CN1885312A - Iris positioning method based on morphology and probability statistic - Google Patents

Abstract

The invention provides an iris positioning method based on the combination of mathematical morphology and probability statistics. It first binarizes the iris image, uses projection to obtain the iris inner circle and roughly locates the center of the circle, and then uses morphology to extract the edge of the iris inner circle, and uses the inner circle to roughly locate the center of the iris (and its surrounding points) to its edge every The distance of the point is used as a sample, so the three parameters of the search circle center and radius are converted into a sample of the minimum mean value, and the mean value of this sample is the radius of the circle. After the inner circle is determined, the edge of the outer circle is obtained by binarization and mathematical morphology, and then the center of the inner circle is used as the center of the outer circle to roughly locate the center of the outer circle, and the method of locating the inner circle is used to locate the outer circle. By adopting the iris positioning method of the invention, the standard error is the smallest, the precision is high, and the speed is fast.

Description

Iris locating method based on mathematical morphology and probability statistics

Technical field

Based on the iris locating method of mathematical morphology and probability statistics, belong to the picture intelligence processing technology field, particularly the iris locating method in the iris recognition technology.

Background technology

Bio-identification is the main direction of current information security developments, is the forward position research topic of present world information security field.The iris biological identification technology then be easy to operate in the present biological identification technology, precision is high, the technology of market outlook is arranged most.It is the important means that solves information security issue, is the application at information security field of computer image processing technology and mode identification technology.This technology is solving the problems of information security field, as access control system, and the information security of network technology, the ecommerce of network economy is widely used in the particular problems such as the accurate authentication of personal identification, has huge economic and realistic meaning.Simultaneously, it occupies critical role in the development of I.D. technology and social safety management.Along with the raising of computing power and the development of image processing techniques, the iris biological recognition system reaches its maturity.See document for details: Clarke R.HumanIdentification System:Management Challenges and Public Policy Issues.InformationTechnology ﹠amp; People, 1994,7 (4): 6～37 and document: the model monarch. the iris recognition biotechnology excites 1,000,000,000 security marketplaces.Golden Card Program, 2003,5:61-64 is described.

In the iris biological identification technology, the location of iris is the emphasis and the difficult point of whole recognition technology.The positioning time of iris and precision directly influence the performance of whole iris biological recognition system.Iris Location is meant accurately locatees the iris portion of people's eyes, this is a most important and crucial step of iris recognition, can can iris locate exactly, be related to extracting exactly of next step iris texture characteristic, thereby analysis result is produced significant effects.And in Iris Location, the outer peripheral extraction of iris then is a difficult point.Therefore, how to have now on all valuable achievements in research, shortening positioning time and improve the main direction that the particularly outer peripheral bearing accuracy of bearing accuracy will become our current research.See document for details: Wang Chun, leaf tiger year.Research on Algorithm of Iris Recognition [J]. Guizhou University of Technology's journal (natural science edition), 2000,29 (3): 4852. and document: Li Qingrong, horse strives. Iris Location algorithm research [J]. and University of Electronic Science and Technology's journal, 2002,31 (1): 79. and document: Daugman.Statisticalrichness of visual phase information Update on recognizing persons by their iris pattern.International Journal of computer Vision, 2001,45 (1): 25-38. is described.

The method of present normally used iris has:

(1) based on the Iris Location algorithm of edge extracting and Hough conversion.It extracts the outer edge of iris by edge detection operator, uses the Hough conversion to locate the inside and outside circle of iris then.Its shortcoming is that the Hough conversion need be provided with threshold value.Because be subjected to the influence of light, the size of pupil changes, thereby the quantity of putting on the iris edge also changes, and for a large amount of iris images, one accurate and appropriate threshold is difficult to obtain.And this method is with the increase of counting, and operand is pushed the speed slack-off.See document for details: Wang Chun, leaf tiger year. Research on Algorithm of Iris Recognition [J]. Guizhou University of Technology's journal (natural science edition), 2000,29 (3): 4852.

(2) two-step approach Iris Location algorithm.It at first does x direction Gray Projection according to the intensity profile characteristics of eye and y direction Gray Projection is carried out coarse positioning, uses operator then Accurately locate.Its shortcoming integral operation amount is big, and speed is slow.See document for details: Daugman.Statistical richness of visual phase informationUpdate on recognizing persons by their iris pattern.International Journal of computer Vision, 2001,45 (1): 25-38.

Summary of the invention

Task of the present invention provides a kind of iris locating method based on mathematical morphology and probability statistics, and it has characteristics such as the high and speed of bearing accuracy is fast.

