidence levelAnd formula
Enter
Row adaptive line merges, thus obtains joint classification probability matrix, wherein,For joint classification probability matrix,
Step S406: according to described joint classification probability matrix, described high spectrum image each neighborhood in two pictures
Class label and default regulatory factor that element is corresponding construct energy function, and according to α-expansion algorithm based on figure hugger opinion
Ask for the minima of described energy function, and obtain classification results image array according to the minima of described energy function.
Classification results image array is utilized to obtain unit 206 according to described joint classification probability matrix, described high-spectrum
Class label that two pixels in each neighborhood of picture are corresponding and default regulatory factor structure energy function, and according to base
α-expansion algorithm in figure hugger opinion asks for the minima of described energy function, and obtains according to the minima of described energy function
Classification results image array.
Specifically, step S406 may include that according to described joint classification probability matrix, described high spectrum image each
Class label that two pixels in neighborhood are corresponding and default regulatory factor structure energy function, and according to based on figure hugger
α-the expansion algorithm of opinion asks for the minima of described energy function, and obtains classification results according to the minima of described energy function
Image array, wherein, the formula of energy function isE(Si) it is energy function value, Y (Si)
For classification results image array, and V(yi,yj)=1-δ (yi,yj), β is regulatory factor, yi, yjFor pixel xiAnd xjThe most corresponding
Class label.
To sum up, the classification hyperspectral imagery apparatus and method that the present invention provides, can effectively promote and high spectrum image is put down
The nicety of grading in territory, skating area, can effectively reduce again the misclassification feelings that the structural texture close quarters to high spectrum image occurs simultaneously
Condition, and can also effectively be lifted at the classification performance under noise and spectral mixing serious conditions.
In several embodiments provided herein, it should be understood that disclosed apparatus and method, it is also possible to pass through
Other mode realizes.Device embodiment described above is only schematically, such as, and the flow chart in accompanying drawing and block diagram
Show the device of multiple embodiments according to the present invention, the architectural framework in the cards of method and computer program product,
Function and operation.In this, each square frame in flow chart or block diagram can represent a module, program segment or the one of code
Part, a part for described module, program segment or code comprises holding of one or more logic function for realizing regulation
Row instruction.It should also be noted that at some as in the implementation replaced, the function marked in square frame can also be to be different from
The order marked in accompanying drawing occurs.Such as, two continuous print square frames can essentially perform substantially in parallel, and they are the most also
Can perform in the opposite order, this is depending on involved function.It is also noted that every in block diagram and/or flow chart
The combination of the square frame in individual square frame and block diagram and/or flow chart, can be with function or the special base of action performing regulation
System in hardware realizes, or can realize with the combination of specialized hardware with computer instruction.
It addition, each functional module in each embodiment of the present invention can integrate one independent portion of formation
Point, it is also possible to it is modules individualism, it is also possible to two or more modules are integrated to form an independent part.
If described function is using the form realization of software function module and as independent production marketing or use, permissible
It is stored in a computer read/write memory medium.Based on such understanding, technical scheme is the most in other words
The part contributing prior art or the part of this technical scheme can embody with the form of software product, this meter
Calculation machine software product is stored in a storage medium, including some instructions with so that a computer equipment (can be individual
People's computer, server, or the network equipment etc.) perform all or part of step of method described in each embodiment of the present invention.
And aforesaid storage medium includes: USB flash disk, portable hard drive, read only memory (ROM, Read-Only Memory), random access memory are deposited
The various media that can store program code such as reservoir (RAM, Random Access Memory), magnetic disc or CD.Need
Illustrate, in this article, the relational terms of such as first and second or the like be used merely to by an entity or operation with
Another entity or operating space separate, and there is any this reality between not necessarily requiring or imply these entities or operating
The relation on border or order.And, term " includes ", " comprising " or its any other variant are intended to the bag of nonexcludability
Contain, so that include that the process of a series of key element, method, article or equipment not only include those key elements, but also include
Other key elements being not expressly set out, or also include the key element intrinsic for this process, method, article or equipment.
