JP2015232911A - Biometric authentication device and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To acquire more biometric information from one image by extracting tone difference in a blood vessel image in a taken image as information to be used for authentication.SOLUTION: A personal authentication device is used for performing personal authentication by using feature information of a blood vessel pattern acquired from a living body. The personal authentication device comprises: an imaging unit for taking an image of an area which is an authentication target in the living body; and an operation unit for acquiring the taken image as an authentication image. The operation unit extracts the blood vessel pattern from the authentication image, and acquires tone degree in the blood vessel pattern as the feature information.

Description

  The present invention relates to an apparatus and a method for identifying an individual using human biometric information, particularly a finger vein pattern.

  As the interest in security technology for safely managing personal information and property increases, biometric authentication using human biometric information is attracting attention. As a conventional biometric authentication technique, an authentication method using a fingerprint, an iris, a voice, a face, a vein on the back of a hand, or a finger vein is known. In particular, vein authentication technology that uses absorption of hemoglobin in the blood of infrared light irradiated to the living body has low psychological resistance because authentication can be performed simply by shining light on the finger, and internal information of the living body is used. Therefore, it has the feature of being excellent in forgery resistance.

  Vein authentication is expected to be applied in various fields, and in the field of entry / exit and attendance management, there is a growing need for large-scale 1-N authentication across multiple sites. Also, in the financial field, for example, utilization in empty-handed payment, in which shopping can be done without presenting a card simply by holding a finger, is drawing attention. In such applications, it is necessary to accurately distinguish tens of thousands to millions of users, and high accuracy exceeding the level of conventional finger vein authentication is required.

  Patent Document 1 discloses a method for improving authentication accuracy by using a plurality of fingers for authentication as compared to when using a single finger. Patent Document 2 discloses that an image is captured using a plurality of lenses, and a three-dimensional structure of a blood vessel is used for authentication.

JP 2010-035560 JP 2010-39534 A

  However, increasing biometric information used for authentication by acquiring a plurality of blood vessel pattern images can be a heavy load on arithmetic processing and storage media. Further increasing biometric information can also cause a problem that the identity matching rate is reduced due to environmental changes caused by external light or the condition of a biometric body that performs authentication. In addition, acquiring a plurality of images to form a three-dimensional structure of blood vessels also requires a lens array, a plurality of imaging devices, etc., which increases the number of parts and may cause an increase in cost and hinder compactness. There is. Furthermore, there is a problem that the identity matching rate is lowered due to unstable blood vessel information in the three-dimensional direction, positional deviation at the time of authentication, and the like.

  In view of such a problem, the present invention extracts a difference in contrast of a blood vessel image in a captured image as information used for authentication in a blood vessel image extraction device or a personal authentication device, and more from one image. The primary purpose is to obtain biological information.

  As an example of the present invention for solving the above-described problem, there is provided a personal authentication device used for performing personal authentication using characteristic information of a blood vessel pattern acquired from a living body, and is an authentication target in the living body. An imaging unit that captures an area; and a calculation unit that acquires the captured image as an authentication image. The calculation unit extracts a blood vessel pattern from the authentication image, and determines the degree of shading of the blood vessel pattern in the feature information. As a personal authentication device.

  According to the present invention, it is possible to acquire more biological information from a single image by using the difference in density of the imaged blood vessel pattern.

An example of the authentication apparatus which implements this invention. The system configuration example of the authentication apparatus which implements this invention. The system configuration example of the authentication apparatus which implements this invention. An example of the flowchart of the registration process which implement | achieves this invention. The relationship figure of coincidence rate and appearance frequency. The relationship figure of an image quality evaluation value and a discriminability evaluation value. An example of the fluctuation | variation of the image quality evaluation value in the fluctuation | variation of a pressure value. An example of the flowchart of the registration process which implement | achieves this invention. An example of the flowchart of the authentication process which implement | achieves this invention. The structural example of the authentication apparatus provided with the elastic member in the finger placement guide part. The structural example of the authentication apparatus provided with the elastic member in the finger placement guide part and the fingertip guide part. The structural example of the authentication apparatus which a finger placement guide part and a fingertip guide part sink in conjunction. The structural example of the authentication apparatus provided with the distance sensor. The structural example of the authentication apparatus which keeps a compression position away from an imaging part. An example of the registration screen which guides a finger to the correct presentation state. An example of the block diagram of the authentication process which implement | achieves this invention. An example of the expression method of brightness undulation information. An example of a finger image, a feature point, and a feature amount used for authentication. An example when blood vessels intersect. Explanatory drawing of the blood vessel identification by the positional relationship from the reference point of a feature point. 2 is a configuration example of an authentication device that performs light amount control. An example of the flowchart of a light quantity adjustment process. An example of the change of the histogram and blood vessel image in light quantity adjustment processing. An example of the authentication apparatus which provided the light-shielding means.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the embodiment of the present application, a finger vein authentication device will be described in particular. However, as shown in FIG. 3, the present invention can be applied to a personal authentication device using another living body such as a palm. Needless to say, the present invention can also be applied to a blood vessel image extracting apparatus which is an apparatus for extracting blood vessel images.

  In order to extract feature information used for authentication from the density of an image, it is important to capture more blood vessel information. However, for example, when the information used for authentication is registered and when authentication verification is performed again later, the method of pressing the presented body part against the device, the way of presentation, the difference in external light amount, etc. The problem is that the reproducibility becomes insufficient due to environmental changes in the body.

