KR20120135381A - Method of biometrics and device by using pupil geometry - Google Patents

Abstract

A new biometric recognition method using unique information taken from a pupil is disclosed. A biometric method includes: obtaining an image comprising a pupil and an iris; Extracting a pupil area from the image; And generating specific biometric information by extracting a specific pattern of the pupil from the pupil region, and further improving the biometric recognition method by combining the unique pupil information and the iris unique information.

Description

Biometric method using pupil boundary information and device applying it {Method of biometrics and device by using pupil geometry}

It is described in relation to a device that recognizes and applies unique information from the shape of the pupil boundary.

During biometrics, many recognition methods using iris have been studied because of their advantages over other biometric information in terms of accuracy, stability, and authentication speed [Non-Patent Documents 1 and 2]. However, noise generated by a point sensitive to the surrounding environment due to ambient light or light reflected light and iris occlusion caused by the eyelid, which is one of the user's movements, causes non-patent document 3 to degrade accuracy of iris recognition. Recently, as a method for further improving the recognition rate, a pupil area detection method that is robust to reflected light [Non-Patent Document 4], a filter method for feature extraction, and an eyelid area detection method have been proposed [Non-Patent Documents 5, 6]. Degradation of recognition rate occurs due to ambient noise generated by the process of acquiring iris image through eye image or iris obstruction by eyelid.

The method disclosed in Patent Literature 1 proposes a method for estimating the pupil region for faster iris recognition. In other words, due to the iris recognition characteristics, there is a change in the recognition rate depending on the influence of the surrounding environment (light reflected by the illumination or the brightness of the ambient light). Can be reduced.

The method disclosed in Patent Document 2 obtains a plurality of biometric information such as pupil size and eye blink rate / frequency through an eye image. This method tracks the user state and performs emotional recognition through interaction between the machine (computer) and the user, but cannot be used for the user's unique information, that is, biometrics.

KR 00572410 B1 KR 2011-0035585 A

G., Anna Poorani, R. Krishnamoorthi, P., Gifty Jeya, S., Petchiammal, “Accurate and Fast Iris Segmentation”, International Journal of Engineering Science and Technology, Vol. 2 (6), pp. 1492-1499, 2010. Zhaofeng He, Tieniu Tan, Zhenan Sun, and Xianchao Qiu, “Towards accurate and Fast Iris segmentation for Iris Biometrics”, IEEE Transactions on PAMI, vol. 31, No. 9, pp. 1670-1684, 2009. Kazuyuki Miyazawa, Koichi Ito, Takafumi Akoki, Koji Kobayashi and Hiroshi Nakajima, "A Phase-Based Iris Recognition Algorithm," Advances in Biometrics: International Conference, ICB 2006, pp. 356-365, Hong Kong, China, January 2006. Cho Min-hwan, Heo Jung-yeon, “A Study on Pupil Center Search Algorithm for Iris Recognition,” Computer Information Society, Vol. 11, 2006. John G. Daugman, "How Iris Recognition Works," IEEE Trans. on Circuits and Systems for Video Technology, Vol. 14, No. 1, pp. 21-29, 2004. Topi Maenpaa, "An Iterative Algorithm for Fast Iris Detection," Advances in Biometric Person Authentication, International Workshop on Biometric Recognition Systems, IWBRS 2005, pp. 127-134, Beiging China, 2005. Cheol Woo Cho, Ji Woo Lee, Eui Chul Lee, Kang Ryoung Park, "A Robust Gaze Tracking Method by Using Frontal Viewing and Eye Tracking Cameras", Optical Engineering, Vol. 48, No. 12, 127202, Dec. 2009. Viola, P. and Jones, M.J. "Robust real-time face detection". Int. J. Comput. Vis. 57, 137-154 (2004) R. C. Gonzalez and R. E. Woods, Digital Image Processing, 2nd ed., Prentice-Hall, Englewood Cliffs, NJ, 2002. N.Otsu, "A Tlreshold Selection Method from Gray-Level Histograms", IEEE Transactions on SMC, Vol. SMC-9, No. 1, pp. 62-66, Jan. 1979. Suykens, J.A.K., Vandewalle, J., "Least Squares Support Vector Machine Classifiers", Neural Processing Letters, 1999-06-01, pp. 293-300, Volume: 9, Issue: 3 Arun Ross, Anil Jain, "Information fusion in biometrics", Pattern Recognition Letters 24 (2003) 2115-2125

The present invention provides a novel method for recognizing biometric information and a device for applying the same, capable of quickly obtaining biometric information without affecting surrounding image noise.

