CN104268582B - The band selection method and device of a kind of high spectrum image - Google Patents
The band selection method and device of a kind of high spectrum image Download PDFInfo
- Publication number
- CN104268582B CN104268582B CN201410426002.7A CN201410426002A CN104268582B CN 104268582 B CN104268582 B CN 104268582B CN 201410426002 A CN201410426002 A CN 201410426002A CN 104268582 B CN104268582 B CN 104268582B
- Authority
- CN
- China
- Prior art keywords
- wave band
- distance
- optimum solution
- distance matrix
- formula
- 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
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
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24143—Distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB
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)
- Complex Calculations (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The band selection method and device of a kind of high spectrum image provided in an embodiment of the present invention, object function is constructed using the distance matrix of wave band to be selected, and determine the optimum solution of object function, because optimum solution is set selected from wave band to be selected, the maximum wave band of distance between any two, so, in the high spectrum image obtained using the wave band in optimum solution, because the distance between wave band is maximum, so, correlation is minimum between the data of adjacent band, therefore, it is possible to improve the nicety of grading of spectrum picture.
Description
Technical field
The present invention relates to the band selection method and device of remote sensing fields, more particularly to a kind of high spectrum image.
Background technology
High-spectrum remote sensing is a kind of common image of remote sensing fields, may be used as classification on a surface target, for example,
The high-spectrum remote sensing in a certain region is obtained, by certain method, the class label of each pixel in image is determined
(such as vegetation, exposed soil, road).
Often there is very strong correlation between the data of the adjacent band of high spectrum image, so, EO-1 hyperion can be reduced
The nicety of grading of pixel in image, for example, the pixel of vegetation should be categorized as, is divided into the class of road one.
The content of the invention
The embodiments of the invention provide a kind of band selection method of high spectrum image and device, it is therefore intended that solves because of phase
The problem of nicety of grading of pixel is low in high spectrum image caused by the strong correlation of adjacent wave segment data.
To achieve these goals, the embodiments of the invention provide following technical scheme:
The embodiments of the invention provide a kind of band selection method of high spectrum image, including:
The distance matrix of wave band to be selected is obtained, the element of the distance matrix is any two ripple in the wave band to be selected
The distance between section;
Object function and its constraints are constructed using the distance matrix;
Optimum solution of the object function under the constraints is determined, the optimum solution is from the ripple to be selected
Set selected in section, the maximum wave band of distance between any two, the optimum solution is used to obtain high spectrum image.
Alternatively, it is described to be included using distance matrix construction object function and its constraints:
Utilize Distance matrix D=[Dij]∈RL×L, constructionWherein,For
Object function, Dij=d (bi,bj),For constraints, its meaning is:Ensure in L xiIn there is the K to be just
1, i.e., there are K just in L wave band and be chosen, xiRepresent wave band biWhether it is chosen, if being chosen, xi=1, otherwise xi=
0。
Alternatively, the optimum solution for determining the object function under the constraints includes:
Initialize this optimal objective function value F0=Inf, optimal wave band collectionAnd selected wave band collection
By way of ant a selects length for K path, selected wave band collection a, the wave band collection a include K ripple
Section, wherein, a=1,2 ... A, A are the integer more than zero, and K is the integer more than zero;
According to A wave band collection, calculating target function value
If f<F0, then F is updated0=f, B0=B1, the B0For the optimum solution of the object function.
Alternatively, it is described by way of ant a selects length for K path, select any one ripple in wave band collection a
The method of section includes:
Calculate ant a transition probabilityWhereinτi=ρ τi+Δ
τi,τiInitial value be τi=1 (i=1,2 ..., L);
Calculate the distribution of ant a transition probability
Generate random number r, the distribution P of minimum transition probability of the selection more than riCorresponding subscript i is selected for this
The numbering i of wave band;
Update B1=B1∪ { i }, B2=B2{ i }, wherein, B2For wave band collection to be selected, its initial value is B2=1,2 ...,
L }, wherein 1,2 ..., L is the sequence number of wave band to be selected.
