CN104504719A - Image edge detection method and equipment - Google Patents

Image edge detection method and equipment Download PDF

Info

Publication number
CN104504719A
CN104504719A CN201510003860.5A CN201510003860A CN104504719A CN 104504719 A CN104504719 A CN 104504719A CN 201510003860 A CN201510003860 A CN 201510003860A CN 104504719 A CN104504719 A CN 104504719A
Authority
CN
China
Prior art keywords
population
individuality
image
initial population
fitness
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
Application number
CN201510003860.5A
Other languages
Chinese (zh)
Inventor
杨振庚
吴楠
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.)
Inspur Beijing Electronic Information Industry Co Ltd
Original Assignee
Inspur Beijing Electronic Information Industry Co Ltd
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 Inspur Beijing Electronic Information Industry Co Ltd filed Critical Inspur Beijing Electronic Information Industry Co Ltd
Priority to CN201510003860.5A priority Critical patent/CN104504719A/en
Publication of CN104504719A publication Critical patent/CN104504719A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Theoretical Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • General Engineering & Computer Science (AREA)
  • Genetics & Genomics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Physiology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses an image edge detection method, relating to the technical field of detection of image edges. The method comprises the steps: generating an initial population according to a training feature set, optimizing the initial population, decoding individuals with highest adaptive fitness in the optimized initial population, acquiring a first division from to-be-detected images to edge images and generating a new feature set; generating a new population based on the new feature set, optimizing the new population, decoding individuals with highest adaptive fitness in the optimized new population, acquiring a second division from to-be-detected images to edge images to obtain an edge detection model for images, and detecting the image edge according to the edge detection model. The invention further discloses image edge detection equipment. According to the image edge detection method and the equipment, learning from a small number of images and edge-marked images, an ordinary model for detection of image edges is obtained, so that the detection of the image edges is realized; and the image edge detection method and the equipment are simple and easy to use.

