JP3991042B2 - Personal authentication method and personal authentication device - Google Patents

Description

  The present invention relates to a technique for personal authentication using an iris image, and particularly to a technique for improving the accuracy of iris authentication under external light such as sunlight.

  In recent years, personal authentication technology using iris images has begun to be used for entrance / exit management of important facilities, ATMs (Automated Teller Machines) such as banks, and PC login applications. Usually, iris authentication is performed in the following steps.

1. 1. Illuminate the iris using a near-infrared LED to obtain an iris image. 2. Analyzing the iris image and extracting the iris code The extracted iris code is compared with the pre-registered iris code, and if the difference (distance) between the two codes is less than the threshold value, the person is judged

  Here, the use of near infrared light for illumination is based on the advantage that the person to be authenticated does not feel dazzling because the near infrared light is not perceived by humans. In addition, most people on Earth have brownish-brown irises (there are few races with blue and gray irises, but few in the world), but brownish irises are visible light This is because it is difficult to see the iris pattern below, but it is possible to shoot with rich contrast with near-infrared light. Note that blue and gray irises can also be photographed with an iris pattern under near-infrared light.

  As described above, iris authentication has begun to be used for entrance / exit management, ATM, PC login, and the like, but most of them are assumed to be used indoors where there is little external light of near infrared components. When performing iris authentication under external light that contains many near-infrared components such as sunlight, a wide range of near-infrared light emitted by the sun, etc., and near-infrared light reflected by objects irradiated on the sun, etc. Reflect in the eyes. In addition, when direct sunlight hits the face, there may be a wrinkle or a wrinkle shadow in the iris area. These factors cause an increase in the rejection rate (FRR). However, it is unlikely that the acquired iris pattern will accidentally become close to another person's iris pattern due to reflections or shadows of the eyelids, and therefore the acceptance rate (FAR) of others increases. The possibility is very low.

  ATM facilities are often installed near street entrances and on streets. In such facilities, Patent Document 1 and Patent Document 2 propose methods for preventing the influence of external light. In Patent Document 1, the first polarizing means is provided outside the lens of the camera that captures the iris, and the second polarizing means is provided between the iris recognition device and the light source of extraneous light (such as a window). Reflection of extraneous light is prevented by changing the polarization direction of the second polarizing means. In Patent Document 2, a non-visible light non-passing filter that reflects or absorbs a component of non-visible light that illuminates the eye among wavelength components of external light is installed at a position (window or the like) where the extraneous light is incident on the eye. This prevents extraneous light from appearing in the eyes.

  As another means for preventing the reflection, there is a method in which a light-shielding means such as an eye cup is provided in front of the camera so that the subject can look into and take an iris image.

  Moreover, although it is not external light, the coping method when the illumination means itself which illuminates eyes is reflected is proposed in patent document 3, patent document 4, patent document 5, and patent document 6.

  In Patent Literature 3, the head of the subject is expected to move while an iris image is captured a plurality of times, and the effect of reflection is reduced by using the obtained plurality of iris images. First, matching between one image and a registered image is performed, and the matched portion is added to the matching image. Thereafter, matching with the registered image is performed for other images, and the matched portions are sequentially added to the matching image. Then, the finally created matching image and the registered image are compared, and authentication is performed.

The techniques disclosed in Patent Document 4, Patent Document 5, and Patent Document 6 all use a plurality of illumination means having different installation positions. That is, a plurality of illumination units are irradiated at different timings, and a plurality of iris images are captured in accordance with the irradiation timing. A plurality of iris images (or feature amounts or collation results) with different illumination reflection positions are combined to create an iris image (or feature amount or collation results) that is not affected by the reflection.
Japanese Patent Laid-Open No. 10-21392 JP 2000-185032 A JP-A-9-212644 Japanese Patent Laid-Open No. 10-162146 JP-A-11-203478 JP 2001-167252 A

  However, the conventional technique has the following problems.

  Since the technologies of Patent Document 1 and Patent Document 2 are premised on facilities such as ATMs, the devices are both large-scale, costly to implement, and applicable applications are limited. There is a problem.

  In addition, the method of providing a light-shielding means such as an eye cup in front of the camera so that the subject can look into and take an iris image loses the advantage of iris authentication that authentication is possible without contact. It is not preferable from the viewpoint of the interface.

  Further, Patent Document 3 is not a measure against external light reflection, but is a measure against reflection of illumination means attached to the apparatus. The subject's head is expected to move naturally, but if the head does not move, the effect of reflection cannot be reduced, and no effect can be obtained. Even if the head moves, since the subject is in front of the device for authentication, it is unlikely that his / her line of sight will be greatly diverted from the device, and therefore the possibility that the position of the reflection will change is low.

  Moreover, although the method using the some illumination means shown by patent document 4, patent document 5, and patent document 6 has a meaning as a countermeasure with respect to the reflection of the illumination means attached to an apparatus, it is reflected under external light. Makes little sense for.

  In view of the above problems, it is an object of the present invention to prevent the authentication accuracy from being lowered even in the case where external light is reflected in the iris in personal authentication using an iris image.

In order to solve the above-mentioned problems, the personal authentication according to the present invention measures the intensity of near-infrared light in a shooting environment, and when the intensity is equal to or higher than a predetermined threshold, the person to be authenticated is converted into an iris by sunlight. A plurality of iris images with different reflection positions are photographed, and authentication is performed using the photographed iris images. On the other hand, when the intensity is less than a predetermined threshold, an iris image is photographed for the person to be authenticated. The authentication is performed using the photographed iris image.

  According to the present invention, when the intensity of near-infrared light is large and reflection is likely to occur, a plurality of iris images are taken, so that it is possible to perform authentication by reducing the influence of sunlight reflection on each other. Thus, even under the outside light, personal authentication with a reduced rejection rate can be executed. In addition, it is possible to release the person to be authenticated from the trouble of constantly shooting a plurality of iris images.

