CN109784286A - Based on multiple target mould because the remote sensing image sky of optimization algorithm composes non-supervised classification - Google Patents
Based on multiple target mould because the remote sensing image sky of optimization algorithm composes non-supervised classification Download PDFInfo
- Publication number
- CN109784286A CN109784286A CN201910056340.9A CN201910056340A CN109784286A CN 109784286 A CN109784286 A CN 109784286A CN 201910056340 A CN201910056340 A CN 201910056340A CN 109784286 A CN109784286 A CN 109784286A
- Authority
- CN
- China
- Prior art keywords
- individual
- population
- class
- value
- remote sensing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of based on multiple target mould because the remote sensing image sky of optimization algorithm composes non-supervised classification, comprising: automatically determines remote sensing image class number using adaptive differential evolution algorithmic；Construct spatial information item, the empty spectrum joint objective function of joint Jm building；Using multiple target mould because optimization algorithm increases the optimization to remote sensing image unsupervised classification result.The present invention improves the automaticity and intelligence of remote sensing image unsupervised classification algorithm, without manually pre-entering the class number of priori, and the comprehensive global and local search capability for using mould because of optimization algorithm, so that the precision of unsupervised classification result is improved.
Description
Technical field
It is the present invention is based on remote sensing images technical treatment field, in particular to a kind of distant because of optimization algorithm based on multiple target mould
Feel image sky spectrum non-supervised classification, system and medium.
Background technique
Classification in Remote Sensing Image plays vital effect in land use mapping, the remote sensing applications such as precision agriculture.Remote sensing
Image unsupervised classification is different from remote sensing image supervised classification, it does not need any priori sample information, and only by digging
The sort operation that the information such as the structure of image itself carry out image is dug, the selection work for obtaining a large amount of high-precision samples is reduced,
To extend the application power of remote sensing images.On this basis, remote sensing image unsupervised classification causes many researchers
Concern.
At the same time, due to the particularity and complexity of remote sensing image structure, remote sensing image unsupervised classification is one difficult
Huge task, traditional remote sensing image non-supervised classification are primarily present disadvantage:
(1) it needs manually to pre-enter the number of image classification, rather than automatically determines, lack automaticity；
(2) the objective function optimization mode in traditional non-supervised classification belongs to gradient descent method, in higher-dimension and complexity
Remote sensing image data space be easily trapped into local extremum, i.e., shortage global optimization ability；
(3) spatial structural form of image is not fully considered, due to the influence of noise and mixed pixel, is easy to cause
Unsupervised classification result is very poor.
Summary of the invention
The present invention is directed to the shortcomings that traditional non-supervised classification, propose it is a kind of based on multiple target mould because of optimization algorithm
Remote sensing image sky composes non-supervised classification, comprising the following steps:
Step 1: the multi-spectrum remote sensing image that input is to be sorted；
Step 2: automatically determine layer, using being automatically determined based on adaptive differential evolution algorithm and export remote sensing image
Best class number；
Step 3: the class number K* obtained in input step two, using adaptive multiple target mould because optimization algorithm is to sky
It composes combined objective function Jm_S and function XB and carries out multiple-objection optimization, obtain optimal class center；
Step 4: carrying out non-supervisory point to the remote sensing image of input using optimal class center obtained in step 3
Class, final output unsupervised classification result.
Preferably, above-mentioned steps two mainly include the operation such as individual and initialization of population, intersection, variation and selection.
Preferably, above-mentioned steps two are accomplished by the following way:
Step 2.1, the individual UVR exposure and initialization of population operation in differential evolution algorithm are carried out, class center is random
Initialization is encoded into individual Ii={ Ii,1,Ii,2,...,Ii,j,...,Ii,d, wherein d=C at randommax×nb, CmaxIt is optional
Maximum class number, nbFor image wave band number；Including multiple populations, there is NP individual in each population, in population
Class number corresponding to body is consistent, but class number K different representated by the individual between population；
Step 2.2, the corresponding objective function Jm value of each individual in each population is calculated separately；
Formula are as follows:Wherein K represents class number, and N represents image pixel number, xjGeneration
J-th of pixel, U on table imageiRepresent i-th of class center, μijRepresent the person in servitude between j-th of pixel and i-th of class center
Category degree, m are fuzzy weighted values index.
Step 2.3, the variation between individual is carried out in each population respectivelyWherein F is scaling
The factor,For three individuals randomly selected from population, r1、r2And r3Three between 1~NP it is unequal with
Machine integer；
Crossover operation is carried out againCR is crossover probability, jrandFor
A random integers between 0~d, finally generate new offspring individual Qi={ Qi,1,Qi,2,...,Qi,j,...,Qi,d}；
In addition, F and CR is very important two parameters in DE algorithm, it is also encoded into individual, carried out adaptive
It evolves, it is as follows to update operation:
Step 2.4, the number of individuals after intersecting and make a variation in each population is 2 × NP, then calculates separately respective target
Function Jm value is ranked up Jm value corresponding to individual in each population and compares, and NP minimum Jm value is corresponding before choosing
Individual is retained；
It is as follows to compare formula:
Step 2.5, the selection for representing optimum solution in the population of different classes of number is respectively compared NP individual, selects one
A individual for possessing minimum Jm value
Step 2.6, XB functional value corresponding to individual selected in step 2.5 is calculated:Wherein most parameter is similar with Jm, by the way that Sep (U) is added, makes
It obtains between class distance to maximize, UkFor the class center of kth class；Individual wherein with minimum XB value is selected againExport the class number corresponding to it；
Step 2.7, step 2.3~2.6 are repeated, until the class number of output is not changing, output is final to be determined
Image class number.
Preferably, above-mentioned steps three specifically include:
Step 3.1, the empty spectrum joint objective function Jm_S of building；
Step 3.2, individual and initialization of population randomly choose K* pixel coder into individual G from imagei={ Gi,1,
Gi,2,...,Gi,j,...,Gi,d, in addition, also F and CR are encoded into individual.The NP bodily form is initialized into one kind simultaneously
Group；
Step 3.3, global search carries out individual in multiple target space using adaptive differential evolution algorithm global
Search, then carry out individual sequence and selection；
Step 3.4, adaptive local is searched for, and randomly chooses some individuals after step 3.2, is carried out Gauss and is locally searched
Rope, then carry out individual sequence and selection；
Step 3.5, step 3.2~3.4 are repeated until reaching maximum number of iterations, exports the individual of final choice and mentions
Take the class center corresponding to it.
