JP2008206536A - Personal authentication system using retina image - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To attain a persona authentication by acquiring information, that is hardly taken secretly and has an individual difference, from a retina image. <P>SOLUTION: The personal authentication is performed on the basis of the retina image of an eye photographed by a fundus camera. The retina image of the eye of an individual to be authenticated is photographed, the positions of an optic disk and an yellow spot part are specified from the photographed retina image, the retina image is normalized on the basis of the determined positions of the optic disk and the yellow spot part and normalized image data are acquired. By comparing a part of the normalized image data with reference data, the individual is authenticated. To put it concretely, there are the system of matching a part of the retina image by a correlation method and performing discrimination, the method of performing discrimination by the intersection position of a measurement ellipsoid for which the optic disk and the yellow spot part are focuses and a blood vessel, and the method of performing discrimination by a color distribution on a straight line connecting the yellow spot part and the optic disk, etc. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to personal authentication, and more particularly to a technique for performing personal authentication using a retinal image taken by a fundus camera.

  For example, many techniques for performing personal authentication using biometric information acquired from external features such as fingerprints, irises, and faces have been proposed. However, each of these techniques has a problem that biometric information can be easily stolen by, for example, extracting an attached fingerprint or taking a picture.

  On the other hand, the retinal blood vessel pattern, unlike these external features, is held in the eye, so it is difficult to shoot in places where the person is not conscious, and it is extremely difficult to take a voyeur. Have advantages.

  However, conventionally, special eye drops for opening the pupils are necessary for taking a retinal image, and the burden on the user is large, making it very difficult to take a picture. For this reason, the biometric authentication method has hardly been put into practical use except that in the past, those using infrared rays have been commercialized in the United States.

  The product manufactured in the United States described above is a product of US Eyedentify Co., Ltd., which was commercialized in 1985, and performs personal authentication using the blood vessel pattern of the retina.

This product uses infrared rays to capture the blood vessel pattern of the retina, and uses the data indicated by the reflected light to compare the registered data and the authentication data to determine the identity. The personal authentication process of this product is generally performed according to the following procedure.
(1) First, the user looks at the light that can be seen in the apparatus and adjusts the line of sight.
(2) Next, circular scanning is performed by irradiating infrared rays with the line of sight fixed.
(3) Since the irradiated infrared rays are absorbed in the blood vessel portion and reflected in portions other than the blood vessel, the reflected light reflects the blood vessel pattern.
(4) The analog signal of the blood vessel pattern read by detecting the reflected light is converted into a digital signal.
(5) At the time of registration, measurement is performed about 5 times, and the average value is stored as registration data.
(6) At the time of authentication, a scan similar to that at the time of registration is executed, and identity determination is performed by comparing whether the authentication data matches the registration data by a certain threshold or more.
US Pat. No. 4,109,237

  However, in the above-described conventional product, the fixation of the line of sight at the time of registration or authentication depends on the user's operation, and if the user shifts the line of sight, the scan area by infrared rays is shifted, which adversely affects the authentication. There was an inconvenience. That is, in the above-described conventional product, fixation of the line of sight is an absolute condition for retinal authentication.

  Further, depending on the degree of face inclination, a rotational deviation occurs in the scan result. Therefore, a process for absorbing this deviation is required during the authentication process, which complicates the authentication process and lowers the authentication speed. This is because the product uses only infrared rays so that only a blood vessel can be specified, and the location used for authentication cannot be automatically determined.

  In recent years, the development and commercialization of devices capable of taking retinal images without applying special eye drops to open the pupils has made it easier to take retinal images. It can be said that the possibility of realizing personal authentication is increasing.

  Therefore, in view of the above-described problems in the prior art, the object of the present invention is based on the concept of identifying the optic nerve head and the macula using retinal images, and does not require the fixation of the user's line of sight as an absolute condition. It is intended to enable accurate personal authentication without depending on the inclination of the user.

  In recent years, the development of a fundus camera capable of taking a retinal image without applying special eye drops for opening the pupil has been developed and put into practical use, and it is becoming easier to take a retinal image. Therefore, in the present invention, the retinal image is used to identify the optic nerve head and the macula, which are characteristic parts that cannot be identified by the conventional infrared method, and these are used as a reference to be used for authentication in the retinal image. Allows the location to be determined automatically. That is, by realizing the automatic positioning using the retina image, it is possible to reduce the dependence on the user's operation and the occurrence of rotational deviation at the time of registration and authentication. That is, according to the present invention, positioning is performed using a retinal image, and personal authentication is performed by acquiring information having individual differences.