Alternatively, obtaining the distance matrix of wave band to be selected includes:
Using range formula, distance between wave band two-by-two is calculated in the wave band to be selected successively, and the range formula includes
Euclidean distance formula, COS distance formula, KL divergence Kullback-Leibler divergence formula or mutual information are public
Formula;
Using the distance between the wave band two-by-two, distance matrix is formed.
The embodiment of the present invention additionally provides a kind of waveband selection device of high spectrum image, including:
Distance matrix acquisition module, the distance matrix for obtaining wave band to be selected, the element of the distance matrix is described
The distance between any two wave band in wave band to be selected;
Function construction module, for constructing object function and its constraints using the distance matrix;
Optimum solution module, it is described optimal for determining optimum solution of the object function under the constraints
Dissolve as set selected from the wave band to be selected, the maximum wave band of distance between any two, the optimum solution is used to obtain
Take high spectrum image.
Alternatively, the function construction module is used to construct object function and its constraints bag using the distance matrix
Include:
The construction of function mold block is used for, and utilizes Distance matrix D=[Dij]∈RL×L, constructionWherein,For object function, Dij=d (bi,bj),
For constraints, its meaning is:Ensure in L xiIn have just during K is 1, i.e., L wave band
Just K are chosen, xiRepresent wave band biWhether it is chosen, if being chosen, xi=1, otherwise xi=0.
Alternatively, the optimum solution module is used to determine optimum solution of the object function under the constraints
Including:
Optimum solution module is specifically for initializing this optimal objective function value F0=Inf, optimal wave band collectionAnd selected wave band collection
By way of ant a selects length for K path, selected wave band collection a, the wave band collection a include K ripple
Section, wherein, a=1,2 ... A, A are the integer more than zero, and K is the integer more than zero;
According to A wave band collection, calculating target function value
If f<F0, then F is updated0=f, B0=B1, the B0For the optimum solution of the object function.
Alternatively, the optimum solution module is used for, and by way of ant a selects length for K path, selectes wave band
The method of any one wave band in collection a includes:
The optimum solution module is specifically for calculating ant a transition probabilityWhereinτi=ρ τi+Δτi,τiInitial value be τi=1 (i=1,2 ..., L);
Calculate the distribution of ant a transition probability
Generate random number r, the distribution P of minimum transition probability of the selection more than riCorresponding subscript i is selected for this
The numbering i of wave band;
Update B1=B1∪ { i }, B2=B2{ i }, wherein, B2For wave band collection to be selected, its initial value is B2=1,2 ...,
L }, wherein 1,2 ..., L is the sequence number of wave band to be selected.
Alternatively, the distance matrix acquisition module includes for obtaining the distance matrix of wave band to be selected:
The distance matrix acquisition module specifically for using range formula, calculating in the wave band to be selected two-by-two successively
Distance between wave band, the range formula includes Euclidean distance formula, COS distance formula, KL divergence Kullback-Leibler
Divergence formula or mutual information formula;Using the distance between the wave band two-by-two, distance matrix is formed.
The band selection method and device of a kind of high spectrum image provided in an embodiment of the present invention, using wave band to be selected away from
From matrix construction object function, and determine the optimum solution of object function because optimum solution be selected from wave band to be selected,
The set of the maximum wave band of distance between any two, so, in the high spectrum image obtained using the wave band in optimum solution, because
Distance between wave band is maximum, so, correlation is minimum between the data of adjacent band, therefore, it is possible to improve the classification of spectrum picture
Precision.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
The accompanying drawing to be used needed for having technology description is briefly described, it should be apparent that, drawings in the following description are only this
Some embodiments of invention, for those of ordinary skill in the art, on the premise of not paying creative work, can be with
Other accompanying drawings are obtained according to these accompanying drawings.