In order to describe content of the present invention easily, at first introduce several notions, and some terms are defined.

Notion one: mathematical morphology.Mathematical morphology is based on graphical analysis, and the form of removing the dimensioned plan picture with " structural element " with certain morphosis is to solve the image understanding problem.Morphologic basis is corrosion and dilation operation, and the open and close computing that produces therefrom.The formula of corrosion and dilation operation is respectively: U=A Θ B={U: B+U  A}, and V=A  B={V: (B+V) ∩ A ≠ Φ }; The formula of open and close computing is respectively: A ο B=(A Θ B)  B, AB=(A  B) Θ B.Wherein, A is an original image, and B is a structural element, and U is the image that original image obtains after corrosion, V is the image that original image obtains after expanding, and Θ is the erosion operation symbol, and  is the dilation operation symbol, ο is the opening operation symbol, is the closed operation symbol, and Φ is the empty set symbol.

Notion two: probability statistics.Suppose that { it is u=2.5 that x} obeys average to array, and variance is σ ²Normal distribution.According to theory of probability statistics, along with σ ²Reduce array { x _iWith the absolute value of the difference of average with also just reduce accordingly.That is to say that { distribution of the value of x} is along with σ for array ²Reduce and more and more concentrate on the both sides of average.Therefore, work as σ ²Be substantially equal at 0 o'clock, array { x _iThe distribution of value also just be substantially equal to average.And if on circle have a few to the distance in the center of circle as an array { x _i, array { x then _iValue be radius r, standard deviation is 0.The present invention is so that certain puts the distance put on its edge as a sample value in the circle.So that certain puts the distance of being had a few on its edge as a sample in the circle.

Notion three: standard error.If the error of n measured value is ε ₁, ε ₂... ε _n, then the standard error σ of this group measured value equals: $\mathrm{\σ}=\sqrt{({\mathrm{\ϵ}}_1^2+{\mathrm{\ϵ}}_2^2+\mathrm{\Λ\Λ}+{\mathrm{\ϵ}}_n^2)/n}=\sqrt{\mathrm{\Σ}{\mathrm{\ϵ}}_i^2/n}.$ Because measured true value is a unknown number, the error of each measured value is not all known yet, therefore can not try to achieve standard error by following formula.That can access during measurement is arithmetic mean (N), and it is near true value (N), and calculates the poor of measured value and arithmetic mean easily, is called residual error and (is designated as v).Theoretical analysis shows the standard error σ that can represent certain measurement result in limited number of time (n time) observation with residual error v, and its computing formula is: $\mathrm{\σ}=\sqrt{[{(N_1-N)}^2+{(N_2-N)}^2+\mathrm{\Λ\Λ}{(N_n-N)}^2]/n}=\sqrt{{\mathrm{\Σ\υ}}_i^2/n}$ It should be noted that standard error is not the actual error of measured value, neither error range, it is just to the estimation of one group of reliable measuring data.Standard error is little, and measuring reliability is big, otherwise, measure just little reliable.

Definition 1: iris.Part in human eye between sclera and the pupil is called iris.Iris between the different people has minutia and texture image at random, and these features keep suitable stability in life the people's, and the protection that it is subjected to eyelid, cornea is malleable not.

Definition 2: two dimension median filter: two dimension median filter (Median filtering) is based on a kind of nonlinear signal processing technology that can effectively suppress noise of sequencing statistical theory, its detailed information of guard signal well in filtering noise (especially impulsive noise).Its ultimate principle is with the replacement of the intermediate value of each point value in the neighborhood of this point any value in digital picture or the Serial No..

Definition 3: binaryzation process.The all values of entire image is changed into the process of having only two kinds of values, and generally these two kinds of values are 0 and 1 or 0 and 255.Its concrete method is: when the value on the image was greater than or equal to the binaryzation threshold values, the value of this point was turned to 1 (or 255) by two-value; When the value on the image less than the binaryzation threshold values time, the value of this point is turned to 0 by two-value.

Definition 4: grey level histogram.The grey level range of image is 0,1 ..., 255, the pixel count of establishing gray level i is n _i, then total pixel of piece image is $N=\underset{i=0}{\overset{255}{\mathrm{\Σ}}}{n}_i,$ The definition of probability that gray level i occurs is $P_i=\frac{n_i}{N}.$ Grey level histogram is the pixel count n of gray level i _iWith the two-dimentional relation of gray level i, it has reflected the statistical property of intensity profile on the piece image, becomes the basis that utilizes pixel grey scale to make the dividing method of attribute.

