CN102930291A - Automatic K adjacent local search heredity clustering method for graphic image - Google Patents

Abstract

The invention discloses an automatic k adjacent local search heredity clustering method for a graphic image, which mainly overcomes the defect of the conventional automatic clustering algorithm that the local optimization is easy to cause. The automatic k adjacent local search heredity clustering method comprises the realization steps of: (1) detecting an outline of an image by utilizing a canny edge detector; (2) describing the outline of the image by utilizing a shape context method and calculating a matched cost matrix of an outline point; (3) matching the outline point by utilizing a dynamic programming method according to the matched cost matrix; (4) converting the matched outline point by utilizing a procrustes analysis method; (5) representing the converted matched outline point and measuring an edit distance between character strings; (6) calculating the distance between the images according to the edit distance of the character strings; (7) clustering the images by utilizing a heredity automatic clustering method; and (8) carrying out k adjacent local search on groups of a heredity method. The automatic k adjacent local search heredity clustering method for the graphic image has the advantages that overall optimization is easy to achieve and an accurate clustering number can be found out.

Description

The k neighbour Local Search heredity automatic clustering method that is used for graph image

Technical field

The invention belongs to technical field of image processing, can be used for the automatic cluster to the obvious image of shape facility.

Background technology

The development of science and technology impels people's research object to become increasingly complex.Vision is as the human knowledge and observe one of the Main Means in the world, and its information is enormous amount but also more complicated not only.It is the inscape of visual information that the spatial relationship of color, shape, texture and things all waits, and can be used as the starting point of research.Shape is one of essential characteristic of having under the visually-perceptible meaning of object, the research of this respect has become the importance of computer vision and pattern-recognition, and utilizes shape facility things to be identified and one of the main tool that also becomes computer vision and pattern identification research of classifying.Research to shape has shape extracting, form fit etc.Shape extracting extracts shape information exactly from image, the more famous method of this respect has canny etc., and form fit can be divided into two aspects such as description, coupling.There is in recent years the method for various shape description and coupling to be suggested.The shape description method has based on the shape edges profile, have based on shape area etc.These methods all respectively have relative merits.Form fit is weighed similarity between shape according to certain measurement criterion.Man-to-man form fit is without any prior imformation, often can not distinguish important and non-important shape difference, in any case design better shape descriptor, all can't address this problem, this just needs the better algorithm of design to go for the internal relation of similar shape in the shape similarity metric space.

Cluster analysis as one of important research direction of data mining, its objective is to allow people's recognition data concentrate intensive and sparse zone, finds the global distributed schema of data, and the mutual relationship between the data attribute.Cluster analysis has been applied to every field widely, comprises data analysis, image processing and market analysis, data mining, statistics, machine learning, spatial database technology, biology and the marketing etc.Traditional clustering algorithm, shifting to an earlier date given cluster numbers, is unpredictable but be subjected to the very multidata cluster numbers of many factors in the actual life, and the during this time effect of automatic cluster highlights, many scholars have proposed various methods, attempt to solve the automatic cluster problem.But these methods often exist defective: algorithm usually can be absorbed in local optimum, can not get correct cluster numbers.Someone utilizes the thought of Gradient Descent to solve the automatic cluster problem, and uses it for the cluster of graph image, but this method still exists above-mentioned said defective, and algorithm is absorbed in local optimum easily, can not obtain correct cluster numbers.

Summary of the invention

The object of the invention is to for above-mentioned existing technical deficiency, propose a kind of k neighbour Local Search heredity automatic clustering method for graph image, to improve the Clustering Effect of image.