Fig. 1 is a kind of flow chart of the band selection method of high spectrum image disclosed in the embodiment of the present invention;
Fig. 2 is the flow chart of the band selection method of another high spectrum image disclosed in the embodiment of the present invention;
Fig. 3 is a kind of structural representation of the waveband selection device of high spectrum image disclosed in the embodiment of the present invention.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation is described, it is clear that described embodiment is only a part of embodiment of the invention, rather than whole embodiments.It is based on
Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under the premise of creative work is not made
Embodiment, belongs to the scope of protection of the invention.
A kind of band selection method of high spectrum image disclosed in the embodiment of the present invention, as shown in figure 1, including:
S101:The distance matrix of wave band to be selected is obtained, the element of the distance matrix is any in the wave band to be selected
The distance between two wave bands;
The one of the main reasons that selection is carried out to the wave band of high spectrum image is that it has stronger phase between wave band
Like property (correlation in other words), redundancy is thereby produced.One of main purpose of waveband selection be exactly find one it is similar
Property small band combination as far as possible.Distance be metric data similitude it is simple and in effective index, the present embodiment, can (but
It is not limited to) Euclidean distance, COS distance, KL divergences and mutual information are used as " distance " of similitude between measurement wave band.
S102:Object function and its constraints are constructed using the distance matrix;
S103:Optimum solution of the object function under the constraints is determined, the optimum solution is from described
Set selected in wave band to be selected, the maximum wave band of distance between any two, the optimum solution is used to obtain high spectrum image.
The purpose of the present embodiment be high spectrum image selection wave band, selected wave band in whole wave bands to be selected,
Distance between any two is maximum, in order that reducing the redundancy between the adjacent band data in high spectrum image, so as to height
When target in spectrum picture is classified, nicety of grading is improved.
In the prior art, the mode for carrying out waveband selection generally has two kinds:The first is by between wave band by some computings
Relative distance is thought in absolute terms, and as each wave band provides the evaluation of a quantification, so as to be selected to band ordering, and in this order
Band subset;It is for second to enter band subset by two farthest wave bands of the first selected distance of Greedy strategy, then gradually will be with
Wave band has been selected adds band subset, the enough quantity until reaching apart from farthest wave band.
The first the disadvantage is that, being easy to occur similar wave band evaluation result also similar feelings during quantitative evaluation
Condition, and then make it that similar wave band is simultaneously chosen, this original intention against waveband selection;The second way the disadvantage is that, first
Selection result below can be influenceed by being selected into the wave band of subset, say that such case is easily trapped into part from the angle of optimization problem
Optimal solution rather than globally optimal solution.
Method described in the present embodiment, using the distance between wave band as foundation, without carrying out quantitative evaluation to wave band,
Avoid the shortcoming of first way.In addition, the present invention constructs the object function of optimization problem using distance matrix, also can
Avoid local optimum.
The band selection method of another high spectrum image disclosed in the embodiment of the present invention, as shown in Fig. 2 including:
S201:Using range formula, distance between wave band two-by-two is calculated in the wave band to be selected successively;
The range formula can include Euclidean distance formula, COS distance formula, Kullback-Leibler
Divergence formula or mutual information formula.
