CN108564592A - Based on a variety of image partition methods for being clustered to differential evolution algorithm of dynamic - Google Patents

Based on a variety of image partition methods for being clustered to differential evolution algorithm of dynamic Download PDF

Info

Publication number
CN108564592A
CN108564592A CN201810178714.XA CN201810178714A CN108564592A CN 108564592 A CN108564592 A CN 108564592A CN 201810178714 A CN201810178714 A CN 201810178714A CN 108564592 A CN108564592 A CN 108564592A
Authority
CN
China
Prior art keywords
population
individual
sub
image
generation
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201810178714.XA
Other languages
Chinese (zh)
Other versions
CN108564592B (en
Inventor
柳培忠
范宇凌
唐加能
骆炎民
邓建华
杜永兆
刘晓芳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Quanzhou City Hongye Mdt Infotech Ltd In Imitation
Huaqiao University
Original Assignee
Quanzhou City Hongye Mdt Infotech Ltd In Imitation
Huaqiao University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Quanzhou City Hongye Mdt Infotech Ltd In Imitation, Huaqiao University filed Critical Quanzhou City Hongye Mdt Infotech Ltd In Imitation
Priority to CN201810178714.XA priority Critical patent/CN108564592B/en
Publication of CN108564592A publication Critical patent/CN108564592A/en
Application granted granted Critical
Publication of CN108564592B publication Critical patent/CN108564592B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]

Abstract

The invention discloses a kind of based on a variety of image partition methods for being clustered to differential evolution algorithm of dynamic, sub- population mechanism is divided and redistributed using initial population in terms of population structure planning, ensure the diversity of Evolution of Population process and avoids the occurrence of local optimum phenomenon；It is combined using local search variation and global search variation on Mutation Strategy, convergent balance to reach population exploration optimal threshold and is accelerated with this；Also by the dynamic incremental variations of the parabolic of the crossover probability factor, the deficiency that standard difference evolution algorithm preset parameter is brought effectively have been directed to.By with comparison result of other evolution algorithms on benchmark test collection this it appears that after improving algorithm optimizing and convergence rate conspicuousness, and improved differential evolution algorithm is applied to the segmentation of image, it is no matter all effective notable in accuracy or speed.