Technical thought of the present invention is: by the profile of canny operator detected image, profile information by shape context method Description Image, by dynamic programming method, procrustes analysis method and the edit distance distance matrix in conjunction with the calculating chart image set, the automatic cluster of image is realized by k neighbour Local Search heredity automatic cluster algorithm.Its specific implementation step is as follows:

(1) input N width of cloth image S to be clustered _i, i=1 ..., N utilizes the canny operator respectively every width of cloth image to be carried out rim detection, obtains the contour images I of each image to be clustered _i

(2) to each contour images I _iCarry out uniform sampling along its outline line, and the point P that represents to sample and obtain with Cartesian coordinates _Ij=(x _Ij, y _Ij), i=1 ..., N, j=1 ..., M, wherein, P _IjJ point of the i width of cloth contour images that the expression sampling obtains, x _Ij, y _IjBe respectively point P _IjThe transverse and longitudinal coordinate of position, M are the number of sampled point;

(3) calculate distance between any two width of cloth images to be clustered:

(3a) with shape context method contour images I is described _i, obtain contour images I _iThe Neighbor Points position distribution histogram of each point, and this histogram carried out normalization;

(3b) with normalized Neighbor Points position distribution histogram, use χ ²Statistical method is calculated contour images I _iThe coupling cost matrix of point;

(3c) according to the coupling cost matrix, with dynamic programming method, remove the outlier of two width of cloth contour images, obtain the Corresponding matching point of contour images;

(3d) to resulting Corresponding matching point, carry out conversion with procrustes analysis method;

(3e) point of the Corresponding matching after the conversion is converted to character string;

(3f) obtain two image S to be clustered through step (3a) to (3e) _i, S _jCharacter string;

(3g) the image S to be clustered to obtaining _iAnd S _jCharacter string, with the Edit between these two character strings of Edit range observation apart from E (S _i, S _j), then calculate image S to be clustered _iAnd S _jBetween distance:

d(S _i,S _j)=λ ^u×E(S _i,S _j)/L，

Wherein λ=1.03 are constant, and u is image S to be clustered _iThe outlier number of contour images, L is image S to be clustered _iThe length of character string;

(3h) to the image to be clustered that obtains apart from d (S _i, S _j), carry out normalization

(4) make evolutionary generation t=0, the initialization population:

(4a) produce at random ρ more than or equal to 2 integer k less than or equal to ζ _i, i=1 ..., ρ, as the chromosomal cluster numbers of ρ bar, ρ is population scale, ζ=15 are maximum cluster numbers;

(4b) to every chromosome produce at random N more than or equal to 1 less than or equal to k _iInteger c _j, j=1 ..., N is as this chromosomal each gene position B _jValue, form this chromosome H _i

(5) calculate all chromosomal fitness value f _i, find out the chromosome of fitness value maximum as current optimum dyeing body H, wherein:

f _i=1/θ _i

${\mathrm{\θ}}_i=\mathrm{\α}\×k_i+\frac{1}{k_i}\×{\mathrm{\Σ}}_{s=1}^{k_i}{\mathrm{\ω}}_s$

${\mathrm{\&omega;}}_s=\frac{2}{n_s}\underset{S_k,S_h,S_h\&Element;C_s,k<h}{\mathrm{\&Sigma;}}d{(S_k,S_h)}^2$

Wherein, α=0.5 is control parameter, k _iBy being calculated chromosomal cluster numbers, d (S _k, S _h) be image S to be clustered _kAnd S _hDistance, n _sBe the secondary number of the contained image to be clustered of s class cluster image, C _sBe the set of s class image to be clustered, S _kAnd S _hIt is the image to be clustered in the s class image to be clustered;

(6) according to resulting chromosomal fitness value f _i, with roulette wheel selection to all chromosome H _iSelect, select to carry out genetic manipulation the centre for chromosome H ' _i

(7) to the centre that obtains for chromosome H ' _i, with probability σ ₁, σ ₁∈ [0,1] carries out the single-point interlace operation;

(8) to through after the single-point interlace operation in the middle of all for chromosome H " _i, with probability σ ₂, σ ₂∈ [0,1] carries out the single-point mutation operation;

(9) all are middle for chromosome H after the interlace operation of calculating process single-point " _iFitness value f _i

(10) to every centre for chromosome H " _i, produce a random number r, r ∈ [0,1], if r＜0.5, execution in step (11), otherwise execution in step (12);