Assuming that high spectrum image { the r comprising N number of pixel, L wave bandij}L×N, wherein rijRepresent j-th of pixel at i-th
Reflectivity (or spoke brightness) on wave band, rj=(r1j,r2j,…,rLj)TRepresent j-th of pixel, bi=(ri1,ri2,…,riN) table
ShowiIndividual wave band, wherein:
Euclidean distance formula is:d(bi,bj)=| | bi-bj||2;
COS distance formula is:
Kullback-Leibler divergence formula are:
Wherein, Xi,XjIt is two stochastic variables, pi(x),pj(x) X is represented respectivelyi,XjTake probability during x.Need explanation
It is that in the present embodiment, the probability distribution of wave band is obtained by 256 grades of grey level histograms.But it is due to KL divergence definition
Symmetry is unsatisfactory for, therefore can not be directly as the distance between two probability distribution.It is often necessary to use a symmetric
Version of the Kullback-Leibler divergence formula, are defined as follows:d(Xi,Xj)=dKL(Xi||Xj)+
dKL(Xj||Xi)。
Mutual information formula is:Wherein, p (xi,xj) it is random
Variable Xi,XjJoint probability, obtained by 256 grades of grey level histograms.But according to mutual informationI meaning, MI
Higher, the correlation of stochastic variable is higher, and this is all opposite with Euclidean distance, COS distance and KL divergence.In order to allow MI
It is consistent with the meaning of other distances, equation below can be used in the present embodiment
Wherein, H (Xi,Xj) it is Xi,XjCombination entropy, be defined as
S202:Using the distance between the wave band two-by-two, Distance matrix D=[D is formedij]∈RL×L, wherein Dij=d (bi,
bj);
S203:Utilize distance matrix, construction
Wherein,For object function, Dij=d (bi,bj),For constraints, its
Meaning is:Ensure in L xiIn there is the K to be to have K selected, x just in 1, i.e., L wave band justiRepresent wave band biWhether by
Selection, if being chosen, xi=1, otherwise xi=0.
Below with L=6, exemplified by K=3, table 1 represents Distance matrix D=[Dij]6×6, its top and the digital of left side represent xi
Value, i.e. x1=x3=x6=0, x2=x4=x5=1.
Table 1
In table 1, only xiAnd xjWhen being equal to 1, corresponding DijSummation can be just included in, otherwise, due to DijCoefficient above
For 0, corresponding DijSummation will not be included in, that is to say, that according to foregoing xiValue condition, the digital meeting only marked in upper table
The calculating of the row of formula 8 the 1st is participated in, object function is obtained.
From another perspective, equivalent to from D by xi=1 corresponding each row, column is extracted, and obtains a new square
Battle array (such as table 2), and all elements in matrix are summed.
Table 2
D22 | D24 | D25 |
D42 | D44 | D45 |
D52 | D54 | D55 |
S204:Loop initialization counter t=1;
S205:If t≤T, S206 is performed, otherwise, S218 is performed;
Wherein, T is maximum cycle.
S206:Ant counter a=1 is initialized, this optimal objective function value F is initialized0=Inf, optimal wave band collectionSelected wave band collectionAnd wave band collection B to be selected2={ 1,2 ..., L }, wherein 1,2 ..., L is wave band to be selected
Sequence number;
S207:If a≤A, S208 is performed, otherwise, S216 is performed;
Wherein, A is the integer more than zero.
S208:Wave band counter k=1 is initialized, if k≤K, S209 is performed, otherwise, S212 is performed;
Wherein, K is the integer more than zero.
S209:Calculate ant a transition probability:
S210:Calculate the distribution of ant a transition probability
S211:Select a wave band.Generate random number r, minimum P of the selection more than riCorresponding subscript i selects for this
Wave band numbering i, update B1=B1∪ { i }, B2=B2{ i }, k=k+1;
S212:Calculating target function value
S213:If f<F0, S214 is performed, S215 is otherwise performed;
S214:Update F0=f, B0=B1;
S215:A=a+1, performs S207;
S216:Fresh information elementτi=ρ τi+Δτi;
S217:T=t+1, performs step S205;
S218:B0For the optimum solution of the object function.
In the present embodiment, will obtain optimum solution the step of circulate T times, using the most the superior in T optimum solution as
Final optimum solution, it is therefore intended that obtain more accurate optimum solution.
In the present embodiment, obtained using ant algorithm in optimum solution, nature, " letter is referred to as by one kind between ant
The chemical substance of breath element (pheromones) " is mutually exchanged, so as to cooperate, and completes complicated task.Single ant exists
When search of food, discharge the pheromones on path by other ants and select the path of oneself, while oneself also discharges
Pheromones, help other ants to select path, and all ants tend to the path for selecting pheromone concentration high, thus shape
Into the system of a positive feedback --- the more routing information elements of the ant that passes through are more and more (more and more easily by other
Ant selects), in addition, the natural dissipation of pheromones strengthens the effect of this positive feedback again --- pass through the fewer road of ant
Footpath pheromones are fewer and feweri (being increasingly difficult to be selected by other ants).By the system of this positive feedback, ant can find nest
Shortest path between cave and food source.
