CN105574865A - Method for extracting eyelashes based on improved ant colony algorithm - Google Patents

Abstract

The invention belongs to the technical field of iris recognition, and particularly relates to a method for extracting eyelashes based on an improved ant colony algorithm. The method is implemented according to the following steps: (1) initializing pheromone in an eyelash extraction area; putting one artificial ant, and setting each parameter of the algorithm; (2) executing step (3) and step (4) in sequence by each artificial ant; (3) selecting the positions in the next step by the artificial ants, wherein the end condition is that the artificial ants walk specified step number s or have no way out; (4) carrying out pheromone updating on the positions where the artificial ants go through; (5) after the traversing of all the artificial ants is ended, obtaining the pheromone concentration at each position of an image, and then using an OTSU algorithm to extract eyelash edges; (6) integrating the obtained eyelash edges to obtain complete eyelashes; and (7) using a sequence sliding window to eliminate noise so as to obtain the final eyelashes. When the distances of the artificial ants take the pixel of about 25, the algorithm can guarantee the speed of eyelash detection and also can guarantee the effect of eyelash detection.

Description

Based on the method improving ant group algorithm extraction eyelashes

Technical field

The invention belongs to iris recognition technology field, particularly relating to a kind of method based on improving ant group algorithm extraction eyelashes.

Background technology

Iris recognition owing to having the advantages such as ubiquity, uniqueness, stability, non-infringement, so be considered to one of biological feather recognition method of current most potentiality.The Main Function of eyelashes is used to the foreign matters such as blocks dust and enters eyes, but when extracting iris feature, eyelashes sometimes may be formed iris to some extent and block, thus have impact on the extraction of true iris feature.Therefore, at the pretreatment stage of iris recognition, detect that eyelashes occlusion area ensures the very important link of iris recognition accuracy rate exactly.

Main method at present about eyelash detection comprises: KongWaikin etc., eyelashes is divided into dispersion and assembles two classes, adopts one dimension Gabor filter and image to do convolution, if result is less than given threshold value, be then judged as eyelashes to dispersion eyelashes; For the detection assembling eyelashes, then use the little rectangular window of a 5*5, if the gray scale difference of window is less than predetermined threshold value, then the center of this rectangular window just can be judged to be an eyelashes point; First HuangJunzhou etc. extract the marginal information of noise according to phase equalization, then the information in jointing edge and region locates eyelash region; First SuhadA etc. carry out greyscale transformation to strengthen the contrast of image to image, then adopt soble operator to detect eyelashes according to given threshold value; First WalidAydi etc. obtain the diagonal angle gradient of image, then extract eyelashes by predetermined threshold value; The people such as Yuan Weiqi propose to adopt multiple diverse ways, extract eyelashes points by choosing corresponding threshold value, and then the eyelashes point that these extract is combined form final eyelashes.Carry out the firm method waiting people to use gray scale morphology, by the image binaryzation after morphology operations, can eyelash be detected.

Above method is adopt the method for artificial predetermined threshold value to judge whether certain pixel belongs to eyelashes mostly.The advantage of this method is simple, efficient, and limitation is often difficult to obtain due to optimal threshold, thus causes the result of eyelash detection satisfactory not.

Summary of the invention

The present invention is intended to overcome the deficiencies in the prior art part and provides a kind of method based on improving ant group algorithm extraction eyelashes.The effect that the method extracts eyelashes is significantly improved compared with other related algorithms; And when human oasis exploited spacing gets about 25 pixel time, algorithm can ensure that the speed of eyelash detection can ensure again the effect of eyelash detection.

For solving the problems of the technologies described above, the present invention is achieved in that

Based on the method improving ant group algorithm extraction eyelashes, can implement as follows:

(1) eyelashes extract the initialization of area information element; Put a human oasis exploited, algorithm parameters is set;

(2) every human oasis exploited performs step (3) and step (4) successively;

(3) human oasis exploited selects next step position, and termination condition covers the step number s of regulation or at the end of one's rope;

(4) Pheromone update is carried out to the position of human oasis exploited process;

(5) after all human oasis exploited traversals terminate, obtain the pheromone concentration of image position, then use OTSU algorithm to extract eyelashes edge;

(6) the eyelashes edge obtained is integrated, obtain complete eyelashes;

(7) adopt sequence moving window stress release treatment, draw final eyelashes.