  In the present embodiment, in view of such a problem, a description will be given of a mode for acquiring information on a blood vessel shape and blood vessel density with higher accuracy.

  FIG. 1 is an example of a personal authentication device that implements the present invention, and FIG. 2 is a schematic diagram of the configuration of the personal authentication device of FIG.

  The user presents the finger 1 to the authentication device 2 at the time of authentication. At this time, the tip of the finger 1 is placed on the fingertip guide portion 5, and the base side of the finger 1 is placed on the finger placement guide portion 6. When the finger 1 is presented to the apparatus, the finger is detected using a detection unit such as a touch sensor (not shown), and authentication processing is started. Other sensors such as a distance sensor and a temperature sensor may be used for finger detection.

  The authentication device 2 is a cross-sectional view from the side. Pressure detectors 7-1 and 7-2 for detecting finger pressure to be presented are provided below the finger placement guide unit 6 and the fingertip guide unit 5, respectively. Further, an elastic member 8 such as a spring is sandwiched between the finger placement guide portion 6 and the pressure detection portion 7-2, and the finger placement guide portion 6 sinks downward due to pressure due to contact with the finger. The elastic member 8 may be any member that relieves pressure, such as a coil spring, a leaf spring, or a rubber member. Although details will be described later, the pressure applied to the guide portion due to the sinking can be reduced. In addition, the direction which sinks may not be a downward | lower direction, but any direction may be used if it is a direction which reduces a press. Alternatively, only one of the pressure detection unit 7-1 and the pressure detection unit 7-2 may be used.

  If the device to be implemented is a table-like device, when the living body places a finger on the guide portion, the finger is generally placed by pressing the guide portion from an obliquely upper side on the living body side. Therefore, when the elastic member 8 is arranged so as to relieve the pressure from directly above the apparatus, the pressure relieved by the elastic member 8 remains only in the vertical component, and the pressure of the horizontal component cannot be reduced. there is a possibility.

In this embodiment, as shown in FIGS. 2 and 3, the elastic member 8 is disposed obliquely on the living body side with respect to the upper side of the apparatus so as to relieve the pressure from the advancing direction when the living body presents the finger to the apparatus. ing. With this configuration, since the elastic member 8 can receive a larger pressure from the finger, the pressing can be more relaxed than an arrangement that relieves the pressure from directly above.
Authentication is performed by the following procedure. Infrared light is irradiated from the back side of the finger 1 from the light source 3. The light passes through the finger 1, passes through the finger presentation unit 9 where the authentication target area of the finger is presented, passes through the optical filter 10 that transmits only infrared wavelength light, and reaches the imaging device 4. The light is converted into an electrical signal by the imaging device 4 and is taken into the computer 12 as an image via the image input unit 11. The captured image is once stored in the memory 14. Then, the finger vein image (registration data) registered in advance is stored in the memory 14 from the storage device 15, and the CPU 16 collates the registered image with the input image by the program stored in the memory 14.

  In the collation processing, a correlation value between the two images to be compared is calculated, and it is determined whether the image matches the registered image according to the value. An individual is authenticated in accordance with the result, and when the authentication is correctly performed, a process at the time of authentication is performed on the control target of the authentication system. Further, the pressure detection units 7-1 and 7-2 provided below the finger placement guide unit 6 and the fingertip guide unit 5 detect the pressure when the finger is pressed, and perform an operation using the detected pressure value. The user can be prompted to relax the pressure value or suppress the fluctuation of the pressure value by using the display unit 17 or the speaker 18 based on the calculation result. A display, a liquid crystal, an LED lamp, or the like can be used as the display unit. Sounds emitted from the speaker 18 include voices and beeps.

  As an authentication mode, 1-N authentication is performed to check all registered images, and an ID number is entered to identify the user in advance, or an IC card is presented to the card reader. In the 1-N authentication mode, authentication is started immediately after a finger is presented to the apparatus in the 1-N authentication mode. Then, after inputting an ID number using the input means 19, a finger is presented and authentication is performed.

  Needless to say, the CPU, the memory, the storage device, the display unit, and the like can be stored in a terminal other than the authentication device 2.

  In the conventional personal authentication device, a guide unit for fixing the position of the finger is installed in the authentication device so that the user can present the finger with good reproducibility. By presenting the finger on the guide unit, the reproducibility of the captured vein image is enhanced, and high-accuracy authentication is possible. However, depending on the contact state between the guide unit and the finger when presenting the finger, the blood flow may be stopped due to the pressure, and the blood vessel pattern may be partially lost. Furthermore, if the finger is turned over and presented, the blood vessel may be compressed due to the tension of the epidermis of the finger, and the blood vessel pattern may be lost. If registration is performed in a state where the finger is strongly pressed or warped, the blood vessel pattern is lost, and an image is registered in a state in which information is insufficient compared to the blood vessel pattern that is originally possessed. Further, the pressing pressure value and the degree of finger warping change each time the finger is presented, and the acquired blood vessel pattern is not stable, so that the reproducibility at the time of authentication is lowered.

  Therefore, in order to improve the accuracy of finger vein authentication, it is necessary to register an image including all blood vessel patterns without any blood vessel pattern defect due to finger pressing or finger warping, and the authentication finger presented at the time of authentication. It is necessary to improve the reproducibility of. In order to improve reproducibility, a finger presentation position is fixed by a guide unit, and a part used for authentication is prevented from touching the apparatus so that a blood vessel pattern is not lost.