Description

A kind of method for detecting image edge and equipment
Technical field
The present invention relates to technique of image edge detection field, specifically one is applied to digital image edge detection model and automatically finds scheme.
Background technology
Have 80% from vision or image information in the information that the mankind receive, have image, figure, animation, video, text data etc.This is the most effective and most important acquisition of information and exchange way.Along with popularizing of computing machine, people utilize computing machine to help the mankind more and more and obtain and processing image information.The edge detecting technology of image is the basis of the picture material understanding technology such as target identification, Images Classification, and good edge detecting technology provides better guarantee for successive image process.
Summary of the invention
Technical matters to be solved by this invention is, provides a kind of method for detecting image edge and equipment, to solve the problem of conventional images edge detection process complexity.
In order to solve the problems of the technologies described above, the invention discloses a kind of method for detecting image edge, the method comprises:
Construct image feature set and training characteristics collection, initial population is generated according to described training characteristics collection, described initial population is optimized to the initial population operating and be optimized, the individuality that in the initial population of decoding optimization, fitness is the highest, obtain first division of image to be detected to edge image, generate new feature set according to first division result;
New population is generated based on described new feature set, described new population is optimized to the new population operating and be optimized, the individuality that in new population after decoding optimization, fitness is the highest, obtain second division of image to be detected to edge image, obtain the Model for Edge Detection of image according to second division result, carry out Image Edge-Detection according to described Model for Edge Detection.
Alternatively, in said method, described initial population is optimized to the process operating the initial population be optimized and comprises:
The fitness of each individuality in assessment initial population;
Fitness size according to individuality each in described initial population selects winning individuality, carries out cross and variation operation to selected winning individuality;
Local Search is carried out to the individuality after cross and variation, Fitness analysis is carried out to individuality each in the population after Local Search, until maximum adaptation degree individual in population is greater than setting value or iterations reaches threshold value, the initial population be optimized.
Alternatively, in said method, winning individuality is selected to comprise according to the fitness size of individuality each in described initial population:
Adopt in championship policy selection initial population and set a number individuality.
Alternatively, in said method, the process of the individuality after cross and variation being carried out to Local Search comprises:
Variation step-length according to setting carries out intensive mutation operation to the population after cross and variation.
Alternatively, in said method, generate new feature set according to first division result and refer to:
Corresponding for the decoded result of individuality the highest for fitness in the initial population of described training characteristics collection and decoding optimization product weightings average computation is obtained new feature set.
The invention also discloses a kind of Image Edge-Detection equipment, at least comprise:
First stage processing module, construct image feature set and training characteristics collection, and generate initial population according to described training characteristics collection, described initial population is optimized to the initial population operating and be optimized, the individuality that in the initial population of decoding optimization, fitness is the highest, obtain first division of image to be detected to edge image, generate new feature set according to this first division result;
Subordinate phase processing module, new population is generated based on described new feature set, described new population is optimized to the new population operating and be optimized, and the individuality that in new population after decoding optimization, fitness is the highest, obtain second division of image to be detected to edge image, obtain the Model for Edge Detection of image according to second division result, and carry out Image Edge-Detection according to described Model for Edge Detection.
Alternatively, in the said equipment, described first stage processing module, is optimized described initial population and operates the population that is optimized and refer to:
The fitness of each individuality in assessment initial population;
Fitness size according to individuality each in described initial population selects winning individuality, carries out cross and variation operation to selected winning individuality;
Local Search is carried out to the individuality after cross and variation, Fitness analysis is carried out to individuality each in the population after Local Search, until maximum adaptation degree individual in population is greater than setting value or iterations reaches threshold value, the initial population be optimized.
Alternatively, in the said equipment, described first stage processing module selects winning individuality to refer to according to the fitness size of individuality each in described initial population:
Adopt in championship policy selection initial population and set a number individuality.
Alternatively, in the said equipment, described first stage processing module carries out intensive mutation operation according to the variation step-length of setting to the population after cross and variation.
Alternatively, in the said equipment, described first stage processing module generates new feature set according to first division result and refers to:
Corresponding for the decoded result of individuality the highest for fitness in the initial population of described training characteristics collection and decoding optimization product weightings average computation is obtained new feature set.
Adopt technical scheme can find image border point and non-edge point mathematical model, by the study of a small amount of image and marker edge image, therefrom draw the universal model of Image Edge-Detection, to realize Image Edge-Detection, and technical scheme is simple and easy to use.
Accompanying drawing explanation
Fig. 1 is FB(flow block) of the present invention;