(11) to the centre for chromosome H " _iEach gene position B _j, produce random number ψ, ψ ≠ c _j, 1≤ψ≤k _i, k _iFor the centre for chromosome H " _iCluster numbers, c _jBe gene position B _jValue, with gene position B _jValue c _jChange ψ into, calculate middle for chromosome H at this moment " ' _iFitness value f _i', and with change before the centre for chromosome fitness value f _iRelatively, if f ' _iF _i, gene position B then _jValue be set to ψ, execution in step (13), otherwise gene position B _jValue be set to c _j, execution in step (13);

(12) to the centre for chromosome H " _iCarry out k neighbour Local Search:

(12a) with the centre for chromosome H " _iEach gene position B _jCorresponding image S to be clustered _j, with other image S to be clustered _lApart from d (S _j, S _l) sort from small to large;

(12b) from sorted distance, choose front K=5 the corresponding image to be clustered of distance, as the secondary arest neighbors image of K S ' _v

(12c) to the secondary arest neighbors image of selected K S ' _vCorresponding gene position B _v, the number of statistics different genes place value finds the maximum gene place value τ of number, if τ=c _j, c _jBe image S to be clustered _jCorresponding gene position B _jValue, execution in step (13) then;

(12d) with the centre for chromosome H " _iGene position B _jValue change τ into, calculate in the middle of this moment for chromosome H " ' _iFitness value f " _i, middle for chromosome H before will changing " _iFitness value f _iIf, f " _iF _i, gene position B then _jValue be set to τ, otherwise gene position B _jValue be set to c _j

(13) relatively all are middle for chromosomal fitness value f _iFind out the dyeing of fitness value maximum, this chromosome and current optimum dyeing body H are compared, if this chromosomal fitness value is greater than the fitness value of current optimum dyeing body H, then this chromosome replication is arrived current optimum dyeing body H, if the fitness value of current optimum dyeing body H greater than this chromosomal fitness value, then copies to this chromosome with current optimum dyeing body H;

(14) evolutionary generation t value is added 1, if t then returns step (6) less than maximum evolutionary generation T=100 at this moment, otherwise, with each gene position B of current optimum dyeing body H _iValue c _iMake gene position B _iCorresponding image S to be clustered _iThe class mark, the image to be clustered that all class marks are identical is divided into a class, the cluster numbers of the image to be clustered behind the output category and current optimum dyeing body H.

The present invention compared with prior art has the following advantages:

1, the present invention so can adjust the evolution process of population among a small circle, has increased population diversity owing to add the Local Search operator after genetic manipulation, is conducive to overcome the defective that is absorbed in local optimum, has improved the Clustering Effect of image.

2, the present invention has improved the direction of Evolution of Population owing to carry out Local Search according to the neighbor information of image, has accelerated speed of convergence.

Description of drawings

Fig. 1 is general flow chart of the present invention;

Fig. 2 is the wherein 4 sub-picture examples among the test pattern image set kimia216;

Fig. 3 is with the cluster simulated effect figure of the present invention to 30 sub-pictures among the test pattern image set kimia216;

Fig. 4 is with the cluster simulated effect figure of contrast algorithm to 30 sub-pictures among the test pattern image set kimia216.

Embodiment

With reference to Fig. 1, realization of the present invention comprises the steps:

Step 1: input N width of cloth image S to be clustered _i, i=1 ..., N utilizes the canny operator respectively every width of cloth image to be carried out rim detection, obtains the contour images I of each image to be clustered _i, N gets 30 but be not limited to 30, canny operator and proposed in 1986 years by John Canny in this example.

Step 2: to each contour images I _iCarry out uniform sampling along its outline line, and the point P that represents to sample and obtain with Cartesian coordinates _Ij=(x _Ij, y _Ij), i=1 ..., N, j=1 ..., M, wherein, P _IjJ point of the i width of cloth contour images that the expression sampling obtains, x _Ij, y _IjBe respectively point P _IjThe transverse and longitudinal coordinate of position, M=100 are the number of sampled point.