Biometric information recognition method according to the present invention:

Obtaining an image comprising the pupil and the iris; And

Extracting a pupil area from the image;

And extracting a specific pattern of the boundary shape of the pupil from the pupil region to generate bio-information information.

Biometric information recognition apparatus according to the present invention:

An image capture unit which acquires an image including a pupil and an iris;

An information processor for extracting a specific pattern of a pupil from the image and calculating biometric information therefrom;

And a storage unit for storing the biometric information; It includes.

According to a specific embodiment of the present invention, the specific pattern is obtained from an outer boundary portion of the pupil area.

According to a more specific embodiment of the present invention, the specific pattern is calculated from the change in the radius of the edge portion of the pupil obtained at regular angular intervals.

According to a more specific embodiment of the present invention, the step of generating the bio-unique information: determining the approximate center of the pupil using the information of the pupil area; Calculating an intersection distance (radius) value from the center to an outer portion of the pupil at a predetermined angle interval (360 degrees / N, N: natural number); And comparing each of the distance values with other adjacent distance values to obtain a difference value, and converting the distance value into a binary bit value based on a sign of the difference value to determine biometric information.

According to an embodiment of the present invention, a difference value obtained from adjacent distance values corresponds to raw data of biometric unique information, and the raw data is converted into a binary bit value corresponding to biometric unique information.

Biometric information recognition method and apparatus according to another embodiment of the present invention:

Generating biometric information from the iris region; And

Combining the unique information obtained from the pupil region and the iris region.

 The combining of the unique information is performed at any one of a feature level, a score level, and a decision level.

1 symbolically illustrates a biometric device according to the present invention.
2 is a flowchart of a method of recognizing biometric information according to the present invention.
Figure 3 shows an example of the approximate pupil center detection using the CED (Circular Edge Detection) method.
4 is a diagram illustrating a process of extracting unique information from a pupil boundary.
5 schematically shows a comparison of original information DB data and newly acquired information New data.
Figure 6 is a photograph showing the change in pupil size of the same phosphorus according to the ambient illuminance difference.
FIG. 7 illustrates distribution of biometric information on a two-dimensional plane formed by an axis of a hamming distance HDp as a result of comparing the pupil boundary information and an axis of a hamming distance HDi as a result of comparing the iris information according to the biometric method of the present invention. Illustrated.
FIG. 8 is a graph illustrating acceptance or rejection when combining pupil boundary information and iris information by applying an AND rule in the distribution diagram illustrated in FIG. 7.
FIG. 9 is a graph illustrating whether biometrics are recognized when the pupil boundary information is combined with the iris information by applying an Or rule in the distribution diagram illustrated in FIG. 7.
10 is a graph illustrating the XOR problem by the linear classifier, with Imposter results having a specific distribution form.
FIG. 11 is a graph illustrating the application of a nonlinear classifier to solve the problem of the linear classifier as shown in FIG. 10.

Hereinafter, with reference to the accompanying drawings will be described in detail a biometric method and an apparatus applying the same according to an embodiment of the present invention.

1 shows a symbolic and schematic structure of a biometric device according to the present invention.

Referring to FIG. 1, an image capture unit 1 including a camera for capturing the eye 6 of a living body sends the captured image information to the storage unit 2, and the information processing unit 3 stores the storage unit 2. Calculate the image information stored in). At this time, the information processed by the information processing unit 3 is basically obtained from the pupil area, and in addition, the iris specific information is also processed from the iris area. Two pieces of unique information obtained from the pupil area and the iris area are fused into one, and this information fusion is performed in a feature level, score level, and decision level described below. It is performed at either level. Through this method or process, the biometric result obtained by the information processing unit 3 is output through the output unit 4. The information processing unit 3 reads the information from the storage unit 2 and stores the required information, that is, the bio-specific information, in the storage unit 2 through a process as described below. The storage unit 2 is a comprehensively defined memory device, and may include a semiconductor memory such as RAM, FRAM, a magnetic memory such as HDD, and a storage medium such as ODD. The storage unit 2 may be provided with a database including a plurality of biometric information. The information processing unit 3 compares the newly acquired bio-information information with bio-information information registered in the database, and performs verification (determination) as to whether the information is pre-registered in the database. All of the above-mentioned parts are controlled by the controller 5, which is responsible for controlling all the elements required in all operations from image acquisition (or capture) to result output.