The embodiment of the present invention using different distance measure wave band between similitude, and using ant algorithm automatically extract to
There is less similitude between the wave band of fixed number amount, the wave band selected.
With above method embodiment accordingly, the embodiment of the present invention additionally provides a kind of waveband selection of high spectrum image
Device, as shown in figure 3, including:
Distance matrix acquisition module 301, the distance matrix for obtaining wave band to be selected, the element of the distance matrix is institute
State the distance between any two wave band in wave band to be selected;
Function construction module 302, for constructing object function and its constraints using the distance matrix;
Optimum solution module 303, for determining optimum solution of the object function under the constraints, it is described most
Optimization solution is set selected from the wave band to be selected, the maximum wave band of distance between any two, and the optimum solution is used for
Obtain high spectrum image.
Alternatively, the function construction module is used to construct object function and its constraints bag using the distance matrix
Include:
The construction of function mold block is used for, and utilizes Distance matrix D=[Dij]∈RL×L, constructionWherein,For object function, wherein, Dij=d (bi,bj),
For constraints, its meaning is:Ensure in L xiIn have just during K is 1, i.e., L wave band
Just K are chosen, xiRepresent wave band biWhether it is chosen, if being chosen, xi=1, otherwise xi=0.
Alternatively, the optimum solution module is used to determine optimum solution of the object function under the constraints
Including:
Optimum solution module is specifically for initializing this optimal objective function value F0=Inf, optimal wave band collectionAnd selected wave band collection
By way of ant a selects length for K path, selected wave band collection a, the wave band collection a include K ripple
Section, wherein, a=1,2 ... A, A are the integer more than zero, and K is the integer more than zero;
According to A wave band collection, calculating target function value
If f<F0, then F is updated0=f, B0=B1, the B0For the optimum solution of the object function.
Alternatively, the optimum solution module is used for, and by way of ant a selects length for K path, selectes wave band
The method of any one wave band in collection a includes:
The optimum solution module is specifically for calculating ant a transition probabilityWhereinτi=ρ τi+Δτi,τiInitial value be τi=1 (i=1,2 ..., L);
Calculate the distribution of ant a transition probability
Generate random number r, the distribution P of minimum transition probability of the selection more than riCorresponding subscript i is selected for this
The numbering i of wave band;
Update B1=B1∪ { i }, B2=B2{ i }, wherein, B2For wave band collection to be selected, its initial value is B2=1,2 ...,
L }, wherein 1,2 ..., L is the sequence number of wave band to be selected.
Alternatively, the distance matrix acquisition module includes for obtaining the distance matrix of wave band to be selected:
The distance matrix acquisition module specifically for using range formula, calculating in the wave band to be selected two-by-two successively
Distance between wave band, the range formula includes Euclidean distance formula, COS distance formula, KL divergence Kullback-Leibler
Divergence formula or mutual information formula;Using the distance between the wave band two-by-two, distance matrix is formed.
Device described in the present embodiment, using the distance between wave band as foundation, without carrying out quantitative evaluation to wave band,
Avoid the shortcoming of first way.In addition, constructing the object function of optimization problem using distance matrix, office can be also avoided
Portion is optimal.
If the function described in present invention method is realized using in the form of SFU software functional unit and is used as independent production
Product are sold or in use, can be stored in a computing device read/write memory medium.Understood based on such, the present invention is real
The part for applying part that example contributes to prior art or the technical scheme can be embodied in the form of software product,
The software product is stored in a storage medium, including some instructions are make it that a computing device (can be personal meter
Calculation machine, server, mobile computing device or network equipment etc.) perform whole or the portion of each of the invention embodiment methods described
Step by step.And foregoing storage medium includes:USB flash disk, mobile hard disk, read-only storage (ROM, Read-Only Memory), with
Machine access memory (RAM, Random Access Memory), magnetic disc or CD etc. are various can be with Jie of store program codes
Matter.