Description

Based on a variety of image partition methods for being clustered to differential evolution algorithm of dynamic
Technical field
The present invention relates to Computer Simulations and optimization field, more particularly to are clustered to differential evolution algorithm based on dynamically a variety of Image partition method.
Background technology
Threshold method is a kind of simple and effective image Segmentation Technology, but threshold method also have the shortcomings that it is apparent, i.e., for The threshold value of high-definition image solves, and calculation amount increases and calculates time growth.Differential evolution algorithm (Differential Evolution, DE) it is a kind of heuristic random searching algorithm by simulating Evolution of Population difference.Storn and Price are initial Imagination is to solve the problems, such as Chebyshev multinomials, finds that, compared with other evolution algorithms, differential evolution algorithm is solving complexity later Global Optimal Problem in terms of performance it is more prominent, process is also more simple, and controlled parameter is few, adaptable.But for The threshold value of high-definition image is solved, there are still some problems for differential evolution algorithm, and population local search ability is weak and is easily absorbed in part Optimal disadvantage.Therefore, the present inventor further explores and studies to standard DE algorithms deficiency point, proposes that a kind of dynamic is a variety of It is clustered to the image partition method of differential evolution algorithm.
Invention content
It is a kind of based on a variety of image partition methods for being clustered to difference algorithm of dynamic it is an object of the invention to propose, it will change Differential evolution algorithm (DE algorithms) after is conjointly employed in the segmentation of image with solution image threshold method, will not only make solution Image threshold is absorbed in local optimum, and can accelerate the solving speed of image threshold.
The present invention is based on a variety of image partition methods for being clustered to differential evolution algorithm of dynamic, include the following steps：
Step 1, setting image segmentation threshold range and the fitness function for assessing segmentation effect：
Image to be split is gray level image, and setting segmentation threshold range is between 0-255；
Fitness function using maximum between-cluster variance as assessment segmentation effect, calculation formula are as follows：
F=w0·w1·(u0-u1)2 (1)
Wherein, w0It is the gray probability of image background, w1It is the gray probability of image object, u0It is the gray scale of image background Mean value, u1It is the gray average of image object；
Step 2, a variety of parameters for being clustered to differential evolution algorithm of setting dynamic, parameter include population scale NP, individual dimension Number D, evolution maximum iteration MaxF, current iteration number iter, evolutionary generation G, maximum evolutionary generation MaxG；
Step 3, initialisation image threshold value population xG：
Wherein, i=1,2 ..., NP, j=1,2 ..., D, population scale NP indicate the quantity of population at individual, individual dimension D Expression solves dimension,It is random number between 0 to 1,For i-th in population individual jth dimension attribute,With RespectivelyLower bound and the upper bound；
Step 4, the fitness that the assessment segmentation effect of each individual threshold value of population is calculated by the fitness function of step 1 Value；
Step 5 divides and distributes sub- population
If the first generation, sub- population only need to be divided；Defect individual ratio will be retained according to parent since the second generation, weight Newly distribute sub- population；
In the Evolution of Population first generation, population Pop is divided into three sub- population Pop according to formula (3)1、Pop2、Pop3, Wherein Pop1Sub- population scale ratio Pop2、Pop3Greatly；
Wherein, NPkIndicate the population scale of k-th of sub- population, λkIndicate the macro ratio of k-th of sub- population, three sons kind Group macro ratio relationship be：λ1＞ λ2=λ3And λk∈[0,1]；
Since the second generation, statistics each sub- population from parent inherit individual ratio, and by the individual ratio of succession into Row ascending sort gives the Mutation Strategy distribution number of individuals more than succession individual few sub- population Pop2、Pop3, to inheriting, individual is few Mutation Strategy distributes the sub- population more than number of individuals；
Step 6 carries out mutation operation to population：