The output unit 4 may include at least one information output device such as a monitor, a printer, a sound device, and the like. The aforementioned elements may be provided by a general computer device or a dedicated computer system based device equipped with a camera for moving picture or still image. Some functions of the information processing unit 3 which processes the image and calculates / calculates the circular data and the final bio-specific information may be supported by software or firmware.

Hereinafter, an embodiment of the biometric information recognition method of the present invention using the above apparatus as a driving environment will be described.

2 is a flowchart of a method of recognizing biometric information.

First, a user's eye image including a pupil and an iris is acquired using a camera (211). A pupil is detected from the eye image (212). If it is determined that there is a face part other than the eye, the face area and the eye area should be detected before the pupil detection, and in this case, a so-called Adaboost (Adaptive Boosting) method may be used. Adaboost is simple, fast and detects faces with relatively high accuracy and works in real time. Non-Patent Document 8 describes such an adaboost, and can extract a face region as described above by the method with reference thereto. After extracting the facial region, before detecting the pupil region, the eye region should be detected, in this case also by the Adaboost method. In order to extract the eye region, information trained through weak classifiers reflecting eye characteristics is used.

On the other hand, in the pupil (area) detection 212, the following circular detection algorithm is used. First, as shown in FIG. 3A, an initial pupil area is determined through circular template matching. The circular template matching method is described by the following equation, and non-patent document [7] can be referred to.

Here, I ( x, y ) represents the brightness of the image at the ( x, y ) position, and ( x ₀ , y ₀ ) and r represent the center and radius of the circular template. In the end, the pupil area is determined at the point where the difference in brightness of each of the two circular templates is maximized. However, since the pupil may appear in the form of an ellipse rather than a circle according to the gaze position and the camera photographing angle, the position determined by the circular template matching method may not be accurate. Therefore, the local binarization process is performed as shown in FIG. 3B based on the determined position. Since the blind area is classified into two types, the pupil area (foreground) and the non-pupillary area (background), the threshold of binarization is the method proposed by Gonzalez (Non-Patent Document 9) and the threshold proposed by Otzu. An automatic determination method (see Non-Patent Document 10) is used. After local binarization, there may be noise areas caused by eyebrows or shadows, and when reflected light is present inside the pupil area, it may appear as punctured. In order to solve this problem, by performing a component labeling method on the binarized areas, by assigning an identity to the areas adjacent to each other, the area having the largest width is removed and the areas having different identities are removed. Remove the noise area. Finally, a morphological closing operation is performed to fill the perforated area. As a result, as shown in FIG.

Pupil center point (Xp, Yp), The distance between the pupil boundary (ie, the portion of the pupil area that changes from the black pixel to the white pixel) is obtained (214). Since the morphology calculation has a characteristic of blurring the boundary of the binarization figure, the image before the morphology calculation is used to determine the pupil boundary information as shown in FIG. The distance at this time is the radius at n (e.g. 256) positions of the pupil boundary 6a, as shown in FIG. That is, the radial straight line (C) disposed equidistant from the pupil centerR ₁ to CR _n ) And the intersection of the pupil boundaryXr ₁ , Yr1) ~ (Xrn _, Yrn) Where n distances from the pupil center to the boundary (D _{i ,} E.g _i = 0 to 255) is obtained (215). Distance to the intersectionD ₀ ~ Dn)After calculating, the radius (distance) change (rate) of the pupil boundary is calculated (216).

From each of the n distance values obtained in the above process, a difference (eg, O _i = D _i - D _i-1 ) of the distance values of any one position is calculated (217). Comparing any I- th point ( x _i ) to its adjacent point ( x _i-1 ), if the difference is equal to or less than zero, binarizes the value to zero, and if it is greater than zero, it is unique. Stored as a binary pattern for the pupil boundary information of (219).