Fig. 2 is the schematic diagram of the interlace operation under the new coded system that proposes of the present invention;
Fig. 3 is the schematic diagram of the mutation operation under the new coded system that proposes of the present invention;
Fig. 4 is the comparison diagram of simulated effect of the present invention and former method.
Embodiment
For making the object, technical solutions and advantages of the present invention clearly understand, hereafter will be described in further detail technical solution of the present invention by reference to the accompanying drawings.It should be noted that, when not conflicting, the feature in the embodiment of the application and embodiment can combine arbitrarily mutually.
Embodiment 1
The present embodiment provides a kind of method for detecting image edge, mainly comprises following operation:
Construct image feature set and training characteristics collection, initial population is generated according to this training characteristics collection, initial population is optimized to the initial population operating and be optimized, the individuality that in the initial population of decoding optimization, fitness is the highest, obtain first division of image to be detected to edge image, generate new feature set according to first division result;
New population is generated based on new feature set, new population is optimized to the new population operating and be optimized, the individuality that in new population after decoding optimization, fitness is the highest, obtain second division of image to be detected to edge image, obtain the Model for Edge Detection of image according to second division result, carry out Image Edge-Detection according to this Model for Edge Detection.
Wherein, the Optimum Operation for initial population and new population can adopt identical operation steps, and particularly, this preferred operations comprises the steps:
The fitness of each individuality in assessment population;
Fitness size according to individuality each in population selects winning individuality, carries out cross and variation operation to selected winning individuality;
Local Search is carried out to the individuality after cross and variation, Fitness analysis is carried out to individuality each in the population after Local Search, until maximum adaptation degree individual in population is greater than setting value or iterations reaches threshold value, the population be optimized.
Preferably, during individuality winning according to the fitness size selection of individuality each in population, can adopt in championship policy selection population and set a number individuality.
When Local Search is carried out to the individuality after cross and variation, also can carry out intensive mutation operation according to the variation step-length of setting to the population after cross and variation.
Also being noted that in said method the new feature set relating to and generate according to first division result, is that the corresponding product weightings average computation of decoded result of the individuality that in the initial population by training characteristics collection and decoding optimization, fitness is the highest obtains.
Below in conjunction with embody rule, the present embodiment is described in further detail.
The specific implementation process of above-mentioned edge detection method, as shown in Figure 1, comprises the steps:
Step 101, according to the image in image library, random choose goes out the image composition training set at three width original images and corresponding handmarking edge, uses Sobel operator, Prewitt operator, Roberts operator and canny operator to generate preliminary edge image training set image.The mode that each image is according to pixels worth to arrange is combined into new eigenmatrix, obtains training characteristics collection;
Wherein, the training characteristics collection in classification is made up of two parts, and a part is sample data, and another part is true and reliable categorical data.Map in building process in sample data and categorical data, preferably, coloured image is converted into gray-scale map, and extract the gray-scale value of each pixel.Setting threshold value ψ, is converted into bianry image by the image crossing edge by handmarking, and will extract the value of each pixel of bianry image.Row merge respective value, the mapping matrix of composition pixel to pixel.In addition, this step is not limited to Sobel operator, Prewitt operator, Roberts operator and canny operator, also can adopt the combination producing preliminary edge image of other operator or operator.
Step 102, the operational character collection of setting first stage full stop collection crossover probability mutation probability population scale variation step factor step, iterations gen 1; The operational character collection of subordinate phase full stop collection crossover probability mutation probability iterations gen 2, population scale
Step 103, according in step 101 training set image and the population scale of first stage of setting produce initial population:
A (t)={ a 1(t), a 2(t), a 3(t) ..., a n(t) | t=0}, wherein a it () is the individuality of i-th in initial population, represent with the tree that the degree of depth is two, i ∈ [1, n].
Step 104, according to the operational character collection of first stage full stop collection initialization the population pop of individuality 1, calculate the fitness of each individuality in initial population;
Preferably, this step can according to the fitness of each individuality in following formulae discovery population:
Individual a ithe fitness of (t) finess = 1 / N × Σ i = 1 N ( precision i + recall i ) / 2 ;
In formula, precision ibe the ratio that the i-th width image border point successfully detects, recall ibe be the ratio of non-edge point by successful classification in the i-th width image, N is the number of image in training set.
Step 105, the fitness size according to each individuality in population is evaluated individual good and bad: the individuality that fitness is high is considered as winning individuality; Championship policy selection population pop can be adopted 1in individuality; To the population of winning individuality composition carry out cross and variation operation;
Particularly, carry out population according to the fitness of individuality and select operation.Can adopt championship strategy, it is 5 that size is often taken turns in championship, produces n*p c± 1 pairing is individual, carries out mutation operation to remaining individuality.Again according to elitism strategy, from population, select n*p eindividual elite is individual.
Cross and variation operates, and completes as follows:
For the individual ind be chosen in population 1with ind 2, the leaf node number of the two is N, to ind 1with ind 2carry out interlace operation.First the random number rand that is positioned at [1, N] is produced, by individual ind 1rand exchange to ind to N number of leaf node 2correspondence position; By individual ind 2rand exchange to ind to N number of leaf node 1correspondence position.As shown in Figure 2, two individual Abs (0.32,0.65,0.51,0.87 ..., 0.12) and Abs (0.48,3.1,0.83,5.0, ... ,-1.2) carry out interlace operation, producing cross-point locations is 3, then exchange all nodes on the right side of two individual point of crossing places, complete interlace operation.Former individual one, individual two are changed to respectively: Abs (0.32,0.65,0.51,5.0 ... ,-1.2) and Abs (0.48,3.1,0.83,0.87 ..., 0.12).Complete interlace operation.
Mutation operation is carried out to certain individual ind3 in population.First determine individual leaf node number N, then produce the random integers R that is positioned at interval [1, N] index, two random number R being positioned at interval [0,1] step, R style:
If 1. R step>=0.4:
If R style< 0.5 x R index = x R index - step ;
If R style>=0.5 x R index = x R index + step ;
If 2. R step< 0.4:
If R style< 0.5 x R index = x R index - step * rand ( ) * 5 ;
If R style>=0.5 x R index = x R index + step * rand ( ) * 5 ; ;
Wherein rand () is the random number between 0 to 1, and x is nodal values.
As shown in Figure 3, individual Abs (0.65,0.65,0.51,0.87,1.5,4.3,0.12) participates in variation, when generation random variation point is 3, and step size controlling parameter R step> 0.4, variation mode controling parameters R stylewhen>=0.5, make a variation by be step-length be 0.5 additivity variation.The value of individual the 3rd sub-Nodes is from left to right converted into 0.51+0.5 that is 1.01, is Abs (0.65,0.65,1.01,0.87,1.5,4.3,0.12) after individual variation.Complete mutation operation.
Step 106, carries out Local Search to the individuality after cross and variation, sets less variation step-length s', according to this step-length to population carry out the mutation operation of comparatively dense;
In this step, fitness calculating is carried out to cross and variation individuality, select individuality more outstanding between two generations according to survival of the fittest rule.Carry out Local Search to individuality, step-size in search step is 0.1 times of step-length in variation; Local Search carries out five times for each individuality, and each generation random variation point, changes dissimilarity place variation method as follows:
Produce random number R ∈ [0,1]; Variable position R index
If R < 0.5, R index=R index+ step;
If R>=0.5, R index=R index-step.
The fitness of the individuality before and after contrast variation, the quantity retaining fitness maximum is Pop sizeindividuality, form new population.
Step 107, carries out Fitness analysis to the population at individual after Local Search, if in population maximum adaptation degree be greater than 0.85 or iterations reach gen 1time, then perform step 108, otherwise, perform step 105;
Step 108, selects the individuality that fitness is the highest, is optimum individual from the population of iteration ends; Decoding optimum individual, obtains the division result of image to be detected to edge image, generates new feature set according to this division result;
Optimum individual in population will be chosen as division result, and the decoding process of optimum individual can be read the N number of leaf node on individual tree successively, and the vector of a composition 1 × N, N is the dimension of characteristics of image simultaneously, as decoded vector { x 1, x 2..., x n;
New latent structure method is: F'=Avg{F 1x 1, F 2x 2..., F nx ni, wherein F' is new feature, F ibe i-th primitive character.Therefore in this step, can { F be used 1, F 2, F 3..., F nrepresent a certain primitive character of certain image, obtain { x after individual decoding 1, x 2, x 3..., x n, then the new feature of this image is Avg{x 1× F 1, x 2× F 2, x 3× F 3..., x k× F k.Namely obtain new feature by the corresponding product weightings average computation of the decoded result of old characteristic sum optimum individual, thus eliminate the difference between each feature.
In addition, in this step, modeling uses the expression tree of various, the change in depth of operational character again.First selection operator, full stop, the depth capacity in evolution, the initial maximum degree of depth, minimum-depth, initialization mode, population scale, evolutionary generation gen n, setting population scale pop ' size, crossover probability p ' c, mutation probability p ' m, elite's Probability p ' e, operational character collection set f, full stop collection set t;
Initialization population, initialization of population mode has two kinds, and a kind of employing constant depth, namely the degree of depth of individual tree is fixed, and all leaf nodes are in the same layer degree of depth.A kind of use growth pattern, the i.e. degree of depth random variation of individual tree.At this, 1:1 mixing method is used to complete initialization.
Step 109, according to the operational character collection of subordinate phase full stop collection with new feature set, initialization the population pop of individuality 2, to population pop 2in each individuality assessment fitness;
Step 110, according to population pop 2in the preferred population pop of assessment fitness result of each individuality 2, obtain preferred population according to mutation probability crossover probability right intersect, mutation operation;
Said cross and variation operation in this step, carry out as follows:
For the individual ind be chosen in population 1, ind 2carry out interlace operation, first calculate ind 1with ind 2the node number N of middle individual expression tree 1, N 2; Produce two random integers r 1, r 2lay respectively at interval [1, N 1], [1, N 2] in; R is found respectively in individual expression tree 1individual and r 2the position of individual node; Exchange the subtree of two positions;
For the individual ind in population 3carry out mutation operation, first calculate the node number N of individual tree; The random integers r that generation is positioned between [1, N] 1; Find the r in individual expression tree 1individual node; Generation is positioned at [0,1] interval random number rand;
If rand < 0.5, from operational character and full stop, random choose operational symbol replaces the r in individual expression tree 1individual operational symbol, and according to the order number of this operational symbol, generate corresponding individual subtree, complete mutation operation;
If rand>=0.5, first obtain r 1the order number T of the operational symbol of individual Nodes; Then from operational symbol, random choose order number is the operational symbol replacement r of T 1the operational symbol of Nodes, completes variation.