Three sub- populations correspond to three kinds of Mutation Strategies respectively, and the Mutation Strategy randomly selected by three kinds is respectively to three sons Population at individualIt carries out mutation operation and obtains variation individualThe Mutation Strategy includes global search Mutation Strategy, locally searches Rope Mutation Strategy and weighting Mutation Strategy；
Step 7 carries out crossover operation to population：
Crossover operation is the process that variation individual is intersected with parent individuality, to generate new experiment individualI.e. Offspring individual；Using binomial random crossover design, crossover operation is as follows：
Wherein, randb is the random number between [0,1], and CR is the crossover probability factor and ranging from [0,1], the crossover probability Factor CR uses the dynamic incremental manner of parabolic, as follows：
Wherein, CRmax、CRminFor the bound of crossover probability factor CR, G is evolutionary generation, and MaxG is maximum evolves generation Number；
Step 8 carries out selection operation to filial generation and parent, chooses the preferable individual of fitness value as population of new generation Body
Selection operation is using the greedy selection mode selected the superior and eliminated the inferior so that the more excellent individual of filial generationSubstitute parent individuality To which population is close towards optimum segmentation threshold value always, selection operation is as follows：
Wherein, f (x) is fitness evaluating function；
Step 9 judges whether to meet iterated conditional iter≤MaxF, and satisfaction then goes to step 5 and carries out next-generation evolution, no Satisfaction then exits evolution and goes to step 10；
Step 10 exports and shows optimum segmentation threshold value
Step 11, by obtained optimum segmentation threshold valueIt is applied in carrying out image threshold segmentation, the figure after being divided Picture.
The global search Mutation Strategy uses DE/current-to-rand/1 algorithms, the following institute of Mutation Strategy formula Show：
The local search Mutation Strategy uses DE/current-to-best/1 algorithms, the following institute of Mutation Strategy formula Show：
The weighting Mutation Strategy uses DE/current-to-pbest/2 or DE/current-to-rbest/2 algorithms, By introducing weighted strategy, by the global search factorWith local search factorIt is grasped into row variation in conjunction with to a certain sub- population Make, Mutation Strategy formula is as follows：
Wherein, w is weighted factor, wmaxAnd wminIt is the bound of w, the value of w gradually becomes with the increase of iterations Greatly, iter is current iteration number, and MaxF is evolution maximum iteration, the global search factorWith local search factor As shown in formula (7)：
Wherein, K is the random number between [0,1], and F is that zoom factor belongs to [0,1],It is individual for local optimum, For global optimum's individual, r1、r2、r3Belong to 1 and arrives NPiInteger.
The present invention is based on a variety of image partition methods for being clustered to differential evolution algorithm (MPWDE) of dynamic to have the following advantages that And good effect：
(1) increase the diversity of search process by multiple sub- populations, selected in the threshold value obtained to each sub- population It is best as a result, conventional threshold values method traversal time is reduced, to accelerate to find the efficiency of optimal threshold；
(2) in mutation operation, weighting Mutation Strategy is introduced to a certain sub- population, which expands search scale early period, Threshold value is avoided to be absorbed in locally optimal solution, the later stage is gradually transitioned into the faster local searching strategy of convergence rate；
(3) standard DE algorithms are directed in crossover operation, constant parameter crossover probability factor CR makes the dynamic of parabolic State incrementally adjusts, and reduces the linear deficiency changed or preset parameter is brought of relative parameter；
(4) it is calculated by intelligent optimization and carrying out image threshold segmentation method effectively combines, letter is assessed by function fitness value The continuous iterative evolution of number, keeps population threshold close towards optimal threshold.
The present invention is divided and is redistributed sub- population mechanism using initial population in terms of population structure planning, is ensured The diversity of Evolution of Population process and avoid the occurrence of local optimum phenomenon；Using local search variation and the overall situation on Mutation Strategy Search variation is combined, and to reach population exploration optimal threshold and accelerates convergent balance with this；Also pass through the crossover probability factor Parabolic dynamic incremental variations, effectively have been directed to the deficiency that standard difference evolution algorithm preset parameter is brought.Pass through With comparison result of other evolution algorithms on benchmark test collection this it appears that algorithm optimizing and convergence rate after improving Conspicuousness, and the segmentation by improved differential evolution algorithm applied to image, it is no matter all effective in accuracy or speed Significantly.