Preferably, above-mentioned steps 3.1 specifically include:
Step 3.1.1 obtains mean value image by calculating surrounding neighbors pixel to the average influence of center pixel；Wherein
Edge pixel is not involved in calculating；
Step 3.1.2 obtains spatial information itemWherein α is weight parameter；Again in Jm
Empty spectrum joint objective function is constructed on the basis of objective function:
Preferably, above-mentioned steps 3.3 specifically include:
Step 3.3.1: determining two objective functions XB and Jm_S that proposed method is optimized, this step carries out adaptive
Variation and intersection, concrete operations are as described in step 2.3；
Step 3.3.2, the Pareto carried out between individual are dominant sequence, and the comparison of multiple target is different from single goal, when expiring
FootIndividual xAIt is dominant in individual xB；
It based on this, carries out Pareto and is dominant sequence, more all individuals do not account for the individual marks of other any individual dominations
Be 1, remaining individual is so compared again, be labeled as 2, and so on have the label of oneself until all individuals；
Step 3.3.3 carries out crowding distance sequence, in order to keep the stabilization of individual amount in population, it would be desirable to plant
Select NP more excellent individuals in group, selected first labeled as 1 individual, but be generally designated as 1 individual amount be more than or
Person just carries out choosing solution in the way of crowding distance sequence at this time less than NP；
Calculate the distance between i-th of individual and two neighboring individualWherein fXB max、fXB min、fJm_S maxAnd fJm_S minPoint
Maximum and minimum value of the respective objects function in current iteration number is not represented；
The individual that finally selection possesses larger crowding distance value carries out next step operation.
Preferably, above-mentioned steps 3.4 specifically include:
Step 3.4.1 selects Gauss local search Gi'=Normal (Gi,δ2) be used as mould because of local search step in algorithm
Suddenly, wherein δ is standard deviation, and be taken in of each dimension for acting on individual class center part meets condition rand [0,1]
When > FLS=0.5；
The standard deviation δ of four different gradients is arranged in step 3.4.21,δ2,δ3,δ4, then stored using Pro_GLS_ δ every
The probability that a gradient standard deviation is selected, initialization are all 0.25；
Step 3.4.3 carries out Gauss local search and obtains new individual to be judged, calculates two target letters corresponding to it
Then numerical value is compared using the thought that Pareto is dominant, newly generated individual can be divided into the individual of the individual of receiving, refusal
With doubtful individual；
Step 3.4.4 calculates the score of each gradient standard deviation
Wherein f (XB) and f (Jm_S) is that the individual does not carry out the functional value before local search,WithTo carry out the functional value after local search；
Later, the average value of each gradient standard deviation is calculatedWherein numiIt represents current
The number that the standard deviation of gradient is used；
Step 3.4.5, update Pro_GLS_ δ, the probability updating that the standard deviation of each gradient is selected forBeing provided with the probability that certain algebra β allows the standard deviation of each gradient to be selected is always
0.25, carry out the update of select probability again later；
Step 3.4.6 is dominant sequence using Pareto and crowding distance sequence carries out population recruitment and it is made to maintain NP
Body.
It is a kind of based on multiple target mould because optimization algorithm remote sensing image sky compose unsupervised classification system, comprising:
Image input unit, for inputting multi-spectrum remote sensing image to be sorted；
Unit is automatically determined, automatically determined for utilization based on adaptive differential evolution algorithm and exports remote sensing image most
Good class number；
Objective optimization unit utilizes adaptive multiple target mould for inputting the class number K* for automatically determining unit acquisition
Because optimization algorithm carries out multiple-objection optimization to sky spectrum combined objective function Jm_S and function XB, optimal class center is obtained；
Taxon, for utilizing the obtained optimal class center of objective optimization unit, to the remote sensing image of input into
Row unsupervised classification, final output unsupervised classification result.
System as described above,
The unit that automatically determines is accomplished by the following way:
The individual UVR exposure and initialization of population operation in differential evolution algorithm are carried out, class center random initializtion is compiled
Code is into individual Ii={ Ii,1,Ii,2,...,Ii,j,...,Ii,d, wherein d=C at randommax×nb, CmaxFor selectable maximum classification
Number, nbFor image wave band number；Including multiple populations, there is NP individual in each population, in population corresponding to individual
Class number is consistent, but class number K different representated by the individual between population；
Calculate separately the corresponding objective function Jm value of each individual in each population；
Formula are as follows:Wherein K represents class number, and N represents image pixel number, xjGeneration
J-th of pixel, U on table imageiRepresent i-th of class center, μijRepresent the person in servitude between j-th of pixel and i-th of class center
Category degree, m are fuzzy weighted values index；
The variation between individual is carried out in each population respectivelyWherein F is zoom factor,For three individuals randomly selected from population, r1、r2And r3Three between 1~NP are unequal random whole
Number；
Crossover operation is carried out againCR is crossover probability, jrandFor
A random integers between 0~d, finally generate new offspring individual Qi={ Qi,1,Qi,2,...,Qi,j,...,Qi,d}；
In addition, F and CR is very important two parameters in DE algorithm, it is also encoded into individual, carried out adaptive
It evolves, it is as follows to update operation:
Number of individuals after intersecting and making a variation in each population is 2 × NP, then calculates separately respective objective function Jm value,
Jm value corresponding to individual is ranked up in each population and is compared, the corresponding individual of NP minimum Jm value carries out before choosing
Retain；
It is as follows to compare formula:
The selection for representing optimum solution in the population of different classes of number is respectively compared NP individual, selects one and possess most
The individual of small Jm value
Calculate XB functional value corresponding to selected individual:Its
Middle major part parameter is similar with Jm, by the way that Sep (U) is added, so that between class distance maximizes, UkFor the class center of kth class；
Individual wherein with minimum XB value is selected againOutput
Class number corresponding to it；
It repeats the above steps, until the image classification number that the class number of output is not changing, and output finally determines
Mesh；
The objective optimization unit is accomplished by the following way:
The empty spectrum joint objective function Jm_S of building, is averaged to center pixel especially by surrounding neighbors pixel is calculated
It influences, obtains mean value image；Wherein edge pixel is not involved in calculating；Obtain spatial information item
Wherein α is weight parameter；Empty spectrum joint objective function is constructed on the basis of Jm objective function again:
Individual and initialization of population randomly choose K* pixel coder into individual G from imagei={ Gi,1,Gi,2,...,
Gi,j,...,Gi,d, in addition, also F and CR are encoded into individual.The NP bodily form is initialized into a population simultaneously；
Global search makes individual carry out global search in multiple target space using adaptive differential evolution algorithm, then
Carry out individual sequence and selection；It specifically includes:
Determine two objective functions XB and Jm_S that proposed method is optimized, this step carries out adaptive variation and friendship
Fork；
The Pareto carried out between individual is dominant sequence, and the comparison of multiple target is different from single goal, works as satisfactionIndividual xAIt is dominant in individual xB；
It based on this, carries out Pareto and is dominant sequence, more all individuals do not account for the individual marks of other any individual dominations
Be 1, remaining individual is so compared again, be labeled as 2, and so on have the label of oneself until all individuals；
Crowding distance sequence is carried out, in order to keep the stabilization of individual amount in population, it would be desirable to select NP in population
A more excellent individual is selected first labeled as 1 individual, but is generally designated as 1 individual amount more or less than NP,
Choosing solution just is carried out in the way of crowding distance sequence at this time；
Calculate the distance between i-th of individual and two neighboring individual
Wherein fXB max、fXB min、 fJm_S max
And fJm_S minRespectively represent maximum and minimum value of the respective objects function in current iteration number；
The individual that finally selection possesses larger crowding distance value carries out next step operation.
