CN110245666B - Multi-target interval value fuzzy clustering image segmentation method based on dual-membership-degree driving - Google Patents
Multi-target interval value fuzzy clustering image segmentation method based on dual-membership-degree driving Download PDFInfo
- Publication number
- CN110245666B CN110245666B CN201910495205.4A CN201910495205A CN110245666B CN 110245666 B CN110245666 B CN 110245666B CN 201910495205 A CN201910495205 A CN 201910495205A CN 110245666 B CN110245666 B CN 110245666B
- Authority
- CN
- China
- Prior art keywords
- interval value
- membership
- image
- local
- function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 90
- 238000003709 image segmentation Methods 0.000 title claims abstract description 16
- 210000000349 chromosome Anatomy 0.000 claims abstract description 42
- 230000011218 segmentation Effects 0.000 claims abstract description 41
- 230000009977 dual effect Effects 0.000 claims abstract description 7
- 239000011159 matrix material Substances 0.000 claims abstract description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 28
- 239000006002 Pepper Substances 0.000 claims description 21
- 238000005457 optimization Methods 0.000 abstract description 3
- 230000035945 sensitivity Effects 0.000 abstract description 3
- 230000000694 effects Effects 0.000 description 8
- 238000004088 simulation Methods 0.000 description 8
- 235000002566 Capsicum Nutrition 0.000 description 3
- 241000722363 Piper Species 0.000 description 3
- 235000016761 Piper aduncum Nutrition 0.000 description 3
- 235000017804 Piper guineense Nutrition 0.000 description 3
- 235000008184 Piper nigrum Nutrition 0.000 description 3
- 150000003839 salts Chemical class 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000001629 suppression Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23211—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with adaptive number of clusters
-
- 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]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention disclosesA multi-target interval value fuzzy clustering image segmentation method based on dual membership degree driving is provided, which mainly solves the problems of sensitivity to noise and easy falling into local optimum in image segmentation, and has the scheme that: inputting an image to be segmented and setting an initial parameter value; constructing an interval value blurred image; constructing a double-membership-driven global interval value fuzzy compactness function JLNAnd dual membership driven interval value fuzzy separability function SLNPerforming multi-objective evolution on the two objective functions to obtain a non-dominated solution set P; calculating an interval value selection solution index W index driven by double membership degrees, and selecting an optimal chromosome from the non-dominated solution set P by using the index to decode the optimal chromosome to obtain an optimal clustering center; and updating the joint membership matrix by using the optimal clustering center, and obtaining a classification result of the pixel points according to the maximum membership principle. The invention can effectively inhibit noise, prevent local optimization, improve the segmentation accuracy and can be used for identifying natural images.
Description
Technical Field
The invention belongs to the field of image processing, and particularly relates to a fuzzy clustering image segmentation method which can be used for identifying natural images.
Background
Image segmentation, as the name implies, is to segment an image into several portions that have characteristics and do not intersect each other, which is a key step from image processing to image analysis. In recent years, researches on image segmentation methods have been receiving high attention from researchers, including a clustering-based method, a threshold-based method, an edge-based method, and a region-based method, wherein the clustering-based method is the focus of the researches. Clustering, as the name implies, is the process of dividing elements in a collection into multiple classes, where similar objects are classified into one class and different objects are classified into different classes. Common clustering methods include a K-means clustering algorithm, a fuzzy clustering algorithm, a spectral clustering algorithm, a hierarchical clustering algorithm and the like, wherein the fuzzy clustering algorithm arouses the interest of researchers.