As a kind of preferred version, in step of the present invention (1), in every m*m region, put a human oasis exploited.

Further, in step of the present invention (3), human oasis exploited selects next step position according to following formula:

Allowed _i=0,1 ... n-1} represents that within the scope of ant i eight neighborhood, next step allows the position selected; Wherein τ _krepresent the pheromone concentration at k place, eight neighborhood position; η _krepresent the local direction factor, η _k=θ _ik/ d _ik, wherein θ _ikrepresent the position r that first of running into along eight neighborhood direction to EA zone boundary from position i is larger than EA region initial information element concentration average ave _kthe initial information element concentration at place,

M, N represent height and the width in EA region respectively; d _ikrepresent position i and r _kbetween distance;

represent the initial information element concentration average of human oasis exploited all pixels larger than ave value from position i along eight neighborhood direction to EA zone boundary;

Parameter alpha, β, μ represent each Factor Weight respectively.

Further, in step of the present invention (4), according to following formula, Pheromone update is carried out to the position of human oasis exploited process:

τ _k(n)＝ρτ _k(n-1)+Δτ _k(n)*ω _k*v _k(n)

Wherein τ _kn () represents n-th human oasis exploited traversal after, the pheromone concentration at k place, position; ρ represents pheromones evaporation coefficient; Δ τ _kn () represents the pheromones that n-th human oasis exploited discharges at k place, position; ω _krepresent the ratio of k place, position initial information element concentration and the plain concentration average of whole EA region initial information.

Further, after the present invention's n-th human oasis exploited traversal, the ratio of k place, position initial information element concentration and the plain concentration averages of all routing informations of n-th human oasis exploited process:

Trail _nrepresent all location of pixels set of n-th human oasis exploited process; N represents trail _nthe number of set interior element.

Further, in step of the present invention (6), adopt

s(i,j)＝s _left(i,j+1)+s _right(i,j-1)

Wherein s _left(i, j) represents left hand edge image, s _right(i, j) represents right hand edge image.

Eyelash detection is an important step of iris recognition pretreatment stage, and the present invention proposes a kind of method that ant group algorithm based on improving extracts eyelashes.First the method makes human oasis exploited energy rapid aggregation to eyelashes edge by introducing the inside and outside direction factor in eyelashes region, and by taking the overall situation and local two kinds of strategies to upgrade pheromones, then use OTSU algorithm to segment the image into eyelashes edge and non-eyelashes edge two parts according to the pheromone concentration of gained image.Finally, the eyelashes edge be partitioned into being integrated, except making an uproar, obtaining final eyelashes.Experimental result shows: the effect that the method extracts eyelashes is significantly improved compared with other related algorithms; And when human oasis exploited spacing gets about 25 pixel time, algorithm can ensure that the speed of eyelash detection can ensure again the effect of eyelash detection.

Accompanying drawing explanation

Below in conjunction with the drawings and specific embodiments, the invention will be further described.Protection scope of the present invention is not only confined to the statement of following content.

Fig. 1 is that eyelashes of the present invention extract area schematic;

Fig. 2 is the comparison schematic diagram of algorithms of different to CASIA-IrisV1 storehouse iris image eyelash detection effect;

Fig. 3 is that algorithms of different is to the comparison schematic diagram from collection iris image eyelash detection effect;

Fig. 4 be ant spacing and working time relation;

Fig. 5-1, Fig. 5-2 and Fig. 5-3 is ant spacing schematic diagram.