Step 3: calculate the distance between any two width of cloth images to be clustered.

3.1) with shape context method contour images I is described _i, obtain contour images I _iThe Neighbor Points position distribution histogram of each point, and this histogram carried out normalization:

3.1.1) to contour images I _i, calculate the ultimate range between all point

d=max _{1≤m＜n≤M}{||P _im-P _in||}

$\left|\right|P_{\mathrm{im}}-P_{\mathrm{in}}\left|\right|=\sqrt{{(x_{\mathrm{im}}-x_{\mathrm{in}})}^2+{(y_{\mathrm{im}}-y_{\mathrm{in}})}^2},$

Ultimate range d is converted to logarithm log (d), and with the b that is quantified as of this logarithmic mean ₁Individual grade, with 360 ° of angles average be quantified as b ₂Individual grade is with point P _IjBe initial point, by this b ₁* b ₂Individual grade classification goes out b ₁* b ₂Individual zone;

3.1.2) to contour images I _iEach point P _Ij, with all the point P except this point _IkBe converted to polar coordinates P ' by Cartesian coordinates _Ik=(r _Ik, θ _Ik), be converted to again log-polar P " _Ik=(r ' _Ik, θ ' _Ik), wherein:

$r_{\mathrm{ik}}=\sqrt{{(x_{\mathrm{ik}}-x_{\mathrm{ij}})}^2+{(y_{\mathrm{ik}}-y_{\mathrm{ij}})}^2}$

${\mathrm{\&theta;}}_{\mathrm{ik}}=\arctan \left(\frac{y_{\mathrm{ik}}-y_{\mathrm{ij}}}{x_{\mathrm{ik}}-x_{\mathrm{ij}}}\right)$

r′ _ik=log(r _ik)

θ′ _ik=θ _ik-θ _O,

${\mathrm{\&theta;}}_0=\arctan \left(\frac{y_0-y_{\mathrm{ij}}}{x_0-x_{\mathrm{ij}}}\right)$

$x_0=\frac{1}{M}\&times;\underset{l=1}{\overset{M}{\mathrm{\&Sigma;}}}{x}_{\mathrm{il}}$

$y_0=\frac{1}{M}\&times;\underset{l=1}{\overset{M}{\mathrm{\&Sigma;}}}{y}_{\mathrm{il}}$

In log-polar system, statistics falls into ready-portioned b ₁* b ₂Point P in the regional in individual zone " _IkNumber c _l, l=1 ..., b ₁* b ₂, recycling experience density formula carries out normalized, obtains point P _IjNeighbor Points position distribution normalization histogram h _j(s):

$h_j\left(s\right)=\frac{c_s}{\underset{l=1}{\overset{b_1\&times;b_2}{\mathrm{\&Sigma;}}}{c}_l},$ s=1,…,b ₁×b ₂；

3.2) Neighbor Points position distribution normalization histogram to obtaining, calculate coupling cost matrix C (i, j) according to formula:

$C(i,j)=\frac{1}{2}\underset{s=1}{\overset{b_1\&times;b_2}{\mathrm{\&Sigma;}}}\frac{{[h_i\left(s\right)-h_j\left(s\right)]}^2}{h_i\left(s\right)+h_j\left(s\right)};$

3.3) according to the coupling cost matrix that obtains, with dynamic programming method, remove the outlier of two width of cloth contour images, obtain the Corresponding matching point of contour images, the implementation step of dynamic programming method sees that M.R.Daliri and V. Torre were published in the article Robust symbolic representation for shape recognition and retrieval on the Pattern Recognition in 2008.

3.4) to resulting Corresponding matching point, carry out conversion with procrustes analysis method, the implementation step of procrustes analsis method sees that Mohammad Reza Daliri and Vincent Torre were published in the article Shape and texture clustering:Best estimate for the clusters number on the Image and Vision Computing in 2009.

3.5) point of the Corresponding matching after the conversion is converted to character string, its implementation step sees that Mohammad Reza Daliri and Vincent Torre were published in the article Shape and texture clustering:Best estimate for the clusters number on the Image and Vision Computing in 2009.