Adaptive local search, some individuals of the random selection after above-mentioned steps, progress Gauss local search, then into
The sequence of row individual and selection；It specifically includes:
It specifically includes:
Select Gauss local search Gi'=Normal (Gi,δ2) be used as mould because of local search step in algorithm, wherein δ is mark
Quasi- poor, each dimension for acting on individual class center part is fooled when meeting condition rand [0,1] > FLS=0.5；
The standard deviation δ of four different gradients is set1,δ2,δ3,δ4, then each gradient standard is stored using Pro_GLS_ δ
The probability that difference is selected, initialization are all 0.25；
It carries out Gauss local search and obtains new individual to be judged, calculate two target function values corresponding to it, then
Be compared using the thought that Pareto is dominant, newly generated individual can be divided into the individual of receiving, refusal individual and fail really
Fixed individual；
Calculate the score of each gradient standard deviation
Wherein f (XB) and f (Jm_S) is that the individual does not carry out the functional value before local search,WithTo carry out the functional value after local search；
Later, the average value of each gradient standard deviation is calculatedWherein numiIt represents current
The number that the standard deviation of gradient is used；
Update Pro_GLS_ δ, the probability updating that the standard deviation of each gradient is selected for
Being provided with the probability that certain algebra β allows the standard deviation of each gradient to be selected is always 0.25, carries out select probability again later
Update；
Be dominant sequence using Pareto and crowding distance sequence carry out population recruitment make its maintain NP it is individual；
It is right that step above-mentioned steps are repeated until reaching maximum number of iterations, export the individual of final choice and extract its institute
The class center answered.
A kind of computer readable storage medium, is stored thereon with computer program, realization when which is executed by processor
The step of above method.
Compared with prior art, present invention has the advantage that
1, image unsupervised classification precision is high, accidentally divides phenomenon less, can be used for remote sensing images analysis；
2, without manually pre-entering remote sensing image class number, the class number of image can directly be automatically determined；
3, spatial information item is merged, spatial structural form generally existing in remote sensing image has been fully considered, by pixel
Between space relationship shown.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, will make below to required in the embodiment of the present invention
Attached drawing is briefly described, for those of ordinary skill in the art, without creative efforts, also
Other drawings may be obtained according to these drawings without any creative labor.
Fig. 1 shows the individual UVR exposure form figure automatically determined in layer in population；
Fig. 2 shows the individual UVR exposure form figures in population in automatic unsupervised classification layer；
Fig. 3 shows population recruitment policy map, and Fig. 3 (a) is that 2 × NP individual schematic diagram, Fig. 3 (b) are in current population
Pareto is dominant the schematic diagram that sorts, and Fig. 3 (c) is crowding distance ordering chart；
Fig. 4 shows the individual figure of newly generated three types after local search；
Fig. 5 shows the result figure of image unsupervised classification；
Fig. 6 shows flow diagram of the invention.
Specific embodiment
The feature and exemplary embodiment of various aspects of the invention is described more fully below, in order to make mesh of the invention
, technical solution and advantage be more clearly understood, with reference to the accompanying drawings and embodiments, the present invention is further retouched in detail
It states.It should be understood that specific embodiment described herein is only configured to explain the present invention, it is not configured as limiting the present invention.
To those skilled in the art, the present invention can be real in the case where not needing some details in these details
It applies.Below the description of embodiment is used for the purpose of better understanding the present invention to provide by showing example of the invention.
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence " including ... ", it is not excluded that including
There is also other identical elements in the process, method, article or equipment of the element.
The present embodiment provides a kind of based on multiple target mould because the remote sensing image sky of optimization algorithm composes non-supervised classification, point
It is carried out for two steps:
Step 1: the multi-spectrum remote sensing image that input is to be sorted；
Step 2: automatically determine layer, using being automatically determined based on adaptive differential evolution algorithm and export remote sensing image
Best class number；
Step 3: the class number K* obtained in input step two, using adaptive multiple target mould because optimization algorithm is to sky
It composes combined objective function Jm_S and function XB and carries out multiple-objection optimization, obtain optimal class center；
Step 4: carrying out non-supervisory point to the remote sensing image of input using optimal class center obtained in step 3
Class, final output unsupervised classification result.
In one embodiment, step 2 mainly includes the operation such as individual and initialization of population, intersection, variation and selection.
In one embodiment, step 2 is accomplished by the following way:
Step 2.1, the individual UVR exposure and initialization of population operation in differential evolution algorithm are carried out, class center is random
Initialization is encoded into individual Ii={ Ii,1,Ii,2,...,Ii,j,...,Ii,d, wherein d=C at randommax×nb, CmaxIt is optional
Maximum class number, nbIt is specific as shown in Fig. 1 for image wave band number；Including multiple populations, in each population
There is NP individual, class number corresponding to individual is consistent in population, but class different representated by the individual between population
Other number K；
Step 2.2, the corresponding objective function Jm value of each individual in each population is calculated separately；
Formula are as follows:Wherein K represents class number, and N represents image pixel number, xjGeneration
J-th of pixel, U on table imageiRepresent i-th of class center, μijRepresent the person in servitude between j-th of pixel and i-th of class center
Category degree, m are fuzzy weighted values index.
Step 2.3, the variation between individual is carried out in each population respectivelyWherein F is scaling
The factor,For three individuals randomly selected from population, r1、r2And r3Three between 1~NP are unequal
Random integers；
Crossover operation is carried out againCR is crossover probability, jrandFor
A random integers between 0~d, finally generate new offspring individual Qi={ Qi,1,Qi,2,...,Qi,j,...,Qi,d}；
In addition, F and CR is very important two parameters in DE algorithm, it is also encoded into individual, carried out adaptive
It evolves, it is as follows to update operation:
Step 2.4, the number of individuals after intersecting and make a variation in each population is 2 × NP, then calculates separately respective target
Function Jm value is ranked up Jm value corresponding to individual in each population and compares, and NP minimum Jm value is corresponding before choosing
Individual is retained；
It is as follows to compare formula:
Step 2.5, the selection for representing optimum solution in the population of different classes of number is respectively compared NP individual, selects one
A individual for possessing minimum Jm value
Step 2.6, XB functional value corresponding to individual selected in step 2.5 is calculated:Wherein most parameter is similar with Jm, by the way that Sep (U) is added, makes
It obtains between class distance to maximize, UkFor the class center of kth class；Individual wherein with minimum XB value is selected againExport the class number corresponding to it；
Step 2.7, step 2.3~2.6 are repeated, until the class number of output is not changing, output is final to be determined
Image class number.