Embodiment

As shown in the figure, based on the method improving ant group algorithm extraction eyelashes, implement as follows:

(1) eyelashes extract the initialization of area information element; Put a human oasis exploited, algorithm parameters is set;

(2) every human oasis exploited performs step (3) and step (4) successively;

(3) human oasis exploited selects next step position, and termination condition covers the step number s of regulation or at the end of one's rope;

(4) Pheromone update is carried out to the position of human oasis exploited process;

(5) after all human oasis exploited traversals terminate, obtain the pheromone concentration of image position, then use OTSU algorithm to extract eyelashes edge;

(6) the eyelashes edge obtained is integrated, obtain complete eyelashes;

(7) adopt sequence moving window stress release treatment, draw final eyelashes.

In step of the present invention (1), in every m*m region, put a human oasis exploited.

In step of the present invention (3), human oasis exploited selects next step position according to following formula:

Allowed _i=0,1 ... n-1} represents that within the scope of ant i eight neighborhood, next step allows the position selected; Wherein τ _krepresent the pheromone concentration at k place, eight neighborhood position; η _krepresent the local direction factor, η _k=θ _ik/ d _ik, wherein θ _ikrepresent the position r that first of running into along eight neighborhood direction to EA zone boundary from position i is larger than EA region initial information element concentration average ave _kthe initial information element concentration at place,

M, N represent height and the width in EA region respectively; d _ikrepresent position i and r _kbetween distance;

represent the initial information element concentration average of human oasis exploited all pixels larger than ave value from position i along eight neighborhood direction to EA zone boundary;

Parameter alpha, β, μ represent each Factor Weight respectively.

In step of the present invention (4), according to following formula, Pheromone update is carried out to the position of human oasis exploited process:

τ _k(n)＝ρτ _k(n-1)+Δτ _k(n)*ω _k*v _k(n)

Wherein τ _kn () represents n-th human oasis exploited traversal after, the pheromone concentration at k place, position; ρ represents pheromones evaporation coefficient; Δ τ _kn () represents the pheromones that n-th human oasis exploited discharges at k place, position; ω _krepresent the ratio of k place, position initial information element concentration and the plain concentration average of whole EA region initial information.

After the present invention's n-th human oasis exploited traversal, the ratio of k place, position initial information element concentration and the plain concentration averages of all routing informations of n-th human oasis exploited process:

Trail _nrepresent all location of pixels set of n-th human oasis exploited process; N represents trail _nthe number of set interior element.

In step of the present invention (6), adopt

s(i,j)＝s _left(i,j+1)+s _right(i,j-1)

Wherein s _left(i, j) represents left hand edge image, s _right(i, j) represents right hand edge image.

The present invention proposes a kind of eyelashes extracting method based on improvement ant group algorithm newly, the method is not judge eyelashes by the method for artificial predetermined threshold value, but first by human oasis exploited search eyelashes edge, the information at eyelashes edge is constantly strengthened in the process of search, weaken non-edge information simultaneously, and then be automatically found optimum eyelashes segmentation threshold by OTSU algorithm.So just can ensure good eyelashes segmentation effect.

Ant group algorithm is that the people such as Italian scholar DorigoM in 1991 gain enlightenment from the process of ant colony foraging behavior, a kind of simulated evolutionary algorithm of simulating ant behavior proposed, we are called Ant Algorithm (AS), there is due to this algorithm the features such as information positive feedback, distribution calculating and heuristic search, be successfully applied to path optimization in the last few years, fault diagnosis and Iamge Segmentation.

The human oasis exploited that the process of carrying out eyelashes extraction to iris image can be understood as the behavior of simulation ant is looked for food process.Image is regarded as two-dimensional matrix, and an element in each location of pixels homography, puts a human oasis exploited in every n*n pixel region.Feature due to eyelashes is that grey scale pixel value is lower and edge gradient is higher, and according to this feature, the search target of human oasis exploited is exactly the little and pixel that gradient is large of gray-scale value.In order to improve the precision that eyelashes extract, introduce the local direction factor at eyelashes intra-zone, this factor guide human oasis exploited to close to him and the eyelashes that pheromone concentration is higher are close; Take global policies and local policy to upgrade pheromones simultaneously respectively, make the pheromone concentration at eyelashes edge obtain enhancing further.Like this by after some step iteration, the pheromone concentration at eyelashes edge significantly higher than the pheromone concentration of non-eyelashes fringe region, and then just can will be partitioned into eyelashes edge according to OTSU algorithm to obtained Image Segmentation Using easily.Finally, the eyelashes edge be partitioned into being integrated, except making an uproar, just obtaining final eyelashes.Step of the present invention is as follows:

3.1 eyelashes extract the determination in region

Find according to observing a large amount of iris image: the scope that the eyelashes of common people block iris is the above region of pupil center, so, we select eyelashes to extract (eyelashabstract, EA) region is in image be bottom line with upper part with the center of pupil, as shown in the above dash area of thick black line in Fig. 1.

3.2 eyelashes extract the determination of area information element initial value

In order to improve the search efficiency of algorithm, the pheromones initial value τ of image _k(0) arrange as follows.

Gray _krepresent the grey scale pixel value at k place, position, grad _krepresent the pixel gradient value at k place, position, expression formula is as shown in formula (2):

grad _k＝|f(i+1,j)-f(i,j)|+|f(i,j+1)-f(i,j)|(2)

Wherein f (i, j) represents the grey scale pixel value at k place.In order to get rid of the impact of noise, the gradient of specified image is less than the threshold gamma of setting

In order to consistent with gradient, the gray scale of image is also normalized to (0 ~ γ).

Like this, then pheromones initial value τ _k(0) scope is (0 ~ 1).

In addition, due to pupil edge place gradient comparatively Datong District time gray scale lower, can interference be formed to human oasis exploited routing, so set to 0 the pheromones initial value of pupil and edge environ thereof.

The definition of 3.3 transition probabilities

Human oasis exploited at every turn from current location i, according to the position k that transition probability selects next step to arrive within the scope of eight neighborhood.Here, transition probability formula is defined as follows:

Allowed _i=0,1 ... n-1} to represent within the scope of ant i eight neighborhood next step position allowing to select (ant accessed position do not allow to select again).

Wherein τ _krepresent the pheromone concentration at k place, eight neighborhood position.

η _krepresent the local direction factor, η _k=θ _ik/ d _ik, wherein θ _ikrepresent the position r that first of running into along eight neighborhood direction to EA zone boundary from position i is larger than EA region initial information element concentration average ave _kthe initial information element concentration at place,

M, N represent height and the width in EA region respectively.D _ikrepresent position i and r _kbetween distance.When human oasis exploited comes into eyelashes intra-zone, parameter η _ijmaking human oasis exploited when searching for eyelashes edge, namely considering to have also contemplated that the edge direction that pheromone concentration is higher the edge direction of the lower but close together of pheromone concentration simultaneously, thus ensure that the precision that eyelashes extract.

represent the initial information element concentration average of human oasis exploited all pixels larger than ave value from position i along eight neighborhood direction to EA zone boundary.Because eyelashes marginal information element concentration is higher than ave, so the direction that average is larger is exactly there is the larger direction of eyelashes possibility.When human oasis exploited at eyelashes region exterior time, this parameter makes human oasis exploited can be fast close to the direction that eyelashes exist.

Parameter alpha, β, μ represent each Factor Weight respectively, and at present, the setting of ant group algorithm correlation parameter does not still have theoretic foundation, and arranging of these parameters is mainly determined by experience.

3.4 pheromone update strategy

After every human oasis exploited covers the step number of regulation, the pheromones of position adjusts according to formula (7):

τ _k(n)＝ρτ _k(n-1)+Δτ _k(n)*ω _k*v _k(n)(7)

τ in formula (7) _kn () represents n-th human oasis exploited traversal after, the pheromone concentration at k place, position.ρ represents pheromones evaporation coefficient, Δ τ _kn () represents the pheromones that n-th human oasis exploited discharges at k place, position, be constant.ω _krepresent the ratio of k place, position initial information element concentration and the plain concentration average of whole EA region initial information, ω _kbe worth larger explanation larger relative to the plain concentration of global information at k place, position.