3.6) through step 3.1) to 3.5) obtain two image S to be clustered _i, S _jCharacter string.

3.7) image S to be clustered to obtaining _iAnd S _jCharacter string, with the Edit between these two character strings of Edit range observation apart from E (S _i, S _j), then calculate image S to be clustered _iAnd S _jBetween distance:

d(S _i,S _j)=λ ^u×E(S _i,S _j)/L，

Wherein λ=1.03 are constant, and u is image S to be clustered _iThe outlier number of contour images, L is image S to be clustered _iThe length of character string.

The measuring process of Edit distance is seen the article Learning string edit distance that Eric SvenRistad and PeterN.Yianilos delivered at IEEE Trans.PAMI in 1998.

3.8) to the image to be clustered that obtains apart from d (S _i, S _j), carry out normalization.

Step 4: make evolutionary generation t=0, the initialization population:

4.1) produce at random ρ more than or equal to 2 integer k less than or equal to ζ _i, i=1 ..., ρ, as the chromosomal cluster numbers of ρ bar, ρ is population scale, ζ=15 are maximum cluster numbers.

4.2) to every chromosome produce at random N more than or equal to 1 less than or equal to k _iInteger c _j, j=1 ..., N is as this chromosomal each gene position B _jValue, form this chromosome H _i

Step 5: calculate all chromosomal fitness value f _i, find out the chromosome of fitness value maximum as current optimum dyeing body H, wherein:

f _i=1/θ _i

${\mathrm{\&theta;}}_i=\mathrm{\&alpha;}\&times;k_i+\frac{1}{k_i}{\mathrm{\&Sigma;}}_{s=1}^{k_i}{\mathrm{\&omega;}}_s$

${\mathrm{\&omega;}}_s=\frac{2}{n_s}\underset{S_k,S_h\&Element;C_s,k<h}{\mathrm{\&Sigma;}}d{(S_k,S_h)}^2$

Wherein, α=0.5 is control parameter, k _iBy being calculated chromosomal cluster numbers, d (S _k, S _h) be image S to be clustered _kAnd S _hDistance, n _sBe the secondary number of the contained image to be clustered of s class cluster image, C _sBe the set of s class image to be clustered, S _kAnd S _hIt is the image to be clustered in the s class image to be clustered.

Step 6: according to resulting chromosomal fitness value f _i, to all chromosome H _iSelect, select to carry out genetic manipulation the centre for chromosome H ' _i

6.1) according to resulting chromosomal fitness value f _i, pass through formula

$F_i=\frac{{\mathrm{\&Sigma;}}_{s=1}^i{f}_s}{{\mathrm{\&Sigma;}}_{j=1}^N{f}_j},$

Calculate all chromosomal cumulative probability F _i

6.2) produce at random ρ number γ _j, γ _j∈ [0,1] is to each random number γ _j, seek subscript value l, so that F _L-1＜γ _j≤ F _lSet up, with l chromosome H _l, copy to middle for chromosome H ' _j

Step 7: to the centre that obtains for chromosome, with probability σ ₁Carry out the single-point interlace operation, σ ₁∈ [0,1].

The concrete steps of single-point interlace operation see that Ujjwal Maulik and Sanghamitra Bandyopadhyay are published in the article Genetic algorithm-based clustering technique on the Pattern Recognition.

Step 8: to through after the single-point interlace operation in the middle of all for chromosome H " _i, with probability σ ₂, σ ₂∈ [0,1] carries out the single-point mutation operation.

The concrete steps of single-point mutation operation see that Ujjwal Maulik and Sanghamitra Bandyopadhyay are published in the article Genetic algorithm-based clustering technique on the Pattern Recognition.

Step 9: all are middle for chromosome H after calculating the interlace operation of process single-point " _iFitness value f _i

Step 10: to every centre for chromosome H " _i, produce a random number r, r ∈ [0,1], if r＜0.5, execution in step 11, otherwise execution in step 12.