As described above, according to the present invention, when the intensity of near- infrared light is large and reflection is likely to occur, the effects of reflection of external light are mutually reflected using a plurality of iris images having different reflection positions of sunlight. Thus, authentication can be performed. Accordingly, personal authentication with sufficiently high authentication accuracy can be executed even under outside light, and the person to be authenticated can be released from the troublesomeness of always taking a plurality of iris images.

According to the first aspect of the present invention, the measurement process for measuring the intensity of near-infrared light in the shooting environment, and when the intensity is equal to or greater than a predetermined threshold, the reflected position on the iris by sunlight is about the person to be authenticated. First authentication processing for photographing a plurality of different iris images, and performing authentication using the plurality of photographed iris images, and when the intensity is less than a predetermined threshold, photographing an iris image for the person to be authenticated And a second authentication process for performing authentication using the iris image.

According to the second aspect of the present invention, the measuring means for measuring the intensity of near-infrared light in the shooting environment, and the reflected position on the iris by sunlight when the intensity is equal to or greater than a predetermined threshold A plurality of iris images that are different from each other, a first authentication unit that performs authentication using the plurality of captured iris images, and when the intensity is less than a predetermined threshold, captures an iris image of the person to be authenticated, Provided is a personal authentication device including a second authentication unit that performs authentication using a photographed iris image.

  Embodiments of the present invention will be described below with reference to the drawings.

  In the present specification, “external light” refers to light other than iris photographing illumination attached to an iris authentication device or photographing device. In addition, “reflection” refers to a luminance pattern of an iris region formed by “external light” that is not spatially uniform in a shot iris image.

(First embodiment)
FIG. 1 is a flowchart showing a personal authentication method using iris recognition according to the first embodiment of the present invention. 2 is a diagram showing an appearance of a mobile phone with an authentication function as an example of a photographing apparatus used when performing the personal authentication method according to the present embodiment, and FIG. 3 is an internal configuration of the mobile phone with an authentication function in FIG. FIG. In this embodiment, it is assumed that the person to be authenticated performs iris authentication under ambient light using a mobile phone with an authentication function as shown in FIG.

  The mobile phone 10 with an authentication function in FIG. 2 is provided with a camera 11 and an illumination 12 for photographing an iris image in a normal mobile phone. In addition to the camera 11 and the illumination 12, a monitor 13, operation buttons 14, a speaker 15, a microphone 16, an antenna 17, and the like are provided. As the illumination 12, one or several near infrared LEDs are used. The reason for using near-infrared light illumination is to prevent the person to be authenticated from feeling dazzling and to capture the iris pattern, which is brownish brown, with high contrast. A visible light cut filter is set in the camera 11 so as to receive only the near-infrared component. On the monitor 13, the iris image being photographed and the authentication result are displayed. Note that when photographing under a light source containing a large amount of near-infrared components such as the sun or an incandescent lamp, it is not always necessary to attach illumination to the photographing apparatus, and even if it is attached, it is not necessary to turn it on.

  In the internal configuration shown in FIG. 3, in the authentication means 21, the camera control unit 22, the illumination control unit 23, and the monitor control unit 24 are connected to the main control unit 27. The camera control unit 22 controls the camera 11 to capture an iris image, and the captured iris image is stored in the image memory 25. The authentication processing unit 26 performs authentication processing using the iris image stored in the image memory 25. The illumination control unit 23 controls the illumination 12 and illuminates the iris region in synchronization with the photographing timing. The monitor control unit 24 controls the display screen of the monitor 13.

  Hereinafter, processing in the personal authentication method according to the present embodiment will be described according to the flow of FIG.

  First, the person to be authenticated has the mobile phone 10 with the authentication function shown in FIG. 2, and takes a plurality (N) of iris images with different positions for reflecting outside light under the outside light (A1). At the time of shooting, the person to be authenticated has the authentication function mobile phone 10 at a predetermined distance from the eye (for example, about 20 cm before the camera 11 is a single focal point), and the iris image captured by the camera 11. Is displayed on the monitor 13, the entire iris is included in the field of view, and after aligning the focus, the shooting button assigned to one of the operation buttons 14 is displayed. Press. This operation is repeated N times so that the external light reflection position is different. Alternatively, N frame images may be acquired continuously while the moving image shooting is performed by pressing the shooting start button once.

  Here, the external light reflection phenomenon will be described in detail.

As shown in FIG. 4, the iris image captured under the external light is reflected in the external light over a wide range. This is because the intensity of the near-infrared component contained in sunlight (sunlight is an electromagnetic wave having a wide wavelength band such as ultraviolet rays, visible light, near infrared rays, and far infrared rays) is high, so that it is irradiated with sunlight as well as direct sunlight This is because the near-infrared component reflected from many objects enters the eye from various angles. (Because the eyeball has a shape close to a sphere, reflection tends to occur.)
The reflection position of the external light is determined by the positional relationship between the light source (including the reflection object), the camera, and the eyeball. Therefore, in order to change the reflection position, one of them may be moved. Of these, the light source is difficult to move, so the camera and / or eyeballs may be moved.

  FIG. 5 shows an example of four iris images captured with the camera position and face position fixed and only the direction of the eyeball (the direction of the line of sight) changed. FIG. 6 shows an example of four iris images taken with the relative positional relationship between the camera and eyeball fixed and the face direction changed from east to west, north and south. As can be seen from FIG. 5, when the eyeball is moved (when the direction of the line of sight is changed), the reflection position based on the iris region changes. Further, as can be seen from FIG. 6, the shape of the reflection changes when the orientation of the face changes.