In order to improve the pheromone concentration of target further, also adopting pheromones local updating strategy, introducing parameter v (n) _k

V _kn () represents n-th human oasis exploited traversal after, the ratio of k place, position initial information element concentration and the plain concentration averages of all routing informations of n-th human oasis exploited process.Trail _nrepresent all location of pixels set of n-th human oasis exploited process.N represents trail _nthe number of set interior element.V _kn, in all location of pixels of the larger explanation of () value n-th human oasis exploited process, the generic pixel element concentration at k place, position is larger.

After n-th human oasis exploited traversal, the pheromones increment at k place, position is according to ω _kand v _kn the carrying out be in proportion of () adjusts accordingly, it is faster that the Messages element that such pheromones is higher increases, and it is slower that the Messages element that pheromones is lower increases, thus be conducive to improving eyelashes extraction effect further.

The segmentation of 3.5 eyelashes

When all ants are according in the process of direction transition strategy and pheromone update strategy traversing graph picture, the pheromone concentration at eyelashes edge can be increasing, after all ants have traveled through, the pheromone concentration at eyelashes edge generally will be obviously high than non-eyelashes marginal information element concentration.At this moment, then adopt OTSU algorithm to Image Segmentation Using, just can be partitioned into eyelashes edge easily.

The integration of 3.6 eyelashes

Target due to ant group algorithm search is that gradient is large and the pixel that gray scale is little, and eyelashes itself exist certain width, so what adopt as above method to extract is the eyelashes that centre exists space.In order to obtain complete eyelashes, need to fill space.Carry out binaryzation to the image after segmentation, edge pixel gray-scale value establishes 255, and non-edge pixels gray-scale value is 0.Observe the gap width in the middle of finding according to great many of experiments and be approximately less than 3 pixels.In order to remove middle space, extract complete eyelashes, we obtain the image at the left and right edge of target two-value eyelashes image respectively, feature due to eyelashes gray scale is that middle low both sides are high, so the method for discrimination of left hand edge is: this position be edge (bianry image gray-scale value 255) and in original-gray image this position right pixels gray-scale value necessarily lower than current location; In like manner the method for discrimination of right hand edge to be this position be edge (bianry image gray-scale value 255) and in original-gray image this position leftmost pixel gray-scale value necessarily lower than current location.Then following formula is adopted to show final eyelashes.

s(i,j)＝s _left(i,j+1)+s _right(i,j-1)(10)

Wherein s _left(i, j) represents left hand edge image, s _right(i, j) represents right hand edge image.Because there is overlapping phenomenon, so return to 255 to the gray-scale value of the pixel of overlap according to formula (11).

3.7 adopt sequence moving window stress release treatment

Can be there are some noises in the eyelashes taking as above algorithm to extract, in order to a nearly step improves eyelashes extraction effect, need to carry out except making an uproar.Find that eyelashes are generally continuous print according to the observation, and these noises are some isolated spots.The method that we take adopts sequence moving window to carry out stress release treatment.The length of side of series of windows is 3 ~ L pixel, thickness window is 1 pixel, first, the window of 3*3 is adopted to slip over image by pixel, in sliding process, if around 8 location of pixels all do not have gray-scale value be 255 pixel, and intermediate pixel is the pixel of 255, then determine that this pixel is noise, its value is changed into 0.Then, the length of side is used to be that the window of 4 ~ L adopts same way to eliminate the noise surrounded by window successively.Attention window is selected can not be too large, not only increase program runtime if excessive, and sometimes eyelashes are mistakened as and do noise and eliminate, thus have impact on the precision that eyelashes extract; If select too small, then some noise can not be removed.Experience shows: L gets 5 ideal.