If it is determined that the boundary information is lost from the reflected light of the infrared light across the boundary between the pupil and the iris in the process of detecting the boundary information as described above, the corresponding information cannot be used. Therefore, when the brightness value in the original image passes through a predetermined number of points, for example, 250 or more, in 256 straight paths having a 1.406 degree interval, information is incorrectly detected by the reflected light. 256 binary patterns storing such validity information (valid: 1, invalid: 0) are configured separately from one boundary information binary pattern.

The binary boundary information obtained as described above is stored as a unique information of a living body in a database and used as a control original in the future, or previously registered by matching with an original of a database having a plurality of bio-specific information previously registered. Determine if it is information or new information.

In order to determine whether the same person or another person through bio-recognition compared to the original of the database, it is necessary to compare original information (DB data) and newly acquired information (New data) as shown in FIG. 5. The stored pupil pattern E _i is identical through a process of comparing the newly acquired pupil N _i with a one-to-one (1: 1) pattern within a predetermined face rotation range (θ), for example, within a range of -20 to +20 degrees. Determine whether or not a person. 5 shows a comparison of two data that match each other. In this case, the newly acquired information (New data) can be compared by circular shift in consideration of the amount of rotation of the face, and as a result, the Hamming distance is selected to be the minimum. In the conventional iris recognition method, Hamming distance measurement is used to compare binary patterns of 2048 bits. In the present invention, a known hamming distance measuring method such as Equation (3) may be used for binary pattern matching.

As is well known, codeA and codeB refer to two pupil boundary information binary patterns to be compared, respectively, and maskA and maskB refer to validity information binary patterns of codeA and codeB . Code A ⓧ code B confirms pattern identity by XOR (Exclusive OR) operation by Hamming distance, and mask A and mask B mean validity of pattern. Therefore, whether or not the pupil pattern coincides with the result of the intersection of the XOR operation result value / mask A, B is determined.

In the above description of the embodiment, the sampling number of the distance of the pupil boundary is limited to 256, but this may be 512 or 1024 for more accurate information extraction, and this sampling number does not limit the present invention. In addition, although a pupil pattern based on a difference value is obtained from a plurality of sampled radius or distance values, binary coding may be performed by another operation or calculation method other than the difference value. That is, a method of calculating an encoded unique pupil pattern from a plurality of sampled distance or radius values may also be tried by other various methods.

Figure 6 is a photograph showing the change in pupil size of the same phosphorus according to the ambient illuminance difference. As a result of comparing a plurality of binarized pupil patterns against the background of various illuminance differences, it was confirmed that the pupil patterns coincide within a certain range regardless of the illuminance, so that information obtained from the pupil boundary of the living body is iris information. It can be used as unique information that is discernible.

By using the method of the present invention as described above, that is, the specific pattern of the pupil as bio-specific information, it is not only prevents the recognition rate deterioration due to the influence of light from the surrounding environment such as ambient light or reflected light, which is a disadvantage of iris recognition, In addition, the recognition rate is reduced using the pupil even in noise caused by the blinking of the user's eyes. In addition, iris recognition can be applied more efficiently to a user with small eyes, for example, a combination of iris and pupil recognition can be expected to improve recognition rate through multi-modal biometrics.

As a method for obtaining pupil boundary information and iris information, it is possible to extract from one image obtained from a living body. In this case, the iris must be captured sufficiently. An eye region is detected from the obtained image and iris information and pupil information are obtained therefrom. Herein, the iris information is obtained by using an existing method, and the pupil information is obtained by the method of the present invention described above. In order to recognize a living body using the two pieces of information thus obtained, a combination of the two pieces of information is required. For information on how to combine the two pieces of information, refer to "Information fusion in biometrics" of Non-Patent Document 12 presented by Arun Ross and Anil Jain. Can be.

The two pieces of information may be combined at various combination methods, for example, feature level and score level decision level, which will be described later, and the combined information may be used for multi-modal biometrics. Can be.