  For this reason, in the present invention, first, the optic nerve head and the macula are identified for the retinal image captured by the fundus camera. Next, normalization such as image rotation deviation is performed using the positions of the optic disc and the macula. Then, biometric information with individual differences is extracted from the normalized retinal image to perform personal authentication.

Various types of biometric information used for personal authentication can be considered. In this specification, the following are proposed as embodiments.
(1) A part of a photographed retinal image is matched by a correlation method to discriminate the person / others.
(2) Draw an ellipse that focuses on the macula and the optic disc, and perform identity determination using the intersection of this ellipse and blood vessel.
(3) Identification is performed using a color distribution on a straight line connecting the macular portion of the retinal image and the optic disc.

  According to the present invention, it is possible to automatically determine a location to be used for authentication in a retinal image by using the retinal image, specifying the optic nerve head and the macula, and using these as a reference. . In other words, by realizing automatic positioning using retinal images, it is possible to reduce the dependence on user actions and the occurrence of rotational deviation during registration and authentication, and to perform personal authentication quickly with higher accuracy. Is possible. In addition, unlike conventional external features such as fingerprints, irises, and faces, the retinal image is held in the eyes, so it is difficult to shoot without being conscious of the person, and take sneak shots of biometric information. It is extremely difficult to perform, and more reliable personal authentication is possible.

Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a functional block diagram showing a schematic configuration for performing a personal authentication method according to the present invention.

  As shown in FIG. 1, a method according to an embodiment of the present invention includes a registration processing stage and an authentication processing stage. First, in the registration process, a retinal image of the individual 101 who is a normal human is photographed by the fundus camera 103. As the fundus camera 103, a camera capable of capturing a color image is used. The retinal image captured by the fundus camera 103 is given to the block 105, and the position of the optic nerve head is specified by a procedure described in detail later. Further, the retinal image obtained by the fundus camera 103 is given to the block 107, and the macula is specified by a procedure described in detail later.

  Next, in block 109, using the position of the optic nerve head and the position of the macula, an angle θ is obtained so that the optic nerve head and the macula are horizontal. Then, the angle θ is input to the function for rotating the image, and the image is rotated to obtain a normalized image in which the optic disc and the macula are horizontal.

  In step 111, feature data to be used for authentication is extracted from the normalized retinal image and stored in the storage unit 113, thereby completing the registration process.

  Next, as an authentication process, the same process as the registration process is performed for an individual to be authenticated. That is, the retinal image of the individual 101a to be authenticated is picked up by the fundus camera 103a, and the optic disc is specified in the block 105a, and the position (coordinate value) of the macula is specified in the block 107a. Then, a retinal image normalized in the block 109a is obtained from the positions of the optic nerve head and the macula determined as described above. In block 111a, feature data to be used for authentication is extracted from the normalized retinal image.

  The feature data extracted from the individual to be authenticated in this way is compared with the feature data registered in the registration process. Then, in block 117, a determination is made by comparing the magnitude of the difference between the registered feature data and the feature data to be authenticated with the threshold 115. By this comparison, as shown in block 119, an authentication result 119 indicating the principal / other is obtained.

  The above processes are usually performed using a computer such as a personal computer or a microprocessor. Also, the individual to be authenticated is usually a human, but other animals are possible.

  Next, with reference to FIG. 2, a method for specifying the optic nerve head in the registration process and the authentication process described above will be described in detail. First, a red (R) component image 201 </ b> R is extracted from a full-color retinal image 201 obtained by a fundus camera to obtain a binarized image 203. Here, the reason why the R component image is used is that the R component image tends to show a remarkable density value around the optic disc as compared with the other color component images.