Definition 5: horizontal projection.A kind ofly add up by horizontal direction, reality is transformed into method in the one-dimensional space to the gradation of image horizontal distribution value in the two-dimensional space, and this transforming function transformation function is $P\left(x\right)=\underset{y}{\mathrm{\Σ}}I(x,y).$ Wherein (P (x) is the capable horizontal projection value of x to I for x, y) the gradation of image value of the capable y of expression x row, and the variation range of y is from 1 to n, the length of n representative image.

Definition 6: vertical projection.A kind ofly add up by vertical direction, reality is transformed into method in the one-dimensional space to the gradation of image vertical distribution value in the two-dimensional space, and this transforming function transformation function is $P\left(y\right)=\underset{x}{\mathrm{\Σ}}I(x,y).$ Wherein (P (y) is the vertical projection value of y row to I for x, y) expression x capable y row gradation of image value, and the variation range of x is from 1 to m, the width of m representative image.

Definition 7: morphology edge extracting.The border of set A is designated as β (A), and can extract the edge by following algorithm: establishing B is a suitable structural element, at first makes A be corroded by B, asks the poor of set A and its corrosion then.Be shown below: β (A)=A-(A Θ B).

Definition 8: cylindrical binary-state threshold.Selected thresholding when the iris image cylindrical is carried out binaryzation.Its computing formula is: $\mathrm{yuzhi}=\left[\frac{1}{2}(\mathrm{aver}1+\mathrm{aver}2)\right]/256.$ Wherein yuzhi is the binaryzation threshold values, (annulus is the center of circle with the interior round heart to aver1 for the average through the annulus behind the two dimension median filter, radius of circle is interior radius of circle+5 in the annulus, the annulus exradius is interior radius of circle+10), aver2 is that (center of square is respectively that the row-coordinate with the interior round heart is a row-coordinate for the average of the square of two 11 * 11 sizes, subtracting 130 with the row coordinate of the interior round heart is the row coordinate and is row-coordinate with the row-coordinate of the interior round heart, adding 140 with the row coordinate of the interior round heart is the row coordinate), / be multiplication sign, the 256th, for normalization yuzhi.

Definition 9: Wiener filtering.S filter is a kind of linear smoothing wave filter, and it is a kind of sef-adapting filter, can adjust the output of wave filter according to the regional variance of image ${\mathrm{\δ}}^2=\frac{1}{L}\underset{(x,y)\∈l}{\mathrm{\Σ}}[F^2(x,y)-G^2(x,y)]$ Wherein, L is the neighborhood M * N that chooses, and (x y) adjusts preceding pixel point value to F, and (x is neighborhood averaging value (as defining shown in 5) y) to G, and δ is a mean square deviation.By neighborhood m * n estimation mean value and standard deviation, to the filtering of image applications pixel smooth adaptive: $F^{*}(x,y)=G(x,y)+\frac{{\mathrm{\δ}}^2+v^2}{{\mathrm{\δ}}^2}[F(x,y)-G(x,y)],$ Wherein, v ²Be noise variance.

Detailed technology scheme of the present invention is:

Based on the iris locating method of mathematical morphology and probability statistics, it comprises the following step:

Step 1, by iris image camera head (special a kind of image pickup device of making) for the iris image that obtains to carry out texture analysis, acquisition can be carried out the eyes image of iris texture analysis.This image is only to comprise monochrome information and without any the gray level image of other colouring informations.

Step 2, the iris image of input is carried out the two dimension median filter of 7 * 7 windows, reach the purpose of level and smooth iris image, weaken the inhomogeneous influence of illumination binaryzation.

The coarse positioning of the round heart in step 3, the iris specifically may further comprise the steps:

Step 1), obtain the extreme value of first crest of the grey level histogram of the iris image behind the two dimension median filter, the gray level i+5 when obtaining extreme value with this is a thresholding, if I (x, y)＞i+5, make I (x, y)=0; If I (x, y)＜i+5, make I that (x, y)=1, wherein (x, y) image intensity value of the capable y row of expression x is carried out binaryzation to iris image to I.Use structural element se (se is that radius is 15 disk) that bianry image is carried out the form opening operation then, reject other isolated points and the face of pupil part, level and smooth iris internal circle edge.