In the case of there is no more restriction, statement " including ... " key element limited, it is not excluded that including described key element
Process, method, article or equipment in there is also other identical element.
The foregoing is only the preferred embodiments of the present invention, be not limited to the present invention, for the skill of this area
For art personnel, the present invention can have various modifications and variations.All within the spirit and principles in the present invention, that is made any repaiies
Change, equivalent, improvement etc., should be included within the scope of the present invention.It should also be noted that similar label and letter exist
Figure below represents similar terms, therefore, the most a certain Xiang Yi accompanying drawing is defined, is then not required in accompanying drawing subsequently
It is defined further and explains.
The above, the only detailed description of the invention of the present invention, but protection scope of the present invention is not limited thereto, and any
Those familiar with the art, in the technical scope that the invention discloses, can readily occur in change or replace, should contain
Cover within protection scope of the present invention.Therefore, protection scope of the present invention should described be as the criterion with scope of the claims.
It should be noted that in this article, the relational terms of such as first and second or the like is used merely to a reality
Body or operation separate with another entity or operating space, and deposit between not necessarily requiring or imply these entities or operating
Relation or order in any this reality.And, term " includes ", " comprising " or its any other variant are intended to
Comprising of nonexcludability, so that include that the process of a series of key element, method, article or equipment not only include that those are wanted
Element, but also include other key elements being not expressly set out, or also include for this process, method, article or equipment
Intrinsic key element.In the case of there is no more restriction, statement " including ... " key element limited, it is not excluded that
Including process, method, article or the equipment of described key element there is also other identical element.
Claims (10)
1. a classification hyperspectral imagery device, it is characterised in that described classification hyperspectral imagery device includes:
Characteristic vector pickup unit, for extracting the Pixel-level structural eigenvector of a high spectrum image, sub-pixel spectrum mixes
Close characteristic vector and super-pixel level sky spectrum similarity feature vector；
Class probability matrix obtains unit, for according to described Pixel-level structural eigenvector, described sub-pixel spectral mixing
Characteristic vector and described super-pixel level sky spectrum similarity feature vector utilize support vector machine to described high spectrum image respectively
Carry out classification pixel-by-pixel to estimate, thus obtain three class probability matrixes；
Class probability sequence obtains unit, for each described class probability matrix sort being obtained three according to default rule
Class probability sequence；
Probability Estimation confidence level obtains unit, puts for obtaining three probability Estimation respectively according to three described class probability sequences
Reliability；
Joint classification probability matrix obtains unit, for three described probability Estimation confidence levels are carried out adaptive line fusion,
Thus obtain joint classification probability matrix；
Classification results image array obtain unit, for according to described joint classification probability matrix, described high spectrum image every
Class label that two pixels in individual neighborhood are corresponding and default regulatory factor structure energy function, and according to cutting based on figure
Theoretical α-expansion algorithm asks for the minima of described energy function, and obtains classification knot according to the minima of described energy function
Really image array.
Classification hyperspectral imagery device the most according to claim 1, it is characterised in that described characteristic vector pickup unit bag
Include:
Morphology sequence obtains subelement, main for the top n utilizing the principal component analysis method described high spectrum image of extraction
Composition, carries out opening operation, closed operation and calculus of differences respectively, thus obtains difference morphology sequence each main component；
Pixel-level structural eigenvector extracts subelement, for extracting Pixel-level architectural feature according to described difference morphology sequence
Vector；
Cluster centre obtains subelement, is used for using minimal noise separation converter technique that high spectrum image is carried out dimensionality reduction, and utilizes
Quickly the high spectrum image after dimensionality reduction is clustered by K-mean algorithm, thus obtains C cluster centre, and wherein, C is to be sorted
The kind number of thing；
Sub-pixel characteristic vector pickup subelement, for obtaining the abundance of C cluster centre according to hybrid modulation filtering method,
And extract sub-pixel characteristic vector according to described abundance；
Basic image forms subelement, forms basic image for the top n main component according to described high spectrum image；
Super-pixel level similarity feature vector extracts subelement, for utilizing entropy rate over-segmentation method to extract in described basic image
Super-pixel, and extract super-pixel level similar features vector according to described super-pixel and mean operator.
