KR101763761B1 - Method of identifying shape of iris and device for identifying iris - Google Patents

Abstract

An object of the present invention is to provide an iris shape detection method which reduces noise of an iris image and reduces a quantization error. An iris shape detection method according to an embodiment of the present invention includes the steps of: receiving an iris image divided into a plurality of pixels; dividing the plurality of pixels into a plurality of groups and dividing the first and second standard deviations Quot; 0 "to" 1 "from the calculated iris feature patterns using a filter having a coefficient corresponding to the difference between the first and second normal distribution curves Wherein the plurality of iris feature patterns are calculated by varying an average of the first and second normal distribution curves.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to a method of recognizing an iris shape,

The present invention relates to an iris shape recognition method and an iris shape recognition apparatus.

Personal passwords and personal identification numbers, which are widely used as traditional methods of identifying individuals, have the disadvantage of being stolen or lost. Therefore, it can not satisfy the demand for stable and accurate personal identification in the information society, which is becoming increasingly sophisticated / advanced.

Personal identification methods based on biometrics have been studied as an alternative method to overcome the disadvantages of personal identification methods using user passwords and personal identification numbers. Biometrics is a method of identifying individuals based on individual biometric characteristics. Specifically, fingerprint, face, and iris are used to identify individuals. Individual identification based on the individual's biometric characteristics is advantageous in that it is neither stolen nor leaked. In addition, since there is no danger of changing the biometric characteristic or loss of the biometric characteristic, it is possible to track who has performed the security violation, so that the audit function can be constructed perfectly.

In particular, among various biometric methods, iris recognition based personal identification is known to be efficient in terms of uniqueness, invariance, and stability in identifying an individual. In addition, personal identification based on iris recognition is applied to areas requiring high security because of low recognition rate.

The iris is a colored part of the eyeball. When the iris is enlarged, it includes various features to the detail part. Therefore, if an individual identification method based on iris recognition is used, an individual can be identified.

An object of the present invention is to provide an iris shape recognition method and an iris shape recognition apparatus which reduce noise of an iris image and reduce a quantization error.

According to an aspect of the present invention, there is provided an iris shape detecting method comprising: receiving an iris image divided into a plurality of pixels; Dividing the plurality of pixels into a plurality of groups and using, in each of the plurality of groups, a filter having a coefficient corresponding to a difference between first and second normal distribution curves having first and second standard deviations, Calculating characteristic patterns; And detecting an iris feature vector corresponding to a value between logical values "0" and "1" from the calculated iris feature patterns, wherein the plurality of iris feature patterns are arranged in the first and second normal distribution It is calculated by varying the average of the curves.

According to the embodiment of the present invention, a filter having the second and third standard deviations [sigma] 2, [sigma] 3 is used to calculate the iris feature pattern, so that an efficient noise cancellation function is provided. And the iris feature vector corresponding to the value between the logical values "0" to "1" is extracted, so that the quantization error is reduced.

1 is a conceptual diagram in which an iris image is associated with a polar coordinate system.
FIG. 2 is a block diagram of rearranging the iris image of FIG. 1 into an iris matrix.
FIG. 3 is a block diagram illustrating a process in which the first track of FIG. 2 is normalized.
4 is a block diagram illustrating a method for calculating an iris feature pattern from a normalized first track.
5 is a graph illustrating an exemplary fuzzy membership function.
6 is a block diagram illustrating an iris recognition apparatus according to an embodiment of the present invention.
7 is a flowchart illustrating an iris shape detecting process according to the first embodiment of the present invention.
8 is a flowchart illustrating an iris shape detection method using a plurality of filters according to a second embodiment of the present invention.
9 is a flowchart showing an iris shape detection method using a plurality of filters according to a third embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, so that those skilled in the art can easily carry out the technical idea of the present invention. .

1 is a conceptual diagram in which an iris image 100 is associated with a polar coordinate system. Referring to FIG. 1, the radius axis p represents distance from the center o of the iris image 100. An angle axis? Represents an angle with respect to the center of the iris image 100.

The iris image 100 is an image including a black character as a whole, including a pupil at the center of the eye and an iris wrinkle at the pupil periphery. Illustratively, since a pattern of iris wrinkles is used, a monochrome image is required instead of a color image.

The iris image 100 may be divided into a plurality of tracks in the direction of the radius axis p. Illustratively, an image captured by a digital camera including a CCD and a CMOS chip may be divided into a plurality of tracks. Illustratively, FIG. 1 shows an iris 100 image divided into eight tracks. And, in each track, a plurality of pixels may be arranged in the direction of the radius axis p (not shown). Also, in each track, a plurality of pixels may be arranged in the direction of the angle axis? (Not shown). That is, the iris image will be received divided into a plurality of pixels.