Step 2), the bianry image of handling through the form opening operation is carried out the coarse positioning that level and vertical projection obtain circle in the iris

The utilization formula $P\left(x\right)=\underset{y}{\mathrm{\Σ}}I(x,y)$ (I (x wherein, y) image intensity value of the capable y row of expression x, P (x) is the capable horizontal projection value of x, the variation range of y is to n from 1, the length of n representative image) obtain the horizontal projection value of the every row of bianry image, with row-coordinate of that row of horizontal projection value maximum row-coordinate a as the round heart in the coarse positioning ₁₁The utilization formula $P\left(y\right)=\underset{x}{\mathrm{\Σ}}I(x,y)$ (I (x wherein, y) image intensity value of the capable y row of expression x, P (y) is the vertical projection value of y row, the variation range of x is to m from 1, the width of m representative image) obtains the vertical projection value of the every row of bianry image, with the row coordinate of that row of vertical projection value maximum row coordinate b as the round heart in the coarse positioning ₁₁

The accurate location of circle in step 4, the iris specifically may further comprise the steps:

Step 1), the bianry image that process form opening operation is handled carry out the morphology edge extracting

Utilization formula β (A)=A-(A Θ B) (wherein β (A) is the border of set A, and B is a suitable structural element) carries out edge extracting to the bianry image of handling through step 3, obtains the inward flange of iris.

Step 2), with the coarse positioning center of circle (a round in the iris ₁₁, b ₁₁) be basic point, every point-to-point (a on the inward flange of calculating iris ₁₁, b ₁₁) distance ${{\stackrel{\mathrm{\ω}}{D}}_1^2=(\stackrel{\mathrm{\ω}}{i}-a_{11}\·E)}^2+{(\stackrel{\mathrm{\ω}}{j}-b_{11}\·E)}^2$ (wherein Every coordinate on the expression inward flange, E is and vector Value with size is 1 vector, Represent a sample).Use formula ${{\stackrel{\mathrm{\ω}}{D}}_2^2=(\stackrel{\mathrm{\ω}}{i}-a_{11}\·E)}^2+{(\stackrel{\mathrm{\ω}}{j}-(b_{11}+1)\·E)}^2$ (wherein

Every coordinate on the expression inward flange, E is and vector

Value with size is 1 vector, Represent a sample) calculate every point-to-point (a on the iris inward flange ₁₁, b ₁₁+ 1) distance.Continue to calculate every point-to-point (a on the iris inward flange ₁₁, b ₁₁+ 2) distance Up to calculating every point-to-point (a on the iris inward flange ₁₁+ 9, b ₁₁+ 9) distance

Step 3), according to formula ${\mathrm{\σ}}^2=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(x_i-u)}^2$ (x wherein _iThe sample value of expression sample, n represents the capacity of sample, u represents sample average), calculate each sample

The variance of (k=[1,100]) is found out the sample of variance minimum According to formula $u=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{x}_i$ (x wherein _iThe sample value of expression sample, n represents the capacity of sample), obtain sample Average r; Obtain the coordinate of corresponding point according to the value of z $(a_{11}+\mathrm{qs}\left(\frac{z}{10}\right),b_{11}+\mathrm{qy}\left(\frac{z}{10}\right))$ (wherein qs represents to ask integer quotient, and qy represents to ask integer remainder).Coordinate then $(a_{11}+\mathrm{qs}\left(\frac{z}{10}\right),b_{11}+\mathrm{qy}\left(\frac{z}{10}\right))$ The center of circle (a of circle in being ₁, b ₁), average r is interior radius of circle.

The location of step 5, iris cylindrical specifically may further comprise the steps:

Step 1), according to formula $\mathrm{yuzhi}=\left[\frac{1}{2}(\mathrm{aver}1+\mathrm{aver}2)\right]/256$ (wherein yuzhi is the binaryzation threshold values, (annulus is the center of circle with the interior round heart to aver1 for the average through the annulus behind the two dimension median filter, radius of circle is interior radius of circle+5 in the annulus, the annulus exradius is interior radius of circle+10), aver2 is that (center of square is respectively that the row-coordinate with the interior round heart is a row-coordinate for the average of the square of two 11 * 11 sizes, subtracting 130 with the row coordinate of the interior round heart is the row coordinate and is row-coordinate with the row-coordinate of the interior round heart, adding 140 with the row coordinate of the interior round heart is the row coordinate), / be multiplication sign, the 256th, for normalization yuzhi), the threshold value of calculating cylindrical binaryzation.Carry out binaryzation with the iris image of this threshold value after then, obtain image B W1 two dimension median filter.