  Next, the sum of the density values of the images of an area 205 having a predetermined constant size in the binarized image 203, for example, an area of 181 pixels × 181 pixels is calculated. The total density value in each region is sequentially calculated while sequentially moving (scanning) the region 205 having a predetermined size in the binarized image. Such movement of the region 205 is performed within a certain range in the binarized image 203. Usually, a certain region including a portion where the optic nerve head is likely to exist is selected. A place showing the maximum total density value among the total density values of the respective areas thus obtained is searched. That is, the brightest area is searched. Then, the coordinate value of the center point of the region where the density value shows the maximum value is recognized as the position of the optic disc of the original image.

  Next, with reference to FIG. 3, a method for specifying a macular portion in the above-described registration process and authentication process will be described. First, a processed image 301 is obtained by replacing an extra portion around the retina 211 of the full-color retinal image 201 obtained by the fundus camera, usually a black portion, with a large density value, in this case, white. A blue (B) component image is extracted from the processed image 301 to obtain a B component image 303. Here, the reason why the B component image is used is that the B component image tends to show a remarkable density value around the macula portion as compared with the other color component images.

  Next, the sum of the density values of the images of an area 305 having a predetermined constant size in the B component image 303, for example, an area of 101 pixels × 101 pixels is calculated. While the region 305 having such a predetermined size is sequentially moved (scanned) in the B component image 303, the total density value in each region is sequentially calculated. Such movement of the region 305 is performed within a certain range in the B component image 303. Usually, a certain region including a portion where a macular portion is likely to exist is selected. A place indicating the minimum total density value among the total density values of the respective areas thus obtained is searched. That is, the darkest area is searched. Then, the coordinate value of the center point of the region where the density value shows the minimum value is specified as the position of the macular portion of the original image.

  Using the coordinate value obtained by the above-described optic nerve head specifying process and the coordinate value obtained by the macular part specifying process, an angle θ is obtained so that the optic disk and the macular part are horizontal. That is, the angle θ formed by the line connecting the coordinate value of the optic disc and the coordinate value of the macula and the horizontal line is obtained. Then, the angle θ obtained in this way is input to a function for rotating the image, and a normalized image rotated so that the line connecting the optic nerve head and the macula is horizontal is obtained. Depending on whether the retinal image is taken from the left eye or the right eye, the positions of the optic disc and the macula are reversed on the retinal image, but in the present invention, the optic disc and the macula are either left or right on the retinal image. It is clear that the present invention can be applied even in the position of.

  Authentication processing is performed using the normalized retinal image obtained as described above. Although various authentication processes can be considered, three embodiments are shown in the present application.

<First Embodiment>
First, an authentication process according to the first embodiment of the present invention will be described with reference to FIGS. The authentication process according to the first embodiment is a method for determining by matching a part of a captured retinal image by a correlation method.

  First, positioning is performed to extract an image of an area used for pattern matching from the captured retina image. That is, the positions of the optic disc and the macula are specified as described above. Then, as described above, the image is rotated and normalized so that the line connecting the optic disc and the macula is horizontal. In other words, the coordinate value of the optic nerve head and the rotation angle for normalization in the original image before rotation are already known as described above. Using these numerical values, the coordinate value of the optic nerve head of the image after rotation is obtained by calculation, and positioning for extracting an image used for pattern matching, that is, positioning of the optic nerve head is performed.

  Next, as shown in detail in FIG. 5, an image used for pattern matching is extracted. First, image data of a green (G) component is extracted from the rotated image. Then, from this G component image data, a certain region, for example, an image of 192 × 192 pixels, centered on the coordinate value of the optic nerve head of the rotated retinal image obtained by the positioning process as described above, is extracted. This is an image of a predetermined area including the optic disc as a center. The above processing is performed on both the registration image and the authentication image. As another embodiment, authentication can be performed using an image of a region including a macular portion as a center.

Next, as shown in FIG. 6, the average value of each pixel is calculated in both the registration image obtained in the registration process and the image extracted from the authentication image obtained in the authentication process. Next, the average value calculated previously is subtracted from each pixel of the image extracted for pattern matching. Here, the extracted image on the authentication image side is S [i, j], the matrix in which all elements are average values of S [i, j] is S _AVG , and the extracted image on the registered image side is F [i, j ], If a matrix in which all elements are average values of F [i, j] is F _AVG , the following formulas 1 and 2 are the characteristic data.