FIG. 2 is a block diagram of rearranging the iris image 100 of FIG. 1 into an iris matrix 200. FIG. Referring to FIG. 2, the iris matrix 200 includes M pixels arranged in the row direction. The row direction corresponds to the angle axis [theta] in Fig. The iris matrix 200 has N tracks arranged in the column direction. The column direction corresponds to the radius axis rho in Fig.

FIG. 3 is a block diagram illustrating a process in which the first track 210 of FIG. 2 is normalized. In the first track 210, a plurality of pixels are arranged in the column direction. For example, as shown in FIG. 3, three pixels in the first track 210 are arranged in the column direction. Then, as described with reference to FIG. 2, M pixels in the row direction of the first track 210 will be arranged.

The pixels included in each of the tracks are divided into a plurality of groups, and a weighted average using a normal distribution for each of a plurality of groups can be calculated. Illustratively, a first scatter 140 having an average value m and a first standard deviation sigma 1 is used to calculate a weighted sum of each of the columns of the first track 210, Each of the pixels of the first track 220 will be derived. Illustratively, the values of the three points on the curve of the first scatter 140 in the fifth column 211 of the first track 210 will be specified. After the values of the respective pixels of the fifth column 211 are multiplied by the values of the specified respective points, the multiplied values are added to yield a normalized fifth pixel 221. Once the weighted sum of each of the columns included in the first track 210 is calculated, the normalized first track 220 is derived. The normalized first track 220 will comprise M pixels.

The mean value m and the first standard deviation sigma 1 of the first scatter 140 are set so as to reduce the noise included in the iris matrix 200. [ Thus, by using the first scatter 140 and normalizing the first track 210, the noise included in the iris image will be reduced. In addition, the normalized first track 220 composed of one row is calculated from the first track 210 composed of a plurality of rows, so that the processing time of the iris image will be shortened.

In FIG. 3, the process of normalizing the first track 210 is shown, but other tracks included in the iris matrix 200 of FIG. 2 will be normalized as in the first track 210.

4 is a block diagram illustrating a method for calculating an iris feature pattern from a normalized first track 220. FIG.

Referring to Fig. 4, the second scatter G2 has a second standard deviation sigma2. The third dispersion G3 has a third standard deviation sigma 3. The third standard deviation sigma 3 may be smaller than the second standard adjustment sigma 2. Illustratively, the second and third distributions G2 and G3 may be Gaussian distributions. The difference (G2-G3) of the second and third dispersions can be used as a coefficient of a Difference of Gaussian (DOG) filter based on the difference of the Gaussian distributions.

The second scatter G2 and the third scatter G3 may be used to calculate the iris feature pattern from the normalized first track 220. [ By convoluting the normalized first track 220 with a filter having coefficients corresponding to the difference of the first and second normal distribution curves having the second and third standard deviations 3 and 3 Iris feature patterns are calculated. Illustratively, a filter corresponding to a difference between a second dispersion G2 having a second standard deviation sigma 2 and a third dispersion G3 having a third standard deviation sigma 3 (hereinafter referred to as a second and a third standard A filter with deviations [sigma] 2, [sigma] 3) may be used to calculate the weighted sum of the normalized first track 220. [ The calculated weighted averages will constitute the iris feature pattern.

4, when a fifth iris feature pattern 231 corresponding to the normalized fifth pixel 221 is calculated, a second dispersion G2 having a second standard deviation 2 , The second average value x2 may be set to correspond to the normalized fifth pixel 221. [ Likewise, a third average value x3 may be set to correspond to the normalized fifth pixel 221 in the third scatter G3 having the third standard deviation 3. Then, a difference between the second dispersion G2 and the third dispersion G3 is calculated, and a filter having the second and third standard deviations? 2 and? 3 will be derived. The points on the filter having the second and third standard deviations 2 and 3 are specified and the respective pixels of each normalized first track 220 are multiplied with each specified point and the multiplied values are added 5 iris feature pattern 231 will be calculated.

Like the fifth iris feature pattern 231, the iris feature pattern will be calculated for each of the pixels of the normalized first track. For example, the second and third average values (x2, x3) may be set to correspond to each of the pixels of the normalized first track (220). And second and third distributions G1 and G2 having second and third average values x2 and x3 for each pixel of the normalized first track 220, Patterns 230 will be calculated. In this case, M iris feature patterns will be generated.