Step 2), image B W1 is carried out the form opening operation, obtain image B W2, thereby obtain image B W3=BW1-BW2 with structural element se (se is that radius is 15 disk).Image B W1 is carried out complementary operation obtain image B W4, obtain image B W5=BW4+BW3 then.Image B W5 is carried out the zone fill, obtain image B W6.Fill the hole in the bianry image, eliminate the influence that the hole in the iris outward flange extracts for the iris outward flange.Image B W6 is carried out the morphology edge extracting.Utilization formula β (A)=A-(A Θ B) (wherein β (A) is the border of set A, and B is a suitable structural element) carries out edge extracting to image B W6, obtains the outward flange of iris.Owing to be subjected to the influence of upper eyelid and lower eyelid, a part of circular arc (outward flange according to the interior round heart and iris obtains) of therefore only getting the right and left positions.

Step 3), with the round heart (a in the iris ₁, b ₁) be basic point, every point-to-point (a on the outward flange of calculating iris ₁, b ₁) distance ${\stackrel{\mathrm{\ω}}{D}}_1^2={(\stackrel{\mathrm{\ω}}{i}-a_1\·E)}^2+{(\stackrel{\mathrm{\ω}}{j}-b_1\·E)}^2$ (wherein

Every coordinate on the expression outward flange, E is and vector

Value with size is 1 vector, Represent a sample).Use formula ${{\stackrel{\mathrm{\ω}}{D}}_2^2=(\stackrel{\mathrm{\ω}}{i}-a_1\·E)}^2+{(\stackrel{\mathrm{\ω}}{j}-(b_1+1)\·E)}^2$ (wherein Every coordinate on the expression outward flange, E is and vector

Value with size is 1 vector, Represent a sample), calculate every point-to-point (a on the iris outward flange ₁, b ₁+ 1) distance.Continue to calculate every point-to-point (a on the iris inward flange ₁, b ₁+ 2) distance

Up to calculating every point-to-point (a on the iris inward flange ₁+ 9, b ₁+ 9) distance

Step 4), each sample is carried out Wiener filtering, to reduce non-iris outward flange point to location influence.Then according to formula ${\mathrm{\σ}}^2=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(x_i-u)}^2$ (x wherein _iThe sample value of expression sample, n represents the capacity of sample, u represents sample average), calculate each sample Variance, find out the sample of variance minimum According to formula $u=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{x}_i$ (x wherein _iThe sample value of expression sample, n represents the capacity of sample), obtain sample Average R; Obtain the coordinate of corresponding point according to the value of z $(a_1+\mathrm{qs}\left(\frac{z}{10}\right),b_1+\mathrm{qy}\left(\frac{z}{10}\right))$ (wherein qs represents to ask integer quotient, and qy represents to ask integer remainder).Coordinate then $(a_1+\mathrm{qs}\left(\frac{z}{10}\right),b_1+\mathrm{qy}\left(\frac{z}{10}\right))$ Be the center of circle (a of cylindrical ₂, b ₂), average R is an exradius.

By above step, just can from the original image that contains iris, extract iris.

Need to prove:

1. the iris image in the step 1 derives from the CASIA1.0 iris database of Institute of Automation, CAS.

2. (se is a disk with structural element se in (1) step by step of step 3, radius is 15) carry out the form opening operation, and without other structural elements, being because it all is 15 disk less than radius that experimental result shows the outer connection object of circle in its iris of binary image, is the influence of other points outside the circle in 15 disc structure element can be eliminated fully with radius.

3. in (2) step by step of step 3, in fact the maximal value of level and vertical projection is all more than one, but adjacent several.This is because view data is to represent with the form of matrix in computing machine, then very short line segment can occur on circumference with the matrix representation circle.And the interior circle after (1) step by step of step 3 processing is not the disk of a rule, and emulation shows that the maximal value of its level and vertical projection is adjacent one section.The present invention chooses the peaked coordinate that occurs at first and adds a side-play amount, as the interior round heart of coarse positioning, and in practice, adjusts side-play amount according to different iris storehouses.

4. in (1) step by step and (2) step by step in the step 5 in the step 4, utilization morphology carry out edge extracting be because utilization morphology to carry out edge extracting faster than the speed that the utilization edge detection operator carries out edge extracting.