  In this embodiment, in order to capture a plurality of iris images having different external light reflection positions, the device as shown in FIG. The line of sight shall be guided. For example, the direction of the face is displayed with an arrow on the monitor 13 (the monitor 13 and its control unit correspond to means for instructing the direction of the face), or the direction of the face is instructed by voice from the speaker 15 (speaker 15 and its control unit correspond to means for instructing the direction of the face), or a specific image such as a character is displayed on the monitor 13 and the image is moved in the direction in which the line of sight is guided ( The monitor 13 and its control unit correspond to means for guiding the line of sight). By adopting such a configuration, the person to be authenticated can easily know which face direction to use when taking a picture, and which direction to look at. Become. A plurality of iris images having different external light reflection positions can be easily captured.

The following processes A2 to A5 are executed by the authentication processing unit 26. Here, it is assumed that the technique disclosed in JP-T-8-504979 (International Publication WO94 / 09446, hereinafter abbreviated as “Reference Document 1”) is used. The outline of the iris authentication method of Reference 1 is as follows.
(1) Cut out the iris area by determining the outer edge of the iris (boundary between the iris and sclera) and the outer edge of the pupil (boundary between the pupil and iris) (2) xy orthogonalize the cut out iris area Convert from coordinate system to rθ polar coordinate system (3) Determine analysis band (radial direction is divided into 8 rings)
(4) Applying a multi-scale 2-d Gabor filter and binarizing the signal after the Gabor filter output is used as an iris code. (5) Pre-registered iris code and iris code at the time of authentication (Exclusive OR) and calculate the Hamming distance between the two codes (6) If the Hamming distance is less than or equal to the threshold, accept as the person, otherwise reject as the other person

  FIG. 7 is a diagram showing the positions of the outer edge of the iris and the outer edge of the pupil in (1), and FIG. 8 is a diagram showing an area surrounded by the outer edge of the iris and the outer edge of the pupil as an iris area and expressed in the xy coordinate system. At this point, the effect of the translation of the iris region is absorbed. FIG. 9 is a diagram in which the iris region is expressed in the rθ polar coordinate system with the pupil center as the center (conversion of (2)). The actual outer edge of the pupil and the outer edge of the iris are not exactly circular. When both are deliberately approximated by a circle, the center of the pupil and the center of the iris are not concentric (eccentric), but by setting the value in the r direction to 0 at the outer edge of the pupil and 1 at the outer edge of the iris, the eccentricity and the opening of the pupil The difference in condition and the influence of scaling can be absorbed.

  FIG. 10 is a diagram showing the analysis area of the 8-ring shape determined in (3), and FIG. 11 is a diagram showing the creation of the iris code in (4). The luminance signal (a) after determining the analysis band in FIG. ), (B) binarization is performed by applying a Gabor filter (c). Although it is actually a two-dimensional signal, it is shown in one dimension here for the sake of simplicity. (A) is an angular direction luminance signal in one of the eight rings. Actually, a multi-scale Gabor filter is applied, and a single-scale Gabor filter also has a real part and an imaginary part, but (b) and (c) are the results of applying a real part of a Gabor filter of one scale. is there. The position of each bit in the binarized iris code (c) can be associated with a certain position on the iris image.

  First, feature extraction is performed from N photographed iris images, and N iris codes are created as iris data, which is a feature amount for authentication (A2). In the process A2, the above-described processes (1) to (4) are applied to N iris images having different reflection positions to create N iris codes. At this time, by performing the processing (1) to (3) in each image, the effects of translation, enlargement / reduction, difference in the degree of opening of the pupil, and eccentricity of the pupil are absorbed for the iris region in the plurality of iris images. An iris code is created.

  Next, the N iris codes and the registered iris code as the registered iris data of the person to be authenticated registered in advance are collated, and N collation results are created (A3). In each embodiment of the present application, corresponding bits of two iris codes are compared with each other, and a bit string indicating matching / mismatching between the compared bits is created as a “collation result”. Since an XOR operation is used to compare bits, “0” is given as a bit value when they match, and “1” is given as a bit value when they do not match. Also, the bit value expressing this match / mismatch is used as the “determination result”.

  In the process A3, the process (5) described above is performed. At this time, there is a possibility that a deviation in the angular direction due to the inclination of the face or the rotation of the eyeball itself exists between the registered iris code and the authentication iris code. In order to compensate for this deviation, matching is performed by rotating the authentication code to a predetermined range, and the one with the smallest Hamming distance is output as the final Hamming distance. This is shown in FIG. In FIG. 12, for simplification of description, analysis is performed using a real part of a Gabor filter of a certain scale, and one of eight rings is expressed.

  FIG. 13 is a diagram showing a result of comparing the registered iris code and N authentication iris codes. Each of the authentication iris codes 1 to N in FIG. 13 is a code after rotation compensation matching has already been performed and shifted to a position where the Hamming distance is minimized. That is, all i-th bit positions correspond to the same position on the iris pattern. In FIG. 13, each authentication iris code is divided into M blocks, and blocks whose hamming distance in the block is larger than a predetermined threshold TH2 (here, 0.30) are shown in black.

  In the comparison result of FIG. 13, the Hamming distances HD of the authentication iris codes 1, 2, N were 0.31, 0.33, and 0.34, respectively. The Hamming distance is 0.5 when the two codes are completely uncorrelated. The hamming distance of another person is distributed around 0.5, but at the time of matching that performs rotation compensation, the hamming distance shifts in a direction that slightly decreases. When the threshold TH1 of the Hamming distance for authenticating with the person is 0.30, all these authentication iris codes are rejected when used alone.

  However, as shown in FIG. 13, although the hamming distance of most blocks is equal to or less than the threshold value TH2, the hamming distance of some blocks (shown in black) is large, so that It can be seen that the Hamming distance HD is larger than the threshold value TH1. Some of these blocks are considered to have a large Hamming distance HD due to external light reflection.