The second standard deviation sigma 2 of the second scatter G2 and the third standard deviation sigma 3 of the third scatter G3 will be set through experiments so that the recognition rate becomes maximum in view of the iris recognition rate. Further, the second and third standard deviations? 2 and? 3 will be set through experiments so that the noise of the iris image can be reduced.

Illustratively, the iris feature pattern can be calculated as shown in equation (1).

In Equation (1), G2 (x2,? 2) represents a second dispersion G2 having a second standard deviation? 2 and a second average value x2. G3 (x3,? 3) represents a third dispersion G3 having a third standard deviation? 3 and a third average value x3. The second mean value x2 and the third mean value x3 may be the same. f (x) represents the value of the pixels included in the normalized first track 220. x will correspond to any one of the pixels included in the first track 220. [ D (x) represents the iris feature pattern value. Illustratively, G2 (x2, sigma 2) can be expressed as: " (2) "

In Equation 2,? 2 will represent the standard deviation of the second scatter G2. x2 will represent the average of the second scatter (G2). Illustratively, if x3 and? 3 are substituted in Equation (2), the third dispersion (G3) may be expressed similarly to Equation (2).

5 is a graph showing the fuzzy membership function s (t). Referring to FIG. 5, the horizontal axis represents the value (t) of the iris feature pattern. And the vertical axis represents the value (s) of the iris feature vector. The value s of the iris feature vector corresponds to a value between logical values "0"

The values of the iris feature patterns calculated using the filter having the second and third standard deviations? 2 and? 3 may correspond to values between 0 and 1 based on the fuzzy membership function. Using the fuzzy membership function s (t), the iris feature vectors are calculated corresponding to values between 0 and 1, rather than being quantized.

Illustratively, it is assumed that the value of the fifth iris feature pattern 231 (see FIG. 5) for the normalized fifth pixel 221 (see FIG. 4) corresponds to to. In this case, the value of the iris feature vector corresponding to the vertical axis will be 0.8.

Illustratively, the fuzzy membership function s (t) can be expressed as: < EMI ID = 3.0 >

In Equation (3), t represents the value of the iris feature pattern. c represents a specified constant. c can be used to change the slope of s (t). Therefore, when c is used, the value (s) of the iris feature vector corresponding to the value (t) of the iris feature pattern can be adjusted.

6 is a block diagram illustrating an iris recognition apparatus 300 according to an embodiment of the present invention. Referring to FIG. 6, the iris recognition apparatus 300 includes an image receiving unit 310, an iris recognition unit 320, a database 330, and a comparison unit 340.

The iris image is acquired in the image receiving unit 310. Illustratively, the image receiving unit 310 may be a digital camera including a CCD and a CMOS chip. The image receiving unit 310 is connected to the iris recognition unit 320 and transmits the acquired iris image to the iris recognition unit 320.

The iris recognition unit 320 is connected to the image receiving unit 310 and the comparison unit 340. The iris feature vector is calculated based on the iris image received by the iris recognition unit 320. Iris feature vectors are calculated as described with reference to Figs. That is, the received iris image is normalized (see FIG. 3), and a normalized iris image is used to calculate iris feature patterns (see FIG. 4). Then, the calculated iris feature pattern is used to calculate the iris feature vector (see FIG. 5).

The database unit 330 is connected to the comparison unit 340. The database unit 330 stores data corresponding to the iris shape of the user. The data may be stored in the database unit 330 before the image receiving unit 310 acquires the iris image.

The comparing unit 340 may identify whether the iris image acquired by the image receiving unit 310 is the user based on the data stored in the database unit 330 and the calculated iris feature vector. Illustratively, the absolute value of the difference between each iris feature vector and the data contained in its corresponding database will be calculated. And each absolute value is summed and compared to a predetermined threshold. If the summed absolute value is larger than the predetermined threshold value, it will be judged that it is not the iris shape of the user. If the summed absolute value is smaller than the predetermined threshold value, it will be judged to be the iris shape of the user.

7 is a flowchart illustrating an iris shape detecting process according to the first embodiment of the present invention.

Referring to FIG. 7, step S110 is a step of normalizing the image of the iris. The measured iris image is divided into a plurality of tracks based on the polar coordinate system, and normalization can be performed on each of the tracks. Illustratively, the measured iris image may be binarized data.