Step 11: to the centre for chromosome H " _iEach gene position B _j, produce random number ψ, ψ ≠ c _j, 1≤ψ≤k _i, k _iFor the centre for chromosome H " _iCluster numbers, c _jBe gene position B _jValue, with gene position B _jValue c _jChange ψ into, calculate middle for chromosome H at this moment " ' _iFitness value f ' _i, and with change before the centre for chromosome fitness value f _iRelatively, if f ' _iF _i, gene position B then _jValue be set to ψ, execution in step 13, otherwise gene position B _jValue be set to c _j, execution in step 13.

Step 12: to the centre for chromosome H " _iCarry out k neighbour Local Search:

12.1) with the centre for chromosome H " _iEach gene position B _jCorresponding image S to be clustered _jWith other image S to be clustered _lApart from d (S _j, S _l), sort from small to large;

12.2) from sorted distance, choose front K=5 the corresponding image to be clustered of distance, as the secondary arest neighbors image of K S ' _v

12.3) to the secondary arest neighbors image of selected K S ' _vCorresponding gene position B _v, the number of statistics different genes place value finds the maximum gene place value τ of number, if τ=c _j, c _jBe image S to be clustered _jCorresponding gene position B _jValue, then execution in step rapid 13;

12.4) with the centre for chromosome H " _iGene position B _jValue change τ into, calculate in the middle of this moment for chromosome H " ' _iFitness value f " _i, middle for chromosome H before will changing " _iFitness value f _iCompare, if f " _iF _i, gene position B then _jValue be set to τ, otherwise gene position B _jValue be set to c _j

Step 13: relatively all are middle for chromosomal fitness value f _iFind out the dyeing of fitness value maximum, this chromosome and current optimum dyeing body H are compared, if this chromosomal fitness value is greater than the fitness value of current optimum dyeing body H, then this chromosome replication is arrived current optimum dyeing body H, if the fitness value of current optimum dyeing body H greater than this chromosomal fitness value, then copies to this chromosome with current optimum dyeing body H.

Step 14: evolutionary generation t value is added 1, at this moment, if t less than maximum evolutionary generation T=100, then returns step (6), otherwise, with each gene position B of current optimum dyeing body H _iValue c _iMake gene position B _iCorresponding image S to be clustered _iThe class mark, the image to be clustered that all class marks are identical is divided into a class, the cluster numbers of the image to be clustered behind the output category and current optimum dyeing body H.

Effect of the present invention can further specify by following experiment:

1, the image of emulation experiment employing:

Experiment has used 30 width of cloth images among the test pattern image set kimia216 as test pattern, and these images are the obvious bianry image of shape facility, and wherein Fig. 2 is 4 pairs in the middle of among the test pattern image set kimia216.

2, the parameter of emulation experiment arranges condition:

Setup parameter is: point number of samples M=100, polar coordinates Neighbor Points position distribution histogram diameter quantized value b ₁=5, polar coordinates Neighbor Points position distribution histogram angular quantification value b ₂=12, the penalty value 0.5 of dynamic programming method, Edit distance measure penalty value 0.5, Edit distance measure consecutive point number 5, Population Size ρ=30, maximum cluster numbers ζ=15, maximum evolutionary generation T=100, the secondary number K=5 of k arest neighbors image, crossover probability σ ₁=0.8, variation probability σ ₂=0.2, control parameter alpha=0.1.

3, emulation experiment environment:

Be that core2 2.4HZ, internal memory 2G, WINDOWS XP system use C++ to carry out emulation at CPU.

4, emulation content

Emulation 1

With the present invention and existing shape image clustering algorithm, independent operating is 20 times respectively, records cluster numbers, three indexs such as ARI, MS of each result, and it is as shown in table 1 to calculate its average and standard deviation.

Table 1 the present invention and the contrast of contrast algorithm complexity index

As can be seen from Table 1, under three kinds of Validity Indexes, performance of the present invention all is better than the contrast algorithm, has proved validity of the present invention.