  Therefore, in the present embodiment, a final matching score is generated by combining and integrating a plurality of matching results (A4). Thereby, when the position of external light reflection differs in a plurality of iris images, the influence of external light reflection on the accuracy of personal authentication can be eliminated. The “collation score” is a scalar value that represents the degree of coincidence as a result of comparing two iris codes. In each embodiment of the present application, the ratio of the number of mismatched bits to the total number of bits compared, that is, the Hamming distance is used as the “matching score”.

  As a collation result integration method, a method of combining in units of bits and a method of combining in units of blocks can be considered.

<Bitwise synthesis>
FIG. 14 is a flowchart showing details of the processing A4 in the case of combining in bit units. In FIG. 14, first, a matching result whose Hamming distance is equal to or smaller than a predetermined threshold TH3 (> TH1) is selected from the N matching results (A401). Next, paying attention to the i-th bit position (A402), if there is a bit indicating the coincidence with the registered iris code at any one of the focused bit positions in the comparison result after selection, the bit position The final determination result is determined to be coincident (A403). Steps A402 and A403 are executed for all the bit positions (A404). The final number of unmatched bits is normalized by the total number of bits, and the Hamming distance is calculated as the final matching score (A405).

  Here, when another person has photographed N (N is many) iris images, the probability that one of the N iris codes and the registration iris code coincide with each other at a certain bit position approaches “1” (correlation). If there is no, there is a match / mismatch probability of 0.5). The same applies to a case where a plurality of persons take turns and take N iris images. For this reason, if all the collation results of a plurality of iris images are used, there is a high probability that another person is erroneously authenticated.

  Therefore, in the process A401, the collation result larger than the threshold value TH3 is excluded from the subsequent processes on the assumption that there is a high probability that it is based on another person's iris image. That is, only the collation result that is larger than the threshold value TH1 for determining as the person and smaller than the predetermined threshold value TH3 is “exceeded the threshold value TH1 because the reflection of external light is present in the person's iris image”. And used for the subsequent processing. Thereby, the improvement of the other person acceptance rate (FAR) which falsely authenticates others can be suppressed. Instead of the Hamming distance, when an index is used such that the value increases when the correlation is high while the value decreases when the correlation is low, a value with a score equal to or greater than a predetermined value is used in process A401. Just choose.

  Further, in the process A403, for example, for the focused bit position, it is possible to take a majority decision of coincidence / non-coincidence of N collation results and use the larger result as the final determination result of the bit position. In addition, for a focused bit position, a final determination result of the bit position may be determined according to whether or not the matching ratio in the N collation results is larger than a predetermined threshold. Furthermore, for the bit position of interest, the percentage of matches in N collation results may be used as the score of the bit position, and the average of the scores at all bit positions may be used as the final collation score.

  In addition, for a focused bit position, a final determination result of the bit position may be obtained using a weighted average of N verification results. A specific method will be described.

ここで、ｘ_ｋ は第ｋの照合結果のハミング距離である。すなわち、ハミング距離が小さい照合結果には、信頼性が高いと判断してより大きな重みを与える。なお、分母は重みｗ_ｋ の総和を１にするための正規化項である。 First, a weight w _k (0 ≦ w _k ≦ 1, k: verification result number) is prepared so as to increase as the reliability of the verification result increases. For example, it is determined as follows.

Here, _xk is the Hamming distance of the kth collation result. In other words, the collation result with a small Hamming distance is judged to have high reliability and is given a larger weight. The denominator is a normalization term for setting the sum of the weights w _k to 1.

ｒ_ｋ，ｉ は、第ｋの照合結果の第ｉビットの値である。この例では、第ｋの照合結果については、全てのビット位置について共通の重みｗ_ｋ が用いられる。 Then, the i-th bit value r _i of the composite verification result is determined by the following equation.

r _{k, i} is the value of the i-th bit of the k-th collation result. In this example, a common weight w _k is used for all bit positions for the k-th matching result.

  Also, different weights may be used for each bit position as follows.

ここで、ｘ_ｋ，ｉ は第ｋの照合結果の第ｉビット付近の局所的なハミング距離である。ここで、局所的とは、虹彩画像上で局所的である、という意味である。すなわち、局所的なハミング距離が小さいビット値ｒ_ｋ，ｉ には、信頼性が高いと判断してより大きな重みを与える。なお、分母は重みｗ_ｋ，ｉ のｋに関する総和を１にするための正規化項である。 That is, first, a weight w _{k, i} (0 ≦ w _{k, i} ≦ 1, k: verification result number, i: bit position) is prepared so as to increase as the reliability of the verification result increases. For example, it is determined as follows.

Here, x _{k, i} is a local Hamming distance near the i-th bit of the k-th collation result. Here, “local” means “local” on the iris image. That is, the bit value rk _{, i} having a small local Hamming distance is determined to have high reliability and is given a larger weight. Note that the denominator is a normalization term for making the sum of weights w _{k, i} related to k 1.

Then, the i-th bit value r _i of the composite verification result is determined by the following equation.

In (Expression 7), x _{k, i} may be a luminance value on the iris image (one point or a local region) corresponding to the i-th bit of the k-th collation result. That is, a bit corresponding to a portion having a large luminance value is given a smaller weight because it is likely to be an indefinite value due to reflection and is not reliable.

  Of course, weights other than those shown here may be used as long as the reliability of the corresponding bit is higher, the value becomes larger.

<Combination of blocks>
FIG. 15 is a flowchart showing details of the processing A4 in the case of combining in block units. In FIG. 15, first, a collation result having a Hamming distance equal to or smaller than a predetermined threshold TH3 (> TH1) is selected from N collation results (A411). Next, the authentication iris code according to each selected matching result is divided into M blocks (A412). Block division is performed as follows. That is, when the iris image is divided in advance and the pixels corresponding to the bits on the iris code belong to the same block on the iris image, the bits on the iris code also belong to the same block. As an iris image dividing method, for example, there are a concentric and radial dividing method as shown in FIG. 16A and a rectangular dividing method as shown in FIG. 16B.