Classification hyperspectral imagery device the most according to claim 2, it is characterised in that
Described morphology sequence obtains subelement for utilizing principal component analysis method to extract the top n master of described high spectrum image
Want composition, according to formulaEach main component is carried out opening operation and obtains morphology
Sequence MPYa, according to formulaEach main component is carried out closed operation, thus obtains shape
State sequence MPYe, to adjacent morphology sequence MPYaOr morphology sequence MPYeCarry out calculus of differences thus obtain based on master
Composition IjDifference morphology sequence.
Described Pixel-level structural eigenvector extracts subelement for according to formulaWherein, xiFor pixel,For pixel xiCorresponding dot structure characteristic vector.
Classification hyperspectral imagery device the most according to claim 3, it is characterised in that described sub-pixel characteristic vector carries
Take subelement for according to formula
Estimate the abundance of C cluster centre, and according to formulaExtraction sub-pixel characteristic vector, wherein, For pixel xiCorresponding
Sub-pixel characteristic vector, EcFor cluster centre, R is correlation matrix,For abundance.
Classification hyperspectral imagery device the most according to claim 4, it is characterised in that
Described probability Estimation confidence level obtains unit, for according to described three described class probability sequences and formula
Obtain three probability Estimation confidence levels, wherein,Represent respectively
With pixel xiFor the class probability sequence of object of classification,For class probability sequence based on Pixel-level structural eigenvector
Probability Estimation confidence level,For the probability Estimation confidence level of class probability sequence based on sub-pixel characteristic vector,
For based on super-pixel level similarity feature vector class probability sequence probability Estimation confidence level.
Described joint classification probability matrix obtains unit, for according to probability Estimation confidence levelProbability Estimation confidence levelProbability Estimation confidence levelAnd formula
Carry out adaptive line fusion, thus obtain joint classification probability matrix, wherein,For joint classification probability matrix,
Classification results image array obtain unit, for according to described joint classification probability matrix, described high spectrum image every
Class label that two pixels in individual neighborhood are corresponding and default regulatory factor structure energy function, and according to cutting based on figure
Theoretical α-expansion algorithm asks for the minima of described energy function, and obtains classification knot according to the minima of described energy function
Really image array, wherein, the formula of energy function isE(Si) it is energy function value, Y
(Si) it is classification results image array, and V(yi,yj)=1-δ (yi,yj), β is regulatory factor, yi, yjFor pixel xiAnd xjThe most corresponding
Class label.
6. a hyperspectral image classification method, it is characterised in that described hyperspectral classification image method includes:
Extract the Pixel-level structural eigenvector of a high spectrum image, sub-pixel spectral mixing characteristic vector and super-pixel level
Empty spectrum similarity feature vector；
Empty according to described Pixel-level structural eigenvector, described sub-pixel spectral mixing characteristic vector and described super-pixel level
Spectrum similarity feature vector utilizes support vector machine that described high spectrum image carries out classification estimation respectively pixel-by-pixel, thus obtains
Three class probability matrixes；
According to default rule, each described class probability matrix sort is obtained three class probability sequences；
Three probability Estimation confidence levels are obtained respectively according to three described class probability sequences；
Three described probability Estimation confidence levels are carried out adaptive line fusion, thus obtains joint classification probability matrix；
According to described joint classification probability matrix, described high spectrum image each neighborhood in classification mark corresponding to two pixels
The regulatory factor structure energy function signed and preset, and ask for described energy letter according to α-expansion algorithm based on figure hugger opinion
The minima of number, and obtain classification results image array according to the minima of described energy function.