However, when the conventional fuzzy clustering algorithm, such as the image fuzzy clustering segmentation method based on the two-dimensional histogram proposed by liujianzhuang in 1992, is applied to image segmentation, there are two problems: the first problem is that the algorithm is sensitive to initial values and tends to fall into local optima. In order to solve the problem, in 2019, Majdi introduces an evolutionary algorithm into a fuzzy clustering algorithm, and provides a fuzzy clustering image segmentation algorithm based on the evolutionary algorithm. The second problem is that the traditional fuzzy clustering algorithm is sensitive to noise and has poor effect on processing images containing noise, so that Zhang introduces spatial information into the fuzzy clustering algorithm in 2018, and provides enhanced spatial constraint remote sensing image segmentation based on the fuzzy local double-neighborhood information C-mean clustering algorithm, so that the effects of removing noise and completely segmenting images are achieved. However, when the algorithms are applied to image segmentation, only a single objective function is considered, and in the real production life, the single objective function cannot meet the requirements of users in various aspects, and in 2011, Mukhopadhyay proposes a fuzzy clustering algorithm based on multi-objective evolution to be applied to segmentation of brain images, and the algorithm optimizes two fitness functions of connectivity and total deviation at the same time, and coordinates and balances among a plurality of objective functions, so that all objective functions are optimized as far as possible. However, the above algorithms all use single-valued data, and cannot better mine the uncertainty of the data. Therefore, De Carvalho provides an interval value fuzzy C-mean IVFCM algorithm, the algorithm replaces single-value data with interval value data, uncertainty of data can be better mined, but the algorithm still has the problems of sensitivity to noise, easiness in falling into local optimization and consideration of only a single objective function, and cannot meet various requirements of users.
Disclosure of Invention
The invention aims to provide a multi-target interval value fuzzy clustering image segmentation method based on dual-membership driving, aiming at overcoming the defects of the prior art, so as to reduce the sensitivity to noise, avoid falling into local optimization, consider the image segmentation problem from multiple aspects and improve the segmentation accuracy.
In order to achieve the above object, the technique of the present invention comprises the steps of:
(1) inputting an image to be segmented, and converting the image to be segmented into a gray image if the image to be segmented is a color image;
(2) setting parameters: setting the population size as 100, the iteration times as 50, the maximum class number as 10 and the variation probability as 0.1, and for Gaussian noisy images, limiting the parameters of local spatial informationAnd the non-local spatial information restriction item parameter psi is 6 and 16 respectively, and for the salt-pepper noisy image, the two parameters are set to 16 and 1 respectively;
(3) method for constructing interval value membership of gray level image by using Gaussian fuzzy numberAnd constructing interval value non-membership degree of gray level image by using m-fuzzy compensation operatorAccording to the degree of membership u of the interval valueI(δi) Sum interval value non-membership vI(δi) Obtaining interval value fuzzy imageWherein: deltaiAre the points of pixels that are to be imaged,u(δi) Andrespectively a left end point and a right end point of the interval value membership degree,v(δi) Andthe left end point and the right end point of the interval value non-membership degree are respectively;
(4) respectively constructing local membership function eta by using local spatial information and non-local spatial informationkiAnd a non-local membership function uki:
Wherein m-2 represents a fuzzy weighting index, Y represents the number of classes,the pixel points of the interval value are represented,represents the clustering center of the interval values, d (-) represents the Euclidean distance,to representLocal spatial information of, NiRepresentsM is the total number of pixels in the neighborhood, a probability functionNkiIs thatThe number of pixels in the neighborhood that belong to the kth class,obtained by iterative non-local mean algorithmPsi is a non-local spatial information restriction item parameter;
(5) calculating a joint membership function Z according to the result of (4)kiAnd designing a double-membership-driven global interval value fuzzy compactness function JLNAnd dual membership driven interval value fuzzy separability function SLN:
Wherein n is the total number of pixel points, xi 1 and ζ 2 are respectively a local membership function limiting item parameter and a non-local membership function limiting item parameter,andare all interval value clustering centers, ηkiIs thatIs subject toOf local membership function ukiIs thatIs subject toIs determined by the non-local membership function of (a),is a local spatial information restriction item parameter, # is a non-local spatial information restriction item parameter, ZpqIs thatIs subject toA joint membership function of (a);
(6) performing multi-objective evolution on the two objective functions obtained in the step (5) to obtain a final group of interval value non-dominated solution sets P;
(7) combining the interval value blurred image obtained in the step (3)Constructing a dual-membership-driven interval value selection solution index W by the pixel points, selecting an optimal chromosome from the interval value non-dominated solution set P obtained in the step (6) by using the index W, and taking an interval value clustering center coded on the chromosome as an optimal clustering center;
(8) updating the joint membership matrix by using the optimal clustering center obtained in the step (7)And classifying the pixel points according to the maximum membership principle to obtain a segmentation result of the gray level image.