5. in (2) step by step in the step 5, the bianry image in (1) step by step in the step 5 being carried out subtraction between opening operation, image and the additive operation between image, is in order further to extract the iris outward flange.The purpose of carrying out filling in the zone then is the hole of eliminating in the iris outward flange, and the zone in the outward flange is communicated with fully, so that the outward flange of the edge that extracts during morphology edge extracting iris only the time, and does not have the influence of other points.And a part of circular arc (exradius according to the interior round heart and estimation obtains) of only getting the right and left positions, be because its upper eyelid of most of iris images or lower eyelid often are to have hidden outer peripheral coboundary of iris or lower boundary, only need the circular arc of the right and left just can accurately locate and carry out iris cylindrical location.Therefore, for the versatility of algorithm with avoid unnecessary calculating, a part of circular arc of only getting the right and left positions.

6. in (4) step by step of step 5, each sample being carried out Wiener filtering, is for level and smooth sample value, and the sample value that reduces non-marginal point is to the outward flange location influence.

7. the hunting zone in (3) step by step of (2) step by step of step 4 and step 5 is according to obtaining after the experiment of the image simulation in the CASIA1.0 iris database.And for different iris storehouses can be suitable the scope of expansion search.

The method that combines based on mathematical morphology and probability statistics that the present invention proposes can improve performances such as the speed of location and bearing accuracy effectively.Utilization form opening operation can smoothed image profile, weaken narrow part, remove thin outstanding.Among the present invention the form opening operation can eliminate in the iris outside the circle and outside the iris cylindrical less than the influence of the non-boundary member of structural element to Iris Location.Thereby the bianry image of step 3 is carried out the form opening operation can be so that it only remains the pupil part, to the bianry image of step 13 carry out the adding of form opening operation and image, subtraction can be so that it only remains pupil part, iris portion and upper eyelid part or lower eyelid part.

The method that method of the present invention adopts mathematical morphology and probability statistics to combine, the extraction of the outer edge of iris has been realized in the interior round zone and the cylindrical zone of at first adopting the method for medium filtering, binaryzation and mathematical morphology to give prominence to iris; Use the standard error principle will locate the circle that obtains then at every turn and see certain measurement result in limited number of time (n time) observation as, the standard error σ of measurement result then is the estimation to one group of reliable measuring data, standard error is little, measuring reliability is big, otherwise, measure just not quite reliably, realized the location of the iris center of circle and radius is converted into the variance of sample and asking for of average according to probabilistic statistical method again.Adopt iris locating method method of the present invention, not only can accurately locate the outer edge of iris, and the computing velocity of method has also satisfied the needs of real-time.

Innovation part of the present invention is:

1, combining form is learned relevant knowledge, utilize medium filtering and carry out the method for binaryzation by the image of choose reasonable binaryzation threshold values after medium filtering, the extraction of iris outer edge has been realized, particularly outer peripheral extraction in the interior circle and the cylindrical position of outstanding iris in original image.The outer peripheral method of extraction of the present invention is not only rapid, and very accurate.

2, utilization standard error principle, see the circle that obtains of location at every turn as limited number of time (n time) certain measurement result in observing, the standard error σ of measurement result then is the estimation to one group of reliable measuring data, standard error is little, measuring reliability is big, otherwise, measure just little reliable.Realized the location of the iris center of circle and radius is converted into the variance of sample and asking for of average according to probabilistic statistical method.

3, the method that combines based on mathematical morphology and probability statistics has improved the speed of Iris Location, has satisfied the needs of iris authentication system real-time.At first utilize binaryzation and morphology methods, extract the outer edge of iris, use standard error principle and probability statistics principle again the outer edge of iris accurately to be located then.

Description of drawings

Fig. 1 is the original iris image synoptic diagram that contains iris;

Wherein, 1 expression people's upper eyelid; The annulus at 2 places is partly represented people's iris; 3 expression people's pupil; 4 expression people's sclera; 5 expression people's lower eyelid.

Fig. 2 is the Iris Location image synoptic diagram that the present invention finally obtains;

Wherein, great circle is represented the outward flange of iris, and roundlet is represented the inward flange of iris.

Fig. 3 is a schematic flow sheet of the present invention.

Specific implementation method

Adopt method of the present invention, at first use Matlab language compilation Iris Location software, the CASIA1.0 iris database of Institute of Automation, CAS is that source data is input in the film positioning software and handles then; Comprise the iris image that can carry out texture analysis through obtaining a width of cloth behind circle location in the iris and the cylindrical location.The iris image that adopts totally 812 different people and different times same person is accurately oriented 800 as source data, and the accuracy rate of location is 98.52%, locatees a width of cloth iris image required time and is no more than 1s.

In sum, method of the present invention makes full use of the characteristic of morphology and probability statistics, in conjunction with the gray distribution features of iris eyes image, thereby realizes orienting iris rapidly and accurately from the iris image that is provided.