Illustratively, the pixels contained in each of the tracks may be divided into a plurality of groups, and a weighted average using a normal distribution for each of a plurality of groups may be calculated. Illustratively, as described with reference to FIG. 3, the pixels included in one track are divided into pixels arranged in the row direction and pixels arranged in the column direction, and each of the pixels arranged in the column direction Lt; / RTI > And a weighted average can be calculated for each of the groups.

Step S120 is a step of calculating the iris feature pattern from each normalized track. As described with reference to FIG. 4, iris feature patterns for each normalized track can be computed using a filter having second and third standard deviations? 2 and? 3. That is, weighted averaging is performed using a filter having the second and third standard deviations? 2 and? 3 for each normalized track so that iris feature patterns can be calculated. Illustratively, a DOG filter may be used to calculate iris feature patterns for each normalized track.

In step S130, the iris feature vector is calculated from each iris feature pattern. Based on the specified function as described with reference to Fig. 5, the values of iris feature vectors may correspond to values between logical values "0" Illustratively, the values of the iris feature patterns may be used as input values for the fuzzy membership function, and the values of the iris feature vectors based on the fuzzy membership function may correspond to values between the logical values "0"

In step S140, it is determined whether the person is a person based on the calculated iris feature vectors. The calculated iris feature vectors are compared with a previously stored database. The previously stored database will correspond to the iris shape of the user. Based on the comparison result, it is determined whether or not the iris shape of the user is correct.

Illustratively, the absolute value of the difference between each iris feature vector and the data contained in its corresponding database will be calculated. And each absolute value is summed and compared to a predetermined threshold. If the summed absolute value is larger than the predetermined threshold value, it will be judged that it is not the iris shape of the user. If the summed absolute value is smaller than the predetermined threshold value, it will be judged to be the iris shape of the user.

8 is a flowchart illustrating an iris shape detection method using a plurality of filters according to a second embodiment of the present invention.

In step S210, the iris image is normalized. As described with reference to Fig. 7, the description thereof is omitted.

In steps S220a, S220b, and S220c, a plurality of filters are used for each of the normalized tracks to calculate iris feature patterns, respectively. In step S220a, a filter having second and third standard deviations [sigma] 2, [sigma] 3 is used to calculate iris feature patterns for each normalized track. In step S220b, a filter having fourth and fifth standard deviations [sigma] 4, [sigma] 5 is used to calculate iris feature patterns for each normalized track. Similarly, in step S220c, a filter having the k-th and k-1th standard deviations? K,? K-1 is used to calculate iris feature patterns for each normalized track.

In step S230, each calculated iris feature pattern is combined. Illustratively, iris feature patterns may be combined by a method such as PCA (Principle Component Analysis), LDA (Linear Discriminant Analysis), SVM-DA (Support Vector Machine-Discriminant Analysis)

In step S240, whether the input iris image is iris shape based on the combined iris feature patterns is identified. That is, a database corresponding to the iris shape of the subject person will be stored in advance. Then, the data of each iris feature pattern and the corresponding database are compared with each other, and it is discriminated whether or not it is the iris shape of the user based on the comparison result.

9 is a flowchart showing an iris shape detection method using a plurality of filters according to a third embodiment of the present invention.

Step S310 is a step of normalizing the image of the iris. In steps S320a, S320b, and S320c, a plurality of filters are used for each of the normalized tracks to calculate iris feature patterns, respectively. Since this is as described with reference to Fig. 7, the description thereof is omitted.

In operation S330a, S330b, and S330c, score data is generated by comparing a previously stored database corresponding to the iris shape of the user and the calculated iris feature patterns.

The filter having the second and third standard deviations [sigma] 2, [sigma] 3 is used, and the calculated iris feature patterns are compared with the previously stored first database to generate score data (S330a). Similarly, the filter having the fourth and fifth standard deviations [sigma] 4, [sigma] 5 is used and the calculated iris feature patterns are compared with the previously stored second database, and score data will be generated (S330b). The filter having the k-th and (k-1) -th standard deviations? K,? K-1 is used and the calculated iris feature patterns are compared with a previously stored nth database to generate score data (S330c). Here, n may be k / 2.

Illustratively, in step S330a, the absolute value of the difference between each iris feature pattern and the data of the corresponding first database will be calculated. And the calculated absolute values will constitute the first score data. The second to nth score data will be calculated similarly.

In step S340, the first to nth score data are combined. Illustratively, the score data is combined by a combination method such as MIN, MAX, SUM, weighted SUM, SVM (Support Vector Machine), and the like.