Compared with the prior art, the invention has the following beneficial technical effects:
first, the invention introduces multi-objective evolution into the interval value fuzzy clustering algorithm, solves the problem that the clustering algorithm of a single clustering criterion is easy to fall into local optimum, and the clustering result can meet various requirements of users.
Secondly, the local space information and the non-local space information of the image are respectively utilized to construct a local membership function and a non-local membership function, two interval value fuzzy fitness functions driven by double membership are designed by fusing the two membership functions, and the influence of noise on the image segmentation process is overcome.
Thirdly, the invention automatically adjusts the cross probability according to the fitness function value, improves the searching performance and the searching goodness and leads the segmentation effect to be more ideal.
Drawings
FIG. 1 is a flow chart of an implementation of the method of the present invention;
FIG. 2 is a comparison of results of a simulation segmentation of an image numbered 238011 in a Berkeley image database using the present invention and a prior art method;
fig. 3 is a comparison graph of the results of simulation segmentation of the image with number 3096 in the Berkeley image database using the present invention and the existing method.
Detailed Description
The following further describes the practice and effects of the invention in detail:
referring to fig. 1, the implementation steps of the present invention are as follows:
step 1: inputting an image to be segmented and setting initial parameter values.
Inputting all images to be segmented, and converting the images to be segmented into gray images if the images to be segmented are color images;
setting the population size as 100, the iteration times as 50, the maximum class number as 10 and the variation probability as 0.1, and for Gaussian noisy images, limiting the parameters of local spatial informationAnd the non-local spatial information restriction parameter psi is 6 and 16, respectively, which are set to 16 and 1, respectively, for noisy salt-and-pepper images.
2.1) construction of the Pixel Point δ by means of Gauss fuzzy numberiDegree of membership u (delta)i,ε,σ):
Wherein epsilon is 150, sigma is 120;
u(δi) Is the left end point of the interval value membership degree,the right end points of the interval value membership degrees are respectively expressed as follows:
wherein e is 2;
2.3) constructing the pixel point delta by using the m-fuzzy compensation operatoriIs not degree of membership v (delta)i) The formula is as follows:
wherein w is 0.2;
v(δi) The left end point of interval value non-membership degree,the right end points of the interval value non-membership degree are respectively expressed as follows:
v(δi)=v(δi)+απ(δi),
and step 3: and performing multi-objective evolution on the two objective functions to obtain a non-dominated solution set P.
3.1) population initialization: coding each chromosome into 2-10 clustering centers by using a variable length coding method, randomly generating 100 parent chromosomes to form a population, and setting the current iteration number to be 1;
3.2) calculating two objective function values of each chromosome in the population;
3.2.1) calculating the clustering center of interval valuesAnd interval value pixel pointEuclidean distance of
Wherein N isiRepresentsThe 3 x 3 neighborhood window of (a),is thatNeighborhood window NiPixel points with interval values in the neighborhood window, wherein M is the number of the total pixel points in the neighborhood window;
3.2.3) calculating the clustering center of interval valuesAnd local spatial informationEuropean distance of
3.2.5) calculating a local membership function eta based on the results of 3.2.3) and 3.2.4)ki:
Wherein d (-) is Euclidean distance, fkiIs a likelihood function, m-2 represents a fuzzy weighting index, and Y represents the number of categories;
3.2.6) calculating interval value pixel points by using iterative non-local mean algorithmNon-local spatial information of
Wherein t is the iteration number of the iterative non-local mean algorithm, Wi rRepresenting pixel points by interval valuesA search window of size 21 x 21 centered,is shown inInterval-value pixel point in centered neighborhood windowThe formula is as follows:
wherein h is a control weight functionThe parameters of the attenuation are such that,is of GaussWeighted Euclidean distance, χ is the standard deviation of the Gaussian kernel, N (-) is the similarity window, ZiTo normalize the constants, the formula is as follows:
3.2.7) calculating the clustering center of the interval value according to the result of 3.2.6)And non-local spatial informationEuropean distance of
3.2.8) calculating a non-local membership function u) based on the results of 3.2.3) and 3.2.7)ki:
Wherein n is the total number of pixel points, and psi is a non-local spatial information restriction item parameter;
3.2.9) calculating a double-membership-driven global interval value fuzzy compactness function J) according to the results of 3.2.4), 3.2.5) and 3.2.8)LNThe value of (c):
3.2.10) calculating a joint membership function Z based on the results of 3.2.5) and 3.2.8)ki:
Where ξ ═ 1 and ζ ═ 2 are the local membership function constraint term parameter and the non-local membership function constraint term parameter, respectively;
3.2.12) according to the results of 3.2.10) and 3.2.11), calculating a dual-membership-driven interval value fuzzy separability function SLNThe value of (c):
wherein,andare all interval value clustering centers, ZpqIs thatIs subject toA joint membership function of (a);
3.2.13) encoding the two objective function values sequentially onto the chromosome;
3.3) carrying out non-dominated sorting on the parent chromosomes by using the target numerical value, and sequentially coding the obtained sequence value and the obtained crowding distance on the chromosomes;
3.4) selecting 50 chromosomes from the population as parents by using a binary tournament method;
3.5) pairing of parent chromosomesAndperforming a Laplace crossover to generate offspring chromosomes, wherein,representing a code on the parent chromosome P1The upper tau interval values cluster the center,representing a code on the parent chromosome P2The upper gamma interval value clustering centers;
3.5.1) calculating the crossover probability pc:
Wherein p iscmax0.9 is the maximum crossover probability, pcmin0.5 is the minimum cross probability, two indices σ1And σ2The calculation formula of (a) is as follows:
wherein,andrespectively representing two objective function values of the first parent chromosome,andrespectively representing two objective function values of the second parent chromosome,andrespectively, the average values of the two objective function values of all chromosomes in the population;
3.5.2) calculating offspring chromosome C(1)Upper interval value clustering centerAnd progeny chromosome C(2)Upper interval value clustering centerThe formula is as follows:
wherein, | - | represents an absolute value,andare all the interval value cluster centers on the first parent chromosome,andis the clustering center of the interval value on the second parent chromosome, and is a random number obeying Laplace distribution, and the calculation formula is as follows:
wherein rand is a random number between 0 and 1 which accords with normal distribution, and the values of the positioning parameter a and the scale parameter b are both 0.5;
3.5.3) obtaining the offspring chromosome C according to the result of 3.5.2)(1)And progeny chromosome C(2)The formula is as follows:
wherein,is encoded in offspring chromosome C(1)The upper theta interval values cluster the center,is encoded in offspring chromosome C(2)P interval value cluster centers;
3.6) selecting the k-th interval value clustering centerD-th dimension value c ofkdCarrying out interval value fuzzy Poisson variation, and obtaining the gene locus value after variationThe following were used:
wherein u (c)kd) Is ckdDegree of membership of, pi (c)kd) Is ckdHesitation degree of;
3.7) circulating from 3.2) to 3.6), and stopping iteration when the maximum iteration times is reached to obtain a filial generation population;
3.8) selecting 100 chromosomes from the parent population and the child population obtained in the step 3.4) and the step 3.7) by using an elite strategy, decoding the 100 chromosomes to obtain an interval value clustering center, wherein the interval value clustering center is the final non-dominated solution set P.
And 4, step 4: the optimal chromosome is selected.
4.1) calculating a double-membership-driven interval value solution index W:
wherein,is the center of the interval value cluster and,is an interval value pixel point, d (-) represents the Euclidean distance, m 2 represents the fuzzy weighting index,is a regionMean value clustering centerAverage value of (2), ZkiIs a joint membership function, nkRepresenting the total number of pixel points belonging to the kth class;
4.2) selecting a solution index W by using an interval value driven by double membership degrees, selecting an optimal chromosome from the interval value non-dominated solution set P obtained in the step 3, and taking an interval value clustering center coded on the chromosome as an optimal clustering center.
And 5: and outputting the result of the image segmentation.
Updating the joint membership matrix by using the optimal clustering center obtained in the step 4And classifying the pixel points according to the maximum membership principle to obtain a segmentation result of the gray level image.
The technical effects of the invention are further explained by combining simulation experiments as follows:
1. simulation conditions are as follows:
the simulation experiment was performed in the software environment of computer Inter (R) core (TM) i5-3230M 2.60GHZ CPU, 4G memory, MATLAB R2014 a.
2. Simulation content:
simulation 1, selecting an image with the number of 238011 in a Berkeley image database, and segmenting the image by using the method of the invention and the existing FCM method, FCM _ S1 method, IFCM method, IT2FCM method, IVFCM method, MOVGA method and MSFCA method respectively, wherein the result is shown in FIG. 2, wherein:
2(a) is an original image of the 238011 image;
2(b) is a standard segmentation map of the 238011 image;
2(c) is a salt-and-pepper noisy image of 238011 images, with a noise intensity of 0.05;
2(d) is the segmentation result of the salt-pepper noisy image of 238011 image by the existing FCM method;
2(e) is the segmentation result of the salt-pepper noisy image of 238011 image by the existing FCM _ S1 method;
2(f) is the segmentation result of the salt-pepper noisy image of the 238011 image by the existing IFCM method;
2(g) is the segmentation result of the salt-pepper noisy image of 238011 image by the existing IT2FCM method;
2(h) is the segmentation result of the salt-pepper noisy image of 238011 image by the existing IVFCM method;
2(i) is the segmentation result of the noisy pepper salt image of 238011 image by the existing MOVGA method;
2(j) is the segmentation result of the salt-pepper noisy image of 238011 image by the existing MSFCA method;
2(k) is the segmentation result of the salt-pepper noisy image of 238011 image by the present invention;
2(l) is a gaussian noise-containing image of 238011 images, with a noise intensity of 0.005;
2(m) is the result of segmentation of the gaussian noisy image of 238011 images using the existing FCM method;
2(n) is the result of segmentation of a gaussian noisy image of 238011 images using the existing FCM _ S1 method;
2(o) is a result of segmenting a gaussian noisy image of the 238011 image by the existing IFCM method;
2(p) is the result of segmentation of the gaussian noisy image of the 238011 image using the existing IT2FCM method;
2(q) is a result of segmenting a gaussian noisy image of the 238011 image by the existing IVFCM method;
2(r) is the result of segmentation of the gaussian noisy image of the 238011 image using the existing MOVGA method;
2(s) is a result of segmenting a gaussian noisy image of the 238011 image by using the existing MSFCA method;
2(t) is the result of the segmentation of the gaussian noisy image of 238011 images using the present invention;
as can be seen from FIG. 2, the background and the target can be separated on the basis of noise suppression, so the segmentation effect of the method on the noisy image is better than that of the existing FCM method, FCM _ S1 method, IFCM method, IT2FCM method, IVFCM method, MOVGA method and MSFCA method.
Simulation 2, selecting an image with the number of 3096 in a Berkeley image database, and segmenting the image by using the method of the present invention and the existing FCM method, FCM _ S1 method, IFCM method, IT2FCM method, IVFCM method, MOVGA method and MSFCA method, respectively, wherein the result is shown in FIG. 3, wherein:
3(a) is an original image of 3096 image;
3(b) is a standard segmentation map of the 3096 image;
3(c) is a noisy pepper salt image of 3096 image with a noise intensity of 0.05;
3(d) is the segmentation result of the salt-pepper noisy image of 3096 image by the existing FCM method;
3(e) is the segmentation result of the salt-pepper noisy image of 3096 image by the existing FCM _ S1 method;
3(f) is the segmentation result of the salt-pepper noisy image of 3096 image by the existing IFCM method;
3(g) is the segmentation result of the salt-pepper noisy image of 3096 image by the existing IT2FCM method;
3(h) is the segmentation result of the salt-pepper noisy image of 3096 image by the existing IVFCM method;
3(i) is the segmentation result of the pepper salt noisy image of 3096 image by the existing MOVGA method;
3(j) is the segmentation result of the salt-pepper noisy image of 3096 image by the existing MSFCA method;
3(k) is the segmentation result of the salt-pepper noisy image of 3096 image by the invention;
3(l) is a Gaussian noisy image of image 3096, with a noise intensity of 0.005;
3(m) is the result of segmentation of the gaussian noisy image of 3096 images using the existing FCM method;
3(n) is the result of segmentation of the gaussian noisy image of 3096 image using the existing FCM _ S1 method;
3(o) is a result of segmenting the gaussian noisy image of 3096 image by the existing IFCM method;
3(p) is the result of segmentation of the gaussian noisy image of 3096 images using the existing IT2FCM method;
3(q) is a result of segmenting the gaussian noisy image of 3096 image by the existing IVFCM method;
3(r) is the result of segmentation of the gaussian noisy image of 3096 image by the existing MOVGA method;
3(s) is the result of segmentation of the gaussian noisy image of 3096 images using the existing MSFCA method;
3(t) is the result of the segmentation of the Gaussian noisy image of 3096 image by the present invention;
as can be seen from fig. 3, the present invention can separate the background and the object based on the noise suppression, and the segmentation effect on the noisy image is better than that of the existing FCM method, FCM _ S1 method, IFCM method, IT2FCM method, IVFCM method, MOVGA method and MSFCA method.
Claims (6)
1. A multi-target interval value fuzzy clustering image segmentation method based on dual membership degree driving is characterized by comprising the following steps:
(1) inputting an image to be segmented, and converting the image to be segmented into a gray image if the image to be segmented is a color image;
(2) setting parameters: setting the population size as 100, the iteration times as 50, the maximum class number as 10 and the variation probability as 0.1, and for Gaussian noisy images, limiting the parameters of local spatial informationAnd the non-local spatial information restriction item parameter psi is 6 and 16 respectively, and for the salt-pepper noisy image, the two parameters are set to 16 and 1 respectively;
(3) method for constructing interval value membership of gray level image by using Gaussian fuzzy numberAnd constructing interval value non-membership degree of gray level image by using m-fuzzy compensation operatorAccording to the degree of membership u of the interval valueI(δi) Sum interval value non-membership vI(δi) Obtaining interval value fuzzy imageWherein: deltaiAre the points of pixels that are to be imaged,u(δi) Andrespectively a left end point and a right end point of the interval value membership degree,v(δi) Andthe left end point and the right end point of the interval value non-membership degree are respectively;
(4) respectively constructing local membership function eta by using local spatial information and non-local spatial informationkiAnd a non-local membership function uki:
Wherein m-2 represents a fuzzy weighting index, Y represents the number of classes,the pixel points of the interval value are represented,an interval value cluster center representing the kth class,an interval value cluster center representing the l-th class; d (-) represents the Euclidean distance,to representThe local spatial information of (a) the local spatial information,is thatNeighborhood window NiPixel points with inner interval values; n is a radical ofiRepresents3 x 3 neighborhood window of (a), M is the total number of pixels in the neighborhood, the probability function of pixel i to class kNkiIs thatNumber of pixels in the neighborhood that belong to the kth class, fliA likelihood function representing pixel i for class i;obtained by iterative non-local mean algorithmPsi is a non-local spatial information restriction item parameter;
(5) calculating a joint membership function Z according to the result of (4)kiAnd designing a double-membership-driven global interval value fuzzy compactness function JLNAnd dual membership driven interval value fuzzy separability function SLN:
Wherein n is the total number of pixel points, xi 1 and ζ 2 are respectively a local membership function limiting item parameter and a non-local membership function limiting item parameter,andare all interval value clustering centers, ηkiIs thatIs subject toOf local membership function ukiIs thatIs subject toIs determined by the non-local membership function of (a),is a local spatial information restriction item parameter, # is a non-local spatial information restriction item parameter, ZpqIs thatIs subject toA joint membership function of (a);
(6) for two objective functions J obtained in (5)LNAnd SLNPerforming multi-objective evolution to obtain a final group of interval value non-dominated solution sets P;
(7) combining the interval value blurred image obtained in the step (3)Constructing a dual-membership-driven interval value selection solution index W by the pixel points, selecting an optimal chromosome from the interval value non-dominated solution set P obtained in the step (6) by using the index W, and taking an interval value clustering center coded on the chromosome as an optimal clustering center;
5. the method of claim 1, wherein the multi-objective evolution of the two objective functions in (6) is implemented as follows:
(6a) population initialization: encoding each chromosome into 2-10 clustering centers by using a variable length encoding method, randomly generating 100 parent chromosomes to form a population, and setting the current iteration number to be 1;
(6b) two objective function values for each chromosome in the population are calculated:
(6b1) calculating interval value clustering centersAnd interval value pixel pointEuclidean distance of
(6b2) Calculating interval value pixel pointsLocal spatial information ofWherein N isiRepresentsThe 3 x 3 neighborhood window of (a),is thatNeighborhood window NiPixel points with interval values in the neighborhood window, wherein M is the number of the total pixel points in the neighborhood window;
(6b3) calculating interval value clustering centersAnd local spatial informationEuropean distance of
(6b4) Computing a likelihood functionWherein N iskiIs thatNeighborhood window NiThe number of pixels belonging to the kth class in the neighborhood window, and M is the number of total pixels in the neighborhood window;
(6b5) will Euclidean distanceLocal spatial informationAnd a probability function fkiSubstituted type<1>Calculating a local membership function etaki;
(6b6) Calculating interval value pixel points by using iterative non-local mean algorithmNon-local spatial information of
(6b7) Calculating interval value clustering centersAnd non-local spatial informationEuropean distance of
(6b8) Will Euclidean distanceAnd Euclidean distanceSubstituted type<2>Calculating a non-local membership function uki;
(6b9) A function η of local membershipkiNon-local membership function ukiAnd a probability function fkiSubstituted type<4>Calculating a fuzzy compactness function J of the global interval value driven by the double membership degreesLNA value of (d);
(6b10) a function η of local membershipkiAnd a non-local membership function ukiSubstituted type<3>Calculating a joint membership function Zki;
(6b12) Will Euclidean distanceAnd joint membership function ZkiSubstituted type<5>Calculating interval value of dual membership degree driveFuzzy separability function SLNA value of (d);
(6b13) encoding the two objective function values onto a chromosome in sequence;
(6c) carrying out non-dominated sorting on the parent chromosomes by using the target numerical values, and sequentially encoding the obtained sequence values and the crowding distances to the chromosomes;
(6d) selecting 50 chromosomes from the population as parents by using a binary tournament method;
(6e) carrying out Laplace crossing and interval value fuzzy Poisson variation on the parent chromosome to generate offspring;
(6f) looping (6b) through (6e), and stopping iteration when the maximum iteration number is reached;
(6g) and (3) selecting 100 chromosomes from the parent population and the child population obtained in the step (6d) and the step (6e) by using an elite strategy, and decoding the 100 chromosomes to obtain an interval value clustering center, wherein the interval value clustering center is the final non-dominated solution set P.
6. The method according to claim 1, wherein the interval value solution index W of the dual membership degree driving constructed in (7) is expressed as follows:
wherein,is the center of the interval value cluster and,is an interval value pixel point, d (-) represents an Euclidean distance, m 2 represents a fuzzy weighting index, Y is the maximum category number, n is the total number of the pixel points,is interval value clustering centerAverage value of (2), ZkiIs a joint membership function, nkRepresenting the total number of pixels belonging to the kth class.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910495205.4A CN110245666B (en) | 2019-06-10 | 2019-06-10 | Multi-target interval value fuzzy clustering image segmentation method based on dual-membership-degree driving |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910495205.4A CN110245666B (en) | 2019-06-10 | 2019-06-10 | Multi-target interval value fuzzy clustering image segmentation method based on dual-membership-degree driving |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110245666A CN110245666A (en) | 2019-09-17 |
CN110245666B true CN110245666B (en) | 2021-06-08 |
Family
ID=67886266
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910495205.4A Active CN110245666B (en) | 2019-06-10 | 2019-06-10 | Multi-target interval value fuzzy clustering image segmentation method based on dual-membership-degree driving |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110245666B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN113723449B (en) * | 2021-07-16 | 2023-08-11 | 西安邮电大学 | Agent-driven multi-target evolutionary fuzzy clustering method based on preference information |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107392921A (en) * | 2017-07-14 | 2017-11-24 | 西安邮电大学 | A kind of semi-supervised multi-object clustering image partition method based on Chebyshev's distance |