[Equation 1]
S [i, j] -S _AVG


[Equation 2]
F [i, j] -F _AVG

Next, a correlation coefficient between the two images extracted from the registration image and the authentication image is calculated. Here, the correlation coefficient is used, but the matching by the correlation method is the simplest pattern matching for the image signal, and is a method of using the density value data as it is as a feature vector. When the feature data of the registered image and the authentication image is used, the correlation coefficient R can be obtained by the following Equation 3.

[Equation 3]

  The correlation coefficient R represents the degree of similarity, and the larger the correlation coefficient, the more similar the two images.

  Next, the maximum correlation coefficient is searched. First, in the retinal image to be authenticated, that is, the authentication image, an image shifted by one pixel from the first extracted image extracted as described above is extracted. A correlation coefficient between the extracted image in the newly extracted authentication image and the extracted image in the registered retinal image is calculated. Then, the correlation coefficient calculated here is compared with the correlation coefficient obtained first. Next, the larger value is stored as a true correlation coefficient.

  Such processing is sequentially repeated in a predetermined scanning range on the authentication image, and a correlation coefficient showing the largest value is set as a final correlation coefficient.

  The maximum correlation coefficient obtained in this way is compared with a preset threshold value. If the maximum correlation coefficient is larger than the threshold value, the person is determined, and if the maximum correlation coefficient is smaller, the person is determined as the other person. According to the first embodiment, it is possible to perform personal authentication with high accuracy without being affected by an image rotation shift or the like.

<Second Embodiment>
Next, an authentication process according to the second embodiment of the present invention will be described with reference to FIGS. The authentication process according to the second embodiment performs authentication by introducing a measurement ellipse. More specifically, an ellipse that focuses on the macula and the optic papilla is drawn, and the intersection point between the ellipse and the blood vessel is determined. It is a method to discriminate using.

  In this embodiment, the person identification is performed by using a part of the retinal image in which individual differences such as a blood vessel pattern appear. As in the case of the first embodiment, feature points of biological information are extracted from a retina image of an individual to be registered in advance in the registration process and registered as a template. Then, in the authentication process, the feature point of the biometric information of the individual to be authenticated, in this case, the intersection position of the ellipse and the blood vessel is extracted, and this is compared with the feature point of the registered biometric information. The authenticity is determined by calculating the degree.

  Also in this embodiment, in both the registration process and the authentication process, the macula and the optic nerve head are extracted from the density difference of the original retinal image in order to determine the location where biometric information is extracted, that is, for positioning.

  Next, in order to identify the blood vessel pattern on the retina, a measurement ellipse focusing on the macula and the optic disc is introduced, and the intersection of this ellipse and the blood vessel is identified. The intersection value on the ellipse is extracted using the characteristic that the luminance value of the blood vessel on the retina image is significantly lower than that of other portions. Using this intersection as a feature point of biometric information, personal identification is performed by comparing the degree of similarity.

  The authentication process according to the second embodiment will be described in further detail. First, in the registration process, in order to extract and identify each feature point of biometric information, a macula and an optic disc are extracted from a retinal image taken from the eyes of an individual to be registered. Then, a measurement ellipse is introduced based on the extracted positions of the macula and the optic disc, and feature points are measured and registered from the intersection of the measurement ellipse and the blood vessel.

  That is, as shown in FIG. 7, the macula and the optic disc are extracted from the original retinal image obtained by photographing the eyes of the individual to be registered. That is, the coordinate values of the macula and the optic disc are obtained.

Next, a predetermined measurement ellipse 701 that focuses on the positions of the macular portion and the optic disc determined in this way is determined. Then, name the intersection of the measurement ellipse 701 and the vessel respectively u1, u2, · · ·, and _{u n.} Specifically, the luminance value on the measurement ellipse 701 is calculated, and the characteristic that the luminance value is remarkably lower than other portions is used. That is, the luminance value of the measurement ellipse intersections u1 significantly lower at the compared to the value of the front and rear, u2, · · ·, extracted as u _n.

  FIG. 8 is a graph in which the horizontal axis is the position along the measurement ellipse as the X coordinate, and the luminance value on the line of the measurement ellipse is plotted on the vertical axis. In the figure, the luminance values are markedly reduced at a1, b1, c1, d1, e1, and f1 along the graph, and these points correspond to the intersection positions of the measurement ellipse and the blood vessel. FIG. 8 shows, as an example, a graph sampled from a retina image taken from an individual's eye as an individual sample 1.