  Next, paying attention to the i-th block (A413), the block with the smallest Hamming distance is selected from the blocks of interest of each iris code (A414). Then, steps A413 and A414 are executed for all blocks (A415).

  Thereafter, all the collation results of the blocks with the minimum Hamming distance are integrated (A416). The block collation result is a bit string indicating match / mismatch between corresponding bits in the block for two iris codes. Then, the final number of unmatched bits is normalized by the total number of bits, and a Hamming distance as a final matching score is calculated (A417).

  In this block unit synthesis, the effect of the process A411 is the same as the process A401 in the bit unit synthesis.

  Then, returning to FIG. 1, authentication is performed using the final matching score (Hamming distance) (A5). When the hamming distance is equal to or less than the predetermined threshold TH1, the person is accepted, and when not, the person is rejected as another person. Then, the final authentication result is displayed on the monitor 13 in the mobile phone with authentication function 10 of FIG.

  In addition, as a process combining the process A417 and the process A5 in the synthesis in units of blocks, in the collation result integrated in the process A416, when the intra-block Hamming distances of the respective blocks are all equal to or less than a predetermined threshold TH2 (≧ TH1) , Accept as the person, otherwise refuse as others.

  As described above, according to the present embodiment, it is possible to photograph a plurality of iris images having different external light reflection positions by moving the face, the eyeball, or the position of the camera. And it becomes possible to provide an apparatus with an excellent user interface by instructing the orientation of the face or line of sight (eyeball) by the photographing apparatus.

  Further, since a final matching score is obtained by integrating a plurality of matching results, it is possible to reduce the influence of external light reflections or shadows of 瞼 / 睫 on authentication accuracy. Furthermore, by integrating the collation results in units of blocks, it is possible to prevent others from being erroneously allowed. Further, by not using a collation result whose Hamming distance is greater than a predetermined threshold, it is possible to prevent others from being erroneously allowed.

  In addition, as a method of photographing a plurality of iris images having different external light reflection positions, there is a method of photographing the iris from a plurality of angles at the same time or at different timings by a photographing device including a plurality of cameras. FIG. 17 is an external view of a mobile phone with an authentication function, which incorporates a plurality of cameras, as a photographing apparatus according to the present invention. In the example of FIG. 17, two cameras 11a and 11b are installed apart in the vertical direction. When this mobile phone is used, two iris images with different external light reflection positions can be taken at a time. This shooting may be repeated once or several times. Of course, if a device incorporating more than two cameras is used, more iris images with different external light reflection positions can be photographed by one photographing.

  FIG. 18 is an external view of a fixed-type door phone with an iris authentication function installed at a front door or the like as a photographing apparatus according to the present invention. The door phone 30 of FIG. 18 is provided with a plurality of line-of-sight guide lights 36a to 36h as means for guiding the line of sight of the person to be authenticated. Then, at the time of shooting, the line-of-sight guide lights are turned on one by one in a predetermined order, and the person to be authenticated takes an iris image with the line of sight directed toward the lighted line-of-sight guide light. In this way, by providing a target for gaze guidance in the imaging device, it becomes easy to know which direction to face, so there is no excessive burden on the person to be authenticated at the time of gaze movement, and a human friendly interface Become.

  Further, the photographing apparatus may instruct the person to be authenticated to change the position of the photographing apparatus itself having the camera. When the person to be authenticated uses, for example, the mobile phone 10 with an authentication function as shown in FIG. 2 as the photographing apparatus, the photographing apparatus instructs the person to be authenticated to move the hand holding the photographing apparatus. In this case, it may be instructed by displaying a message, an arrow, or the like on the monitor 13 (the monitor 13 and its control unit correspond to means for instructing the moving direction of the hand), or instructed by voice from the speaker 15. (The speaker 15 and its control unit correspond to means for instructing the moving direction of the hand).

  Furthermore, a camera whose arrangement position can be changed is provided in the imaging apparatus, and the position of the camera may be changed automatically or manually by an authenticated person when an iris image is taken. As a result, it is possible to capture iris images from different positions depending on the movement of the camera itself, once the imaging device is held. Therefore, a plurality of iris images can be easily captured.

  In addition, when moving only the eyeball and shooting from the front of the face, if the movement of the eyeball is large, the iris will be shot from an oblique direction, and the iris outer edge (pupil outer edge) shape in the obtained iris image is elliptical. It becomes. In such a case, it is necessary to perform distortion correction processing on the input image, convert it to an iris image viewed from the front, and perform subsequent processing. Distortion correction can be realized using primary transformation (matrix calculation).

  The present embodiment is particularly effective at the time of authentication under the external light. However, under the external light, the near infrared light intensity in the external light is much stronger than the power of the illumination 12, so the effect of the illumination 12 is rare. Therefore, when the apparatus is provided with an on / off switch for the illumination 12 and authentication is performed under the external light, the illumination 12 may be turned off for photographing. Further, a near-infrared light intensity sensor may be provided in the apparatus so that the illumination 12 does not emit light when photographing when the near-infrared light intensity in the outside light exceeds a predetermined threshold.

  In the present embodiment, all the processes A1 to A5 are performed on the terminal side, but the process A1 for capturing a plurality of iris images is performed on the terminal side, and the captured plurality of images are transmitted to the server via the network. Then, the processes A2 to A5 may be performed on the server. In this case, the authentication result is calculated on the server, the calculation result is transmitted again to the terminal via the network, and the authentication result is displayed on the monitor 13 of the terminal.