Description of the drawings
Fig. 1 is the flow chart of the present invention；
Fig. 2 be in the present invention dynamic differential evolution algorithm on multiple populations for benchmark test function in the case of D=30 Evolution curve graph；
Fig. 3 be in the present invention dynamic differential evolution algorithm on multiple populations for benchmark test function in the case of D=50 Evolution curve graph；
The present invention is described in further detail in the following with reference to the drawings and specific embodiments.
Specific implementation mode
As shown in Figure 1, the present invention is based on a variety of image partition methods for being clustered to differential evolution algorithm of dynamic, including it is as follows Step：
Step 1, setting image segmentation threshold range and the fitness function for assessing segmentation effect：
Image to be split is gray level image, and setting segmentation threshold range is between 0-255；
To obtain optimal threshold, fitness function of the present invention using maximum between-cluster variance as assessment segmentation effect.Most Big inter-class variance is a validity evaluation index of carrying out image threshold segmentation, can effectively obtain optimal threshold f, calculation formula As follows：
F=w0·w1·(u0-u1)2 (1)
Wherein, w0It is the gray probability of image background, w1It is the gray probability of image object, u0It is the gray scale of image background Mean value, u1It is the gray average of image object；
Step 2, a variety of parameters for being clustered to differential evolution algorithm (MPWDE algorithms) of setting dynamic, parameter include population rule Mould NP, individual dimension D, evolution maximum iteration MaxF, current iteration number iter, evolutionary generation G, maximum evolutionary generation MaxG；
Step 3, initialisation image threshold value population xG：
Wherein, i=1,2 ..., NP, j=1,2 ..., D, population scale NP indicate the quantity of population at individual, individual dimension D Expression solves dimension,It is random number between 0 to 1,For i-th in population individual jth dimension attribute,With RespectivelyLower bound and the upper bound；
Step 4, the fitness that the assessment segmentation effect of each individual threshold value of population is calculated by the fitness function of step 1 Value fit；
Step 5 divides and distributes sub- population
If the first generation, sub- population only need to be divided；Defect individual ratio will be retained according to parent since the second generation, weight Newly distribute sub- population；
In the Evolution of Population first generation, population Pop is divided into three sub- population Pop according to formula (3)1、Pop2、Pop3, Wherein Pop1Sub- population scale ratio Pop2、Pop3Greatly；
Wherein, NPkIndicate the population scale of k-th of sub- population, λkIndicate the macro ratio of k-th of sub- population, three sons kind Group macro ratio relationship be：λ1＞ λ2=λ3And λk∈[0,1]；
Since the second generation, needs to count each sub- population and inherit individual ratio, and the individual ratio for passing through succession from parent Rate carries out ascending sort, and the individual of succession represents the fewer of sub- Population Variation strategy influence, it should give this variation plan Population slightly distributes number of individuals few sub- population Pop2、Pop3；Son kind more than the Mutation Strategy distribution number of individuals few to individual is inherited Group；
Step 6 carries out mutation operation to population：
Three sub- populations correspond to three kinds of Mutation Strategies respectively, and the Mutation Strategy randomly selected by three kinds is respectively to three sons Population at individualIt carries out mutation operation and obtains variation individualThe Mutation Strategy includes global search Mutation Strategy, locally searches Rope Mutation Strategy and weighting Mutation Strategy：
Global search Mutation Strategy uses DE/current-to-rand/1 algorithms, Mutation Strategy formula as follows：
Local search Mutation Strategy uses DE/current-to-best/1 algorithms, Mutation Strategy formula as follows：
It weights Mutation Strategy and uses DE/current-to-pbest/2or DE/current-to-rbest/2 algorithms, lead to Introducing weighted strategy is crossed, by the global search factorWith local search factorMutation operation is carried out in conjunction with to a certain sub- population, Mutation Strategy formula is as follows：