  FIG. 9 shows the luminance value characteristic collected from retinal images taken at different times from the same individual's eyes as the personal sample 2. Also in FIG. 9, the luminance values are extremely low at points a2, b2, c2, d2, e2, and f2, and these points are considered to be intersections of the measurement ellipse and the blood vessel.

  Further, FIG. 10 shows a change in luminance value on a measurement ellipse taken from a retinal image taken from the eyes of an individual different from FIGS. 8 and 9 as another person's sample. In FIG. 10, a3, b3, c3, d3, e3, and f3, in which the luminance value is remarkably reduced, are considered to be the intersections of the measurement ellipse and the blood vessel.

  As shown in FIGS. 8 to 10, the luminance and the X coordinate of the portion with the low luminance value substantially correspond to each other, that is, the principal sample 1 and the principal sample 2. On the other hand, it can be seen that the low luminance value portion of the other person sample does not correspond very much to the low luminance value portion and the luminance value and X coordinate of the own sample 1 and the own sample 2. The person / other is determined in consideration of the correspondence between the feature points of the biological information.

  The actual determination work is performed as follows, for example. That is, as described above, the positions of the macular portion and the optic disc are extracted from the original retinal image, and the image is corrected or normalized. And the measurement ellipse which has a fixed magnitude | size which makes the position of the extracted macular part and optic nerve head a focus is determined. Then, the luminance value on the line of the measurement ellipse is calculated, and a point whose luminance value is significantly lower than the surrounding luminance is extracted as the luminance between the measurement ellipse and the blood vessel, that is, a feature point. The position of this feature point varies depending on the individual. Personal authentication is performed by comparing the coordinates of the feature points.

  As an example, the coordinates of feature points collected with respect to two individuals are shown in Table 1 and Table 2 below as sample coordinates 1 and sample coordinates 2, respectively.

[Table 1]
Sample coordinates 1
Sample number (1) (2) (3) (4) (5) (6) (7) (8)
X coordinate 500 524 553 769 737 799 876 927
Y coordinate 562 659 718 850 848 848 822 786

[Table 2]
Sample coordinates 2
Sample number (1) (2) (3) (4) (5) (6) (7) (8)
X coordinate 498 525 557 761 736 790 703 867
Y coordinate 541 661 724 850 848 849 840 826

  Looking at the example of sample coordinates 1 and sample coordinates 2 shown in Table 1 and Table 2, the coordinates of the sample numbers (2), (3), (4), (5) and (6) in both sample coordinates are as follows. It can be seen that they are almost identical to each other. Therefore, in this way, the number of feature points whose coordinates coincide with each other in the sample coordinates is obtained. And if this feature point is larger than a predetermined threshold value, it will determine with the person, and if it is smaller than a predetermined threshold value, it will determine with others. In order to determine the accuracy with which the coordinates match, a certain threshold value is set, and when the difference between the coordinate values of both is smaller than the threshold value, it is determined that the coordinates match. In addition, it is possible to determine the coincidence of coordinates using a certain threshold value for each of the X coordinate and the Y coordinate, and to determine that the X and Y coordinates are coincident with each other. Alternatively, the determination may be made using the fact that the difference between the total values of the X and Y coordinates of both sample coordinates is smaller than a certain threshold value. In any case, a threshold value is set in order to determine the coincidence of the coordinates of both samples, and the number of feature points whose coordinates match is calculated based on this threshold value.

  Furthermore, in order to determine whether or not the number of coincident coordinates of both samples is sufficient, another threshold value for determining the number of coincident coordinates is set. That is, this threshold value is determined in advance according to the required authentication accuracy.

Therefore, the threshold value to be set is
(1) a threshold value for determining the degree of displacement between coordinates; and (2) a second value for determining whether the number of coordinates determined by the threshold value in (1) is equal to or greater than a predetermined number. Set the threshold.

  In the above description, the measurement ellipse is used to extract the intersection point with the blood vessel in the retinal image. This measurement ellipse has a different size, for example, focusing on the same macular portion and the optic disc. Of these, two ellipses may be used. It is also possible to use three or more ellipses. Furthermore, it is possible to introduce a measurement curve or a straight line figure other than an ellipse. For example, instead of an ellipse, a circle, rectangle, rhombus or other measurement curve can be used.