In step S350, it is determined whether or not the input iris image is the iris shape of the user based on the combined score data. That is, a database corresponding to the iris shape of the subject will be stored in advance. Then, the score data combined with the previously stored databases are compared, and it is discriminated whether or not the iris image of the user is based on the comparison result.

According to the embodiment of the present invention, a DOG filter is used to provide an efficient noise canceling function. Since the values of the iris feature vectors corresponding to the logical values "0" to "1" are extracted using the fuzzy membership function, no quantization error is generated and an improved iris detection function is provided.

It will be apparent to those skilled in the art that various modifications and variations can be made in the structure of the present invention without departing from the scope or spirit of the invention. In view of the foregoing, it is intended that the present invention cover the modifications and variations of this invention provided they fall within the scope of the following claims and equivalents.

100: iris image
200: iris procession
210: first track
220: normalized first track
230: First iris feature patterns

Claims (14)

A method for recognizing an iris shape of an iris recognition device,
Receiving a plurality of pixels including iris information and dividing the plurality of pixels into a plurality of groups;
Dividing each of the plurality of groups into subgroups;
Calculating normalized pixels by calculating a weighted average of the pixels included in each of the subgroups;
Collecting the normalized pixels corresponding to each of the subgroups to generate a normalized first track;
Filtering the normalized first track to generate iris feature patterns corresponding to the normalized pixels;
Detecting iris feature vectors through logical operations from the iris feature patterns; And
Comparing the iris feature vectors with previously stored data, and determining whether the iris feature vectors are personal according to the comparison result. delete The method according to claim 1,
Wherein the logical operation corresponds to the iris feature vectors corresponding to values between logical values "0" to "1 ". The method according to claim 1,
Wherein generating the iris feature patterns comprises:
A first normal distribution curve having a first mean value and a first standard deviation of the normalized pixels and a second normal distribution curve having a second mean value of the normalized pixels and a second normal distribution curve having a second standard deviation, And filtering the normalized first track using a filter. A method for recognizing an iris shape of an iris recognition device,
Dividing a plurality of pixels including iris information into a plurality of groups;
Dividing the plurality of groups into subgroups;
Combining the pixels included in each of the subgroups to generate normalized first tracks corresponding to the plurality of groups;
Filtering the normalized first tracks using filters having filter coefficients depending on the plurality of groups to generate iris feature patterns corresponding to the plurality of groups; And
Comparing the data generated using the iris feature patterns with previously stored data, and determining whether the iris feature pattern is personal according to the comparison result. 6. The method of claim 5,
Wherein the generated data is iris feature vectors generated as a result of a logical operation of the iris feature patterns. 6. The method of claim 5,
The first score data is generated by comparing the first iris feature patterns filtered using the first filter coefficient among the filter coefficients and the first data stored in advance,
Comparing the second iris feature patterns filtered using the second filter coefficient among the filter coefficients and the second data stored in advance to generate second score data,
Wherein the generated data is generated by combining the first score data and the second score data. An image receiving unit configured to receive a plurality of pixels including iris information;
An iris recognition unit configured to divide the plurality of pixels into a plurality of groups, divide the plurality of groups into subgroups, and generate iris feature patterns based on the subgroups; And
And a comparison unit for determining whether the user is an individual using data generated based on the iris feature patterns and data corresponding to the iris image stored in advance,
Wherein the iris recognition unit comprises:
Combining the pixels included in each of the subgroups to generate normalized pixels,
Collecting the normalized pixels corresponding to each of the subgroups to generate normalized tracks corresponding to each of the plurality of groups,
And the iris feature patterns are generated by filtering the normalized tracks using a filter having different filter coefficients. 9. The method of claim 8,
Wherein the iris recognition unit is configured to generate the tracks by calculating weighted averages of the pixels included in each of the subgroups. 9. The method of claim 8,
Wherein the data is values in which the iris feature patterns correspond to values between logical values "0" to "1 ". 9. The method of claim 8,
Wherein the iris recognition unit corresponds to a difference between a first normal distribution curve having a first average value and a first standard deviation of the normalized pixels and a second normal distribution curve having a second average value and a second standard deviation of the normalized pixels Wherein the iris recognition unit generates the iris feature patterns by filtering the normalized first track using a filter having a coefficient. 9. The method of claim 8,
Wherein coefficients of the filter used for the filtering vary depending on the plurality of groups, respectively. 9. The method of claim 8,
Wherein the data are values obtained by combining the iris feature patterns. 13. The method of claim 12,
Wherein the comparing unit generates score data by comparing the iris feature patterns with previously stored data, and combines the score data to generate the data.

Images