Family Cites Families (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103473786B (en) * | 2013-10-13 | 2017-01-18 | 西安电子科技大学 | Gray level image segmentation method based on multi-objective fuzzy clustering |
CN105678798A (en) * | 2016-03-10 | 2016-06-15 | 西安邮电大学 | Multi-target fuzzy clustering image segmentation method combining local spatial information |
CN106127794A (en) * | 2016-07-29 | 2016-11-16 | 天津大学 | Based on probability FCM algorithm MRI tumor image dividing method and system |
CN106373129A (en) * | 2016-09-20 | 2017-02-01 | 辽宁工程技术大学 | FCM remote sensing image segmentation method based on dual degree of membership |
CN107977952A (en) * | 2016-10-21 | 2018-05-01 | 冯原 | Medical image cutting method and device |
KR101866522B1 (en) * | 2016-12-16 | 2018-06-12 | 인천대학교 산학협력단 | Object clustering method for image segmentation |
CN107220978B (en) * | 2017-06-09 | 2020-06-09 | 西安邮电大学 | Multi-target threshold image segmentation method fusing interval fuzzy information and statistical information |
CN107730516B (en) * | 2017-10-16 | 2020-03-24 | 江南大学 | Brain MR image segmentation method based on fuzzy clustering |
CN109544570B (en) * | 2018-11-30 | 2022-08-02 | 南京邮电大学 | Brain MRI image segmentation method based on fuzzy multi-threshold and regional information |
-
2019
- 2019-06-10 CN CN201910495205.4A patent/CN110245666B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107392921A (en) * | 2017-07-14 | 2017-11-24 | 西安邮电大学 | A kind of semi-supervised multi-object clustering image partition method based on Chebyshev's distance |
Also Published As
Publication number | Publication date |
---|---|
CN110245666A (en) | 2019-09-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110689086B (en) | Semi-supervised high-resolution remote sensing image scene classification method based on generating countermeasure network | |
Pare et al. | Image segmentation using multilevel thresholding: a research review | |
Mukhopadhyay et al. | A survey of multiobjective evolutionary clustering | |
Zhao et al. | A multiobjective spatial fuzzy clustering algorithm for image segmentation | |
Nakane et al. | Application of evolutionary and swarm optimization in computer vision: a literature survey | |
Bi et al. | A survey on evolutionary computation for computer vision and image analysis: Past, present, and future trends | |
CN110929679B (en) | GAN-based unsupervised self-adaptive pedestrian re-identification method | |
Fang et al. | Confident learning-based domain adaptation for hyperspectral image classification | |
Zhao et al. | Intuitionistic fuzzy set approach to multi-objective evolutionary clustering with multiple spatial information for image segmentation | |
CN113408605A (en) | Hyperspectral image semi-supervised classification method based on small sample learning | |
CN113469270B (en) | Semi-supervised intuitive clustering method based on decomposition multi-target differential evolution superpixel | |
Chin-Wei et al. | Multiobjective optimization approaches in image segmentation–the directions and challenges | |
CN111611972B (en) | Crop leaf type identification method based on multi-view multi-task integrated learning | |
Zhang et al. | Sparse feature clustering network for unsupervised SAR image change detection | |
CN110245666B (en) | Multi-target interval value fuzzy clustering image segmentation method based on dual-membership-degree driving | |
Ramadas et al. | Metaheuristics and data clustering | |
CN109902808B (en) | Method for optimizing convolutional neural network based on floating point digital variation genetic algorithm | |
CN114118369B (en) | Image classification convolutional neural network design method based on group intelligent optimization | |
CN109840512A (en) | A kind of Facial action unit recognition methods and identification device | |
Shin et al. | Incremental deep learning for robust object detection in unknown cluttered environments | |
CN112509017B (en) | Remote sensing image change detection method based on learnable differential algorithm | |
Huang et al. | Self-supervised graph attention networks for deep weighted multi-view clustering | |
CN113723449A (en) | Preference information-based agent-driven multi-objective evolutionary fuzzy clustering method | |
CN113222072A (en) | Lung X-ray image classification method based on K-means clustering and GAN | |
CN115905855A (en) | Improved meta-learning algorithm MG-copy |
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 |