The embodiment of each in this specification is described by the way of progressive, what each embodiment was stressed be with it is other
Between the difference of embodiment, each embodiment same or similar part mutually referring to.
The foregoing description of the disclosed embodiments, enables professional and technical personnel in the field to realize or using the present invention.
A variety of modifications to these embodiments will be apparent for those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, it is of the invention
The embodiments shown herein is not intended to be limited to, and is to fit to and principles disclosed herein and features of novelty phase one
The most wide scope caused.
Claims (8)
1. a kind of band selection method of high spectrum image, it is characterised in that including:
Obtain the distance matrix of wave band to be selected, the element of the distance matrix be any two wave band in the wave band to be selected it
Between distance;
Object function and its constraints are constructed using the distance matrix;
Optimum solution of the object function under the constraints is determined, the optimum solution is from the wave band to be selected
Set select, the maximum wave band of distance between any two, the optimum solution is used to obtain high spectrum image;
Wherein, it is described to be included using distance matrix construction object function and its constraints:
Utilize Distance matrix D=[Dij]∈RL×L, constructionWherein,For target letter
Number, Dij=d (bi,bj),For constraints, its meaning is:Ensure in L xiIn there is the K to be 1, i.e. L just
There are K just in individual wave band to be chosen, xiRepresent wave band biWhether it is chosen, if being chosen, xi=1, otherwise xi=0.
2. according to the method described in claim 1, it is characterised in that described to determine the object function under the constraints
Optimum solution include:
Initialize this optimal objective function value F0=Inf, optimal wave band collectionAnd selected wave band collection
By way of ant a selects length for K path, selected wave band collection a, the wave band collection a include K wave band, its
In, a=1,2 ... A, A are the integer more than zero, and K is the integer more than zero;
According to A wave band collection, calculating target function value
If f < F0, then F is updated0=f, B0=B1, the B0For the optimum solution of the object function.
3. method according to claim 2, it is characterised in that described to pass through side of the ant a selection length for K path
The method of any one wave band in formula, selected wave band collection a includes:
Calculate ant a transition probabilityWhereinτi=ρ τi+Δτi,
τiInitial value be τi=1 (i=1,2 ..., L);
Calculate the distribution of ant a transition probability
Generate random number r, the distribution P of minimum transition probability of the selection more than riCorresponding subscript i is this selected wave band
Numbering i;
Update B1=B1∪ { i }, B2=B2{ i }, wherein, B2For wave band collection to be selected, its initial value is B2={ 1,2 ..., L }, its
In 1,2 ..., L be wave band to be selected sequence number.
4. the method according to any one of claims 1 to 3, it is characterised in that obtaining the distance matrix of wave band to be selected includes:
Using range formula, distance between wave band two-by-two is calculated in the wave band to be selected successively, and the range formula includes Euclidean
Range formula, COS distance formula, KL divergence Kullback-Leibler divergence formula or mutual information formula;
Using the distance between the wave band two-by-two, distance matrix is formed.
5. a kind of waveband selection device of high spectrum image, it is characterised in that including:
Distance matrix acquisition module, the distance matrix for obtaining wave band to be selected, the element of the distance matrix is described to be selected
The distance between any two wave band in wave band;
Function construction module, for constructing object function and its constraints using the distance matrix;
Optimum solution module, for determining optimum solution of the object function under the constraints, the optimum solution
For set selected from the wave band to be selected, the maximum wave band of distance between any two, the optimum solution is used to obtain height
Spectrum picture;
Wherein, the function construction module is used to include using distance matrix construction object function and its constraints:
The function construction module is specifically for utilizing Distance matrix D=[Dij]∈RL×L, constructionIts
In,For object function, Dij=d (bi,bj),
For constraints, its meaning is:Ensure in L xiIn there is the K to be lucky in 1, i.e., L wave band just
There are K to be chosen, xiRepresent wave band biWhether it is chosen, if being chosen, xi=1, otherwise xi=0.