Step 111, carries out Fitness analysis to the individuality after cross and variation.If evolutionary generation is less than gen 2time, then return and perform step 110; Otherwise the optimum individual selected according to the maximum principle of fitness in existing population, export the decoded result of optimum individual, obtain new division result, result obtains the Model for Edge Detection of image accordingly.
To the individuality of initialization assessment fitness, assessment mode calculates according to precision and the recall of classify of image element mean value, identical with step 4; According to the fitness of each individuality, carry out population selection, the individuality be chosen to carries out cross and variation operation;
Again Fitness analysis is carried out to the population after cross and variation, and selects to be about to according to assessment result the individuality that remains, form new population, and it is of future generation to evolve; Until reach evolutionary generation gen ntill.
Decoding optimum individual in this step, obtains final Model for Edge Detection.
From final population, select the highest individuality of fitness, preorder traversal is carried out to it, be decoded as function analytic expression f (I), be training net result model.Wherein I is the image treating rim detection.Utilize this model, calculate corresponding function result, the order returned results according to same is arranged again, namely obtains the edge detection results of image.
As can be seen from above-described embodiment, technical scheme overcomes the deficiencies in the prior art, the genetic programming algorithm proposing a kind of improvement, to realize Image Edge-Detection model, is divided into the value of safety pin to each pixel Computation function model of different images, obtains edge contour figure.The effect once learning to use can be realized well everywhere.Further, for different images, respond well.A row image in the middle of Fig. 4 is the edge contour image of handmarking.
Embodiment 2
The present embodiment provides a kind of Image Edge-Detection equipment, can realize the method for above-described embodiment 1, and it at least comprises as lower module.
First stage processing module, construct image feature set and training characteristics collection, and generate initial population according to training characteristics collection, initial population is optimized to the initial population operating and be optimized, the individuality that in the initial population of decoding optimization, fitness is the highest, obtain first division of image to be detected to edge image, generate new feature set according to this first division result;
Wherein, first stage processing module, when being optimized to initial population the population operating and be optimized, performing and operates as follows:
The fitness of each individuality in assessment initial population;
Fitness size according to individuality each in described initial population selects winning individuality, carries out cross and variation operation to selected winning individuality;
Local Search is carried out to the individuality after cross and variation, Fitness analysis is carried out to individuality each in the population after Local Search, until maximum adaptation degree individual in population is greater than setting value or iterations reaches threshold value, the initial population be optimized.
Preferably, first stage processing module can adopt championship policy selection, selects a setting number winning individuality from initial population.
In addition, first stage processing module also can carry out intensive mutation operation according to the variation step-length of setting to the population after cross and variation.
And first stage processing module is when generating new feature set according to first division result, mainly the corresponding product weightings average computation of decoded result of individuality the highest for fitness in the initial population of training characteristics collection and decoding optimization is obtained new feature collection.
Subordinate phase processing module, new population is generated based on new feature set, new population is optimized to the new population operating and be optimized, and the individuality that in new population after decoding optimization, fitness is the highest, obtain second division of image to be detected to edge image, obtain the Model for Edge Detection of image according to second division result, and carry out Image Edge-Detection according to Model for Edge Detection.
Same, subordinate phase processing module, when being optimized to new population the new population operating and be optimized, can adopt the mode of first stage processing module, namely performs following operation:
The fitness of each individuality in assessment new population;
Fitness size according to individuality each in new population selects winning individuality, carries out cross and variation operation to selected winning individuality;
Local Search is carried out to the individuality after cross and variation, Fitness analysis is carried out to individuality each in the population after Local Search, until maximum adaptation degree individual in population is greater than setting value or iterations reaches threshold value, thus the new population be optimized.
Preferably, championship policy selection can be adopted, from new population, select a setting number winning individuality.
Also intensive mutation operation can be carried out according to the variation step-length of setting to the population after cross and variation.The equipment provided due to the present embodiment can implement the method for above-described embodiment 1, so other of equipment operate in detail, comprising adopted specific algorithm etc. see the corresponding contents of above-described embodiment 1, can not repeat them here.
The all or part of step that one of ordinary skill in the art will appreciate that in said method is carried out instruction related hardware by program and is completed, and described program can be stored in computer-readable recording medium, as ROM (read-only memory), disk or CD etc.Alternatively, all or part of step of above-described embodiment also can use one or more integrated circuit to realize.Correspondingly, each module/unit in above-described embodiment can adopt the form of hardware to realize, and the form of software function module also can be adopted to realize.The application is not restricted to the combination of the hardware and software of any particular form.
The above, be only preferred embodiments of the present invention, be not intended to limit protection scope of the present invention.Within the spirit and principles in the present invention all, any amendment made, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (10)