  An example until registration processing of registration data used for authentication in this embodiment will be described with reference to the flowchart of FIG. First, it is detected in step S100 that a finger has been presented to the apparatus, and an initial image quality evaluation value is calculated from the blood vessel image photographed in step S110. The image quality evaluation value is, for example, a method using an edge emphasis filter or a matched filter for emphasizing a line segment, a method of extracting a line pattern by tracking a line component, and a local indentation position of a luminance value in a cross-sectional profile of an image. The content rate of a blood vessel pattern obtained by binarizing into a blood vessel pixel and other pixels using an extraction method, the content rate of a highly specific region such as a bent region or a branched region of a blood vessel, and the like can be given. The content ratio of the blood vessel pattern can be represented by the ratio of the blood vessel pixels to the total number of pixels in the image after binarizing the blood vessel image and dividing the blood vessel into other portions. In addition, the content ratio of highly specific areas is the total number of pixels in the image and the number of pixels in the specific area when the blood vessel image is binarized into a specific area where the blood vessel is bent or branched and other areas. It can express with the ratio of.

  Next, in step S120, the user is prompted to relax the finger pressure presented through the display unit 17 and the like, and in step S130, the image quality evaluation value is calculated again. Then, in step S140, the amount of increase / decrease in the image quality evaluation value before and after the pressure is reduced is calculated.If the image quality evaluation value increases, an instruction is given to further reduce the pressure. In S150, it is checked whether or not the pressure value has changed. If the pressure value fluctuates and is not stable, an instruction is given to stop the finger in order to suppress the fluctuation of the pressure value in step S160, and the image quality evaluation value is calculated again in step S170. If the pressure value no longer fluctuates, it is checked in step S180 whether the image quality evaluation value exceeds the threshold value TH. If it falls below the threshold value TH, it is instructed to present another finger in step S190, and if it exceeds, it is registered as registration data used at the time of authentication in step S200.

  In addition to using a pressure sensor for finger presentation detection, a temperature sensor, a distance sensor, a blood flow sensor using an optical Doppler, or the like can be used to detect the presentation of a finger while performing living body determination.

  As a method for determining the image quality evaluation value threshold TH, for example, if a large amount of blood vessel image database can be prepared in advance, the image quality evaluation value threshold can be determined using this database and the discriminability evaluation value. The discriminability evaluation value is an index representing how accurately a vein image of a finger can be distinguished from a vein image of the same finger and another finger. For example, suppose that a vein image of a finger is collated with N databases of at least one identical finger and other finger vein images.

  5A and 5B are graphs in which the matching rate when a finger 1 and a finger 2 are collated with the database is represented on the horizontal axis, and the appearance frequency of the matching rate is represented on the vertical axis. If the finger 1 has high reproducibility of the blood vessel pattern and the blood vessel pattern can be acquired stably, the matching rate between the same fingers increases. Furthermore, if the finger 1 has a low matching rate with an average of another finger, the finger 1 has a high discriminability because there is a large difference between the matching rate between the same finger and another finger. On the other hand, since the finger 2 has a low matching rate between the same fingers and a high matching rate with the average other finger, the difference in the matching rate between the same finger and the other finger is small and the discrimination performance is low. It turns out that.

  Therefore, specifically, the distinction evaluation value is determined such that the lower the average of the matching rate with another finger, the higher the distinguishing evaluation value, or the matching rate between the same finger and another finger There is a method in which the magnitude of the average difference of the rates is used as a discrimination evaluation value. Plotting all registered data in the database with the vertical axis as the discriminant evaluation value and the horizontal axis as the image quality evaluation value, images with a high discriminant evaluation value have a large amount of blood vessels and many unique regions where the blood vessels are bent or branched As a result, the image quality evaluation value is also increased, and the image is distributed in an upwardly rising manner as shown in FIG. From this distribution, it is possible to statistically know how much the image quality evaluation value is, and whether the matching degree between the same fingers is high and the matching degree between the different fingers is low, that is, whether high discrimination can be maintained. By setting the image quality evaluation value corresponding to the discriminability evaluation value to be secured as a threshold value, when newly registering a blood vessel image, the image quality evaluation value of the blood vessel image is obtained without actually performing an authentication experiment. By simply calculating and registering a blood vessel image exceeding the threshold value, a highly discriminating blood vessel image can be registered.

  When there is a blood vessel pattern defect caused by pressing a finger against the device, the blood vessel pattern extracted from the image captured at each pressure increases as the user relaxes the pressure. The image quality evaluation value also increases as shown in the graph of FIG. If the characteristic value exceeds a preset value without considering the pressure value, the blood vessel image identification performance is guaranteed. If it is determined that registration is possible, the pressure in the graph of FIG. Registration is possible even if the finger is continuously presented with the value 2a. Alternatively, the image quality evaluation value of the image after the pressure has been relaxed is compared with the image quality evaluation value of the image before the pressure has been relaxed. It may be a value. The predetermined rate of change is, for example, the minimum value of the rate of change of the image quality evaluation value in the process of releasing the pressure and shifting the finger to a stationary state in a preliminary experiment, and the minimum value of the rate of change is the largest. The change rate of the subject can be set to a predetermined change rate. In addition, a predetermined change rate can be determined for each person to be presented.

  Actually, a blood vessel pattern defect occurs at the pressure value 2a, and all the blood vessel pattern defects caused by pressing cannot be eliminated unless the force is reduced to the pressure value 1a. By letting the user relax the pressure until there is no increase in the image quality evaluation value, it is possible to acquire a stable blood vessel image without a blood vessel pattern defect due to pressing for each finger, regardless of the amount of blood vessel patterns possessed by the finger. it can.