Wherein, w is weighted factor, wmaxAnd wminIt is the bound of w, the value of w gradually becomes with the increase of iterations Greatly, iter is current iteration number, and MaxF is evolution maximum iteration, the global search factorWith local search factor As shown in formula (7)：
Wherein, K is the random number between [0,1], and F is that zoom factor belongs to [0,1],It is individual for local optimum, For global optimum's individual, r1、r2、r3Belong to 1 and arrives NPiInteger；
Step 7 carries out crossover operation to population：
Crossover operation is the process that variation individual is intersected with parent individuality, to generate new experiment individualI.e. Offspring individual；Dynamically a variety of differential evolution algorithm MPWDE that are clustered to use binomial random crossover design, the following institute of crossover operation Show：
Wherein, randb is the random number between [0,1], and CR is the crossover probability factor and ranging from [0,1], MPWDE algorithms Crossover probability factor CR uses the dynamic incremental manner of parabolic, as follows：
Wherein, CRmax、CRminTo define the bound of crossover probability factor CR, G is evolutionary generation, and MaxG is maximum evolves Algebraically；
Step 8 carries out selection operation to filial generation and parent, chooses the preferable individual of fitness value as population of new generation Body
Selection operation is using the greedy selection mode selected the superior and eliminated the inferior so that the more excellent individual of filial generationSubstitute parent individuality To which population is close towards optimum segmentation threshold value always, selection operation is as follows：
Wherein, f (x) is fitness evaluating function；
Step 9 judges whether to meet iterated conditional iter≤MaxF, and satisfaction then goes to step 5 and carries out next-generation evolution, no Satisfaction then exits evolution and goes to step 10；
Step 10 exports and shows optimum segmentation threshold value
Obtained optimum segmentation threshold value is applied in carrying out image threshold segmentation, the image after being divided by step 11.
MPWDE algorithms are applied to before image segmentation threshold optimization, first use it for 30 peacekeepings 50 dimension of optimal inspection collection Benchmark test function.Compared with other evolution algorithms, Fig. 2, Fig. 3 can be seen that inventive algorithm for handle unimodal function, The best results of Solving Multimodal Function and offset function can quickly find global optimum.The present invention using dynamic difference on multiple populations into Change algorithm (MPWDE), correct segmentation can not only be obtained to increase solution threshold value paces by variation, intersection, selection operation As a result, and splitting speed faster.
The present invention introduces the thinking of the sub- population of population dividing in steps of 5, passes through the evolution of iteration neutron population each time As a result next-generation sub- population is distributed, multiple sub- population methods more can guarantee the diversity of Evolution of Population process, to avoid planting The unicity of group promotes algorithm precocity to be absorbed in locally optimal solution, to which segmentation effect be not achieved.
In step 6, to overcome the shortcomings of standard DE algorithms, Mutation Strategy mechanism is weighted to a certain sub- population foundation, point Not Wei local search variation and global search variation be combined, reach population exploration optimal threshold with this and accelerate convergent flat Weighing apparatus.
In step 7, for standard DE algorithms in crossover operation, constant parameter crossover probability factor CR makes parabola The dynamic of formula incrementally adjusts, and this strategy targetedly changes the value of crossover probability factor CR, reduces the line of relative parameter It sexually revises or deficiency that preset parameter is brought.
The present invention is calculated by intelligent optimization and carrying out image threshold segmentation method effectively combines, and is assessed by function fitness value The continuous iterative evolution of function, keeps population threshold close towards optimal threshold.This technology for combining two fields not only embodies Intelligent optimization calculates the importance in optimization application, and certain effect is also provided for image processing field.
The above is only present pre-ferred embodiments, is not intended to limit the scope of the present invention, therefore It is every according to the technical essence of the invention to any subtle modifications, equivalent variations and modifications made by above example, still belong to In the range of technical solution of the present invention.