  As an advantage obtained by the second embodiment, feature points can be extracted simply by reading the intersection of a fixed line such as a measurement ellipse of a retinal image and a blood vessel pattern, simplifying an identification algorithm for personal authentication, In addition, the authentication speed can be made relatively fast.

<Third Embodiment>
Next, an authentication process according to the third embodiment of the present invention will be described with reference to FIGS. The authentication process according to the third embodiment is a method of discriminating using a color distribution on a straight line connecting the macular portion of a captured retinal image and the optic disc.

  Also in this embodiment, the positions of the macular portion and the optic disc are specified from the captured retinal image. Then, the image is rotated and normalized so that the line connecting the macula and the optic disc is horizontal. That is, the coordinate value of the optic nerve head in the original retinal image before rotation and the rotation angle for normalization are already known as described above. Using these numerical values, the coordinate values of the macular portion and optic disc of the image after rotation are obtained.

  As shown in FIG. 11, from the position of the macular portion 1105 and the optic nerve head 1103 obtained in this way, a straight line portion 1107 connecting both of them, that is, image data of a measurement line is extracted. That is, the density values of red (R), green (G), and blue (B) images on a straight line connecting the macula and the optic disc are extracted as image data.

  FIG. 12 shows an example of RGB and luminance (Y) value distributions on a measurement line connecting the macula and the optic disc in a color retinal image obtained by photographing an eye of an individual. In FIG. 12, the horizontal axis represents coordinates on a straight line from the macula to the optic disc, and the vertical axis represents the density value of each image data.

  Referring to FIG. 12, the values of R, G, B, and Y are similarly shifted along the measurement line. Therefore, although it is possible to perform personal authentication using all the values of RGB and Y, since these values are shifted in the same way, an example of performing personal authentication using one of these RGBY values. It is shown below. However, the present invention is not limited to the case where only one of RGBY is used, and when all of these values are used, there are also cases where data of a plurality of colors among these values RGBY are used in any combination. Including.

  When authentication is performed with one value of RGBY, referring to FIG. 12, R reaches the maximum density value of 255 on the way, while B starts with 0 at the beginning of the density value. . For this reason, it is considered that the values of G and Y can provide the most feature data over the measurement line. In the present embodiment, it is assumed that authentication is performed using the value G as an example.

  FIG. 13 shows a comparison of G (green) values in retinal image data of the same person. That is, the curves indicated by GM1, GM2, and GM3 in FIG. 13 show part of the G image data extracted from retinal images obtained by photographing the eyes of the same person at different dates and times. Also in FIG. 13, the horizontal axis represents coordinates on a straight line connecting the macula and the optic disc, and the vertical axis represents the G density value.

  As can be seen from FIG. 13, the curves GM1, GM2, and GM3 have almost the same change tendency, although there are some deviations in the vertical and horizontal directions.

  In this case, even if the same retina (the same individual) is photographed, a slight shift occurs in the vertical and horizontal directions for each photographing, and it is necessary to allow such a shift. For this reason, as shown in FIG. 14, it is necessary to set a value allowing a deviation in the vertical and horizontal directions, that is, an error, that is, a threshold value (threshold value). FIG. 14 shows an example of two G image data extracted from retinal images obtained by photographing the same retina (individual) at different dates and times.

  In the present embodiment, when performing personal authentication, the G value extracted from the image data taken from the eyes of a certain individual in the registration process is extracted from the retina image obtained by photographing the eyes of the individual in the authentication process. Authentication is performed by comparing with the value of G. That is, at each coordinate on the measurement line, authentication is performed by comparing the value of the registered G image data with the G value of the individual to be authenticated. It is determined whether or not the G values of all the pixels on the measurement line or the points set at intervals on the measurement line coincide with each other. You can authenticate yourself and others. For this reason, it is also necessary to set a threshold value (threshold value) in which the person is recognized as the person when the G values of the two match.

  That is, in the present embodiment, it is convenient to use three threshold values that allow deviation of image data in the vertical and horizontal directions and a threshold value for identifying the person. However, the threshold value for allowing the vertical and horizontal shifts may not be used by correcting the image data as will be described later.