6. device according to claim 5, it is characterised in that the optimum solution module is used to determine the object function
Optimum solution under the constraints includes:
Optimum solution module is specifically for initializing this optimal objective function value F0=Inf, optimal wave band collectionAnd
Selected wave band collection
By way of ant a selects length for K path, selected wave band collection a, the wave band collection a include K wave band, its
In, a=1,2 ... A, A are the integer more than zero, and K is the integer more than zero;
According to A wave band collection, calculating target function value
If f < F0, then F is updated0=f, B0=B1, the B0For the optimum solution of the object function.
7. device according to claim 6, it is characterised in that the optimum solution module is used for, selects to grow by ant a
The mode in the path for K is spent, the method for any one wave band in selected wave band collection a includes:
The optimum solution module is specifically for calculating ant a transition probabilityWhereinτi=ρ τi+Δτi,τiInitial value be τi=1 (i=1,2 ..., L);
Calculate the distribution of ant a transition probability
Generate random number r, the distribution P of minimum transition probability of the selection more than riCorresponding subscript i is this selected wave band
Numbering i;
Update B1=B1∪ { i }, B2=B2{ i }, wherein, B2For wave band collection to be selected, its initial value is B2={ 1,2 ..., L }, its
In 1,2 ..., L be wave band to be selected sequence number.
8. the device according to any one of claim 5 to 7, it is characterised in that the distance matrix acquisition module is used to obtain
Taking the distance matrix of wave band to be selected includes:
The distance matrix acquisition module specifically for using range formula, calculating in the wave band to be selected wave band two-by-two successively
Between distance, the range formula include Euclidean distance formula, COS distance formula, KL divergence Kullback-Leibler
Divergence formula or mutual information formula;Using the distance between the wave band two-by-two, distance matrix is formed.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410426002.7A CN104268582B (en) | 2014-08-26 | 2014-08-26 | The band selection method and device of a kind of high spectrum image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410426002.7A CN104268582B (en) | 2014-08-26 | 2014-08-26 | The band selection method and device of a kind of high spectrum image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104268582A CN104268582A (en) | 2015-01-07 |
CN104268582B true CN104268582B (en) | 2017-07-14 |
Family
ID=52160102
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410426002.7A Active CN104268582B (en) | 2014-08-26 | 2014-08-26 | The band selection method and device of a kind of high spectrum image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104268582B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106408010A (en) * | 2016-09-09 | 2017-02-15 | 河南理工大学 | Matrix pattern waveband selecting method based on spectrum information divergence |
CN106529563B (en) * | 2016-09-19 | 2019-03-26 | 西安电子科技大学 | EO-1 hyperion band selection method based on the sparse Non-negative Matrix Factorization of digraph |
CN108154094B (en) * | 2017-12-14 | 2020-04-24 | 浙江工业大学 | Hyperspectral image unsupervised waveband selection method based on subinterval division |
CN108446599B (en) * | 2018-02-27 | 2021-11-05 | 首都师范大学 | Hyperspectral image band rapid selection method of p-value statistical modeling independence |
CN109191443B (en) * | 2018-08-28 | 2022-03-04 | 广东工业大学 | Hyperspectral image waveband selection method based on sequence information and waveband quality |
CN109871768B (en) * | 2019-01-18 | 2022-04-29 | 西北工业大学 | Hyperspectral optimal waveband selection method based on shared nearest neighbor |
CN110969125B (en) * | 2019-12-02 | 2022-10-25 | 哈尔滨工业大学 | Detection spectrum band optimization method for target identification task |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102289673A (en) * | 2011-06-22 | 2011-12-21 | 复旦大学 | Method for selecting hyperspectral remote sensing image bands based on partial least squares |