Claims (1)

1, based on the iris location method of mathematical morphology and probability statistics, it is characterized in that, it comprises the following steps: Step 1. Obtain an eye image capable of iris texture analysis through an iris image camera device; Step 2. Carry out two-dimensional median filtering of 7×7 windows on the input iris image to achieve the purpose of smoothing the iris image and weaken the influence of uneven illumination on binarization; Step 3, coarse positioning of the center of the inner circle of the iris, specifically includes the following steps: Step 1), obtain the extremum of the first peak of the grayscale histogram of the iris image after the two-dimensional median filtering, take the grayscale i+5 when obtaining the extremum as the threshold, if I(x, y)>i+5, let I(x, y)=0; if I(x, y)<i+5, let I(x, y)=1, where I(x, y) means row x The image intensity value in the yth column is used to binarize the iris image; then, the disc structure element se with a radius of 15 is used to perform morphological opening operation on the binary image, and other isolated points and surfaces in the pupil part are eliminated to smooth the inner circle of the iris edge; Step 2), perform horizontal and vertical projection on the binary image processed by the morphological opening operation to obtain the rough positioning of the inner circle of the iris using the formula $P\left(x\right)=\underset{\mathrm{the\; y}}{\mathrm{\&Sigma;}}I(x,\mathrm{the\; y})$ (where I(x, y) represents the image intensity value of row x and column y, P(x) is the horizontal projection value of row x, the range of y is from 1 to n, and n represents the length of the image) Get the horizontal projection value of each row of the binary image, and use the row coordinate of the row with the largest horizontal projection value as the row coordinate a ₁₁ of the center of the inner circle for rough positioning; use the formula $P\left(\mathrm{the\; y}\right)=\underset{x}{\mathrm{\&Sigma;}}I(x,\mathrm{the\; y})$ (where I(x, y) represents the image intensity value of row x and column y, P(y) is the vertical projection value of column y, the range of x is from 1 to m, and m represents the width of the image) Get the vertical projection value of each column of the binary image, and use the column coordinate of the column with the largest vertical projection value as the column coordinate _b11 of the center of the inner circle for rough positioning; Step 4, precise positioning of the inner circle of the iris, specifically includes the following steps: Step 1), perform morphological edge extraction on the binary image processed by morphological opening operation Use the formula β(A)=A-(AΘB) (wherein β(A) is the boundary of set A, and B is a suitable structural element) to perform edge extraction on the binary image processed in step 3 to obtain the inner edge of the iris ; Step 2), taking the rough positioning center (a ₁₁ , b ₁₁ ) of the inner circle of the iris as the base point, calculate the distance from each point on the inner edge of the iris to the point (a ₁₁ , b ₁₁ ) ${\stackrel{\mathrm{\&omega;}}{\mathrm{D.}}}_1^2={(\stackrel{\mathrm{\&omega;}}{i}-a_{11}\&CenterDot;\mathrm{E.})}^2+{(\stackrel{\mathrm{\&omega;}}{j}-b_{11}\&Center\; Dot;\mathrm{E.})}^2$ (in

Represents the coordinates of each point on the inner edge, E is the vector a vector of 1s of the same size,

represents a sample); use the formula ${\stackrel{\mathrm{\&omega;}}{\mathrm{D.}}}_2^2={(\stackrel{\mathrm{\&omega;}}{i}-a_{11}\&Center\; Dot;\mathrm{E.})}^2+{(\stackrel{\mathrm{\&omega;}}{j}-(b_{11}+1)\&CenterDot;\mathrm{E.})}^2$ (in

Represents the coordinates of each point on the inner edge, E is the vector a vector of 1s of the same size,