  If the above three threshold values are increased too much, there is a possibility that erroneous authentication for authenticating the person himself / herself may occur even in the case of a different retina (other person). On the other hand, if the threshold value is decreased, even the same retina (person) may be authenticated with another person. Accordingly, it is necessary to determine these threshold values by actually performing the authentication process on a large number of samples by experiments or the like.

  FIG. 15 shows a distribution of G values extracted from image data obtained by photographing from the eyes of the same individual. Curves GM6 and GM7 in FIG. 15 show the distribution of G values extracted from image data obtained by photographing the same individual at different dates and times. As is clear from the figure, image data obtained from the same individual may rarely deviate greatly.

In such a case, the above-described threshold value that allows the vertical shift may not be absorbed. However, referring to FIG. 15, the image data GM6 and GM7 are largely shifted in the vertical direction, but the transition patterns are almost the same. In such a case, some correction is considered necessary. Various correction methods are conceivable, but the correction can be performed by applying the following formula as an example.

[Equation 4]
GM7 value = GM7 value
− ((Sum of values of GM7 / number of coordinates) − (sum of values of GM6 / number of coordinates))

  Formula 4 above corrects by subtracting the difference between the average values of the two image data from the value of the image data showing a higher density value as a whole. The result of such correction is shown in FIG. As is apparent from FIG. 16, it can be seen that the image data of both images are substantially the same by such correction. Therefore, when performing the authentication process, if the registered image data and the image data to be authenticated are greatly deviated, they can be compared after the above correction.

  In this embodiment, authentication is performed by comparing the distribution of the color signal and / or luminance signal data on the measurement line connecting the macula and the optic disc. Therefore, since the comparison is directly performed using RGBY values without extracting blood vessels, it is considered that the authentication processing speed becomes faster.

  In the above embodiment, the image data on the measurement line connecting the macula and the optic disc is used, but some curve, broken line or the like can be used instead of this measurement line. In addition, any line figure based on the macula and the optic disc can be used.

  The present invention can be used for immigration control, identity verification in bank ATMs, and the like, and uses biometric information that is difficult to voyeurize. Therefore, highly reliable authentication is possible.

It is a functional block diagram which shows the schematic structure for performing the personal authentication method concerning this invention. It is explanatory drawing which shows the specific procedure of an optic disc. It is explanatory drawing which shows the specific procedure of a macular part. It is explanatory drawing which shows the procedure of the outline of the authentication method in connection with 1st Embodiment. It is explanatory drawing which shows the extraction procedure of the image used for pattern matching in 1st Embodiment. It is explanatory drawing which shows the procedure until it reaches | attains the search of the largest correlation coefficient in 1st Embodiment. It is explanatory drawing which shows an example of the measurement ellipse in the authentication process in connection with the 2nd Embodiment of this invention, and the intersection of this measurement ellipse and a blood vessel. It is a graph which shows the luminance value data of the principal sample 1 in 2nd Embodiment. It is a graph which shows the luminance value data of the principal sample 2 in 2nd Embodiment. It is a graph which shows the brightness value data of the other person sample in 2nd Embodiment. It is explanatory drawing which shows the measurement straight line for acquiring the color information for authentication, etc. in the authentication method in connection with the 3rd Embodiment of this invention. It is a graph which shows distribution of the data of each color on the measurement straight line in 3rd Embodiment. It is a graph which shows distribution of the value of G (green) of the same person. It is explanatory graph which shows the threshold value which accept | permits the shift | offset | difference of the vertical and horizontal direction in 3rd Embodiment. It is a graph which shows an example of the image data of the same individual with a big gap. It is a graph which shows distribution of the image data after correcting the data of FIG.