Hyperspectral image classification method the most according to claim 6, it is characterised in that described extraction one high spectrum image
Pixel-level structural eigenvector, sub-pixel spectral mixing characteristic vector and the step of super-pixel level sky spectrum similarity feature vector
Suddenly include:
Utilize principal component analysis method to extract the top n main component of described high spectrum image, respectively each main component is entered
Row opening operation, closed operation and calculus of differences, thus obtain difference morphology sequence, extract according to described difference morphology sequence
Pixel-level structural eigenvector；
Use minimal noise separation converter technique that high spectrum image carries out dimensionality reduction, and after utilizing quick K-mean algorithm to dimensionality reduction
High spectrum image clusters, thus obtains C cluster centre, and wherein, C is the kind number of thing to be sorted, adjusts according to mixing
Filtering method processed and the abundance of C cluster centre of acquisition, and extract sub-pixel characteristic vector according to described abundance；
Form basic image according to the top n main component of described high spectrum image, utilize entropy rate over-segmentation method to extract described base
Super-pixel in image, and extract super-pixel level similar features vector according to described super-pixel and mean operator.
Hyperspectral image classification method the most according to claim 7, it is characterised in that
The described top n main component utilizing principal component analysis method to extract described high spectrum image, respectively to each main one-tenth
Divide and carry out opening operation, closed operation and calculus of differences, thus obtain difference morphology sequence, according to described difference morphology sequence
The step extracting Pixel-level structural eigenvector includes:
Principal component analysis method is utilized to extract the top n main component of described high spectrum image, according to formulaEach main component is carried out opening operation and obtains morphology sequence MPYa, according to calculating
FormulaEach main component is carried out closed operation, thus obtains morphology sequence MPYe, to phase
Adjacent morphology sequence MPYaOr morphology sequence MPYeCarry out calculus of differences thus obtain based on main constituent IjDifference morphology
Sequence, according to formulaWherein, xiFor pixel,For pixel xiCorresponding dot structure characteristic vector.
Hyperspectral image classification method the most according to claim 8, it is characterised in that described according to hybrid modulation filtering side
Method obtains the abundance of C cluster centre, and includes according to the step of described abundance extraction sub-pixel characteristic vector:
According to formulaEstimate the abundance of C cluster centre, and foundation
FormulaExtraction sub-pixel characteristic vector, wherein, For pixel xiRight
The sub-pixel characteristic vector answered, EcFor cluster centre, R is correlation matrix,For abundance.
Hyperspectral image classification method the most according to claim 9, it is characterised in that
The step that the described class probability sequence of described foundation three obtains three probability Estimation confidence levels respectively includes: according to described
Three described class probability sequences and formula
Obtain three probability Estimation confidence levels, wherein,Represent respectively
With pixel xiFor the class probability sequence of object of classification,For class probability sequence based on Pixel-level structural eigenvector
Probability Estimation confidence level,For the probability Estimation confidence level of class probability sequence based on sub-pixel characteristic vector,
For based on super-pixel level similarity feature vector class probability sequence probability Estimation confidence level.