The present invention's experiment adopts the CASIA-IrisV1 iris picture library of Institute of Automation, CAS and this research department oneself to gather iris picture library respectively and verifies.Wherein iris image resolution in CASIA-IrisV1 storehouse is 320*280, and this research department is 800*600 from gathering iris image resolution.The dominant frequency of experiment service machine is 2.71GHz, and inside save as 2G, operating system is windowsxp, and programming used tool is visual c++ 2010.Due to document 1 (WANIKINK, DAVIDZ.Detectingeyelashandreflectionforaccurateirissegme ntation [J] .InternationalJournalofPatternRecognitionandArtificialIn telligence, 2003, 17 (6): 1025-1034) and document 4 (AYDIW, KAMOUNL, MASMOUDIN.AFastandAccurateEyelidsandEyelashesDetectionAp proachforIrisSegmentation [J] .JournalofMultimediaProcessingandTechnologies, 2012, 3 (4): 166-173.) be all iris image for not carrying out standardization processing.So select these two algorithms and the present invention to contrast.In addition, all can produce noise when document 1 and document 4 extract eyelashes, in order to objectively compare algorithm, these two kinds of algorithms are also adopt and method the same herein to the filtration of noise.The typical case that table 1 provides major parameter of the present invention is arranged.

Table 1 algorithm parameter is arranged

The extraction effect of 4.1 eyelashes compares

As can be seen from Figure 2: Fig. 2 (a) is original image; Fig. 2 (b) is document [1] algorithm; Fig. 2 (c) is document [4] algorithm; Fig. 2 (d) is the present invention.There is a lot of phenomenon of interrupting in the eyelashes that Fig. 2 (b), Fig. 2 (c) algorithm extract, this gray-scale value mainly due to same eyelashes different parts is different often, and decision threshold is constant causing.By contrast, the eyelashes continuity that Fig. 2 (d) of the present invention is extracted is significantly improved than algorithm pattern 2 (b), Fig. 2 (c)., it can also be seen that from figure: for the region (as eyelashes upper left) that eyelashes are closeer, other algorithms relatively, the effect that Fig. 2 (d) of the present invention is extracted is also more satisfactory meanwhile.What still have the end of some eyelashes to extract in Fig. 2 (d) is imperfect, mainly because the gray-scale value of end of these eyelashes and background more close, and Grad is less, so the more difficult identification of human oasis exploited.

As shown in Figure 3: Fig. 3 (a) is original image; Fig. 3 (b) is document [1] algorithm; Fig. 3 (c) is document [4] algorithm; Fig. 3 (d) is the present invention.The present invention is also better than other algorithms to this research department from the Detection results gathering iris image eyelashes.In Fig. 3 (d), also there is a few place by mistake the patch edge on iris image as eyelash detection out, this is mainly due to the edge feature closely eyelashes of these patches, so cause erroneous judgement.In addition, because this visible ray iris picture color is comparatively dark, closely, this also causes some part of part eyelashes and eyelashes to detect to the gray-scale value of its gray-scale value and eyelashes.

The relation of 4.2 ant spacing and working time and eyelash detection effect

In order to analyze the relation of working time and human oasis exploited spacing, a human oasis exploited is put respectively at different pixels spacing place, obtain the relation of the average operating time of algorithm and human oasis exploited spacing as shown in Figure 4, as shown in Figure 4: human oasis exploited spacing is larger, Riming time of algorithm is shorter; And human oasis exploited spacing is less, Riming time of algorithm is longer.When human oasis exploited spacing is less than about 20 pixel time, working time of program declines very fast along with the increase of human oasis exploited spacing; And when human oasis exploited spacing is greater than about 20 pixel time, program runtime declines slower;

In addition, from experiment: human oasis exploited spacing is larger, the effect that eyelashes extract is poorer; Spacing is less, and extraction effect is better, is the working time and operational effect adopting the present invention to use different interval number human oasis exploited extraction eyelashes to spend for Fig. 2 (a), Fig. 5-1, Fig. 5-2 and Fig. 5-3 respectively.Can find out: Fig. 5-1 is ant spacing: 5 pixels, working time: 11.273s; Fig. 5-2 is ant spacing, 15 pixels, working time: 1.502s; Fig. 5-3 is ant spacing: 40 pixels, working time: 0.458s.Fig. 5-1 and Fig. 5-2 effect difference not obvious, Fig. 5-3 extract eyelashes then there is obvious fracture.