1. a method for detecting image edge, is characterized in that, the method comprises:
Construct image feature set and training characteristics collection, initial population is generated according to described training characteristics collection, described initial population is optimized to the initial population operating and be optimized, the individuality that in the initial population of decoding optimization, fitness is the highest, obtain first division of image to be detected to edge image, generate new feature set according to first division result;
New population is generated based on described new feature set, described new population is optimized to the new population operating and be optimized, the individuality that in new population after decoding optimization, fitness is the highest, obtain second division of image to be detected to edge image, obtain the Model for Edge Detection of image according to second division result, carry out Image Edge-Detection according to described Model for Edge Detection.
2. the method for claim 1, is characterized in that, described initial population is optimized to the process operating the initial population be optimized and comprises:
The fitness of each individuality in assessment initial population;
Fitness size according to individuality each in described initial population selects winning individuality, carries out cross and variation operation to selected winning individuality;
Local Search is carried out to the individuality after cross and variation, Fitness analysis is carried out to individuality each in the population after Local Search, until maximum adaptation degree individual in population is greater than setting value or iterations reaches threshold value, the initial population be optimized.
3. method as claimed in claim 2, is characterized in that, select winning individuality to comprise according to the fitness size of individuality each in described initial population:
Adopt in championship policy selection initial population and set a number individuality.
4. method as claimed in claim 2, it is characterized in that, the process of the individuality after cross and variation being carried out to Local Search comprises:
Variation step-length according to setting carries out intensive mutation operation to the population after cross and variation.
5. the method as described in any one of Claims 1-4, is characterized in that, generates new feature set refer to according to first division result:
Corresponding for the decoded result of individuality the highest for fitness in the initial population of described training characteristics collection and decoding optimization product weightings average computation is obtained new feature set.
6. an Image Edge-Detection equipment, is characterized in that, at least comprises:
First stage processing module, construct image feature set and training characteristics collection, and generate initial population according to described training characteristics collection, described initial population is optimized to the initial population operating and be optimized, the individuality that in the initial population of decoding optimization, fitness is the highest, obtain first division of image to be detected to edge image, generate new feature set according to this first division result;
Subordinate phase processing module, new population is generated based on described new feature set, described new population is optimized to the new population operating and be optimized, and the individuality that in new population after decoding optimization, fitness is the highest, obtain second division of image to be detected to edge image, obtain the Model for Edge Detection of image according to second division result, and carry out Image Edge-Detection according to described Model for Edge Detection.
7. equipment as claimed in claim 6, is characterized in that, described first stage processing module, is optimized operates the population be optimized and refer to described initial population:
The fitness of each individuality in assessment initial population;
Fitness size according to individuality each in described initial population selects winning individuality, carries out cross and variation operation to selected winning individuality;
Local Search is carried out to the individuality after cross and variation, Fitness analysis is carried out to individuality each in the population after Local Search, until maximum adaptation degree individual in population is greater than setting value or iterations reaches threshold value, the initial population be optimized.
8. equipment as claimed in claim 7, it is characterized in that, described first stage processing module selects winning individuality to refer to according to the fitness size of individuality each in described initial population:
Adopt in championship policy selection initial population and set a number individuality.
9. equipment as claimed in claim 7, is characterized in that, described first stage processing module carries out intensive mutation operation according to the variation step-length of setting to the population after cross and variation.
10. the equipment as described in any one of claim 6 to 9, is characterized in that, described first stage processing module generates new feature set according to first division result and refers to:
Corresponding for the decoded result of individuality the highest for fitness in the initial population of described training characteristics collection and decoding optimization product weightings average computation is obtained new feature set.
CN201510003860.5A 2015-01-05 2015-01-05 Image edge detection method and equipment Pending CN104504719A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510003860.5A CN104504719A (en) 2015-01-05 2015-01-05 Image edge detection method and equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510003860.5A CN104504719A (en) 2015-01-05 2015-01-05 Image edge detection method and equipment