CN103150577A (en) * | 2013-03-26 | 2013-06-12 | 河海大学 | Method for selecting high spectrum remotely sensed image self-adaptive wave band based on particle swarm optimization |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8406469B2 (en) * | 2009-07-20 | 2013-03-26 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | System and method for progressive band selection for hyperspectral images |
-
2014
- 2014-08-26 CN CN201410426002.7A patent/CN104268582B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102289673A (en) * | 2011-06-22 | 2011-12-21 | 复旦大学 | Method for selecting hyperspectral remote sensing image bands based on partial least squares |
CN103150577A (en) * | 2013-03-26 | 2013-06-12 | 河海大学 | Method for selecting high spectrum remotely sensed image self-adaptive wave band based on particle swarm optimization |
Non-Patent Citations (3)
Title |
---|
Ant colony optimization-based supervised and unsupervised band selections for hyperspectral urban data classification;Jianwei Gao等;《Journal of Applied Remote Sensing》;20140813;全文 * |
一种新的高光谱遥感图像降维方法;刘春红等;《中国图象图形学报》;20050228;第10卷(第2期);全文 * |
高光谱遥感数据解译的最佳波段选择方法研究;刘建平等;《中国科学院研究生院学报》;19991031;第16卷(第2期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN104268582A (en) | 2015-01-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104268582B (en) | The band selection method and device of a kind of high spectrum image | |
Amini et al. | Object-based classification of hyperspectral data using Random Forest algorithm | |
Dormann et al. | A method for detecting modules in quantitative bipartite networks | |
Herold et al. | The use of remote sensing and landscape metrics to describe structures and changes in urban land uses | |
Tomppo et al. | A sampling design for a large area forest inventory: case Tanzania | |
Zhang et al. | Object-oriented method for urban vegetation mapping using IKONOS imagery | |
Bergseng et al. | Assessing forest inventory information obtained from different inventory approaches and remote sensing data sources | |
Surabuddin Mondal et al. | Modeling of spatio-temporal dynamics of land use and land cover in a part of Brahmaputra River basin using Geoinformatic techniques | |
Gao et al. | A novel unsupervised segmentation quality evaluation method for remote sensing images | |
Gao et al. | Object-based classification with features extracted by a semi-automatic feature extraction algorithm–SEaTH | |
Meraj et al. | Assessing the yield of wheat using satellite remote sensing-based machine learning algorithms and simulation modeling | |
Xu et al. | Urban climate zone classification using convolutional neural network and ground-level images | |
Kindu et al. | Monitoring of urban growth patterns in rapidly growing Bahir Dar City of Northwest Ethiopia with 30 year Landsat imagery record | |
Pulugurtha et al. | Modeling AADT on local functionally classified roads using land use, road density, and nearest nonlocal road data | |
Karydas | Optimization of multi-scale segmentation of satellite imagery using fractal geometry | |
Zhou et al. | A similarity-based community detection method with multiple prototype representation | |
Chen et al. | A subpixel mapping algorithm combining pixel-level and subpixel-level spatial dependences with binary integer programming | |
Helen et al. | Land use change, carbon stocks and tree species diversity in green spaces of a secondary city in Myanmar, Pyin Oo Lwin | |
Couto et al. | Integrating environmental, geographical and social data to assess sustainability in hydrographic basins: The ESI approach | |
Kazakeviciute-Januskeviciene et al. | Assessment of the segmentation of RGB remote sensing images: a subjective approach | |
Strimas‐Mackey et al. | Reserve design to optimize the long‐term persistence of multiple species | |
Powell et al. | Mapping wetland types in semiarid floodplains: a statistical learning approach | |
Malhi et al. | An integrated spatiotemporal pattern analysis model to assess and predict the degradation of protected forest areas | |
Machicao et al. | A deep-learning method for the prediction of socio-economic indicators from street-view imagery using a case study from Brazil | |
Lati et al. | Three-dimensional image-based modelling of linear features for plant biomass estimation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |