CN105787935A - Fuzzy cluster SAR image segmentation method based on Gamma distribution - Google Patents
Fuzzy cluster SAR image segmentation method based on Gamma distribution Download PDFInfo
- Publication number
- CN105787935A CN105787935A CN201610096321.5A CN201610096321A CN105787935A CN 105787935 A CN105787935 A CN 105787935A CN 201610096321 A CN201610096321 A CN 201610096321A CN 105787935 A CN105787935 A CN 105787935A
- Authority
- CN
- China
- Prior art keywords
- function
- sar image
- fcm
- fuzzy
- gamma distribution
- 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
Links
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Image Processing (AREA)
Abstract
The invention provides a fuzzy cluster SAR image segmentation method based on Gamma distribution. The method comprises the following steps: taking ray values of pixels of an SAR image to be segmented as sampling points, and constructing an FCM object function with a Gamma distribution function; determining solution formulas of parameters in the FCM object function; carrying out fuzzy clustering on the SAR image to be segmented by use of the FCM object function with the Gamma distribution function to obtain fuzzy membership degree matrixes when the gray value of each pixel of the SAR image to be segmented belongs to each ground object type; and performing defuzzification on the fuzzy membership degree matrixes according to a maximum membership degree principle so as to realize SAR image segmentation. According to the invention, a dissimilarity measure from pixel points to clusters is described by use of negative logarithms of a Gamma distribution probability density function, through accurate fitting of SAR image distribution features, the influence exerted by noise in the SAR image on a segmentation result is further overcome, the segmentation precision is improved, and the fitting and segmentation precision of the SAR image is effectively improved.
Description
Technical field
The invention belongs to image processing field, be specifically related to a kind of fuzzy clustering SAR image segmentation method based on Gamma distribution.
Background technology
Synthetic aperture radar (SyntheticApertureRadar, SAR) image segmentation is the key technology during SAR image processes, the precision of its segmentation directly affects follow-up interpretation and processes, and therefore, the research for SAR image cutting techniques is significant.But, due to the factors impact such as image-forming principle and human vision is runed counter to, high-resolution brings the details become apparent from and imaging process to cause intrinsic speckle noise of SAR, it is difficult to obtain high-precision segmentation result.
At present, the image partition method based on SAR can be largely classified into: threshold method, statistical method and clustering method.Method And Principle based on threshold value is simple, but is difficult to the SAR image segmentation finding rational threshold value and the method can only carry out limited class;Statistics-Based Method can model noise, has good noise immunity, but it solves difficulty;Its maximum feature of method based on cluster is modeling and solves easily, but can not for noise modeling, therefore to noise-sensitive.In said method, clustering method particularly fuzzy clustering method is most widely used due to features such as principle is simple, good, the fast convergence rates of algorithm stability, such as the most frequently used SAR image segmentation method based on FuzzyC-means (FCM), but, owing to the method Euclidean distance definition non-similarity is estimated and does not account for neighborhood territory pixel effect, therefore the speckle noise in SAR image and exceptional value is sensitive, therefore based on the segmentation result of the SAR image of traditional F CM exists substantial amounts of noise.In order to overcome above-mentioned dividing method to noise-sensitive problem, (the Zhao Xuemei such as Zhao Xuemei, Li Yu, Zhao Quanhua. based on the fuzzy clustering high-resolution remote sensing image partitioning algorithm of Hidden Markov gaussian random field model: electronics and information journal, 2014,36 (11): 2730-2736) in conjunction with statistical method and clustering method characteristic, Hidden Markov random field and Gauss regression model is utilized to set up Characteristic Field and the neighborhood relationships of label field, and realize SAR image segmentation by FCM method, thus efficiently solve classical way segmentation problem inaccurate, that point pixel is many by mistake.But, the method thinks that SAR image is distributed as Gauss distribution and does not correspond with the SAR image obedience Gamma feature being distributed, it is impossible to solve the segmentation problem of all SAR image.
Summary of the invention
For the deficiencies in the prior art, the present invention proposes a kind of fuzzy clustering SAR image segmentation method based on Gamma distribution.
The technical scheme is that
A kind of fuzzy clustering SAR image segmentation method based on Gamma distribution, comprises the following steps:
Step 1: read SAR image to be split;
Step 2: using the gray value of each for SAR image to be split pixel as sample point, build the FCM object function with Gamma distribution function, this object function is to overcome noise in SAR image on the impact of segmentation result for target, obey the feature of Gamma distribution according to SAR image intensity profile, estimate as object function non-similarity using the negative logarithm of Gamma distribution function;The prior probability in FCM object function is determined by the potential-energy function of defined label field.This FCM objective function parameters includes the form parameter in fuzzy factor, fuzzy membership function, prior probability, Gamma distribution function and scale parameter;
Step 3: determine the solution formula of each parameter in FCM object function;
Step 4: utilizing the FCM object function with Gamma distribution function that SAR image to be split is carried out fuzzy clustering, the gray value obtaining each pixel of SAR image to be split belongs to the fuzzy membership matrix of each atural object classification;
Step 5: above-mentioned fuzzy membership matrix is carried out anti fuzzy method by maximum membership grade principle, it is achieved SAR image is split.
Specifically comprising the following steps that of described step 3
Step 3.1: determine the prior probability formula in FCM object function;
Step 3.1.1: definition potential-energy function: it is 0 that the gray value of neighborhood territory pixel and the gray value of center pixel belong to potential energy during same atural object classification, and otherwise potential energy is 1;
Step 3.1.2: define the prior probability in FCM object function according to potential-energy function and be normalized;
Step 3.2: determine the numerical solution formula of fuzzy membership: add Lagrange multiplier in FCM object function, then the fuzzy membership function in FCM object function is sought local derviation, and make its first derivative equal to 0, obtain the fuzzy membership function with Lagrange coefficient, reelimination Lagrange coefficient;
Step 3.3: determine the numerical solution formula of form parameter in Gamma distribution function, scale parameter:
Determine the numerical solution formula of form parameter: form parameter is carried out Equivalent Infinitesimal replacement;
Determine the numerical solution formula of scale parameter: make FCM object function that the single order local derviation of form parameter is equal to zero.
Specifically comprising the following steps that of described step 4
Step 4.1: cluster centre number is set, initializes fuzzy membership matrix, fuzzy factor and iteration stopping condition;Atural object classification number in cluster centre number and SAR image to be split, fuzzy factor represents the confusion degree of SAR image segmentation result;
Step 4.2: random initializtion FCM target function value, the scale parameter of Gamma distribution function and form parameter;
Step 4.3: calculate the prior probability for constrained clustering yardstick in FCM object function;
Step 4.4: estimate new form parameter and scale parameter, and then try to achieve non-similarity and estimate;
Step 4.5: estimate according to the prior probability in FCM object function, non-similarity, the gray value calculating each pixel belongs to the fuzzy membership of each atural object classification;
Step 4.6: according to fuzzy membership, non-similarity is estimated, prior probability calculates current FCM target function value;
Step 4.7: if the difference of the FCM target function value of current FCM target function value and previous calculating is more than setting threshold value, using present Fuzzy degree of membership as initial fuzzy membership, return step 4.3;If the difference of the FCM target function value of current FCM target function value and previous calculating is less than setting threshold value, then iteration stopping;
Step 4.8: present Fuzzy subordinated-degree matrix is the optimum fuzzy membership matrix of each grey scale pixel value of SAR image to be split.
Beneficial effect:
The present invention utilizes Gamma to be distributed accurate description SAR image intensity distributions feature, proposes a kind of fuzzy clustering SAR image segmentation method based on Gamma distribution in conjunction with FCM method.The method utilize the negative logarithm of Gamma distribution probability density function describe pixel to cluster non-similarity estimate, the impact on segmentation result of the noise in SAR is overcome further by Accurate Curve-fitting SAR image distribution characteristics, thus improving segmentation precision, it is effectively increased matching and the segmentation precision of SAR image.Good stability of the present invention, fast convergence rate, the automatic interpretation for SAR image provides new approaches.
Accompanying drawing explanation
Fig. 1 is the fuzzy clustering SAR image segmentation method flow chart in the specific embodiment of the invention based on Gamma distribution;
Fig. 2 is the particular flow sheet of step 3 in the specific embodiment of the invention;
Fig. 3 is the particular flow sheet of step 4 in the specific embodiment of the invention;
Fig. 4 is the emulating image in the specific embodiment of the invention, and wherein (a) is template, and (b) is analog image, and (c) is true SAR image, by bright to being secretly followed successively by ice, ice-melt and water;
Fig. 5 applies the inventive method, existing FCM method and based on the split-run test to analog image of the FCM method of Gauss regression model in the specific embodiment of the invention, wherein (a) is standard FCM method segmentation result, b () is the FCM method segmentation result based on Gauss regression model, (c) is the inventive method segmentation result;
Fig. 6 applies the inventive method, existing FCM method and based on the split-run test to true SAR image of the FCM method of Gauss regression model in the specific embodiment of the invention, wherein (a) is standard FCM method segmentation result, b () is the FCM method segmentation result based on Gauss regression model, (c) is the inventive method segmentation result;
Fig. 7 is Gauss regression model and the corresponding true histogrammic fitting result chart of SAR image segmentation result in the specific embodiment of the invention;
Fig. 8 is Gamma distribution function and the corresponding true histogrammic fitting result chart of SAR image segmentation result in the specific embodiment of the invention.
Detailed description of the invention
Below in conjunction with accompanying drawing, the specific embodiment of the present invention is elaborated.
A kind of fuzzy clustering SAR image segmentation method based on Gamma distribution, as it is shown in figure 1, comprise the following steps:
Step 1: read SAR image to be split;
In present embodiment, define SAR image territory X={x to be split1,…,xi,…,xN, wherein, xiBeing that the gray scale of ith pixel is estimated, i is pixel index, and N is total pixel number, and X is 128 × 128 pixels, total pixel number N=16384.
Step 2: using the gray value of each for SAR image to be split pixel as sample point, build the FCM object function with Gamma distribution function, this object function is to overcome noise in SAR image on the impact of SAR image segmentation result for target, obey the feature of Gamma distribution according to SAR image intensity profile, estimate as object function non-similarity using the negative logarithm of Gamma distribution function;The prior probability in FCM object function is determined by the potential-energy function of defined label field.This FCM objective function parameters includes the form parameter in fuzzy factor, fuzzy membership function, prior probability, Gamma distribution and scale parameter;
This FCM object function determines the prior probability in object function by the potential-energy function of defined label field, to consider the neighborhood relationships impact on segmentation result of pixel.
The FCM object function that definition has Gamma distribution function is as follows:
Wherein, c is cluster centre number, uijFor fuzzy membership function, the gray value of expression ith pixel belongs to the subjection degree of j atural object classification, meets 0≤uij≤ 1 constraint, λ is fuzzy factor, represents the confusion degree of SAR image segmentation result, πijFor the variable prior probability of constrained clustering yardstick, for c × N matrix, πijFor posterior probability.
In the present embodiment, definition is estimated using minor function as the non-similarity of FCM object function:
Wherein αj, βjFor the characterising parameter of Gamma distribution function, αjFor form parameter, βjFor scale parameter.
Step 3: that determines each parameter in FCM object function solves mode;
As in figure 2 it is shown, specifically comprise the following steps that
Step 3.1: determine the prior probability formula in FCM object function;
Step 3.1.1: definition potential-energy function Vc: it is 0 that the gray value of neighborhood territory pixel and the gray value of center pixel belong to potential energy during same atural object classification, and otherwise potential energy is 1;
In the present embodiment, defining neighborhood territory pixel relation in label field, definition potential-energy function formula is as follows:
In the present embodiment, if L={l1,l2,…,lNIt is the label field of X, li∈ 1,2 ..., j ..., c} be ith pixel gray value belonging to the label of atural object classification, if being divided into 3 classes, then li∈ 1,2,3}, in 3 × 3 windows, define potential-energy function, liˊRepresent the label of atural object classification, l belonging to the gray value of neighborhood territory pixeliCentered by pixel gray value belonging to the label of atural object classification, when the gray value of neighborhood territory pixel and the gray value of center pixel have identical belonging to the label of atural object classification time, reach steady statue, potential energy is 0, and otherwise potential energy is 1, V in formulac∈{0,1,2,…,8}。
Step 3.1.2: define the prior probability π in FCM object function according to potential-energy functionijAnd be normalized;
Normalization πijFormula is as follows:
In above formula, μ is the parameter in energy function, μ=0.5 during enforcement;
Step 3.2: determine the numerical solution formula of fuzzy membership function: add Lagrange multiplier in FCM object function, then the fuzzy membership function in FCM object function is sought local derviation, and make its first derivative equal to 0, obtain the fuzzy membership function with Lagrange multiplier, reelimination Lagrange multiplier;
In the present embodiment, in order to solve FCM object function J, it is necessary to fuzzy membership function u in (1) formula of derivingij, non-similarity estimate dij;uijThe gray value of the expression ith pixel subjection degree to jth atural object classification, it is necessary to meet the following conditions;
In the present embodiment, in order to obtain fuzzy membership function, it is necessary to add Lagrange multiplier m in FCM object function, obtain following FCM object function:
Then the fuzzy membership function in FCM object function is asked local derviation, and makes its first derivative equal to 0, obtain the fuzzy membership function u containing mij:
Step (8) is brought into (5) formula, eliminates Lagrange multiplier m, obtain following fuzzy membership functions:
Step 3.3: determine the numerical solution formula of form parameter in Gamma distribution function, scale parameter;
In the present embodiment, make FCM object function that the single order local derviation of form parameter is equal to zero, it is determined that scale parameter βjNumerical solution formula as follows:
Abbreviation obtains following βjNumerical solution formula:
By form parameter is carried out Equivalent Infinitesimal replacement, it is determined that form parameter αjNumerical solution formula;
In the present embodiment because of Γ (αj) existence, to α in FCM object functionjDerivation is relatively difficult, so first to Γ (αj) carry out Equivalent Infinitesimal replacement, it may be assumed that
Bring formula (12) into formula (2), do not comprised Γ (αj) the non-similarity Measure Formula of approximate Gamma profile:
Formula (13) is brought into formula (1), and makes FCM object function that the single order local derviation of form parameter is equal to zero, thus trying to achieve form parameter solution formula αj;
In the present embodiment, detailed process is as follows:
Step 4: utilizing the FCM object function with Gamma distribution function to carry out SAR image fuzzy clustering to be split, the gray value obtaining each pixel of SAR image to be split belongs to the fuzzy membership matrix of each atural object classification;
As it is shown on figure 3, specifically comprise the following steps that
Step 4.1: cluster centre number c is set, initializes fuzzy membership matrix uij, fuzzy factor λ and iteration stopping condition e;Atural object classification number in cluster centre number and SAR image to be split, fuzzy factor represents the confusion degree of SAR image segmentation result;
In present embodiment, the SAR image atural object classification number c=3 to be split of simulation, real SAR image atural object classification number c=3 to be split;Initial fuzzy subordinated-degree matrix uijStochastic generation, fuzzy factor λ is obtained about 2.3 by experience;Iteration stopping condition is: the minima e < 0.02 of the difference of twice FCM object function in iterative process;
Step 4.2: random initializtion FCM target function value J0, Gamma distribution function scale parameter βj 0And form parameter αj 0;
Step 4.3: utilize formula (4) to calculate the prior probability π of the constrained clustering yardstick in FCM object functionij 0;
In formula (4), μ is the parameter in potential-energy function, μ=0.5;
Step 4.4: estimate new form parameter βj 1And scale parameter αj 1, and then try to achieve non-similarity and estimate;
Step 4.5: estimate according to the prior probability in FCM object function, non-similarity, the gray value calculating each pixel belongs to the fuzzy membership u of each atural object classificationij 1;
Step 4.6: according to fuzzy membership, non-similarity is estimated, prior probability calculates current FCM target function value J1;
Step 4.7: if the difference of the FCM target function value of current FCM target function value and previous calculating is more than setting threshold value, using present Fuzzy degree of membership as initial fuzzy membership, return step 3.3;If the difference of the FCM target function value of current FCM target function value and previous calculating is less than setting threshold value, then iteration stopping;
In the present embodiment, arrange | J1-J0| > e, military order J1=J0, βj1=βj 0, αj 1=αj 0, repeat step 4.3~4.7;If | J1-J0|≤e iteration stopping.
Step 4.8: present Fuzzy subordinated-degree matrix is the optimum fuzzy membership matrix of each grey scale pixel value of SAR image to be split.
Step 5: above-mentioned fuzzy membership matrix is carried out anti fuzzy method by maximum membership grade principle, it is achieved SAR image is split.Maximum membership grade principle: the fuzzy membership angle value that pixel to be divided in SAR image is corresponding in fuzzy membership matrix is compared, it is believed that the atural object classification corresponding to maximum membership degree value is the affiliated atural object classification of current pixel.
Zj=argi{max{UijJ=1 ..., N;I=1 ..., m (16)
Wherein, ZjRepresent the affiliated atural object classification of the gray value of jth pixel, and use Z={Z1,Z2,…,ZNRepresent the result that SAR image is split.
The present invention can be use MATLAB7.12.0 software programming to realize emulation in Core (TM) i5-34703.20GHz, internal memory 4GB, Windows7 Ultimate system at CPU.
In present embodiment, the atural object classification simulation SAR image with 3 Gamma distributions of the design packet true SAR image containing 3 atural object classifications and generation is as emulating image, and the Gamma distributed constant wherein simulating SAR image is known.Table 1 lists simulates the Gamma distributed constant of each atural object classification in SAR image in present embodiment.
SAR image Gamma distributed constant simulated by table 1
Fig. 4 is emulating image, and wherein (a) is template, and (b) is analog image, and (c) is true SAR image, by bright to being secretly followed successively by ice, ice-melt and water.
Fig. 5 be in present embodiment apply the inventive method, existing FCM method and based on Gauss regression model FCM method to simulation SAR image split-run test, wherein (a) is standard FCM method segmentation result, b () is the FCM method segmentation result based on Gauss regression model, the SAR image segmentation result that (c) is the inventive method.It can be seen that FCM segmentation result is worst, the segmentation result (Fig. 5 (c)) of the present invention is substantially better than the FCM segmentation result based on the distribution of Gauss regression model.
Fig. 6 applies the inventive method, existing FCM method and based on the split-run test to true SAR image of the FCM method of Gauss regression model in present embodiment, wherein (a) is standard FCM method segmentation result, b () is the FCM method segmentation result based on Gauss regression model, (c) is the inventive method segmentation result.It can be seen that the segmentation result noise of the inventive method reduces in a large number, segmentation effect is greatly increased.
Table 2 is in present embodiment, with the quantitative assessment that analog image segmentation result is carried out by template image for the standard FCM method to the inventive method and based on Gauss regression model.Owing to containing much noise in standard FCM segmentation effect, segmentation effect is poor, therefore, does not account for standard FCM precision index when precision evaluation, only latter two method is carried out quantitative assessment.It can be seen that the FCM method overall segmentation accuracy based on Gauss regression model is 61.1%, and the dividing method overall segmentation accuracy of the present invention reaches 99.1%, relative to control methods, the inventive method makes segmentation precision improve 38%, and segmentation effect is good, and segmentation precision improves notable.
Table 2 carries out quantitative assessment based on FCM method and the inventive method segmentation result of Gauss regression model
Fig. 7 is Gauss regression model and the corresponding true histogrammic fitting result chart of SAR image segmentation result in present embodiment, wherein, discrete point represents the rectangular histogram of four atural object classifications in the segmentation result of Fig. 6 (b), and black curve represents the Gauss regressing fitting model of four atural object classifications.It can be seen that Gauss regression model can not accurate its corresponding rectangular histogram of matching.
Fig. 8 is Gamma distribution function and the corresponding true histogrammic fitting result chart of SAR image segmentation result in present embodiment, wherein, discrete point represents the rectangular histogram of four atural object classifications in the segmentation result of Fig. 6 (c), and black curve represents the Gamma fitting of distribution model of four classifications.It can be seen that the Gamma distribution function used in the present invention can accurately matching SAR image distribution characteristics, use Gamma distribution function to model the reasonability of FCM method of SAR image model thus demonstrating the present invention.
The above; it is only in the present invention most basic detailed description of the invention; but protection scope of the present invention is not limited thereto; any the art personage is in the technical scope that disclosed herein; the replacement being understood that; what all should be encompassed in the present invention comprises within scope, for instance the classification based on other remotely-sensed data type of the inventive method processes, feature extraction etc..Therefore, protection scope of the present invention should be as the criterion with the protection domain of claims.
Claims (3)
1. the fuzzy clustering SAR image segmentation method based on Gamma distribution, it is characterised in that comprise the following steps:
Step 1: read SAR image to be split;
Step 2: using the gray value of each for SAR image to be split pixel as sample point, build the FCM object function with Gamma distribution function, this object function is to overcome noise in SAR image on the impact of segmentation result for target, obey the feature of Gamma distribution according to SAR image intensity profile, estimate as object function non-similarity using the negative logarithm of Gamma distribution function;The prior probability in FCM object function is determined by the potential-energy function of defined label field;This FCM objective function parameters includes fuzzy factor, fuzzy membership function, prior probability, form parameter in Gamma distribution function and scale parameter;
Step 3: determine the solution formula of each parameter in FCM object function;
Step 4: utilizing the FCM object function with Gamma distribution function to carry out SAR image fuzzy clustering to be split, the gray value obtaining each pixel of SAR image to be split belongs to the fuzzy membership matrix of each atural object classification;
Step 5: above-mentioned fuzzy membership matrix is carried out anti fuzzy method by maximum membership grade principle, it is achieved SAR image is split.
2. the fuzzy clustering SAR image segmentation method based on Gamma distribution according to claim 1, it is characterised in that specifically comprising the following steps that of described step 3
Step 3.1: determine the prior probability formula in FCM object function;
Step 3.1.1: definition potential-energy function: it is 0 that the gray value of neighborhood territory pixel and the gray value of center pixel belong to potential energy during same atural object classification, and otherwise potential energy is 1;
Step 3.1.2: define the prior probability in FCM object function according to potential-energy function and be normalized;
Step 3.2: determine the numerical solution formula of fuzzy membership: add Lagrange multiplier in FCM object function, then the fuzzy membership function in FCM object function is sought local derviation, and make its first derivative equal to 0, obtain the fuzzy membership function with Lagrange multiplier, reelimination Lagrange multiplier;
Step 3.3: determine the numerical solution formula of form parameter in Gamma distribution function, scale parameter:
Determine the numerical solution formula of form parameter: form parameter is carried out Equivalent Infinitesimal replacement;
Determine the numerical solution formula of scale parameter: make FCM object function that the single order local derviation of form parameter is equal to zero.
3. the fuzzy clustering SAR image segmentation method based on Gamma distribution according to claim 1, it is characterised in that specifically comprising the following steps that of described step 4
Step 4.1: cluster centre number, initial fuzzy membership, fuzzy factor and iteration stopping condition are set;Atural object classification number in cluster centre number and SAR image to be split, fuzzy factor represents the confusion degree of SAR image segmentation result;
Step 4.2: random initializtion FCM target function value, the scale parameter of Gamma distribution function and form parameter;
Step 4.3: calculate the prior probability for constrained clustering yardstick in FCM object function;
Step 4.4: estimate new form parameter and scale parameter, and then try to achieve non-similarity and estimate;
Step 4.5: estimate according to the prior probability in FCM object function, non-similarity, the gray value calculating each pixel belongs to the fuzzy membership of each atural object classification;
Step 4.6: according to fuzzy membership, non-similarity is estimated, prior probability calculates current FCM target function value;
Step 4.7: if the difference of the FCM target function value of current FCM target function value and previous calculating is more than setting threshold value, using present Fuzzy degree of membership as initial fuzzy membership, return step 4.3;If the difference of the FCM target function value of current FCM target function value and previous calculating is less than setting threshold value, then iteration stopping;