Explanation of symbols

101, 101a Individual 103, 103a Fundus camera 105, 105a Optic papilla specifying part 107, 107a Macular part specifying part 109, 109a Image normalization part 111, 111a Feature data extraction part 113 Registered data storage part 115 Threshold setting part 117 Determination part 119 Authentication result output unit 201 Original retina image 201R R component image 203 Binarized image 205 Density value total calculation area 207 Optic nerve head 209 Macula part 211 Fundus 301 Processed image 303 B component image 305 Concentration value total calculation area 701 Measurement ellipse 703 705 Optic disc 1101 Retinal image 1103 Optic disc 1105 Macula 1107 Measurement line

Claims (16)

A personal authentication method for performing personal authentication based on a retina image of an eye photographed by a fundus camera,
(A) taking a retina image of the eye of an individual to be authenticated, and determining the positions of the optic nerve head and the macula from the taken retina image;
(B) normalizing a retinal image based on the determined positions of the optic disc and the macula to obtain normalized image data;
(C) performing personal authentication by comparing a part of the normalized image data with reference image data;
A personal authentication method comprising:   The reference image data is obtained by taking a retinal image of an eye of an individual to be registered, determining positions of the optic disc and the macula from the taken retinal image, and obtaining a retinal image based on the determined location of the optic disc and the macula. The personal authentication method according to claim 1, wherein the personal authentication method is obtained from normalized image data obtained by normalization. Extracting red (R) component image data from the photographed retinal image and performing binarization processing;
In the binarized image data, a region having a certain size is determined, and the density value sum of the pixels of the image in each region is calculated while sequentially moving the region;
Searching for a region where the total concentration value is maximized; and determining a center point of the region where the total concentration value is maximum as the position of the optic disc.
The personal authentication method according to claim 1, further comprising: In the captured retinal image, a step of replacing an extra black portion around the retinal image with a large density value, that is, white, and extracting blue (B) component image data from the obtained image data;
In the extracted B component image data, a region having a certain size is determined, and the density value sum of pixels in each region is obtained while sequentially moving the region;
Identifying a region where the total density value is minimized, and determining a center point of the region where the total density value is minimized as the position of the macula,
The personal authentication method according to claim 1, further comprising:   By comparing the images of a predetermined partial region based on the position of the optic nerve head or the macular region among the retinal images obtained for the eyes of the individual to be authenticated and the eyes of the individual to be registered, by comparing each other with a correlation method The personal authentication method according to claim 1, wherein authentication is performed.   6. The personal authentication method according to claim 5, wherein personal authentication is performed using an image of a predetermined area including the optic nerve head.   The personal authentication method according to claim 6, wherein an image of a square area of 192 × 192 pixels centering on the position of the optic disc is used.   The personal authentication method according to any one of claims 5 to 7, wherein personal authentication is performed using an inner green (G) component of image data of a predetermined area including the optic nerve head.   For the eyes of the individual to be authenticated and the eyes of the individual to be registered, a predetermined measurement ellipse focusing on the optic disc and the macula is determined, and the individual authentication is performed by comparing the intersection positions of the measurement ellipse and the blood vessel image with each other. The personal authentication method according to claim 1, wherein the personal authentication method is performed.   The personal authentication method according to claim 9, wherein the intersection position between the measurement ellipse and the blood vessel image is obtained using a characteristic that the luminance value of the blood vessel portion of the retinal image is significantly lower than the luminance value of the other portion. .   The coordinates of the intersection position obtained from the retina image of the eye of the individual to be authenticated are compared with the coordinates of the registered intersection position obtained from the retina image of the eye of the individual to be authenticated. The personal authentication method according to claim 9, wherein personal authentication is performed based on a number or a ratio.   Personal identification is performed by comparing the intersection positions of a predetermined measurement curve and blood vessel image with respect to the eyes of the individual to be authenticated and the eyes of the individual to be registered with reference to the optic disc and the macula. The personal authentication method according to claim 1.   The personal authentication is performed by comparing color distributions on a straight line connecting the optic disc and the macula in the retinal image obtained with respect to the eyes of the individual to be authenticated and the reference individual. 1. The personal authentication method according to 1.   The color distribution uses a density value distribution characteristic on the straight line of one or more colors selected from red (R), green (G), blue (B), and luminance (Y). Item 14. The personal authentication method according to Item 13.   The personal authentication method according to claim 14, wherein the color distribution uses a density value distribution characteristic on the straight line of green (G).   Individual authentication is performed based on a single or a plurality of predetermined curves and / or color distributions on a straight line connecting the optic disc and the macula in the retinal image obtained for the eyes of the individual to be authenticated and the eyes of the individual to be used as a reference. The personal authentication method according to claim 1, wherein the personal authentication method is performed.

Images