  In this embodiment, an authentication method is adopted after N images (fixed number) of images having different external light reflection positions are captured, but the above-described processing is performed every time one image is captured. If the Hamming distance is calculated and the Hamming distance is equal to or smaller than a predetermined threshold, the authentication process is terminated. If not, the next image may be captured. In this case, an upper limit of the number of shootings may be set.

  In addition, the present invention is a measure for when external light reflection occurs, and originally, when no external light reflection occurs, it is not necessary to shoot a plurality of iris images. In addition, a system that always shoots a plurality of times is troublesome for the person to be authenticated. Therefore, the following method may be adopted.

  That is, in FIG. 25, after a single iris image is captured (D1), normal authentication (the method disclosed in Reference Document 1 above) is performed (D2). At this time, the photographed iris image and the registered iris data of the person to be authenticated registered in advance are collated, and a collation result is obtained. Then, when the authentication is successful, for example, when the Hamming distance obtained in step D2 is equal to or smaller than a predetermined threshold, the process is terminated, while when not, for example, when the Hamming distance exceeds a predetermined threshold, Proceed to step D4 (D3).

  In step D4, the collation result already obtained in step D2 is divided into a plurality of blocks, and a Hamming distance as a collation score is calculated for each of the divided blocks. Then, the number of blocks M2 whose Hamming distance is equal to or greater than a predetermined first threshold is counted (D5). When the number of blocks M2 is equal to or smaller than the second threshold in step D6, it is highly possible that the person is himself. Then, it is determined that external light reflection has occurred, and the processing (D7) of this embodiment is performed. On the other hand, when the number of blocks M2 exceeds the predetermined second threshold value, it is rejected that it is not the person (D8), and the process is terminated. That is, the process of determining whether or not the reflected iris image has occurred in the captured iris image is realized by steps D2 to D6.

Similarly, as a condition for performing the processing of the present embodiment,
When the iris image has a pixel whose luminance is equal to or higher than a predetermined first threshold value or higher than a predetermined second threshold value, or
A first threshold value based on the average luminance of the iris image may be calculated, and a pixel having a luminance equal to or higher than the threshold value in the iris image may be equal to or higher than a predetermined second threshold value. This is equivalent to determining whether or not the reflection is generated in the iris image at the image level based on the iris image.
Similarly, as a condition for performing the processing of the present embodiment,
・ When the lighting of the device is turned off, or
A near-infrared light intensity sensor may be provided in the apparatus, and the near-infrared intensity in outside light when the illumination is off may exceed a predetermined threshold. This corresponds to determining whether or not iris authentication is performed under external light based on the illumination switch setting or near-infrared intensity. For example, as shown in FIG. 26, first, the intensity of near-infrared light in the shooting environment is measured (E1), and when the measured intensity is equal to or greater than a predetermined threshold (Yes in E2), the processing of this embodiment is performed. (E3). On the other hand, when the measured intensity is lower than the predetermined threshold (No in E2), only one iris image is taken (E4), and the authentication process similar to the conventional one is performed (E5).

  Note that when shooting multiple images by moving the eyeball, the direction of movement may not be indicated appropriately, but the direction in which the reflection will disappear from the area where external light reflection is present may be obtained and indicated. Is possible. A region where external light reflection is present can be estimated as a region corresponding to a block having a large Hamming distance in the feature amount divided into blocks. And if the area | region where external light reflection exists can be estimated, it is easy to calculate which eyeball should be moved to. Further, as described above, when iris authentication is performed with the photographing apparatus in hand, an area where external light reflection is present may be estimated in the same manner to determine the moving direction of the hand. Furthermore, it is possible to determine the movement distance of the hand based on the position of the area where the external light reflection is present and to instruct this.

(Second Embodiment)
In the first embodiment, a plurality of matching results are created by collating a plurality of iris codes and a registered iris code, and the plurality of matching results are integrated to obtain a final matching score for authentication. It was. In contrast, in the present embodiment, an integrated iris code is created by integrating a plurality of iris codes, and authentication is performed by comparing the integrated iris code with a registered iris code.

  FIG. 19 is a flowchart showing a personal authentication method using iris recognition according to the second embodiment of the present invention. In FIG. 19, first, N iris images with different external light reflection positions are photographed (B1), and feature extraction is performed from the photographed N iris images to create N iris codes (B2). . For the processes B1 and B2, the same method as the processes A1 and A2 in the first embodiment can be applied.

  Next, the N iris codes are integrated to create an integrated iris code (B3). FIG. 20 is a flowchart showing details of the process B3.

  First, for the N iris codes ICi (i = 1 to N), the Hamming distance between all the codes is calculated (B301). There are N (N-1) / 2 combinations of codes. At this time, as shown in FIG. 12, matching that compensates for the rotation of the iris code is performed.

  Next, an iris code ICj that minimizes the sum of the Hamming distances with all other (N−1) codes ICi (i ≠ j) is selected (B302). Then, in the selected iris code ICj and the other code ICi (i ≠ j), the majority of the kth bit value (0 or 1) is taken to determine the kth bit value of the integrated iris code (B303). At this time, with respect to the reference iris code ICi, the other iris code ICj uses the kth position after being rotated by the rotation compensation angle when the Hamming distance is calculated in B301.

  By such processing B3, an integrated iris code is generated from the N iris codes. As a result, if the number of bits affected by external light reflection is less than N / 2 at any bit position of the iris code, an integrated iris code that is not affected by external light reflection can be generated. It is.

  Then, the integrated iris code and the registered iris code are collated to obtain a collation result (B4), and authentication is performed using the collation result (B5). The processes B4 and B5 may be the same as the method disclosed in Reference Document 1, that is, the processes (5) and (6) described in the first embodiment.