represents a sample) Calculate the distance from each point on the inner edge of the iris to point (a ₁₁ , b ₁₁ +1); continue to calculate the distance from each point on the inner edge of the iris to point (a ₁₁ , b ₁₁ +2) Until the distance from each point on the inner edge of the iris to point (a ₁₁ +9, b ₁₁ +9) is calculated Step 3), according to the formula ${\mathrm{\&sigma;}}^2=\frac{1}{\mathrm{no}}\underset{i=1}{\overset{\mathrm{no}}{\mathrm{\&Sigma;}}}{(x_i-u)}^2$ (where _xi represents the sample value of the sample, n represents the capacity of the sample, and u represents the sample mean), calculate each sample ${\stackrel{\mathrm{\&omega;}}{\mathrm{D.}}}_k(k=[\mathrm{1,100}\left]\right)$ The variance of , find the sample with the smallest variance According to the formula $u=\frac{1}{\mathrm{no}}\underset{i=1}{\overset{\mathrm{no}}{\mathrm{\&Sigma;}}}{x}_i$ (where x _i represents the sample value of the sample, and n represents the capacity of the sample), to find the sample The mean value r; find the coordinates of the corresponding point according to the value of z $(a_{11}+\mathrm{qs}\left(\frac{z}{10}\right),b_{11}+\mathrm{qy}\left(\frac{z}{10}\right))$ (where qs means to seek integer quotient, and qy means to seek integer remainder); then the coordinates of the point $(a_{11}+\mathrm{qs}\left(\frac{z}{10}\right),b_{11}+\mathrm{qy}\left(\frac{z}{10}\right))$ That is, the center of the inner circle (a ₁ , b ₁ ), and the mean value r is the radius of the inner circle; Step 5, the positioning of the outer circle of the iris, specifically includes the following steps: Step 1), according to the formula $\mathrm{yuzhi}=\left[\frac{1}{2}(\mathrm{aver}1+\mathrm{aver}2)\right]/256$ (where yuzhi is the binarization threshold, aver1 is the mean value of the ring after two-dimensional median filtering (the center of the inner circle of the ring is the center of the circle, the radius of the inner circle of the ring is the radius of the inner circle + 5, and the radius of the outer circle of the ring is is the inner circle radius + 10), aver2 is the average value of two 11×11 squares (the center of the square is the row coordinate of the center of the inner circle as the row coordinate, the column coordinate of the center of the inner circle minus 130 as the column coordinate and the inner circle The row coordinate of the center of the circle is the row coordinate, the column coordinate of the center of the inner circle plus 140 is the column coordinate), / is the multiplication sign, 256 is for normalizing yuzhi), and the threshold value of the binarization of the outer circle is calculated; The iris image after dimensional median filtering is binarized to obtain image BW1; Step 2), use the disc structure element se with a radius of 15 to perform morphological opening operation on image BW1 to obtain image BW2, thereby obtaining image BW3=BW1-BW2; performing complement operation on image BW1 to obtain image BW4, and then obtaining image BW5= BW4+BW3; area filling is carried out to image BW5, and image BW6 is obtained; use formula β (A)=A-(AΘB) (wherein β (A) is the boundary of set A, and B is a suitable structural element) to image BW6 Perform edge extraction to obtain the outer edge of the iris; Step 3), taking the center of the inner circle of the iris (a ₁ , b ₁ ) as the base point, calculate the distance from each point on the outer edge of the iris to the point (a ₁ , b ₁ )

(in Represents the coordinates of each point on the outer edge, E is the vector

a vector of 1s of the same size,

represents a sample); use the formula

(in Represents the coordinates of each point on the outer edge, E is the vector a vector of 1s of the same size, represents a sample), calculate the distance from each point on the outer edge of the iris to point (a ₁ , b ₁ +1); continue to calculate the distance from each point on the inner edge of the iris to point (a ₁ , b ₁ +2) Until the distance from each point on the inner edge of the iris to point (a ₁ +9, b ₁ +9) is calculated Step 4), carry out Wiener filtering to each sample, to reduce the impact of non-iris outer edge points on positioning; then according to the formula ${\mathrm{\&sigma;}}^2=\frac{1}{\mathrm{no}}\underset{i=1}{\overset{\mathrm{no}}{\mathrm{\&Sigma;}}}{(x_i-u)}^2$ (where _xi represents the sample value of the sample, n represents the capacity of the sample, and u represents the sample mean), calculate each sample

The variance of , find the sample with the smallest variance

According to the formula $u=\frac{1}{\mathrm{no}}\underset{i=1}{\overset{\mathrm{no}}{\mathrm{\&Sigma;}}}{x}_i$ (where x _i represents the sample value of the sample, and n represents the capacity of the sample), to find the sample The mean value R; find the coordinates of the corresponding point according to the value of z $(a_1+\mathrm{qs}\left(\frac{z}{10}\right),b_1+\mathrm{qy}\left(\frac{z}{10}\right))$ (where qs means to seek integer quotient, and qy means to seek integer remainder); then the coordinates of the point $(a_1+\mathrm{qs}\left(\frac{z}{10}\right),b_1+\mathrm{qy}\left(\frac{z}{10}\right))$ That is, the center of the outer circle (a ₂ , b ₂ ), and the mean value R is the radius of the outer circle.

Images