Claims (2)

1. based on a variety of image partition methods for being clustered to differential evolution algorithm of dynamic, it is characterised in that include the following steps：
Step 1, setting image segmentation threshold range and the fitness function for assessing segmentation effect：
Image to be split is gray level image, and setting segmentation threshold range is between 0-255；
Fitness function using maximum between-cluster variance as assessment segmentation effect, calculation formula are as follows：
F=w0·w1·(u0-u1)2 (1)
Wherein, w0It is the gray probability of image background, w1It is the gray probability of image object, u0It is the gray average of image background, u1It is the gray average of image object；
Step 2, a variety of parameters for being clustered to differential evolution algorithm of setting dynamic, parameter include population scale NP, individual dimension D, Evolution maximum iteration MaxF, current iteration number iter, evolutionary generation G, maximum evolutionary generation MaxG；
Step 3, initialisation image threshold value population xG：
Wherein, i=1,2 ..., NP, j=1,2 ..., D, population scale NP indicate that the quantity of population at individual, individual dimension D indicate Dimension is solved,It is random number between 0 to 1,For i-th in population individual jth dimension attribute,WithRespectively ForLower bound and the upper bound；
Step 4, the fitness value that the assessment segmentation effect of each individual threshold value of population is calculated by the fitness function of step 1；
Step 5 divides and distributes sub- population
If the first generation, sub- population only need to be divided；Defect individual ratio will be retained according to parent since the second generation, divided again Gamete population；
In the Evolution of Population first generation, population Pop is divided into three sub- population Pop according to formula (3)1、Pop2、Pop3, wherein Pop1Sub- population scale ratio Pop2、Pop3Greatly；
Wherein, NPkIndicate the population scale of k-th of sub- population, λkIndicate the macro ratio of k-th of sub- population, three sub- populations Macro ratio relationship is：λ1＞ λ2=λ3And λk∈[0,1]；
Since the second generation, each sub- population of statistics inherits individual ratio from parent, and is risen by the individual ratio of succession Sequence sorts, and gives the Mutation Strategy distribution number of individuals more than succession individual few sub- population Pop2、Pop3, give succession individual few variation Sub- population more than strategy distribution number of individuals；
Step 6 carries out mutation operation to population：
Three sub- populations correspond to three kinds of Mutation Strategies respectively, and the Mutation Strategy randomly selected by three kinds is respectively to three sub- populations IndividualIt carries out mutation operation and obtains variation individualThe Mutation Strategy includes global search Mutation Strategy, local search change Different strategy and weighting Mutation Strategy；
Step 7 carries out crossover operation to population：
Crossover operation is the process that variation individual is intersected with parent individuality, to generate new experiment individualThat is filial generation Individual；Using binomial random crossover design, crossover operation is as follows：
Wherein, randb is the random number between [0,1], and CR is the crossover probability factor and ranging from [0,1], the crossover probability factor CR uses the dynamic incremental manner of parabolic, as follows：
Wherein, CRmax、CRminFor the bound of crossover probability factor CR, G is evolutionary generation, and MaxG is maximum evolutionary generation；
Step 8 carries out selection operation to filial generation and parent, chooses the preferable individual of fitness value as population at individual of new generation
Selection operation is using the greedy selection mode selected the superior and eliminated the inferior so that the more excellent individual of filial generationSubstitute parent individualityTo Population is close towards optimum segmentation threshold value always, and selection operation is as follows：
Wherein, f (x) is fitness evaluating function；
Step 9 judges whether to meet iterated conditional iter≤MaxF, and satisfaction then goes to step 5 and carries out next-generation evolution, is unsatisfactory for It then exits evolution and goes to step 10；
Step 10 exports and shows optimum segmentation threshold value
Step 11, by obtained optimum segmentation threshold valueIt is applied in carrying out image threshold segmentation, the image after being divided.
2. according to claim 1 based on a variety of image partition methods for being clustered to differential evolution algorithm of dynamic, feature It is：
The global search Mutation Strategy uses DE/current-to-rand/1 algorithms, Mutation Strategy formula as follows：
The local search Mutation Strategy uses DE/current-to-best/1 algorithms, Mutation Strategy formula as follows：
The weighting Mutation Strategy uses DE/current-to-pbest/2 or DE/current-to-rbest/2 algorithms, passes through Weighted strategy is introduced, by the global search factorWith local search factorMutation operation is carried out in conjunction with to a certain sub- population, is become Different strategy formula is as follows：
Wherein, w is weighted factor, wmaxAnd wminIt is the bound of w, the value of w becomes larger with the increase of iterations, Iter is current iteration number, and MaxF is evolution maximum iteration, the global search factorWith local search becauseSuch as formula (7) shown in：
Wherein, K is the random number between [0,1], and F is that zoom factor belongs to [0,1],It is individual for local optimum,It is complete Office's optimum individual, r1、r2、r3Belong to 1 and arrives NPiInteger.
CN201810178714.XA 2018-03-05 2018-03-05 Image segmentation method based on dynamic multi-population integration differential evolution algorithm Active CN108564592B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810178714.XA CN108564592B (en) 2018-03-05 2018-03-05 Image segmentation method based on dynamic multi-population integration differential evolution algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810178714.XA CN108564592B (en) 2018-03-05 2018-03-05 Image segmentation method based on dynamic multi-population integration differential evolution algorithm