In one embodiment, step 3 specifically includes:
Step 3.1, the empty spectrum joint objective function Jm_S of building；
Step 3.2, individual and initialization of population randomly choose K* pixel coder into individual G from imagei={ Gi,1,
Gi,2,...,Gi,j,...,Gi,d, in addition, also F and CR are encoded into individual, as shown in Fig. 2.NP are initialized simultaneously
Body forms a population；
Step 3.3, global search carries out individual in multiple target space using adaptive differential evolution algorithm global
Search, then carry out individual sequence and selection；
Step 3.4, adaptive local is searched for, and randomly chooses some individuals after step 3.2, is carried out Gauss and is locally searched
Rope, then carry out individual sequence and selection；
Step 3.5, step 3.2~3.4 are repeated until reaching maximum number of iterations, exports the individual of final choice and mentions
Take the class center corresponding to it.
Step 3.1 specifically includes:
Step 3.1.1 obtains mean value image by calculating surrounding neighbors pixel to the average influence of center pixel；Wherein
Edge pixel is not involved in calculating；
Step 3.1.2 obtains spatial information itemWherein α is weight parameter；Again in Jm
Empty spectrum joint objective function is constructed on the basis of objective function:
Step 3.3 specifically includes:
Step 3.3.1: determining two objective functions XB and Jm_S that proposed method is optimized, this step carries out adaptive
Variation and intersection, concrete operations are as described in step 2.3；
Step 3.3.2, the Pareto carried out between individual are dominant sequence, and the comparison of multiple target is different from single goal, when expiring
FootIndividual xAIt is dominant in individual xB；
It based on this, carries out Pareto and is dominant sequence, more all individuals do not account for the individual marks of other any individual dominations
Be 1, remaining individual is so compared again, be labeled as 2, and so on have the label of oneself until all individuals, such as attached drawing 3
It is shown；
Step 3.3.3 carries out crowding distance sequence, in order to keep the stabilization of individual amount in population, it would be desirable to plant
Select NP more excellent individuals in group, selected first labeled as 1 individual, but be generally designated as 1 individual amount be more than or
Person just carries out choosing solution in the way of crowding distance sequence at this time, as shown in Fig. 3 less than NP；
Calculate the distance between i-th of individual and two neighboring individualWherein fXB max、fXB min、fJm_S maxAnd fJm_S minPoint
Maximum and minimum value of the respective objects function in current iteration number is not represented；
The individual that finally selection possesses larger crowding distance value carries out next step operation.
Step 3.4 specifically includes:
Step 3.4.1 selects Gauss local search Gi'=Normal (Gi,δ2) be used as mould because of local search step in algorithm
Suddenly, wherein δ is standard deviation, and be taken in of each dimension for acting on individual class center part meets condition rand [0,1]
When > FLS=0.5；
The standard deviation δ of four different gradients is arranged in step 3.4.21,δ2,δ3,δ4, then stored using Pro_GLS_ δ every
The probability that a gradient standard deviation is selected, initialization are all 0.25；
Step 3.4.3 carries out Gauss local search and obtains new individual to be judged, calculates two target letters corresponding to it
Then numerical value is compared using the thought that Pareto is dominant, newly generated individual can be divided into the individual of the individual of receiving, refusal
With doubtful individual, as shown in Fig. 4；
Step 3.4.4 calculates the score of each gradient standard deviation
Wherein f (XB) and f (Jm_S) is that the individual does not carry out the functional value before local search,With
To carry out the functional value after local search；
Later, the average value of each gradient standard deviation is calculatedWherein numiIt represents current
The number that the standard deviation of gradient is used；
Step 3.4.5, update Pro_GLS_ δ, the probability updating that the standard deviation of each gradient is selected forBeing provided with the probability that certain algebra β allows the standard deviation of each gradient to be selected is always
0.25, carry out the update of select probability again later；
Step 3.4.6 is dominant sequence using Pareto and crowding distance sequence carries out population recruitment and it is made to maintain NP
Body.
In addition, additionally providing a kind of remote sensing image sky spectrum unsupervised classification system based on multiple target mould because of optimization algorithm
Embodiment, comprising:
Image input unit, for inputting multi-spectrum remote sensing image to be sorted；
Unit is automatically determined, automatically determined for utilization based on adaptive differential evolution algorithm and exports remote sensing image most
Good class number；
Objective optimization unit utilizes adaptive multiple target mould for inputting the class number K* for automatically determining unit acquisition
Because optimization algorithm carries out multiple-objection optimization to sky spectrum combined objective function Jm_S and function XB, optimal class center is obtained；
Taxon, for utilizing the obtained optimal class center of objective optimization unit, to the remote sensing image of input into
Row unsupervised classification, final output unsupervised classification is as a result, as shown in Fig. 5.