Step 4.8: present Fuzzy subordinated-degree matrix is the optimum fuzzy membership matrix of each grey scale pixel value of SAR image to be split.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610096321.5A CN105787935B (en) | 2016-02-22 | 2016-02-22 | A kind of fuzzy clustering SAR image segmentation method based on Gamma distributions |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610096321.5A CN105787935B (en) | 2016-02-22 | 2016-02-22 | A kind of fuzzy clustering SAR image segmentation method based on Gamma distributions |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105787935A true CN105787935A (en) | 2016-07-20 |
CN105787935B CN105787935B (en) | 2018-05-08 |
Family
ID=56403612
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610096321.5A Expired - Fee Related CN105787935B (en) | 2016-02-22 | 2016-02-22 | A kind of fuzzy clustering SAR image segmentation method based on Gamma distributions |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105787935B (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106570880A (en) * | 2016-10-28 | 2017-04-19 | 中国人民解放军第三军医大学 | Brain tissue MRI image segmentation method based on fuzzy clustering and Markov random field |
CN107169962A (en) * | 2017-05-16 | 2017-09-15 | 西安电子科技大学 | The gray level image fast partition method of Kernel fuzzy clustering is constrained based on space density |
CN107808380A (en) * | 2016-12-28 | 2018-03-16 | 中国测绘科学研究院 | One kind is based on G0With the multiple dimensioned High-resolution SAR Images water segmentation method of Gamma Joint Distributions |
CN108182436A (en) * | 2017-12-29 | 2018-06-19 | 辽宁工程技术大学 | A kind of method for segmentation of high resolution remote sensing image |
CN109300137A (en) * | 2018-09-20 | 2019-02-01 | 北京航空航天大学 | A kind of two type fuzzy clustering magnetic resonance brain image segmentation method of more curved surface estimation intervals |
CN110853064A (en) * | 2019-11-12 | 2020-02-28 | 吉林大学 | Image collaborative segmentation method based on minimum fuzzy divergence |
CN111968131A (en) * | 2020-07-24 | 2020-11-20 | 哈尔滨理工大学 | Medical image segmentation algorithm based on wolf colony algorithm |
CN113362337A (en) * | 2021-05-21 | 2021-09-07 | 桂林理工大学 | Automatic fuzzy interval determination method for object-oriented segmentation weight parameters of remote sensing image |
CN114266321A (en) * | 2021-12-31 | 2022-04-01 | 广东泰迪智能科技股份有限公司 | Weak supervision fuzzy clustering algorithm based on unconstrained prior information mode |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102903102A (en) * | 2012-09-11 | 2013-01-30 | 西安电子科技大学 | Non-local-based triple Markov random field synthetic aperture radar (SAR) image segmentation method |
CN104850867A (en) * | 2015-06-10 | 2015-08-19 | 中国人民武装警察部队工程大学 | Object identification method based on intuitive fuzzy c-means clustering |
CN105096315A (en) * | 2015-06-19 | 2015-11-25 | 西安电子科技大学 | Method for segmenting heterogeneous super-pixel SAR (Synthetic Aperture Radar) image based on Gamma distribution |
-
2016
- 2016-02-22 CN CN201610096321.5A patent/CN105787935B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102903102A (en) * | 2012-09-11 | 2013-01-30 | 西安电子科技大学 | Non-local-based triple Markov random field synthetic aperture radar (SAR) image segmentation method |
CN104850867A (en) * | 2015-06-10 | 2015-08-19 | 中国人民武装警察部队工程大学 | Object identification method based on intuitive fuzzy c-means clustering |
CN105096315A (en) * | 2015-06-19 | 2015-11-25 | 西安电子科技大学 | Method for segmenting heterogeneous super-pixel SAR (Synthetic Aperture Radar) image based on Gamma distribution |
Non-Patent Citations (7)
Title |
---|
DELIANG XIANG 等: "A Kernel Clustering Algorithm With Fuzzy Factor: Application to SAR Image Segmentation", 《IEEE GEOSCIENCE AND REMOTE SENSING LETTERS》 * |
T.CHAIRA 等: "Segmentation using fuzzy divergence", 《PATTERN RECOGNITION LETTERS》 * |
张耀天 等: "一种基于二维Gamma分布的SAR图像变化检测方法", 《系统工程与电子技术》 * |
王丽芳: "可见光遥感图像分割算法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
王春艳 等: "融入空间关系的二型模糊模型高分辨率遥感影像分割", 《遥感学报》 * |
王玉 等: "利用RJMCMC算法的可变类SAR图像分割", 《信号处理》 * |
赵雪梅 等: "结合高斯回归模型和隐马尔可夫随机场的模糊聚类图像分割", 《电子与信息学报》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106570880B (en) * | 2016-10-28 | 2019-10-11 | 中国人民解放军第三军医大学 | In conjunction with the brain tissue MRI image dividing method of fuzzy clustering and markov random file |
CN106570880A (en) * | 2016-10-28 | 2017-04-19 | 中国人民解放军第三军医大学 | Brain tissue MRI image segmentation method based on fuzzy clustering and Markov random field |
CN107808380A (en) * | 2016-12-28 | 2018-03-16 | 中国测绘科学研究院 | One kind is based on G0With the multiple dimensioned High-resolution SAR Images water segmentation method of Gamma Joint Distributions |
CN107169962A (en) * | 2017-05-16 | 2017-09-15 | 西安电子科技大学 | The gray level image fast partition method of Kernel fuzzy clustering is constrained based on space density |