Publications (2)

Publication Number Publication Date
CN108564592A true CN108564592A (en) 2018-09-21
CN108564592B CN108564592B (en) 2021-05-11

Family

ID=63531310

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810178714.XA Active CN108564592B (en) 2018-03-05 2018-03-05 Image segmentation method based on dynamic multi-population integration differential evolution algorithm

Country Status (1)

Country Link
CN (1) CN108564592B (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110009088A (en) * 2019-01-22 2019-07-12 西安理工大学 A kind of Rail crack detection algorithm based on Population Entropy, population variance improvement
CN110222300A (en) * 2019-05-21 2019-09-10 北京科技大学 A method of calculating cubic System Material parent phase and the alternate orientation relationship of son
CN110780938A (en) * 2019-09-18 2020-02-11 河海大学 Computing task unloading method based on differential evolution in mobile cloud environment
CN111340823A (en) * 2020-02-24 2020-06-26 华侨大学 Mammary image segmentation method based on fuzzy entropy and differential evolution
CN113392566A (en) * 2021-07-13 2021-09-14 柳培忠 Simulation optimization design method based on difference
CN113826368A (en) * 2019-04-18 2021-12-21 甲骨文国际公司 Detecting behavioral anomalies for cloud users for outlier actions
CN116258090A (en) * 2023-05-16 2023-06-13 中国地质大学（武汉） Differential evolution deep space orbit design method and system based on double-stage information migration
CN116883672A (en) * 2023-09-05 2023-10-13 山东省工业技术研究院 Image segmentation method based on clustering division differential evolution algorithm and OTSU algorithm
CN117173061A (en) * 2023-10-27 2023-12-05 山东省工业技术研究院 Image enhancement method based on self-adaptive double-variation differential evolution algorithm

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040101191A1 (en) * 2002-11-15 2004-05-27 Michael Seul Analysis, secure access to, and transmission of array images
CN101866490A (en) * 2010-06-30 2010-10-20 西安电子科技大学 Image segmentation method based on differential immune clone clustering
CN103514457A (en) * 2013-07-29 2014-01-15 北京师范大学 Multi-target improvement differential evolution model for land use structure optimization under ecological constraint conditions
CN105678401A (en) * 2015-12-29 2016-06-15 浙江工业大学 Global optimization method based on strategy adaptability differential evolution

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040101191A1 (en) * 2002-11-15 2004-05-27 Michael Seul Analysis, secure access to, and transmission of array images
CN101866490A (en) * 2010-06-30 2010-10-20 西安电子科技大学 Image segmentation method based on differential immune clone clustering
CN103514457A (en) * 2013-07-29 2014-01-15 北京师范大学 Multi-target improvement differential evolution model for land use structure optimization under ecological constraint conditions
CN105678401A (en) * 2015-12-29 2016-06-15 浙江工业大学 Global optimization method based on strategy adaptability differential evolution

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
LIU J 等: "A image segmentation algorithm based on differential evolution particle swarm optimization fuzzy c-means clustering", 《COMPUTER SCIENCE AND INFORMATION SYSTEMS》 *
张锦华 等: "加权变异策略动态差分进化算法", 《计算机工程与应用》 *