In one embodiment, unit is automatically determined to be accomplished by the following way:
The individual UVR exposure and initialization of population operation in differential evolution algorithm are carried out, class center random initializtion is compiled
Code is into individual Ii={ Ii,1,Ii,2,...,Ii,j,...,Ii,d, wherein d=C at randommax×nb, CmaxFor selectable maximum classification
Number, nbFor image wave band number；Including multiple populations, there is NP individual in each population, in population corresponding to individual
Class number is consistent, but class number K different representated by the individual between population；
Calculate separately the corresponding objective function Jm value of each individual in each population；
Formula are as follows:Wherein K represents class number, and N represents image pixel number, xjGeneration
J-th of pixel, U on table imageiRepresent i-th of class center, μijRepresent the person in servitude between j-th of pixel and i-th of class center
Category degree, m are fuzzy weighted values index；
The variation between individual is carried out in each population respectivelyWherein F is zoom factor,For three individuals randomly selected from population, r1、r2And r3Three between 1~NP are unequal random whole
Number；
Crossover operation is carried out againCR is crossover probability, jrandFor
A random integers between 0~d, finally generate new offspring individual Qi={ Qi,1,Qi,2,...,Qi,j,...,Qi,d}；
In addition, F and CR is very important two parameters in DE algorithm, it is also encoded into individual, carried out adaptive
It evolves, it is as follows to update operation:
Number of individuals after intersecting and making a variation in each population is 2 × NP, then calculates separately respective objective function Jm value,
Jm value corresponding to individual is ranked up in each population and is compared, the corresponding individual of NP minimum Jm value carries out before choosing
Retain；
It is as follows to compare formula:
The selection for representing optimum solution in the population of different classes of number is respectively compared NP individual, selects one and possess most
The individual of small Jm value
Calculate XB functional value corresponding to selected individual:
Wherein most parameter is similar with Jm, by the way that Sep is added
(U), so that between class distance maximizes, UkFor the class center of kth class；Individual wherein with minimum XB value is selected againExport the class number corresponding to it；
It repeats the above steps, until the image classification number that the class number of output is not changing, and output finally determines
Mesh；
In one embodiment, objective optimization unit is accomplished by the following way:
The empty spectrum joint objective function Jm_S of building, is averaged to center pixel especially by surrounding neighbors pixel is calculated
It influences, obtains mean value image；Wherein edge pixel is not involved in calculating；Obtain spatial information item
Wherein α is weight parameter；Empty spectrum joint objective function is constructed on the basis of Jm objective function again:
Individual and initialization of population randomly choose K* pixel coder into individual G from imagei={ Gi,1,Gi,2,...,
Gi,j,...,Gi,d, in addition, also F and CR are encoded into individual.The NP bodily form is initialized into a population simultaneously；
Global search makes individual carry out global search in multiple target space using adaptive differential evolution algorithm, then
Carry out individual sequence and selection；It specifically includes:
Determine two objective functions XB and Jm_S that proposed method is optimized, this step carries out adaptive variation and friendship
Fork；
The Pareto carried out between individual is dominant sequence, and the comparison of multiple target is different from single goal, works as satisfactionIndividual xAIt is dominant in individual xB；
It based on this, carries out Pareto and is dominant sequence, more all individuals do not account for the individual marks of other any individual dominations
Be 1, remaining individual is so compared again, be labeled as 2, and so on have the label of oneself until all individuals；
Crowding distance sequence is carried out, in order to keep the stabilization of individual amount in population, it would be desirable to select NP in population
A more excellent individual is selected first labeled as 1 individual, but is generally designated as 1 individual amount more or less than NP
It is a, choosing solution just is carried out in the way of crowding distance sequence at this time；
Calculate the distance between i-th of individual and two neighboring individual
Wherein fXB max、fXB min、 fJm_S max
And fJm_S minRespectively represent maximum and minimum value of the respective objects function in current iteration number；
The individual that finally selection possesses larger crowding distance value carries out next step operation.
Adaptive local search, some individuals of the random selection after above-mentioned steps, progress Gauss local search, then into
The sequence of row individual and selection；It specifically includes:
It specifically includes:
Select Gauss local search Gi'=Normal (Gi,δ2) be used as mould because of local search step in algorithm, wherein δ is mark
Quasi- poor, each dimension for acting on individual class center part is fooled when meeting condition rand [0,1] > FLS=0.5；
The standard deviation δ of four different gradients is set1,δ2,δ3,δ4, then each gradient standard is stored using Pro_GLS_ δ
The probability that difference is selected, initialization are all 0.25；
It carries out Gauss local search and obtains new individual to be judged, calculate two target function values corresponding to it, then
Be compared using the thought that Pareto is dominant, newly generated individual can be divided into the individual of receiving, refusal individual and fail really
Fixed individual；
Calculate the score of each gradient standard deviation
Wherein f (XB) and f (Jm_S) is that the individual does not carry out the functional value before local search,With
To carry out the functional value after local search；
Later, the average value of each gradient standard deviation is calculatedWherein numiIt represents current
The number that the standard deviation of gradient is used；
Update Pro_GLS_ δ, the probability updating that the standard deviation of each gradient is selected for
Being provided with the probability that certain algebra β allows the standard deviation of each gradient to be selected is always 0.25, carries out select probability again later
Update；
Be dominant sequence using Pareto and crowding distance sequence carry out population recruitment make its maintain NP it is individual；
It is right that step above-mentioned steps are repeated until reaching maximum number of iterations, export the individual of final choice and extract its institute
The class center answered.
In addition, additionally providing a kind of computer readable storage medium, it is stored thereon with computer program, the program is processed
The step of above method is realized when device executes.
It may, furthermore, provide a kind of server, including memory, processor and storage on a memory and can handled
The step of computer program run on device, the processor realizes the above method when executing described program.
Compared with prior art, present invention has the advantage that
1, image unsupervised classification precision is high, accidentally divides phenomenon less, can be used for remote sensing images analysis；
2, without manually pre-entering remote sensing image class number, the class number of image can directly be automatically determined；
3, spatial information item is merged, spatial structural form generally existing in remote sensing image has been fully considered, by pixel
Between space relationship shown.
For convenience of description, it is divided into various units when description apparatus above with function to describe respectively.Certainly, implementing this
The function of each unit can be realized in the same or multiple software and or hardware when application.
It should be understood by those skilled in the art that, the embodiment of the present invention can provide as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the present invention
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the present invention, which can be used in one or more,
The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces
The form of product.
The present invention be referring to according to the method for the embodiment of the present invention, the process of equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
The application can describe in the general context of computer-executable instructions executed by a computer, such as program
Module.Generally, program module includes routines performing specific tasks or implementing specific abstract data types, programs, objects, group
Part, data structure etc..The application can also be practiced in a distributed computing environment, in these distributed computing environments, by
Task is executed by the connected remote processing devices of communication network.In a distributed computing environment, program module can be with
In the local and remote computer storage media including storage equipment.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
In a typical configuration, calculate equipment include one or more processors (CPU), input/output interface,
Network interface and memory.
Memory may include the non-volatile memory in computer-readable medium, random access memory (RAM) and/
Or the forms such as Nonvolatile memory, such as read-only memory (ROM) or flash memory (flash RAM).Memory is computer-readable medium
Example.
Computer-readable medium includes permanent and non-permanent, removable and non-removable media can be by any method
Or technology come realize information store.Information can be computer readable instructions, data structure, the module of program or other data.
The example of the storage medium of computer includes, but are not limited to phase change memory (PRAM), static random access memory (SRAM), moves
State random access memory (DRAM), other kinds of random access memory (RAM), read-only memory (ROM), electrically erasable
Except programmable read only memory (EEPROM), flash memory or other memory techniques, read-only disc read only memory (CD-ROM) (CD-
ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, tape magnetic disk storage or other magnetism are deposited
Equipment or any other non-transmission medium are stored up, can be used for storage can be accessed by a computing device information.According to boundary herein
Fixed, computer-readable medium does not include temporary computer readable media (transitory media), such as the data-signal of modulation and
Carrier wave.
It should also be noted that, the terms "include", "comprise" or its any other variant are intended to nonexcludability
It include so that the process, method, commodity or the equipment that include a series of elements not only include those elements, but also to wrap
Include other elements that are not explicitly listed, or further include for this process, method, commodity or equipment intrinsic want
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including described want
There is also other identical elements in the process, method of element, commodity or equipment.