Emulation 2

With the present invention 30 sub-pictures among the test pattern image set kimia216 are carried out cluster, obtain cluster result such as Fig. 3, the image that is positioned at same row among Fig. 3 belongs to same class, and the image that is positioned at different rows belongs to inhomogeneity.

Emulation 3

With the contrast algorithm 30 sub-pictures among the test pattern image set kimia216 are carried out cluster, obtain cluster result such as Fig. 4, the image that is positioned at same row among Fig. 4 belongs to same class, and the image that is positioned at different rows belongs to inhomogeneity.

Can find out that by above-mentioned emulation experiment the present invention can obtain preferably Clustering Effect to test pattern.Relatively going up numerically, and relatively the going up of visual effect are all verified rationality of the present invention and validity effectively.

Contrast algorithm from emulation 1 and emulation 3 sees that Mohammad Reza Daliri and Vincent Torre were published in the article Shape and texture clustering:Best estimate for the clustersnumber on the Image and Vision Computing in 2009.

Claims (3)

1. a k neighbour Local Search heredity automatic clustering method that is used for graph image comprises the steps:

(1) input N width of cloth image S to be clustered _i, i=1 ..., N utilizes the canny operator respectively every width of cloth image to be carried out rim detection, obtains the contour images I of each image to be clustered _i

(2) to each contour images I _iCarry out uniform sampling along its outline line, and the point P that represents to sample and obtain with Cartesian coordinates _Ij=(x _Ij, y _Ij), i=1 ..., N, j=1 ..., M, wherein, P _IjJ point of the i width of cloth contour images that the expression sampling obtains, x _Ij, y _IjBe respectively point P _IjThe transverse and longitudinal coordinate of position, M are the number of sampled point;

(3) calculate distance between any two width of cloth images to be clustered;

(4) make evolutionary generation t=0, the initialization population:

(4a) produce at random ρ more than or equal to 2 integer k less than or equal to ζ _i, i=1 ..., ρ, as the chromosomal cluster numbers of ρ bar, ρ=30 are population scale, ζ=15 are maximum cluster numbers;

(4b) to every chromosome produce at random N more than or equal to 1 less than or equal to k _iInteger c _j, j=1 ..., N is as this chromosomal each gene position B _jValue, form this chromosome H _i

(5) calculate all chromosomal fitness value f _i, find out the chromosome of fitness value maximum as current optimum dyeing body H;

(6) according to resulting chromosomal fitness value f _i, with roulette wheel selection to all chromosome H _iSelect, select to carry out genetic manipulation the centre for chromosome H ' _i

(7) to the centre that obtains for chromosome H ' _i, with probability σ ₁, σ ₁∈ [0,1] carries out the single-point interlace operation;

(8) to through after the single-point interlace operation in the middle of all for chromosome H " _i, with probability σ ₂, σ ₂∈ [0,1] carries out the single-point mutation operation;

(9) all are middle for chromosome H after the interlace operation of calculating process single-point " _iFitness value f _i

(10) to every centre for chromosome H " _i, produce a random number r, r ∈ [0,1], if r＜0.5, execution in step (11), otherwise execution in step (12);

(11) to the centre for chromosome H " _iEach gene position B _j, produce random number ψ, ψ ≠ c _j, 1≤ψ≤k _i, k _iFor the centre for chromosome H " _iCluster numbers, c _jBe gene position B _jValue, with gene position B _jValue c _jChange ψ into, calculate middle for chromosome H at this moment " ' _iFitness value f ' _i, and with change before the centre for chromosome fitness value f _iRelatively, if f ' _iF _i, gene position B then _jValue be set to ψ, execution in step (13), otherwise gene position B _jValue be set to c _j, execution in step (13);

(12) to the centre for chromosome H " _iCarry out k neighbour Local Search:

(12a) with the centre for chromosome H " _iEach gene position B _jCorresponding image S to be clustered _j, with other image S to be clustered _lApart from d (S _j, S _l) sort from small to large;