Described three described probability Estimation confidence levels are carried out adaptive line fusion, thus obtain joint classification probability matrix
Step includes:
According to probability Estimation confidence levelProbability Estimation confidence levelProbability Estimation confidence levelAnd formula
Described according to described joint classification probability matrix, described high spectrum image each neighborhood in class corresponding to two pixels
Distinguishing label and default regulatory factor structure energy function, and ask for described energy according to α-expansion algorithm based on figure hugger opinion
The minima of flow function, and include according to the step of the minima acquisition classification results image array of described energy function:
According to described joint classification probability matrix, described high spectrum image each neighborhood in classification mark corresponding to two pixels
The regulatory factor structure energy function signed and preset, and ask for described energy letter according to α-expansion algorithm based on figure hugger opinion
The minima of number, and obtain classification results image array, wherein, the formula of energy function according to the minima of described energy function
ForE(Si) it is energy function value, Y (Si) it is classification results image array, and
V(yi,yj)=1-δ (yi,yj), β is regulatory factor, yi, yjFor pixel xiAnd xjThe most right
The class label answered.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610571286.8A CN106228186B (en) | 2016-07-20 | 2016-07-20 | Classification hyperspectral imagery apparatus and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610571286.8A CN106228186B (en) | 2016-07-20 | 2016-07-20 | Classification hyperspectral imagery apparatus and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106228186A true CN106228186A (en) | 2016-12-14 |
| CN106228186B CN106228186B (en) | 2019-11-15 |
Family
ID=57531782
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610571286.8A Active CN106228186B (en) | 2016-07-20 | 2016-07-20 | Classification hyperspectral imagery apparatus and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106228186B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107133633A (en) * | 2017-03-17 | 2017-09-05 | 广东工业大学 | The algorithm of classification hyperspectral imagery based on transfinite learning machine and confidence spread |
| CN107590472A (en) * | 2017-09-19 | 2018-01-16 | 中国科学院西安光学精密机械研究所 | Spectral classification device and sorting technique based on push-broom type optical spectrum imagers |
| CN111144488A (en) * | 2019-12-27 | 2020-05-12 | 之江实验室 | Pathological section visual field classification improving method based on adjacent joint prediction |
| CN111814727A (en) * | 2020-07-22 | 2020-10-23 | 吉林大学 | Hyperspectrum-based potential fingerprint detection method |
| CN112235291A (en) * | 2020-10-13 | 2021-01-15 | 陕西师范大学 | Pixel-by-pixel texture generation type information hiding and recovering method, equipment and medium |
| CN113470036A (en) * | 2021-09-02 | 2021-10-01 | 湖南大学 | Hyperspectral image unsupervised waveband selection method and system based on knowledge distillation |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102708373A (en) * | 2012-01-06 | 2012-10-03 | 香港理工大学 | Method and device for classifying remote images by integrating space information and spectral information |
| CN104331880A (en) * | 2014-10-20 | 2015-02-04 | 西安电子科技大学 | Hyper-spectral mixed pixel decomposition method based on geometric spatial spectral structure information |
| US20150043796A1 (en) * | 2013-08-07 | 2015-02-12 | Toshiba Medical Systems Corporation | Sinogram (data) domain pansharpening method and system for spectral ct |
| CN104952050A (en) * | 2015-07-07 | 2015-09-30 | 西安电子科技大学 | Self-adaptive hyperspectral image unmixing method based on region segmentation |
| CN105654117A (en) * | 2015-12-25 | 2016-06-08 | 西北工业大学 | Hyperspectral image spectral-spatial cooperative classification method based on SAE depth network |
-
2016
- 2016-07-20 CN CN201610571286.8A patent/CN106228186B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102708373A (en) * | 2012-01-06 | 2012-10-03 | 香港理工大学 | Method and device for classifying remote images by integrating space information and spectral information |
| US20150043796A1 (en) * | 2013-08-07 | 2015-02-12 | Toshiba Medical Systems Corporation | Sinogram (data) domain pansharpening method and system for spectral ct |
| CN104331880A (en) * | 2014-10-20 | 2015-02-04 | 西安电子科技大学 | Hyper-spectral mixed pixel decomposition method based on geometric spatial spectral structure information |
| CN104952050A (en) * | 2015-07-07 | 2015-09-30 | 西安电子科技大学 | Self-adaptive hyperspectral image unmixing method based on region segmentation |