Coupling at the feature level simply combines, for example, 256 bits of information extracted from the pupil boundary with 2048 bits of information extracted from the iris, thus corresponding to an arithmetic sum, for example a total of 2304 bits of biometrics. You can get information. Such 2304-bit biometric information is compared with a previously registered biometric information DB of 2304-bit by the method as described above, thereby determining whether or not the same.

The combination at the score level is a result of comparing the Hamming distance (HDp) between the pupil boundary information (new captured pupil boundary information and the pupil boundary information registered in the DB) and the Hamming distance (HDi) between the iris information. When defining the score for the field, each Hamming distance is summed (HDp + HDi) or multiplied (HDp * HDi) and expressed as one representative value. The resulting value can be used to read new biometric information.

In the case of binding at the decision level, the results of Genuine matching (compare multiple information obtained from the same living body) and Imposter matching (compare multiple information obtained from different living bodies) are compared with the pupil boundary information comparison result Hamming distance (HDp). As a result of comparing the axis with the iris information on the two-dimensional plane formed by the axis about the Hamming distance HDi, the distribution of the shape shown in FIG. 7 appears. In the case of combining the two results through AND rule, when a value below the threshold (Ti, Tp) appears for both the iris information and the pupil information, the accept, i.e. the identity of the living body, is confirmed. Since the rest is rejected as unconfirmed information, it can be expressed as shown in FIG. In the case of combining through an Or rule, if only one of the iris information and the pupil is less than the threshold value Ti or Tp, the value is an accept and thus may be expressed as shown in FIG. 9. On the other hand, when having a distribution as shown in Figures 10 and 11, the linear classifier (linear classifier) may be a serious error due to the XOR problem (Exclusive OR problem) (false acceptance error of Figure 10), SVM (Support Nonlinear classification techniques such as Vector Machine (see Non-Patent Paper 11) can be used to determine the appropriate classifier.

Claims (14)

Obtaining an image comprising the pupil and the iris;
Extracting a pupil area from the image; And
And extracting a specific pattern of a pupil from the pupil area to generate biometric information. The method of claim 1,
And the unique information is obtained from a boundary of the pupil area. The method of claim 2,
And the unique information is calculated from a plurality of distance values sampled from a boundary of the pupil area. The method of claim 2,
And the unique information is calculated from a change value of a distance value sampled from a boundary of the pupil area. 5. The method according to any one of claims 1 to 4,
Generating the bio-specific information:
Determining a center of gravity of the pupil using information of the pupil area; Calculating an intersection distance (radius) value from the center of gravity to the outer portion of the pupil at a predetermined angle interval (360 degrees / N, N: natural number);
Comparing each of the distance values with other adjacent distance values to obtain an absolute difference, and converting the absolute difference into a binarization value based on an arbitrary reference value to determine bio-specific information. Information recognition method. The method of claim 5, wherein
Generating biometric information from the iris region; And
And combining the unique information obtained from the pupil area and the iris area. The method according to claim 6,
The combining of the unique information may be performed at any one of a feature level, a score level, and a decision level. An image capture unit which acquires an image including a pupil and an iris;
An information processor for extracting a specific pattern of a pupil from the image and calculating biometric information therefrom;
And a storage unit for storing the biometric information. The method of claim 8,
And the information processing unit obtains the unique information from a boundary of the pupil area. The method of claim 8,
And the information processing unit calculates the unique information from a plurality of distance values sampled from a boundary of the pupil area. The method of claim 8,
And the information processing unit calculates the unique information from a change value of a distance value sampled from a boundary of the pupil area. The method according to any one of claims 6 to 9,
The information processing unit:
The center of gravity of the pupil is determined using the information of the pupil area, and the intersection distance (radius) value from the center of gravity to the outer portion of the pupil is calculated as a predetermined angular interval (360 degrees / N, N: natural number). ,
And comparing each of the distance values with other adjacent distance values to obtain a difference value, and converting the distance value into a binarization value based on a sign of the difference value to determine biometric information. 13. The method of claim 12,
The information processing unit:
Generate bio-specific information from the iris region, and
And a unique information obtained from the pupil area and the iris area. The method of claim 13,
The information processing unit combines the unique information obtained from the pupil area and the iris area at any one of a feature level, a score level, and a decision level. Biometric Information Recognition Method.

Images