All the embodiments in this specification are described in a progressive manner, same and similar portion between each embodiment
Dividing may refer to each other, and each embodiment focuses on the differences from other embodiments.Especially for system reality
For applying example, since it is substantially similar to the method embodiment, so being described relatively simple, related place is referring to embodiment of the method
Part explanation.
The above description is only an example of the present application, is not intended to limit this application.For those skilled in the art
For, various changes and changes are possible in this application.All any modifications made within the spirit and principles of the present application are equal
Replacement, improvement etc., should be included within the scope of the claims of this application.
Claims (10)
1. it is a kind of based on multiple target mould because optimization algorithm remote sensing image sky compose non-supervised classification, it is characterised in that the side
Method includes:
Step 1: the multi-spectrum remote sensing image that input is to be sorted；
Step 2: automatically determine layer, using being automatically determined based on adaptive differential evolution algorithm and export the best of remote sensing image
Class number；
Step 3: the class number K* obtained in input step two, using adaptive multiple target mould because optimization algorithm is to sky spectrum knot
It closes objective function Jm_S and function XB and carries out multiple-objection optimization, obtain optimal class center；
Step 4: carrying out unsupervised classification to the remote sensing image of input, most using optimal class center obtained in step 3
Output unsupervised classification result eventually.
2. it is according to claim 1 based on multiple target mould because optimization algorithm remote sensing image sky compose non-supervised classification,
It is characterized in that, the step 2 mainly includes the operation such as individual and initialization of population, intersection, variation and selection.
3. it is according to claim 2 based on multiple target mould because optimization algorithm remote sensing image sky compose non-supervised classification,
It is characterized in that, the step 2 is accomplished by the following way:
Step 2.1, the individual UVR exposure and initialization of population operation in differential evolution algorithm are carried out, class center is initial at random
Change is encoded into individual Ii={ Ii,1,Ii,2,...,Ii,j,...,Ii,d, wherein d=C at randommax×nb, CmaxFor selectable maximum
Class number, nbFor image wave band number；Including multiple populations, there is NP individual in each population, individual institute is right in population
The class number answered is consistent, but class number K different representated by the individual between population；
Step 2.2, the corresponding objective function Jm value of each individual in each population is calculated separately；
Formula are as follows:Wherein K represents class number, and N represents image pixel number, xjRepresent shadow
As upper j-th of pixel, UiRepresent i-th of class center, μijThe degree of membership between j-th of pixel and i-th of class center is represented,
M is fuzzy weighted values index.
Step 2.3, the variation between individual is carried out in each population respectivelyWherein F is zoom factor,For three individuals randomly selected from population, r1、r2And r3Three between 1~NP are unequal random whole
Number；
Crossover operation is carried out againCR is crossover probability, jrandFor 0~d
Between a random integers, finally generate new offspring individual Qi={ Qi,1,Qi,2,...,Qi,j,...,Qi,d}；
In addition, F and CR is very important two parameters in DE algorithm, it is also encoded into individual, carries out adaptive evolution,
It is as follows to update operation:
Step 2.4, the number of individuals after intersecting and make a variation in each population is 2 × NP, then calculates separately respective objective function
Jm value is ranked up Jm value corresponding to individual in each population and compares, the NP corresponding individual of minimum Jm value before choosing
Retained；
It is as follows to compare formula:
Step 2.5, the selection for representing optimum solution in the population of different classes of number is respectively compared NP individual, selects one and gather around
There is the individual of minimum Jm value
Step 2.6, XB functional value corresponding to individual selected in step 2.5 is calculated:Wherein most parameter is similar with Jm, by the way that Sep (U) is added, so that
Between class distance maximizes, UkFor the class center of kth class；Individual wherein with minimum XB value is selected againExport the class number corresponding to it；
Step 2.7, step 2.3~2.6 are repeated, until the shadow that the class number of output is not changing, and output finally determines
As class number.
4. it is according to claim 2 based on multiple target mould because optimization algorithm remote sensing image sky compose non-supervised classification,
It is characterized in that, the step 3 specifically includes:
Step 3.1, the empty spectrum joint objective function Jm_S of building；
Step 3.2, individual and initialization of population randomly choose K* pixel coder into individual G from imagei={ Gi,1,
Gi,2,...,Gi,j,...,Gi,d, in addition, also F and CR are encoded into individual.The NP bodily form is initialized into a population simultaneously；
Step 3.3, global search makes individual carry out the overall situation in multiple target space and searches using adaptive differential evolution algorithm
Rope, then carry out individual sequence and selection；
Step 3.4, adaptive local is searched for, and randomly chooses some individuals after step 3.2, carries out Gauss local search,
Individual sequence and selection are carried out again；
Step 3.5, step 3.2~3.4 are repeated until reaching maximum number of iterations, exports the individual of final choice and extracts it
Corresponding class center.
5. it is according to claim 4 based on multiple target mould because optimization algorithm remote sensing image sky compose non-supervised classification,
It is characterized in that, the step 3.1 specifically includes:
Step 3.1.1 obtains mean value image by calculating surrounding neighbors pixel to the average influence of center pixel；Wherein edge
Pixel is not involved in calculating；
Step 3.1.2 obtains spatial information itemWherein α is weight parameter；Again in Jm target
Empty spectrum joint objective function is constructed on the basis of function:
6. it is according to claim 4 based on multiple target mould because optimization algorithm remote sensing image sky compose non-supervised classification,
It is characterized in that, the step 3.3 specifically includes:
Step 3.3.1: determining two objective functions XB and Jm_S that proposed method is optimized, this step carries out adaptive variation
And intersection, concrete operations are as described in step 2.3；
Step 3.3.2, the Pareto carried out between individual are dominant sequence, and the comparison of multiple target is different from single goal, works as satisfactionIndividual xAIt is dominant in individual xB；
Based on this, carrying out Pareto and be dominant sequence, more all individuals, not accounting for the individual marks that other any individuals dominate is 1,
Remaining individual is so compared again, is labeled as 2, and so on have the label of oneself until all individuals；
Step 3.3.3 carries out crowding distance sequence, in order to keep the stabilization of individual amount in population, it would be desirable in population
NP more excellent individuals are selected, are selected first labeled as 1 individual, but is generally designated as 1 individual amount and is more than or few
In NP, choosing solution just is carried out in the way of crowding distance sequence at this time；
Calculate the distance between i-th of individual and two neighboring individualWherein fXB max、fXB min、fJm_S maxAnd fJm_S minPoint
Maximum and minimum value of the respective objects function in current iteration number is not represented；
The individual that finally selection possesses larger crowding distance value carries out next step operation.