Publications (1)

Publication Number Publication Date
CN104504719A true CN104504719A (en) 2015-04-08

Family

ID=52946113

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510003860.5A Pending CN104504719A (en) 2015-01-05 2015-01-05 Image edge detection method and equipment

Country Status (1)

Country Link
CN (1) CN104504719A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107507210A (en) * 2017-09-27 2017-12-22 上海斐讯数据通信技术有限公司 A kind of method for detecting image edge and device based on genetic algorithm
CN110020725A (en) * 2019-03-22 2019-07-16 北京航空航天大学 A kind of test design method for serving Weapon Equipment System operation emulation
CN111066061A (en) * 2017-09-11 2020-04-24 富士通株式会社 Information processing apparatus, information processing method, and information processing program

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6473525B1 (en) * 1998-01-30 2002-10-29 Compaq Computer Corporation Method for detecting an image edge within a dithered image
CN103325123A (en) * 2013-07-09 2013-09-25 江南大学 Image edge detection method based on self-adaptive neural fuzzy inference systems
CN103942571A (en) * 2014-03-04 2014-07-23 西安电子科技大学 Graphic image sorting method based on genetic programming algorithm

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6473525B1 (en) * 1998-01-30 2002-10-29 Compaq Computer Corporation Method for detecting an image edge within a dithered image
CN103325123A (en) * 2013-07-09 2013-09-25 江南大学 Image edge detection method based on self-adaptive neural fuzzy inference systems
CN103942571A (en) * 2014-03-04 2014-07-23 西安电子科技大学 Graphic image sorting method based on genetic programming algorithm

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
吴晓琳: "基于遗传算法实现数字图像边缘检测", 《舰船电子工程》 *
杨振庚: "遗传规划算法研究及其在分类分析上的应用", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *
董云影 等: "遗传模糊聚类算法在图像边缘检测中的应用", 《模糊***与数学》 *
赵彤洲 等: "基于机器学习的边缘检测方法研究", 《湖北大学学报（自然科学版）》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111066061A (en) * 2017-09-11 2020-04-24 富士通株式会社 Information processing apparatus, information processing method, and information processing program
CN107507210A (en) * 2017-09-27 2017-12-22 上海斐讯数据通信技术有限公司 A kind of method for detecting image edge and device based on genetic algorithm
CN110020725A (en) * 2019-03-22 2019-07-16 北京航空航天大学 A kind of test design method for serving Weapon Equipment System operation emulation

Similar Documents

Publication Publication Date Title
CN105512289B (en) Image search method based on deep learning and Hash
CN109978807B (en) Shadow removing method based on generating type countermeasure network
Jain et al. Autoencoders for level generation, repair, and recognition
Bodhwani et al. Deep residual networks for plant identification
US9449253B2 (en) Learning painting styles for painterly rendering
CN103942571B (en) Graphic image sorting method based on genetic programming algorithm
Ulhaq et al. Efficient diffusion models for vision: A survey
CN109712165A (en) A kind of similar foreground picture image set dividing method based on convolutional neural networks
Cai et al. Multi-objective evolutionary 3D face reconstruction based on improved encoder–decoder network
CN109858487A (en) Weakly supervised semantic segmentation method based on watershed algorithm and image category label
CN107240085A (en) A kind of image interfusion method and system based on convolutional neural networks model
CN107545271A (en) Image-recognizing method, device and system
CN109711411B (en) Image segmentation and identification method based on capsule neurons
CN107730536A (en) A kind of high speed correlation filtering object tracking method based on depth characteristic
Xu et al. AutoSegNet: An automated neural network for image segmentation
KR20200094938A (en) Data imbalance solution method using Generative adversarial network
Bounsaythip et al. Genetic algorithms in image processing-a review
CN104504719A (en) Image edge detection method and equipment
CN106600610A (en) FCM image segmentation method and apparatus
CN115482380A (en) Multi-level highway three-dimensional point cloud target segmentation method based on deep learning
CN104680190B (en) Object detection method and device
CN113554656B (en) Optical remote sensing image example segmentation method and device based on graph neural network
CN107274425A (en) A kind of color image segmentation method and device based on Pulse Coupled Neural Network
WO2019219949A1 (en) Image processing using generative graphical models
CN109658508A (en) A kind of landform synthetic method of multiple dimensioned details fusion

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20150408

RJ01 Rejection of invention patent application after publication