(12b) from sorted distance, choose front K=5 the corresponding image to be clustered of distance, as the secondary arest neighbors image of K S ' _v

(12c) to the secondary arest neighbors image of selected K S ' _vCorresponding gene position B _v, the number of statistics different genes place value finds the maximum gene place value τ of number, if τ=c _j, c _jBe image S to be clustered _jCorresponding gene position B _jValue, execution in step (13) then;

(12d) with the centre for chromosome H " _iGene position B _jValue change τ into, calculate in the middle of this moment for chromosome H " ' _iFitness value f " _i, middle for chromosome H before will changing " _iFitness value f _iIf, f " _iF _i, gene position B then _jValue be set to τ, otherwise gene position B _jValue be set to c _j

(13) relatively all are middle for chromosomal fitness value f _iFind out the dyeing of fitness value maximum, this chromosome and current optimum dyeing body H are compared, if this chromosomal fitness value is greater than the fitness value of current optimum dyeing body H, then this chromosome replication is arrived current optimum dyeing body H, if the fitness value of current optimum dyeing body H greater than this chromosomal fitness value, then copies to this chromosome with current optimum dyeing body H;

(14) evolutionary generation t value is added 1, if t then returns step (6) less than maximum evolutionary generation T=100 at this moment, otherwise, with each gene position B of current optimum dyeing body H _iValue c _iMake gene position B _iCorresponding image S to be clustered _iThe class mark, the image to be clustered that all class marks are identical is divided into a class, the cluster numbers of the image to be clustered behind the output category and current optimum dyeing body H.

2. the k neighbour Local Search heredity automatic clustering method for graph image according to claim 1, the distance between the described calculating of step (3) any two width of cloth images to be clustered wherein, carry out as follows:

(3a) with shape context method contour images I is described _i, obtain contour images I _iThe Neighbor Points position distribution histogram of each point, and this histogram carried out normalization;

(3b) with normalized Neighbor Points position distribution histogram, use χ ²Statistical method is calculated contour images I _iThe coupling cost matrix of point;

(3c) according to the coupling cost matrix, with dynamic programming method, remove the outlier of two width of cloth contour images, obtain the Corresponding matching point of contour images;

(3d) to resulting Corresponding matching point, carry out conversion with procrustes analysis method;

(3e) point of the Corresponding matching after the conversion is converted to character string;

(3f) obtain two image S to be clustered through step (3a) to (3e) _iAnd S _jCharacter string;

(3g) according to the image S to be clustered that obtains _iAnd S _jCharacter string, calculate image S to be clustered _iAnd S _jBetween distance:

d(S _i,S _j)=λ ^u×E(S _i,S _j)/L，

E (S wherein _i, S _j) be image S to be clustered _iAnd S _jCharacter string between the Edit distance, λ=1.03 are constant, u is image S to be clustered _iThe outlier number of contour images, L is image S to be clustered _iThe length of character string;

(3h) to the distance between all images to be clustered that obtain, carry out normalization.

3. the heredity of k neighbour Local Search automatic clustering method, wherein all chromosomal fitness value f of the described calculating of step (5) for graph image according to claim 1 _i, calculate by following formula:

f _i=1/θ _i

${\mathrm{\&theta;}}_i=\mathrm{\&alpha;}\&times;k_i+\frac{1}{k_i}\&times;{\mathrm{\&Sigma;}}_{s=1}^{k_i}{\mathrm{\&omega;}}_s$

${\mathrm{\&omega;}}_s=\frac{2}{n_s}\underset{S_k,S_h\&Element;C_s,k<h}{\mathrm{\&Sigma;}}d{(S_k,S_h)}^2$

Wherein, α=0.5 is control parameter, k _iBy being calculated chromosomal cluster numbers, d (S _k, S _h) be image S to be clustered _kAnd S _hDistance, n _sBe the secondary number of the contained image to be clustered of s class cluster image, C _sBe the set of s class image to be clustered, S _kAnd S _hIt is the image to be clustered in the s class image to be clustered.

Images