7. it is according to claim 4 based on multiple target mould because optimization algorithm remote sensing image sky compose non-supervised classification,
It is characterized in that, the step 3.4 specifically includes:
Step 3.4.1 selects Gauss local search Gi'=Normal (Gi,δ2) as mould because of local search step in algorithm,
Middle δ is standard deviation, and be taken in of each dimension for acting on individual class center part meets condition rand [0,1] > FLS
When=0.5；
The standard deviation δ of four different gradients is arranged in step 3.4.21,δ2,δ3,δ4, then each gradient is stored using Pro_GLS_ δ
The probability that standard deviation is selected, initialization are all 0.25；
Step 3.4.3 carries out Gauss local search and obtains new individual to be judged, calculates two objective functions corresponding to it
Value, be then compared using the thought that Pareto is dominant, newly generated individual can be divided into the individual of receiving, refusal individual and
Doubtful individual；
Step 3.4.4 calculates the score of each gradient standard deviation
Wherein f (XB) and f (Jm_S) is that the individual does not carry out the functional value before local search,WithFor into
Functional value after row local search；
Later, the average value of each gradient standard deviation is calculatedWherein numiRepresent current gradient
The number that is used of standard deviation；
Step 3.4.5, update Pro_GLS_ δ, the probability updating that the standard deviation of each gradient is selected forBeing provided with the probability that certain algebra β allows the standard deviation of each gradient to be selected is always
0.25, carry out the update of select probability again later；
Step 3.4.6, be dominant sequence using Pareto and crowding distance sequence carry out population recruitment make its maintain NP it is individual.
8. it is a kind of based on multiple target mould because optimization algorithm remote sensing image sky compose unsupervised classification system, comprising:
Image input unit, for inputting multi-spectrum remote sensing image to be sorted；
Unit is automatically determined, for utilizing the optimum kind that remote sensing image is automatically determined and exported based on adaptive differential evolution algorithm
Other number；
Objective optimization unit, for inputting the class number K* for automatically determining unit acquisition, using adaptive multiple target mould because excellent
Change algorithm and multiple-objection optimization is carried out to sky spectrum combined objective function Jm_S and function XB, obtains optimal class center；
Taxon carries out the remote sensing image of input non-for utilizing the obtained optimal class center of objective optimization unit
Supervised classification, final output unsupervised classification result.
9. system according to claim 8, which is characterized in that
The unit that automatically determines is accomplished by the following way:
Carry out differential evolution algorithm in individual UVR exposure and initialization of population operation, by class center random initializtion encode into
Individual Ii={ Ii,1,Ii,2,...,Ii,j,...,Ii,d, wherein d=C at randommax×nb, CmaxFor selectable maximum classification number
Mesh, nbFor image wave band number；Including multiple populations, there is NP individual in each population, class corresponding to individual in population
Other number is consistent, but class number K different representated by the individual between population；
Calculate separately the corresponding objective function Jm value of each individual in each population；
Formula are as follows:Wherein K represents class number, and N represents image pixel number, xjRepresent shadow
As upper j-th of pixel, UiRepresent i-th of class center, μijThe degree of membership between j-th of pixel and i-th of class center is represented,
M is fuzzy weighted values index；
The variation between individual is carried out in each population respectivelyWherein F is zoom factor,
For three individuals randomly selected from population, r1、r2And r3Three unequal random integers between 1~NP；
Crossover operation is carried out againCR is crossover probability, jrandFor 0~d
Between a random integers, finally generate new offspring individual Qi={ Qi,1,Qi,2,...,Qi,j,...,Qi,d}；
In addition, F and CR is very important two parameters in DE algorithm, it is also encoded into individual, carries out adaptive evolution,
It is as follows to update operation:
Number of individuals after intersecting and making a variation in each population is 2 × NP, then calculates separately respective objective function Jm value, every
Jm value corresponding to individual is ranked up in a population and is compared, the NP corresponding individual of minimum Jm value is retained before choosing；
It is as follows to compare formula:
The selection for representing optimum solution in the population of different classes of number is respectively compared NP individual, selects one and possess minimum Jm
The individual of value
Calculate XB functional value corresponding to selected individual:
Wherein most parameter is similar with Jm, by the way that Sep (U) is added,
So that between class distance maximizes, UkFor the class center of kth class；Individual wherein with minimum XB value is selected againExport the class number corresponding to it；
It repeats the above steps, until the image class number that the class number of output is not changing, and output finally determines；
The objective optimization unit is accomplished by the following way:
The empty spectrum joint objective function Jm_S of building, especially by calculating surrounding neighbors pixel to the average shadow of center pixel
It rings, obtains mean value image；Wherein edge pixel is not involved in calculating；Obtain spatial information itemIts
Middle α is weight parameter；Empty spectrum joint objective function is constructed on the basis of Jm objective function again:
Individual and initialization of population randomly choose K* pixel coder into individual G from imagei={ Gi,1,Gi,2,...,
Gi,j,...,Gi,d, in addition, also F and CR are encoded into individual.The NP bodily form is initialized into a population simultaneously；
Global search makes individual carry out global search in multiple target space, then carries out using adaptive differential evolution algorithm
Individual sequence and selection；It specifically includes:
Determine two objective functions XB and Jm_S that proposed method is optimized, this step carries out adaptive variation and intersection；
The Pareto carried out between individual is dominant sequence, and the comparison of multiple target is different from single goal, works as satisfactionIndividual xAIt is dominant in individual xB；
Based on this, carrying out Pareto and be dominant sequence, more all individuals, not accounting for the individual marks that other any individuals dominate is 1,
Remaining individual is so compared again, is labeled as 2, and so on have the label of oneself until all individuals；
Crowding distance sequence is carried out, in order to keep the stabilization of individual amount in population, it would be desirable to selected in population NP compared with
Excellent individual is selected first labeled as 1 individual, but is generally designated as 1 individual amount more or less than NP, at this time
Just choosing solution is carried out in the way of crowding distance sequence；
Calculate the distance between i-th of individual and two neighboring individual
Wherein fXB max、fXB min、fJm_S maxWith
fJm_S minRespectively represent maximum and minimum value of the respective objects function in current iteration number；
The individual that finally selection possesses larger crowding distance value carries out next step operation.