| CN105654117A (en) * | 2015-12-25 | 2016-06-08 | 西北工业大学 | Hyperspectral image spectral-spatial cooperative classification method based on SAE depth network |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107133633A (en) * | 2017-03-17 | 2017-09-05 | 广东工业大学 | The algorithm of classification hyperspectral imagery based on transfinite learning machine and confidence spread |
| CN107590472A (en) * | 2017-09-19 | 2018-01-16 | 中国科学院西安光学精密机械研究所 | Spectral classification device and sorting technique based on push-broom type optical spectrum imagers |
| CN107590472B (en) * | 2017-09-19 | 2024-04-05 | 中国科学院西安光学精密机械研究所 | Spectrum classifier and classification method based on push-broom spectrum imager |
| CN111144488A (en) * | 2019-12-27 | 2020-05-12 | 之江实验室 | Pathological section visual field classification improving method based on adjacent joint prediction |
| CN111144488B (en) * | 2019-12-27 | 2023-04-18 | 之江实验室 | Pathological section visual field classification improving method based on adjacent joint prediction |
| CN111814727A (en) * | 2020-07-22 | 2020-10-23 | 吉林大学 | Hyperspectrum-based potential fingerprint detection method |
| CN111814727B (en) * | 2020-07-22 | 2022-06-24 | 吉林大学 | Hyperspectrum-based potential fingerprint detection method |
| CN112235291A (en) * | 2020-10-13 | 2021-01-15 | 陕西师范大学 | Pixel-by-pixel texture generation type information hiding and recovering method, equipment and medium |
| CN112235291B (en) * | 2020-10-13 | 2023-04-07 | 陕西师范大学 | Pixel-by-pixel texture generation type information hiding and recovering method, equipment and medium |
| CN113470036A (en) * | 2021-09-02 | 2021-10-01 | 湖南大学 | Hyperspectral image unsupervised waveband selection method and system based on knowledge distillation |
| CN113470036B (en) * | 2021-09-02 | 2021-11-23 | 湖南大学 | Hyperspectral image unsupervised waveband selection method and system based on knowledge distillation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106228186B (en) | 2019-11-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106228186A (en) | Classification hyperspectral imagery apparatus and method | |
| CN111738318B (en) | Super-large image classification method based on graph neural network | |
| CN105825238B (en) | A kind of vision significance mesh object detection method | |
| CN109447169A (en) | The training method of image processing method and its model, device and electronic system | |
| CN106372648A (en) | Multi-feature-fusion-convolutional-neural-network-based plankton image classification method | |
| CN107392968B (en) | The image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure | |
| CN105138987B (en) | A kind of vehicle checking method based on converging channels feature and estimation | |
| CN107730515A (en) | Panoramic picture conspicuousness detection method with eye movement model is increased based on region | |
| CN103985130B (en) | A kind of saliency analysis method for complex texture image | |
| CN104408429A (en) | Method and device for extracting representative frame of video | |
| CN102968782A (en) | Automatic digging method for remarkable objects of color images | |
| CN109948593A (en) | Based on the MCNN people counting method for combining global density feature | |
| CN105868708A (en) | Image object identifying method and apparatus | |
| CN105869175A (en) | Image segmentation method and system | |
| CN102592132A (en) | System and method for classifying digital image data | |
| CN106295498A (en) | Remote sensing image target area detection apparatus and method | |
| CN105976376A (en) | High resolution SAR image target detection method based on part model | |
| Mahmood et al. | Automatic threshold selection for morphological attribute profiles | |
| Bora et al. | AERASCIS: An efficient and robust approach for satellite color image segmentation | |
| CN113920378B (en) | Bupleurum seed identification method based on attention mechanism | |
| CN110232317A (en) | Hyperspectral image classification method based on super-pixel segmentation and two phase classification strategy | |
| Wang et al. | A region-line primitive association framework for object-based remote sensing image analysis | |
| CN106228553A (en) | High-resolution remote sensing image shadow Detection apparatus and method | |
| CN104008374B (en) | Miner's detection method based on condition random field in a kind of mine image | |
| Khodadadzadeh et al. | A new framework for hyperspectral image classification using multiple spectral and spatial features |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230815 Address after: 410001 No. 002, Floor 5, Building B, No. 10, Zone 2, CSCEC Smart Industrial Park, No. 50, Jinjiang Road, Yuelu Street, Yuelu District, Changsha, Hunan Province Patentee after: Hunan Xinxin Xiangrong Intelligent Technology Co.,Ltd. Address before: 410082 Yuelu District Lushan South Road Lushan Gate, Changsha City, Hunan Province Patentee before: HUNAN University |
|
| TR01 | Transfer of patent right |