Balance as above two aspects factor and through experimental verification repeatedly: getting about 25 pixels when choosing human oasis exploited skip number n, good eyelashes extraction effect can be ensured, make again the working time of calling program shorter.

Be understandably, above about specific descriptions of the present invention, the technical scheme described by the embodiment of the present invention is only not limited to for illustration of the present invention, those of ordinary skill in the art is to be understood that, still can modify to the present invention or equivalent replacement, to reach identical technique effect; Needs are used, all within protection scope of the present invention as long as meet.

Claims (6)

1. based on the method improving ant group algorithm extraction eyelashes, it is characterized in that, implement as follows:

(1) eyelashes extract the initialization of area information element; Put a human oasis exploited, algorithm parameters is set;

(2) every human oasis exploited performs step (3) and step (4) successively;

(3) human oasis exploited selects next step position, and termination condition covers the step number s of regulation or at the end of one's rope;

(4) Pheromone update is carried out to the position of human oasis exploited process;

(5) after all human oasis exploited traversals terminate, obtain the pheromone concentration of image position, then use OTSU algorithm to extract eyelashes edge;

(6) the eyelashes edge obtained is integrated, obtain complete eyelashes;

(7) adopt sequence moving window stress release treatment, draw final eyelashes.

2. the method based on improving ant group algorithm extraction eyelashes according to claim 1, is characterized in that: in described step (1), puts a human oasis exploited in every m*m region.

3. the method based on improving ant group algorithm extraction eyelashes according to claim 2, is characterized in that: in described step (3), human oasis exploited selects next step position according to following formula:

Allowed _i=0,1 ... n-1} represents that within the scope of ant i eight neighborhood, next step allows the position selected; Wherein τ _krepresent the pheromone concentration at k place, eight neighborhood position; η _krepresent the local direction factor, η _k=θ _ik/ d _ik, wherein θ _ikrepresent the position r that first of running into along eight neighborhood direction to EA zone boundary from position i is larger than EA region initial information element concentration average ave _kthe initial information element concentration at place,

$ave=\frac{\underset{k=1}{\overset{M*N}{\Σ}}{\mathrm{\τ}}_k\left(0\right)}{M*N}$

M, N represent height and the width in EA region respectively; d _ikrepresent position i and r _kbetween distance;

represent the initial information element concentration average of human oasis exploited all pixels larger than ave value from position i along eight neighborhood direction to EA zone boundary;

Parameter alpha, β, μ represent each Factor Weight respectively.

4. the method based on improving ant group algorithm extraction eyelashes according to claim 3, is characterized in that: in described step (4), carry out Pheromone update: τ according to following formula to the position of human oasis exploited process _k(n)=ρ τ _k(n-1)+Δ τ _k(n) * ω _k* v _k(n)

Wherein τ _kn () represents n-th human oasis exploited traversal after, the pheromone concentration at k place, position; ρ represents pheromones evaporation coefficient; Δ τ _kn () represents the pheromones that n-th human oasis exploited discharges at k place, position; ω _krepresent the ratio of k place, position initial information element concentration and the plain concentration average of whole EA region initial information.

5. according to claim 4ly extracting the method for eyelashes based on improving ant group algorithm, it is characterized in that: after n-th human oasis exploited traversal, the ratio of k place, position initial information element concentration and the plain concentration averages of all routing informations of n-th human oasis exploited process:

$v_k\left(n\right)=\frac{{\mathrm{\τ}}_k\left(0\right)}{{\underset{s\∈trail}{\Σ}}_n{\mathrm{\τ}}_s\left(0\right)/N}$

Trail _nrepresent all location of pixels set of n-th human oasis exploited process; N represents trail _nthe number of set interior element.

6. the method based on improving ant group algorithm extraction eyelashes according to claim 5, is characterized in that: in described step (6), adopts

s(i,j)＝s _left(i,j+1)+s _right(i,j-1)

Wherein s _left(i, j) represents left hand edge image, s _right(i, j) represents right hand edge image.