Adaptive local search, randomly chooses some individuals after above-mentioned steps, carries out Gauss local search, then carries out a
Body sequence and selection；It specifically includes:
It specifically includes:
Select Gauss local search Gi'=Normal (Gi,δ2) be used as mould because of local search step in algorithm, wherein δ is standard
Difference, each dimension for acting on individual class center part are taken in when meeting condition rand [0,1] > FLS=0.5；
The standard deviation δ of four different gradients is set1,δ2,δ3,δ4, it is selected that each gradient standard deviation then is stored using Pro_GLS_ δ
The probability selected, initialization are all 0.25；
It carries out Gauss local search and obtains new individual to be judged, calculate two target function values corresponding to it, then utilize
The thought that Pareto is dominant is compared, newly generated individual can be divided into the individual of receiving, refusal it is individual and doubtful
Individual；
Calculate the score of each gradient standard deviation
Wherein f (XB) and f (Jm_S) is that the individual does not carry out the functional value before local search,WithFor into
Functional value after row local search；
Later, the average value of each gradient standard deviation is calculatedWherein numiRepresent current gradient
The number that is used of standard deviation；
Update Pro_GLS_ δ, the probability updating that the standard deviation of each gradient is selected forWherein
It is always 0.25 that the probability that certain algebra β allows the standard deviation of each gradient to be selected, which is arranged, carries out select probability more again later
Newly；
Be dominant sequence using Pareto and crowding distance sequence carry out population recruitment make its maintain NP it is individual；
Step above-mentioned steps are repeated until reaching maximum number of iterations, the individual of final choice is exported and extracts corresponding to it
Class center.
10. a kind of computer readable storage medium, is stored thereon with computer program, power is realized when which is executed by processor
Benefit requires the step of any one of 1-7 the method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910056340.9A CN109784286A (en) | 2019-01-22 | 2019-01-22 | Based on multiple target mould because the remote sensing image sky of optimization algorithm composes non-supervised classification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910056340.9A CN109784286A (en) | 2019-01-22 | 2019-01-22 | Based on multiple target mould because the remote sensing image sky of optimization algorithm composes non-supervised classification |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109784286A true CN109784286A (en) | 2019-05-21 |
Family
ID=66501961
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910056340.9A Pending CN109784286A (en) | 2019-01-22 | 2019-01-22 | Based on multiple target mould because the remote sensing image sky of optimization algorithm composes non-supervised classification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109784286A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111126467A (en) * | 2019-12-16 | 2020-05-08 | 武汉大学 | Remote sensing image space spectrum clustering method based on multi-target sine and cosine algorithm |
| CN112183459A (en) * | 2020-10-20 | 2021-01-05 | 安徽大学 | Remote sensing water quality image classification method based on evolution multi-objective optimization |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104021425A (en) * | 2014-05-19 | 2014-09-03 | 中国人民解放军国防科学技术大学 | Meme evolutionary algorithm for solving advancing-delay scheduling problem |
| CN105869053A (en) * | 2016-03-23 | 2016-08-17 | 西安电子科技大学 | Two-stage memetic based social network influence maximizing method |
-
2019
- 2019-01-22 CN CN201910056340.9A patent/CN109784286A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104021425A (en) * | 2014-05-19 | 2014-09-03 | 中国人民解放军国防科学技术大学 | Meme evolutionary algorithm for solving advancing-delay scheduling problem |
| CN105869053A (en) * | 2016-03-23 | 2016-08-17 | 西安电子科技大学 | Two-stage memetic based social network influence maximizing method |
Non-Patent Citations (1)
| Title |
|---|
| 马爱龙: "遥感影像分类中的演化计算理论及其应用方法研究", 《中国博士学位论文全文数据库 基础科学辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111126467A (en) * | 2019-12-16 | 2020-05-08 | 武汉大学 | Remote sensing image space spectrum clustering method based on multi-target sine and cosine algorithm |
| CN111126467B (en) * | 2019-12-16 | 2022-02-15 | 武汉大学 | Remote sensing image space spectrum clustering method based on multi-target sine and cosine algorithm |
| CN112183459A (en) * | 2020-10-20 | 2021-01-05 | 安徽大学 | Remote sensing water quality image classification method based on evolution multi-objective optimization |
| CN112183459B (en) * | 2020-10-20 | 2022-10-04 | 安徽大学 | Remote sensing water quality image classification method based on evolution multi-objective optimization |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pawlak | Rough sets and decision analysis | |
| CN107402955A (en) | The method and apparatus for determining the index grid of geography fence | |
| CN108038576A (en) | Based on the logistics distribution routing resource and system for improving dijkstra's algorithm | |
| ARGIALAS et al. | Quantitative description and classification of drainage patterns | |
| CN110347881A (en) | A kind of group's discovery method for recalling figure insertion based on path | |
| CN110097529A (en) | A kind of farmland Grading unit division methods and system based on semantic rules | |
| CN109784286A (en) | Based on multiple target mould because the remote sensing image sky of optimization algorithm composes non-supervised classification | |
| Rodzin et al. | New computational models for big data and optimization | |
| Wong et al. | Hierarchical fuzzy signature structure for complex structured data | |
| Cordón et al. | A multiobjective genetic learning process for joint feature selection and granularity and contexts learning in fuzzy rule-based classification systems | |
| CN108614932B (en) | Edge graph-based linear flow overlapping community discovery method, system and storage medium | |
| Liu et al. | PRUC: P-regions with user-defined constraint | |
| Nourzad et al. | Ensemble methods for binary classifications of airborne LiDAR data | |
| CN108491505A (en) | A kind of DSATUR figure vertex coloring methods based on the sequence of Topology Potential value | |
| Aliabad et al. | Comparison of neural network methods (fuzzy ARTMAP, Kohonen and Perceptron) and maximum likelihood efficiency in preparation of land use map | |
| CN110189230B (en) | Construction method of analytic model of dynamic partition | |
| Verma et al. | A novel framework for neural architecture search in the hill climbing domain | |
| Lee et al. | Rule discovery using hierarchical classification structure with rough sets | |
| Sahraei et al. | Cluster-based multi-objective optimization for identifying diverse design options: Application to water resources problems | |
| Bauer et al. | Finite loop spaces are manifolds | |
| Ojaghi et al. | A new method for semi-automatic classification of remotely sensed images developed based on the cognitive approaches for producing spatial data required in geomatics applications | |
| Holmgren et al. | Mapping biased higher-order walks reveals overlapping communities | |
| Torabi et al. | Multi-objective interior search algorithm for optimization: A new multi-objective meta-heuristic algorithm | |
| Tran | Content-based retrieval using a multi-objective genetic algorithm | |
| US20220188714A1 (en) | System and Method for Automatic Learning and Selecting of a Candidate Workflow Representative of Machine Learning Flow Set |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 430071 No. 27 Justice Road, Wuchang District, Wuhan City, Hubei Province Applicant after: Hubei Provincial Institute of land and resources (real estate registration center of Hubei Provincial Department of land and resources) Address before: 430071 No. 27 Justice Road, Wuchang District, Wuhan City, Hubei Province Applicant before: HUBEI PROVINCIAL INSTITUTE OF LAND AND RESOURCES |
|
| CB02 | Change of applicant information | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190521 |
|
| WD01 | Invention patent application deemed withdrawn after publication |