CN107169962B (en) * | 2017-05-16 | 2020-02-07 | 西安电子科技大学 | Gray level image fast segmentation method based on space density constraint kernel fuzzy clustering |
CN108182436A (en) * | 2017-12-29 | 2018-06-19 | 辽宁工程技术大学 | A kind of method for segmentation of high resolution remote sensing image |
CN109300137A (en) * | 2018-09-20 | 2019-02-01 | 北京航空航天大学 | A kind of two type fuzzy clustering magnetic resonance brain image segmentation method of more curved surface estimation intervals |
CN109300137B (en) * | 2018-09-20 | 2021-05-07 | 北京航空航天大学 | Two-type fuzzy clustering magnetic resonance brain image segmentation method for multi-surface estimation interval |
CN110853064A (en) * | 2019-11-12 | 2020-02-28 | 吉林大学 | Image collaborative segmentation method based on minimum fuzzy divergence |
CN110853064B (en) * | 2019-11-12 | 2022-03-25 | 吉林大学 | Image collaborative segmentation method based on minimum fuzzy divergence |
CN111968131A (en) * | 2020-07-24 | 2020-11-20 | 哈尔滨理工大学 | Medical image segmentation algorithm based on wolf colony algorithm |
CN113362337A (en) * | 2021-05-21 | 2021-09-07 | 桂林理工大学 | Automatic fuzzy interval determination method for object-oriented segmentation weight parameters of remote sensing image |
CN114266321A (en) * | 2021-12-31 | 2022-04-01 | 广东泰迪智能科技股份有限公司 | Weak supervision fuzzy clustering algorithm based on unconstrained prior information mode |
Also Published As
Publication number | Publication date |
---|---|
CN105787935B (en) | 2018-05-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105787935A (en) | Fuzzy cluster SAR image segmentation method based on Gamma distribution | |
CN111461038B (en) | Pedestrian re-identification method based on layered multi-mode attention mechanism | |
CN113033520B (en) | Tree nematode disease wood identification method and system based on deep learning | |
CN107300698B (en) | Radar target track starting method based on support vector machine | |
CN106127784B (en) | A kind of high-resolution remote sensing image dividing method | |
CN104091216A (en) | Traffic information predication method based on fruit fly optimization least-squares support vector machine | |
CN106203521A (en) | Based on disparity map from the SAR image change detection of step study | |
CN110082738B (en) | Radar target identification method based on Gaussian mixture and tensor recurrent neural network | |
CN106778838A (en) | A kind of method for predicting air quality | |
CN107301430A (en) | Broad sense Multivariable Fuzzy c means clustering algorithms | |
CN108447057A (en) | SAR image change detection based on conspicuousness and depth convolutional network | |
CN101598783A (en) | Based on distance by radar under the strong noise background of PPCA model as statistical recognition method | |
CN103456017B (en) | Image partition method based on the semi-supervised weight Kernel fuzzy clustering of subset | |
CN104850867A (en) | Object identification method based on intuitive fuzzy c-means clustering | |
CN111562571A (en) | Maneuvering multi-target tracking and track maintaining method for unknown new-born strength | |
CN116187835A (en) | Data-driven-based method and system for estimating theoretical line loss interval of transformer area | |
CN105678790A (en) | High-resolution remote sensing image supervised segmentation method based on variable Gaussian hybrid model | |
Li et al. | State of Health and Charge Estimation Based on Adaptive Boosting integrated with particle swarm optimization/support vector machine (AdaBoost-PSO-SVM) Model for Lithium-ion Batteries | |
CN115311449A (en) | Weak supervision image target positioning analysis system based on class reactivation mapping chart | |
CN106408571A (en) | Variable class remote sensing image segmentation method based on optimal fuzzy factor selection | |
CN108846845B (en) | SAR image segmentation method based on thumbnail and hierarchical fuzzy clustering | |
CN104463227A (en) | Polarimetric SAR image classification method based on FQPSO and target decomposition | |
CN114594440A (en) | Radar high-resolution one-dimensional range profile target identification method and system based on double parallel networks | |
Kurosawa et al. | A statistical hypothesis testing strategy for adaptively blending particle filters and ensemble Kalman filters for data assimilation | |
CN111950646A (en) | Hierarchical knowledge model construction method and target identification method for electromagnetic image |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180508 Termination date: 20210222 |