  In particular, in registered data that is required to be the most reliable data after registration and is required to be the most reliable data, if the pressure value is smaller than the pressure value 1c in FIG. Since the image quality evaluation value is the maximum, it is desirable to register with a pressure value smaller than the pressure value 1c. For example, the pressure value 1c can be determined as a pressure value when the image evaluation value does not increase even though the pressure value continues to decrease from the frame two frames before or after.

  Further, in the state where the image quality evaluation value is insufficient due to the blood vessel pattern defect when the finger is presented at the pressure value 3a in the graph of FIG. 7 (a) and the pressure is relaxed as in the pressure value 1b in the graph of FIG. 7 (b). It is possible to distinguish the shortage of image quality evaluation values. Therefore, by encouraging the pressure to be relaxed when the image quality evaluation value is insufficient due to a blood vessel pattern defect, the image quality evaluation value is increased to create a registerable image, and the image quality evaluation value is insufficient even when the pressure is relaxed. For such a finger having a small blood vessel pattern from the beginning, the recognizability of the registered image is enhanced by taking measures such as prompting another finger to re-present.

  By recording the pressure value in the storage device 5 at the time of registration, in the case of 1-1 authentication, reproducibility is improved by prompting the user to present a finger with the same pressure as at the time of registration through the display unit 17 or the like. Can do.

  In the first embodiment, as a means for further improving the authentication accuracy, a contact detection means for ensuring that the finger is in contact with the living body and a correlation is generated between the pressure from the finger and the image quality evaluation value is provided. Is effective. Details will be described in the second embodiment.

  In the first embodiment, a case where a contact detection unit is further provided will be described with reference to the flowchart of FIG. First, presentation of a finger is detected in step S200, and an image quality evaluation value calculation in step S210 is performed. Next, in step S220, whether or not the image quality evaluation value exceeds the threshold value TH is checked. If it exceeds, registration is performed in step S230. If it is lower, in step S240, it is checked whether or not the pressure value is lower than the threshold value TH2. For the method of determining the pressure value threshold TH2, as with the above example of the image quality evaluation value threshold TH, plotting all registered data in the database with the vertical axis as the distinguishability evaluation value and the horizontal axis as the pressure value at the time of registration. Thus, it is possible to determine a pressure value at the time of registration that can statistically maintain the distinguishability evaluation value. If the threshold value TH2 is exceeded, the display unit 17 or the like is instructed to relax the pressure in step S250, and if it is less than the threshold value TH2, finger contact detection is performed in steps S260 and S270. If the finger is not touching the device, an instruction is given to touch the device in step S280, and if the finger is touching, the presentation of another finger is instructed in step S290.

  With regard to detection of finger contact, an instruction is given to eliminate the pressure of the finger against the device and increase the image quality evaluation value, but some people may release their hands from the device. In this state, not only can the blood vessel pattern be stably extracted, but also the image quality evaluation value may be lowered, and even a finger having a blood vessel pattern that can be registered may be instructed to present another finger. There is. Therefore, by detecting the finger floating, it is possible to maintain the state where the pressure is relaxed and the finger is touching the device, so that it is possible to provide guidance for correctly instructing the presentation of another finger. . In addition to using the output value of the pressure sensor, a distance sensor or image processing can be used as the finger floating detection means. As for the detection of finger floating by image processing, a method of detecting by a change in the contour of a finger, a method of detecting by deformation of a blood vessel pattern, and the like can be considered.

  In the first and second embodiments, an example of the steps until the registration data is authenticated is shown. However, the present invention can also be used when determining the accuracy of the authentication data to be verified against the registration data. An example is shown in Example 3.

An example of the authentication process in the present invention will be described with reference to FIG. First, a finger presented in step S400 is detected, and an initial image quality evaluation value is calculated in step S410. Next, in step S420, collation with the registered image is performed. In the case of 1-N authentication, collation is performed with all N registered images, and in the case of 1-1 authentication, collation is performed with one specific registered image. In step S430, it is determined whether or not the blood vessel image input and the registered image match. If they match, post-authentication processing is performed in step S440. If they do not match, an instruction is given to release the pressure in step S450, and the image quality evaluation value is calculated again in step S460. In step S470, it is determined whether the image quality evaluation value does not increase. If the image quality evaluation value does not increase, collation with the registered image is performed again.
When the finger placement guide part 6 sinks downward by the action of the elastic member 8, the pressure applied to the finger is relieved, and the loss of the blood vessel pattern can be suppressed. The direction in which the finger placement guide 6 sinks may not be vertical. When sinking forward diagonally toward the fingertip, as shown in FIG. 10 (a), when the base of the finger first touches the finger placement guide, even if the finger is warped, as shown in FIG. 10 (b). Since the finger placement guide portion sinks diagonally forward, the finger is more easily bent than when the finger placement guide portion sinks vertically, and the loss of the blood vessel pattern due to warping can be prevented. The elastic member 8 may be sandwiched between the fingertip guide part and the pressure detection part. As shown in FIGS. 11 (a) and 11 (b), the fingertip guide portion sinks obliquely downward on the finger base side, whereby the finger is more easily bent and warping is less likely to occur. In addition, as described above, the ability of the living body to relieve the pressure from the advancing direction when presenting the finger to the device is also an effect obtained by being disposed obliquely on the living body side with respect to the upper side of the device. .

  A large number of subjects are collected and a preliminary experiment is performed, and the pressure value at the time of most registration is set to a pressure value at which the elastic member 8 starts to sink, and is used so that the elastic member 8 does not sink as much as possible. By prompting, it is possible to guide to a specific pressure state without requiring a special instruction during registration and authentication, thereby improving reproducibility. Alternatively, the optimum finger state can be easily reproduced by setting the limit position where the elastic member 8 sinks to an appropriate position in which no bending or excessive bending occurs in many fingers. Moreover, if it tries to sink deeper than said appropriate position, the repulsive force of the elastic member 8 will change, and it can also be made difficult to sink from an appropriate position.

  Further, as shown in FIGS. 12A and 12B, the fingertip guide portion 5 and the finger placement guide portion 6 may be sunk together.

The vein authentication apparatus of FIG. 2 does not have to use the elastic member 8. When the elastic member 8 is not used, finger warping is likely to occur. However, it is possible to detect the bending of the finger by means other than the elastic member 8. For example, by using a distance detection unit 21 such as a distance sensor to measure the distance between the finger and the sensor, it is possible to detect the bowing of the finger. As shown in FIGS. 13A and 13B, three or more distance measurement points 22-1, 22-2,..., 22 on a straight line extending in the major axis direction of the finger on the abdomen or dorsal side of the finger. The distance between -N and the distance sensor is measured, and from the deviation from the measured distance when a flat object is placed, the state of finger warping, bending, and finger floating can be detected. As the distance sensor, an optical type, an ultrasonic type, or the like can be used.
Moreover, you may provide the pressure detection part 7-1 and the pressure detection part 7-2 in the lower part of the whole apparatus. By providing pressure detection at three places, the finger placement guide section 6, the fingertip guide section 5 and the lower part of the apparatus, it is possible to know the breakdown of pressure applied to the entire apparatus. Therefore, the reproducibility can be further improved during the 1-1 authentication by recording the breakdown of the pressure at the time of registration.

  Even if guidance is provided to relax the pressure, the blood vessel pattern may be lost because there is no more pressure than the finger and the device are in contact. Therefore, the blood vessel pattern by pressing is arranged by disposing the contact positions of the finger placement guide unit 6 and the fingertip guide unit 5 with the fingers away from the imaging device 4 as shown in FIG. 14B than FIG. 14A. Reduce the effects of defects.

  FIG. 15 is an example of a registration screen that displays the presentation state of the user's finger during registration and guides the user to the correct presentation state. Since most users are not familiar with the authentication device at the time of registration, it is necessary to guide the finger presentation method. The presentation state of the finger 1 is displayed on a display unit 17 such as a monitor, and the user changes the presentation state of the finger according to the guidance while viewing the display unit 17.

  Information displayed on the display unit 17 includes an output value of the pressure detection unit 7, an image quality evaluation value, detection information of finger floating, a degree of finger bending, a degree of finger bending, a degree of sinking of the elastic member 8, and the like. is there. The information displayed on the display unit 17 is switched in real time according to the presentation state of the finger.

For example, when the pressure due to pressing is too high, an indicator indicating the pressure value and an image diagram or animation of a state where the finger is pressed strongly are displayed together, and a guidance sentence such as loosen the pressing pressure is displayed on the display unit 17. To display. When there is a finger lift, finger bend, finger warp, or elastic member 8 sinking, the finger floated state, finger bent state, finger warped state, elasticity, etc. The state where the member 8 sinks is expressed, and is displayed on the display unit 17 together with a guidance sentence such as do not lift the finger, do not bend, do not warp.
Based on these pieces of information, the user can know how to present the finger.

  In addition, it is possible to immediately confirm whether or not the finger placement is correctly corrected while looking at the indicator. Correct registration work according to the guidance on the screen eliminates the problems of blood vessel pattern defects and blood vessel pattern changes caused by finger pressing and finger floating that are difficult to detect even when an administrator is present at the time of registration. be able to.

In the registration process according to the present invention, even if the user presents a finger as instructed, if the image quality evaluation value is low, the user may be prompted to present another finger. Therefore, some people may present a plurality of fingers. Since the blood vessel image of the second and subsequent fingers presented does not necessarily have a higher image quality evaluation value than the first finger, the image quality evaluation value is the highest among the photographed blood vessel images when prompted to present multiple fingers. Register a finger with a high. Alternatively, all registrants may be prompted to present a plurality of fingers, and the finger having the highest image quality evaluation value of the blood vessel image among the plurality of fingers may be registered. It may be instructed to register blood vessels images of all fingers and present a finger with the highest image quality evaluation value at the time of authentication.

  Thus, by imaging a clearer blood vessel pattern and extracting more blood vessel information by image quality evaluation value and pressure control, information based on the density of the blood vessel can be extracted more accurately.

  In the fourth embodiment, a means for extracting feature information from the density of a blood vessel pattern of an image acquired as an authentication image will be described.

  An embodiment of an authentication processing procedure using the feature point matching method will be described using the processing block of the authentication device in FIG. First, the blood vessel image 31 stored in the memory 14 is used as an input, and processing for detecting the presentation of a finger is performed in step S500. When the presentation of the finger is detected, the light amount of the light source 3 is adjusted so that the clearest blood vessel image is obtained in step S510. For this purpose, the average luminance value of the image is constantly monitored, the light amount is feedback-controlled according to that value, and an evaluation function that determines the sharpness of the pattern is applied to the finger vein pattern image, and the result is optimized. For example, a method for adjusting the amount of light can be used.

  As an example of determining the sharpness of a vein pattern, the evaluation function can be determined so that the pattern becomes clearer as the amount of extracted blood vessels increases.

Next, pre-processing in step S520 is performed. The photographed blood vessel image contains a lot of noise other than blood vessels such as rough finger surfaces and wrinkles, and the blood vessels are not always clearly displayed. Therefore, processing for emphasizing only blood vessels and suppressing noise is necessary. As the preprocessing, a technique such as edge holding smoothing filtering or unsharp masking can be used.
After the preprocessing, the feature point 51 is detected directly from the image in which the blood vessel is emphasized in step S530. FIGS. 17A and 17B are image diagrams of three points 41-1, 41-2, and 41-3 on a certain blood vessel image and a luminance profile on a straight line connecting these three points.

  The feature point is detected by, for example, a curvature or an angle that can be expressed by a height difference of three or more points on a straight line on a blood vessel image as shown in FIG. 17A, and a gradient that can be expressed by a difference in luminance between two points. Use undulations such as brightness. Hereinafter, information based on the luminance difference will be referred to as luminance undulation information. A branching point or bending point in a blood vessel in which the undulating shape of the luminance is stable and specific is detected from the grayscale image of one blood vessel as shown in FIG. 18A by using a luminance gradient, a luminance curvature, or the like. . Thereafter, feature quantities are extracted from the feature points 51 detected in step S540. A method of converting a luminance gradient, a luminance curvature, or the like into a feature amount as luminance undulation information around the feature point 51, a method of combining the luminance undulation information around the feature point 51 and the blood flow direction of the blood vessel around the feature point, etc. Can be used. As an example of a method for converting the luminance undulation information around the feature point 51 into a feature amount, a square area of D pixels × D pixels centering around the feature point 51 is divided into 9 areas of 3 × 3, and created for each area. There is a method in which a luminance direction histogram in eight directions (FIG. 16B) is used as a feature amount. The feature amount of one feature point is position information of the feature point 51 and 72-dimensional luminance relief information. The luminance gradient histogram of each of the nine regions is obtained by calculating the difference in luminance value as a gradient with respect to each of the reference pixels 53 in the eight directions separated by the distance d with reference to the reference pixel 52 located at the center of each region.

  Finally, the feature amount matching process in step S550 is performed. All feature points of the registered image and all feature points of the collation image are compared in a brute force manner to perform association. The association is performed when the similarity between the feature points of the closest feature amounts exceeds a preset threshold value. When the ratio of the number of feature points associated with the total number of feature points of the registered image and the collation image is set as the coincidence rate and the coincidence rate exceeds a preset threshold value, it is determined that the fingers of both images are the same person.

  By using the positional relationship between coordinates corresponding to each corresponding point obtained by associating the feature points of the registered image and the input image, accurate alignment between images to be collated can be performed. In addition to rotational movement and parallel movement on the image plane, three-dimensional geometric distortion is corrected and aligned, and then matching is performed by a method other than feature point matching, such as a template matching method. Compared with collation using a point matching method or a template matching method alone, the matching rate of vein images of the same finger can be increased.

  In an authentication device that has a high degree of freedom when presenting a finger and is highly convenient and easy to use, finger misalignment and rolls are likely to occur, and the reproducibility of the presented finger is reduced. Therefore, in a matching method using a blood vessel pattern that represents only the presence or absence of blood vessels, such as a template matching method, when the deviation of the finger presentation position increases, the matching area between the images to be matched narrows, and information for identifying an individual Is insufficient, and the accuracy of authentication decreases. Therefore, even when the matching region between the images to be collated is narrow, the authentication accuracy can be maintained high by using more information than the conventional information such as the blood vessel pattern representing only the presence or absence of the blood vessel.

Two blood vessel images in which two blood vessels intersect as shown in FIGS. 19A and 19B are determined as the same blood vessel in a blood vessel pattern in which the blood vessel and the other blood vessels are expressed in binary, and cannot be distinguished. However, it is possible to grasp the front-rear relationship of two intersecting blood vessels by using the depth information of the shading due to luminance undulations, and FIG. 19 (a) and FIG. 19 (b) can be distinguished as different blood vessels. is there. Accordingly, in FIG. 19A, of the five points 61a, 62a, 63a, 64a, and 65a, three points 63a, 61a, and 65a on the same blood vessel have close brightness, and on the blood vessel that runs behind them. The two points 62a and 64a are close in luminance. In FIG. 19B, the front and rear of the two blood vessels are reversed, so that the three points 62b, 61b, and 64b have close luminance, and the two points 63b and 65b have close luminance. Therefore, the region where the blood vessels intersect has higher specificity than the other regions, and has more characteristic information, and therefore has high discrimination. Therefore, preferentially register fingers that contain a lot of blood vessel crossing areas, and prioritize matching by matching the weights of the crossing area with a large weight. As a result, it is possible to achieve both high accuracy and high speed authentication. Further, not only blood vessel information but also information such as fingerprints and wrinkles existing on the surface of the finger is used as a feature amount, so that discrimination can be further improved. For example, after detecting a feature point from a branch point or bending point on a blood vessel using the above method, not only the luminance gradient information of the blood vessel around the feature point but also the finger surface corresponding to the detection position of the feature point A method using fingerprint information and wrinkle information on a finger surface around a point as a feature amount can be considered.
Furthermore, the extracted feature points can be compared, and each feature point can identify the positional relationship of each feature point in the depth direction of the blood vessel pattern.

  For example, for two types of blood vessels as shown in FIG. 20A, it is assumed that feature points A to B having the same feature information exist on each blood vessel. In such a case, the two blood vessels are determined to be the same blood vessel only by comparing the characteristic information of A and A, B and B, and C and C, respectively. Therefore, as shown in FIG. 20B, a predetermined reference value is provided for the luminance of the image, and the luminance and the luminance of each feature point are compared, whereby the relationship between the luminance values of the respective feature points, that is, the image. Information about the shape in the depth direction of the blood vessel imaged on a plane can be acquired, and the two blood vessels in FIG. 20 (a) can be identified.

  In addition, a plurality of characteristic points that are always detected at the same position are extracted from the blood vessel of the vein image using luminance undulation information (minimum curvature, etc.), regardless of the illumination intensity of the light source or the position of the light source, It is possible to reconstruct the luminance relief structure of the blood vessel pattern while connecting the extracted feature points.

  By using this reconstructed blood vessel shape in the depth direction for authentication in the same way as two-dimensional template matching, the computational load can be significantly reduced compared to when comparing feature information one by one. It becomes possible.

In addition, since the reconstructed blood vessel shape is generated from stable feature points of depth information, it is more stable and more accurate than the case of constructing a 3D structure directly from the captured image. A high blood vessel pattern can be obtained.
At the time of reconstruction, it is possible to cope with a change in the irradiation intensity of the light source by estimating and correcting the luminance undulation state of a desired brightness with respect to the luminance undulation in an extremely bright / dark region.
In addition, if the position of the light source is different, the overall luminance gradient of the vein image changes. From this luminance gradient, the relative position of the finger and the light source is estimated, and correction is made so that the luminance gradient becomes flat. Thus, it is possible to cope with a change in the position of the light source.
When the finger vein authentication method based on the feature point matching method described in the present embodiment is set as method A and the authentication method using the conventional template matching method is set as method B, the authentication results of methods A and B are integrated, By using the final authentication result, higher authentication accuracy can be realized.

  Although it is possible to perform the processing of the fourth embodiment alone, using the image quality evaluation value described in the first embodiment, the processing of the first embodiment is performed on the captured image, and a sufficient blood vessel pattern is captured. By using an image determined to be used as an authentication image, the luminance difference between pixels constituting the image used for authentication becomes clearer, and authentication accuracy can be improved by obtaining more accurate luminance undulation information. it can.

  In order to acquire more brightness relief information more accurately, it is effective to control the amount of light emitted from the light source more precisely, and authentication accuracy can be further improved. Further, in extracting the image quality evaluation values shown in the first and second embodiments, it is possible to obtain a clear blood vessel pattern image and a highly accurate content rate by appropriately controlling the light amount. An example of a method for appropriately controlling the amount of light is shown in the fifth embodiment.

  In this embodiment, as shown in FIG. 21, the upper part of the finger to be presented is opened, and the device shape does not make the user feel a sense of pressure. Due to the limitations of the apparatus, the entire finger is irradiated with two light sources arranged obliquely in front of the fingertip.

  When the amount of light emitted from the light source is large and luminance saturation such as overexposure occurs, the luminance saturation area loses the undulation information, and therefore, the presence of luminance saturation is not desirable in the authentication method using the luminance undulation. However, if the irradiation intensity is weak and the amount of light is insufficient, uneven brightness occurs, the blood vessel image becomes dark as a whole, and a clear blood vessel image cannot be obtained. Therefore, particularly for an authentication method that uses luminance undulations, it is desirable to acquire a clear blood vessel image without luminance saturation and luminance unevenness. Such control is also effective in obtaining a large amount of brightness undulation information in the fourth embodiment and accurately obtaining the information.

  FIG. 22 is an example of a flowchart of the two-stage light amount adjustment. First, in step S600, the initial light amount is turned on. Next, in steps S610 and S620, the light intensity ratio of the plurality of light sources 3 is fixed at the ratio of the initial light intensity, the irradiation intensity is changed, and the optimum irradiation intensity is determined. The optimum irradiation intensity represents the irradiation intensity that has been brightened to the limit at which luminance saturation does not occur under a certain light quantity ratio. The method of determining the light amount ratio of the initial light amount can be determined by, for example, performing a light amount control experiment on a plurality of fingers in advance and selecting the light amount ratio set as the optimum with the most fingers. When the light intensity ratio is fixed and the irradiation intensity is changed, the luminance histogram of the blood vessel image is shifted while maintaining the shape (transition from state 1 to state 2 in FIG. 23A). Once irradiation is performed at a certain appropriate irradiation intensity, a correlation between the irradiation intensity and the position of the histogram can be obtained, and the optimum irradiation intensity can be estimated. Next, a plurality of light sources are turned on at different light intensity ratios of several patterns determined in advance at the optimum irradiation intensity in step S630, and the optimum light intensity ratio is set so that the entire image is uniformly brightened in step S640. decide. The optimum light quantity ratio is determined based on an index that indicates how well the luminance histogram is organized. For example, when the index is luminance variance, the base of the histogram becomes smaller from the average luminance as the variance becomes smaller. Therefore, the light quantity ratio that minimizes the variance is set as the optimum light quantity ratio (from state 2 in FIG. 23B). Transition to state 3).

  In the arrangement of the light sources in the apparatus structure of FIG. 21, there is an area 73 where the irradiation range 71 of the finger base light source that is intended to irradiate the finger base side overlaps with the irradiation range 72 of the finger tip light source that is intended to irradiate the fingertip side. Changes depending on the thickness and length of the finger. For this reason, when each light source is controlled independently, a luminance saturation region or a dark region due to insufficient light amount is formed, and luminance unevenness may occur. Therefore, it is necessary to simultaneously determine the optimum irradiation intensity at which luminance saturation does not occur and the optimum light quantity ratio that uniformly brightens the entire finger uniformly.

  First, determine either the optimal irradiation intensity or the optimal light intensity ratio, and then perform a two-level light intensity control method that determines the other. It is possible to acquire a clear blood vessel image. In this example, after obtaining the optimum irradiation intensity, the optimum light intensity ratio was obtained under the optimum irradiation intensity. However, after obtaining the optimum light intensity ratio, the optimum irradiation intensity was obtained under the optimum light intensity ratio. Also good. Further, it is not always necessary to adjust the light amount through two stages, and the light intensity adjustment may be performed by determining the optimal irradiation intensity and the optimal light amount ratio by simultaneously changing the irradiation intensity and the light amount ratio.

  When this embodiment is combined with the methods shown in the first to fourth embodiments, it is possible to obtain a more accurate image quality evaluation value and luminance undulation information and obtain higher authentication accuracy. It can also be implemented as a single approach by example only.

  In carrying out the present embodiment, it is possible to set a more accurate and efficient light height and light quantity ratio by providing a light-shielding portion that prevents the overlapping of irradiation ranges between each of a plurality of light sources. An example is shown in Example 6.

  FIG. 24A is a configuration example of an input interface of a personal authentication device provided with a light-shielding unit 81 for preventing overlapping of irradiation ranges of a plurality of light sources. 24A, 24B, and 24C are side views of the input interface.

When the irradiation ranges of the two light sources overlap as shown in FIG. 21, the light shielding unit 81 blocks a part of the irradiation light of the light source that irradiates the finger base side as shown in FIG. Since there is no need to consider the mutual relationship between the light sources, a plurality of light sources can be controlled independently.
However, since the light shielding portion 81 is installed as shown in FIG. 24B, the irradiation range may not be spread over the entire finger depending on the size of the finger. Therefore, by making the light shielding portion 81 movable as shown in FIG. 24C, it is possible to irradiate the entire finger brightly in response to individual differences in finger size.

It can be used as a personal authentication device or a blood vessel pattern extraction device using a blood vessel pattern.

DESCRIPTION OF SYMBOLS 1 ... Finger, 2 ... Authentication apparatus, 3 ... Light source, 4 ... Imaging part, 5 ... Finger tip guide part, 6 ... Finger placement guide part, 7 ... Pressure detection part , 8 ... elastic member, 9 ... finger presentation unit, 10 ... shading filter, 11 ... image input unit, 12 ... computer, 13 ... interface, 14 ... memory, 15・ ・ ・ Storage device, 16 ... CPU, 17 ... Display unit, 18 ... Speaker, 19 ... Input unit, 21 ... Distance detection unit, 22 ... Distance measuring point, 41. .. Pixels in blood vessel image, 31... Blood vessel image, 51... Feature point, 52... Reference pixel, 53... Reference pixel, 71.・ Lighting range of fingertip light source, 73... Overlap area of light quantity, 81.

Claims (6)

A personal authentication device that performs personal authentication using characteristic information of a blood vessel pattern acquired based on light transmitted through a living body,
An imaging unit for imaging an area to be authenticated in the living body;
An acquisition unit that acquires an image including the blood vessel pattern captured by the imaging unit as an authentication image;
An authentication unit that performs authentication based on the authentication image acquired by the acquisition unit,
The personal authentication apparatus, wherein the authentication unit preferentially authenticates a crossing region where blood vessels intersect. The personal authentication device according to claim 1,
The personal authentication apparatus, wherein the authentication unit extracts the intersection area from a luminance difference of the authentication image. The personal authentication device according to claim 2,
The authentication unit acquires a plurality of feature points having a predetermined feature on the blood vessel pattern based on the luminance of the authentication image,
Based on the brightness of each of the plurality of feature points, a brightness difference between the plurality of feature points is calculated,
Obtaining the luminance of each of the plurality of feature points on the blood vessel pattern as depth information derived from the shape of the blood vessel pattern in a direction perpendicular to the plane forming the authentication image;
A personal authentication device characterized in that, from the depth information, a relative position of each of the plurality of feature points in the vertical direction is calculated as position information, and the intersection area is extracted from the position information and the luminance difference. A personal authentication method for performing personal authentication using characteristic information of a blood vessel pattern acquired based on light transmitted through a living body,
An imaging step of imaging an area to be authenticated in the living body;
An acquisition step of acquiring the imaged image including the blood vessel pattern as an authentication image;
An authentication step of performing authentication based on the acquired authentication image,
In the authentication step, a personal authentication method characterized by preferentially authenticating a crossing region where blood vessels intersect. The personal authentication method according to claim 4,
In the authentication step, the intersection region is extracted from a luminance difference of the authentication image. The personal authentication method according to claim 5,
In the authentication step, a plurality of feature points having a predetermined feature on the blood vessel pattern are acquired based on the luminance of the authentication image,
Based on the brightness of each of the plurality of feature points, a brightness difference between the plurality of feature points is calculated,
Obtaining the luminance of each of the plurality of feature points on the blood vessel pattern as depth information derived from the shape of the blood vessel pattern in a direction perpendicular to the plane forming the authentication image;
A personal authentication method, wherein a relative position of each of the plurality of feature points in the vertical direction is calculated as position information from the depth information, and the intersection area is extracted from the position information and the luminance difference.

Images