  Here, the reference iris code ICj is selected under the condition that the sum of the Hamming distances with all the other (N−1) codes ICi (i ≠ j) is minimized. The iris code may be compared with the registered iris code, and the iris code that minimizes the Hamming distance may be selected as a reference.

  In the present embodiment, the majority vote is used for integrating a plurality of iris codes. Instead, for example, a weighted average value may be used.

  Similarly to the first embodiment, the Hamming distance between the N iris codes and the registered iris code is calculated, and the integrated iris code is created using only the iris code whose Hamming distance is equal to or less than a predetermined threshold TH3. It doesn't matter if you do.

(Modification of the second embodiment)
The method described above is to integrate the binarized iris code as shown in FIG. On the other hand, the signals after applying the Gabor filter as shown in FIG. 11B may be integrated and then binarized. Below, the method is demonstrated as a modification of this embodiment.

  FIG. 21 is a flowchart showing a personal authentication method using iris recognition according to this modification. In FIG. 21, the same processes as those in FIG. 19 are denoted by the same reference numerals as those in FIG. 19, and detailed description thereof is omitted here.

  After shooting N iris images having different external light reflection positions (B1), the brightness level of the N iris images is adjusted (B15). When the iris images are taken in different directions, the brightness levels of the N iris images may differ depending on the aperture control of the camera itself or AGC (Auto Gain Control). In that case, an aperture value or an AGC gain value is acquired from the camera, and the luminance level is adjusted.

Then, feature extraction is performed from N iris images, and N intermediate iris codes are created as iris data, which is a feature amount for authentication (B21). Here, the method disclosed in the above-mentioned reference document 1, that is, the processing of (1) to (3) described in the first embodiment is performed, and the 2-d Gabor filter of (4) is further applied. The multi-value signal as shown in FIG. 11B is obtained. This multilevel signal is used as an intermediate iris code.

  Then, N intermediate iris codes are integrated to create an integrated iris code (B31). FIG. 22 is a flowchart showing details of the process B31.

ただし、ｘは角度方向のずらし量であり、ＩＣ２_ｊ ^ｘ （ｋ）は、中間虹彩コードＩＣ２_ｊ を角度方向にｘだけずらした後のｋ番目の値を示す。最もユークリッド距離が小さくなる位置でのユークリッド距離ＥＤ_ｉｊは、（数１）を用いて（数２）のように表せる。

First, for the N intermediate iris codes IC2i (i = 1 to N), Euclidean distances between all codes are calculated (B311). There are N (N-1) / 2 combinations of codes. At this time, as in the case of collating the binary code in the first embodiment, the collation is performed while shifting in the angular direction, and the Euclidean distance at the shift position where the Euclidean distance becomes the smallest is calculated. (Equation 1) is the Euclidean distance when the intermediate iris codes IC2i and IC2j are shifted by x in the angular direction.

However, x is a shift amount in the angular direction, and IC2 _j ^x (k) indicates a kth value after the intermediate iris code IC2 _j is shifted by x in the angular direction. The Euclidean distance ED _ij at the position where the Euclidean distance becomes the smallest can be expressed as (Equation 2) using ( _Equation 1).

  Then, when an intermediate iris code IC2j is used as a reference, the intermediate iris code IC2j is selected such that the sum of the Euclidean distances with all other (N-1) intermediate codes IC2i (i ≠ j) is minimized. (B312). Then, in the selected intermediate iris code IC2j and the other intermediate code IC2i (i ≠ j), the median value of the kth signal value of each intermediate code is taken to determine the kth signal value of the integrated intermediate iris code. (B313). At this time, with respect to the reference intermediate iris code IC2i, the other intermediate iris code IC2j uses the k-th position after being rotated by the rotation compensation angle when the Euclidean distance is calculated in B311.

In this way, an integrated intermediate iris code is generated from the N intermediate iris codes. In the integrated intermediate iris code, a Gabor filter output signal in which the influence of external light reflection is reduced is obtained by using the median value. Instead of using the median value, a weighted average value may be used. (It is possible to use a simple average value, but it is easily affected by the outlier due to the reflection of external light.)
Then, the integrated intermediate iris code is binarized to create an integrated iris code (B314). As a binarization method, a method similar to that for creating an iris code (FIG. 11C) from a normal intermediate iris code (FIG. 11B) may be used.

  Returning to FIG. 21, the integrated iris code and the registered iris code are collated, a collation result is obtained (B4), and authentication is performed using the collation result (B5).

(Third embodiment)
In the first embodiment, a plurality of matching results are integrated, and in the second embodiment, a plurality of iris codes or intermediate iris codes are integrated. In contrast, in the present embodiment, a plurality of iris images are integrated to create an integrated iris image, an iris code is extracted from the integrated iris image, the iris code and the registered iris code are collated, and authentication is performed. Assumed to be performed.

  FIG. 23 is a flowchart showing a personal authentication method using an iris image according to the third embodiment of the present invention. In FIG. 23, first, N iris images having different external light reflection positions are photographed (C1), and the brightness level of the N iris images is adjusted (C15). The processes C1 and C15 may be executed in the same manner as the processes B1 and B15 in the second embodiment.

  Then, the N iris images are integrated to generate an integrated iris image (C2). FIG. 24 is a flowchart showing details of the process C2.

  First, one image is selected from N iris images (C200). Then, the iris outer edge and the pupil outer edge are determined for the selected iris image (C201), and the iris region is converted from the xy orthogonal coordinate system to the rθ polar coordinate system (C202). Processes C201 and C202 are executed using the technique disclosed in the above-mentioned Reference Document 1. The processes C201 and C202 are executed for all N iris images (C203). The number of pixels of the converted iris image (polar coordinate image) is unified as r = R and θ = T for N images.

ただし、θ＋ｘ＞Ｔの場合は、θ＋ｘをθ＋ｘ−Ｔとする。（数３）は顔の傾きや眼球自体の回転を吸収するために角度方向にｘだけずらした場合のユークリッド距離である。予め定めたｘの範囲（許容回転範囲）で（数３）のユークリッド距離を計算し、距離が最小となるｘの距離を最終的な距離ＥＤｉｊとして決定する。

Then, for the N polar coordinate images, the Euclidean distance between all the polar coordinate images is calculated (C204). There are N (N-1) / 2 combinations of images. The Euclidean distance between the image Ii and the image Ij is defined as (Equation 3) as the sum of squares of the difference in luminance value of each pixel.

However, when θ + x> T, θ + x is θ + x−T. (Equation 3) is the Euclidean distance when the angle is shifted by x in order to absorb the tilt of the face and the rotation of the eyeball itself. The Euclidean distance of (Equation 3) is calculated within a predetermined range of x (allowable rotation range), and the distance of x that minimizes the distance is determined as the final distance EDij.

  Next, when a certain polar coordinate image Ij is used as a reference, a polar coordinate image Ij that minimizes the sum of the Euclidean distances from all the other (N−1) polar coordinate images Ii (i ≠ j) is selected ( C205). Then, in the selected polar coordinate image Ij and the other polar coordinate image Ii (i ≠ j), the median value of the luminance value of the pixel (r, θ) is obtained, and the luminance value of the pixel (r, θ) of the integrated polar coordinate image is obtained. Is determined (C206). The coordinate θ in the angular direction of each polar coordinate image Ii is one that has been rotationally corrected so that the Euclidean distance from the selected polar coordinate image Ij is minimized. In addition, when integrating a plurality of images, a weighted average value may be used instead of using the median value. (It is possible to use a simple average value, but it is easily affected by the outlier due to the reflection of external light.)

  Then, returning to FIG. 23, an iris code is created from the integrated iris image (C3), the iris code and the registered iris code are collated, and a collation result is obtained (C4). Thereafter, authentication is performed using the collation result (C5). Processes C4 and C5 use the method disclosed in Reference 1.

  In the present embodiment, the polar image after adjusting the luminance level is rotationally corrected, and the median value of each pixel is obtained to generate a composite polar image. By using the median value, a polar coordinate image in which the influence of reflection is reduced can be obtained. On the other hand, when the average value is used, it is easily affected by the outlier value due to the reflection. Therefore, in the subsequent processing, it is possible to perform processing with reduced influence of reflection.

  Needless to say, the processes of the second and third embodiments may also be used as the process of step D7 in the flow of FIG. 25 or step E3 of the flow of FIG.

It is a flowchart which shows the personal authentication method using the iris image which concerns on the 1st Embodiment of this invention. 1 is a mobile phone with an authentication function as an example of a photographing apparatus according to the present invention. It is a figure which shows typically the internal structure of the mobile telephone of FIG. This is an iris image in which external light is reflected in the iris region. It is the iris image image | photographed by changing the direction of an eyeball (gaze direction). It is an iris image taken by changing the direction of the face. It is a figure which shows a pupil outer edge and an iris outer edge. It is the figure which expressed the iris image by xy rectangular coordinate system. It is the figure which expressed the iris image by the rtheta polar coordinate system. It is a figure which shows the analysis zone | band which divided the iris into 8 rings. It is a figure which shows the preparation method of an iris code. It is a figure which shows the method of collating two iris codes, carrying out rotation compensation. It is a figure which shows the result of collating N iris code and registration iris code. It is a flowchart which shows the process in the case of synthesize | combining per bit about process A4 of FIG. It is a flowchart which shows the process in the case of synthesize | combining per block about process A4 of FIG. It is a figure which shows the method of dividing | segmenting the iris area | region in an iris image into a block. It is an external view of the mobile phone with an authentication function incorporating a plurality of cameras. It is an external view of the door phone with an authentication function incorporating a gaze guidance light. It is a flowchart which shows the personal authentication method using the iris recognition which concerns on the 2nd Embodiment of this invention. It is a flowchart which shows the detail of the process B3 of FIG. It is a flowchart which shows the personal authentication method using the iris recognition which concerns on the modification of the 2nd Embodiment of this invention. It is a flowchart explaining in detail the process B31 of FIG. It is a flowchart which shows the personal authentication method using the iris recognition which concerns on the 3rd Embodiment of this invention. It is a flowchart explaining the process C2 of FIG. 23 in detail. It is a flowchart which shows the personal authentication method including the process which judges the presence or absence of external light reflection in an image level. It is a flowchart which shows the personal authentication method including the process which judges whether it is the authentication under external light.

Explanation of symbols

10,10A Mobile phone with authentication function (photographing device)
11, 11a, 11b Camera 13 Monitor 15 Speaker 30 Door phone with authentication function (photographing device)
31 Cameras 36a to 36h Line-of-sight guide lights

Claims (2)

A measurement process that measures the intensity of near-infrared light in the shooting environment;
When the intensity is equal to or greater than a predetermined threshold, a plurality of iris images having different reflection positions on the iris due to sunlight are photographed for the person to be authenticated, and authentication is performed using the plurality of photographed iris images. Authentication process,
A personal authentication method comprising: a second authentication process for photographing an iris image of the person to be authenticated and performing authentication using the photographed iris image when the intensity is less than a predetermined threshold. . Measuring means for measuring the intensity of near-infrared light in the shooting environment;
When the intensity is equal to or greater than a predetermined threshold, a plurality of iris images having different reflection positions on the iris due to sunlight are photographed for the person to be authenticated, and authentication is performed using the plurality of photographed iris images. Authentication means,
A personal authentication device comprising: a second authentication unit that captures an iris image of the person to be authenticated when the intensity is less than a predetermined threshold, and performs authentication using the captured iris image; .

Images