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110009088A (en) * 2019-01-22 2019-07-12 西安理工大学 A kind of Rail crack detection algorithm based on Population Entropy, population variance improvement
CN113826368A (en) * 2019-04-18 2021-12-21 甲骨文国际公司 Detecting behavioral anomalies for cloud users for outlier actions
CN110222300A (en) * 2019-05-21 2019-09-10 北京科技大学 A method of calculating cubic System Material parent phase and the alternate orientation relationship of son
CN110780938A (en) * 2019-09-18 2020-02-11 河海大学 Computing task unloading method based on differential evolution in mobile cloud environment
CN111340823B (en) * 2020-02-24 2023-04-07 华侨大学 Mammary image segmentation method based on fuzzy entropy and differential evolution
CN111340823A (en) * 2020-02-24 2020-06-26 华侨大学 Mammary image segmentation method based on fuzzy entropy and differential evolution
CN113392566A (en) * 2021-07-13 2021-09-14 柳培忠 Simulation optimization design method based on difference
CN116258090A (en) * 2023-05-16 2023-06-13 中国地质大学（武汉） Differential evolution deep space orbit design method and system based on double-stage information migration
CN116258090B (en) * 2023-05-16 2023-08-18 中国地质大学（武汉） Differential evolution deep space orbit design method and system based on double-stage information migration
CN116883672A (en) * 2023-09-05 2023-10-13 山东省工业技术研究院 Image segmentation method based on clustering division differential evolution algorithm and OTSU algorithm
CN116883672B (en) * 2023-09-05 2024-01-16 山东省工业技术研究院 Image segmentation method based on clustering division differential evolution algorithm and OTSU algorithm
CN117173061A (en) * 2023-10-27 2023-12-05 山东省工业技术研究院 Image enhancement method based on self-adaptive double-variation differential evolution algorithm
CN117173061B (en) * 2023-10-27 2024-03-19 山东省工业技术研究院 Image enhancement method based on self-adaptive double-variation differential evolution algorithm

Also Published As

Publication number Publication date
CN108564592B (en) 2021-05-11

Similar Documents

Publication Publication Date Title
CN108564592A (en) Based on a variety of image partition methods for being clustered to differential evolution algorithm of dynamic
CN108846472A (en) A kind of optimization method of Adaptive Genetic Particle Swarm Mixed Algorithm
CN107316099A (en) Ammunition Storage Reliability Forecasting Methodology based on particle group optimizing BP neural network
CN108898154A (en) A kind of electric load SOM-FCM Hierarchical clustering methods
Eskandari et al. FastPGA: A dynamic population sizing approach for solving expensive multiobjective optimization problems
CN105488528A (en) Improved adaptive genetic algorithm based neural network image classification method
CN113300380B (en) Load curve segmentation-based power distribution network reactive power optimization compensation method
CN108805193A (en) A kind of power loss data filling method based on mixed strategy
CN106845642A (en) A kind of self adaptation multi-target evolution method of belt restraining cloud workflow schedule
CN109271427A (en) A kind of clustering method based on neighbour's density and manifold distance
CN107330902A (en) Chaos-Genetic BP neural network image partition method based on Arnold conversion
CN108665322A (en) The construction method of grain ration Consumption forecast model, Consumption forecast method and device
CN108062720A (en) A kind of load forecasting method based on similar day selection and random forests algorithm
CN109861211A (en) A kind of power distribution network dynamic reconfiguration method based on data-driven
CN110751121A (en) Unsupervised radar signal sorting method based on clustering and SOFM
CN108805346A (en) A kind of hot continuous rolling force forecasting method based on more hidden layer extreme learning machines
CN108846405A (en) Uneven medical insurance data classification method based on SSGAN
CN115525038A (en) Equipment fault diagnosis method based on federal hierarchical optimization learning
CN114625506A (en) Edge cloud collaborative task unloading method based on adaptive covariance matrix evolution strategy
CN111008790A (en) Hydropower station group power generation electric scheduling rule extraction method
CN108388961A (en) Self-adapting random neighbours' community detecting algorithm based on modularity optimization
CN109919374A (en) Prediction of Stock Price method based on APSO-BP neural network
CN107577896B (en) Wind power plant multi-machine aggregation equivalent method based on hybrid Copula theory
CN109670634A (en) A kind of gas flow data predication method based on non-isometric granular feature
CN109858667A (en) It is a kind of based on thunder and lightning weather to the short term clustering method of loading effects

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant