JP2012128868A - Feature image photographing device and personal identification device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve highly accurate personal identification while using an unclear finger vein pattern image with position deviation concerning personal identification under an environment where non-contact property is requested.SOLUTION: This feature image photographing device includes: means for acquiring a finger vein pattern image in a non-contact state; means for performing spin compensation by using the outline of the finger as a method of extracting a vein pattern included in the acquired image; means for normalizing the position of a finger image by using the fingertip as a reference; means for acquiring a statistically overall vein pattern by repeatedly tracing the dark section of a luminance place only by arbitrary length from the arbitrary position of the image; matching means for comparing only the sections where strong features are shown in a blood vessel pattern; and means for dividing the image into small areas, and for independently performing matching for every small area, and for evaluating a position deviation value in the case of matching.

Description

The present invention relates to an apparatus and method for performing personal authentication using a vein pattern obtained by imaging light transmitted through a human finger.

In personal authentication technology, there are methods based on fingerprints, irises, voice, veins on the back of the hand, and the like. Personal authentication devices based on fingerprints have already been commercialized and sold by multiple companies. These products perform fingerprint authentication by touching the fingerprint sensor on the fingerprint sensor, recognize fingerprint endpoints and branching points as feature points, and perform personal authentication by collating them with registered fingerprint feature points. Yes.

Japanese Patent Laid-Open No. 10-295674 discloses a personal authentication device based on the veins on the back of the hand. According to this, the back of the hand is directed to the imaging camera, and the blood vessel pattern is captured using the reflected light emitted from the imaging camera, and authentication is performed. At this time, it is devised so that the position of the hand for imaging is not shifted for each authentication by grasping the fixed rod-shaped guide.

Japanese Patent Application Laid-Open No. 7-21373 discloses a personal authentication device using a finger vein, and in particular, in order to reduce a loss of light quantity when photographing, a finger image is photographed with an optical fiber in close contact with the finger. It is disclosed.

JP-A-10-295664 JP 7-21373 A

In the conventional technology, a method having a great psychological resistance such as collecting a fingerprint at the time of authentication or putting light in the eyes is adopted. Further, in the conventional personal authentication device, it is necessary to bring the human body into contact with the authentication device, and there are cases where it is not appropriate to use it in a medical field that places importance on hygiene. It can also be counterfeited because it uses features that are exposed outside the body.

An object of this invention is to construct a security system in the environment where non-contact property is calculated | required, such as a medical field. Therefore, the present invention provides an apparatus and method for photographing a finger image in a non-contact manner, extracting a finger vein pattern from the finger image, and performing personal authentication.
Furthermore, in the present invention, when a finger is photographed in a non-contact manner, attention is paid to a new problem that rotation and luminance unevenness are likely to occur and it is difficult to obtain a highly accurate personal authentication result. Therefore, the present invention provides a personal authentication apparatus and method for performing high-accuracy personal authentication while using a finger vein pattern image that is likely to cause rotation and uneven brightness due to non-contact.

In order to achieve the above object, a personal authentication device of the present invention includes a storage device that stores a vein pattern of a registered finger image, an interface including a light source and a camera that captures finger transmitted light, and captured finger transmitted light. Means for extracting a vein pattern included in the image and collating the extracted vein pattern with the vein pattern of the registered finger image for personal authentication, and the interface has a groove for inserting the finger in a non-contact manner. The light source and the camera are arranged to face each other with the groove interposed therebetween.

Further, the means for performing personal authentication corrects a rotation on the imaging plane that occurs when the finger is inserted into the interface with respect to the photographed finger image, and a vein pattern included in the rotation-corrected finger image. Extracting and performing personal authentication.

ADVANTAGE OF THE INVENTION According to this invention, the authentication using the characteristic inside a living body which does not require the contact to an apparatus, has a low psychological resistance at the time of authentication, and is hard to be counterfeited can be performed. In addition, high-precision personal authentication can be realized while using a non-contact-specific positional shift and a blurred image.

It is an example of the system configuration | structure for implement | achieving this invention. It is an example of the input interface structure for acquiring the vein of a finger. It is an example of the input interface structure for acquiring the vein of a finger. It is an example of the finger vein pattern input interface configuration in consideration of safety. It is an example of arrangement | positioning of the light source and CCD camera in an input interface for imaging a vein pattern from multiple directions. It is an example of a system configuration in which entry / exit including authentication can be performed without contact. This is an example of a system configuration that performs authentication by combining personal feature information such as a personal identification number, fingerprint, iris, voice, handwriting, and face in addition to a vein pattern. It is an example of a system configuration for acquiring a template image of a vein pattern using an IC card. It is the flowchart which showed the outline | summary of the process by the software for implement | achieving this invention. It is the image figure which showed the outline tracking method of a finger image. It is the image figure which showed the method of carrying out the rotation correction | amendment of the inclination of a finger image. It is the image figure which showed the method of normalizing the cutout part of a finger image. It is an example of the flowchart for taking out a vein pattern from a finger image. It is an example of the flowchart for calculating | requiring the mismatch rate of two vein patterns. It is an example of the flowchart for calculating | requiring the correlation of a vein pattern using the partial image of two vein patterns. It is a performance comparison with the method by this invention, and another method in the relationship between the other party acceptance rate (FAR) of an authentication system, and a principal rejection rate (FRR).

Hereinafter, an embodiment of the present invention will be described in detail. FIG. 1 is a schematic block diagram of a system configuration for realizing the present invention. In the vein pattern input interface 1 corresponding to the finger insertion portion, there are a light source 2, an optical filter 3, and a CCD camera 4. A vein pattern is acquired by inserting a finger between the light source 2 and the optical filter 3. The light transmitted through the finger is imaged by the CCD camera 4 through the optical filter 3. The image signal picked up by the CCD camera 4 is taken into the PC 6 using the image capture board 5. Inside the PC 6, the captured image signal is stored in the memory 8 through the interface 7. A registered image stored in the external storage device 10 is stored in the memory 8. Then, according to the program stored in the memory 8, it is determined whether or not the image captured by the CPU 9 matches the registered image. This program may be supplied to the apparatus using an external storage medium. As the storage medium, for example, a flexible disk, hard disk, optical disk, magneto-optical disk, CD-ROM, CD-R, magnetic tape, nonvolatile memory card, ROM, or the like can be used.

FIG. 2 is an example of the structure of the finger vein pattern input interface 1 that acquires a finger vein pattern image by a non-contact method. The interface has a finger insertion opening 22 which is a groove-like gap, and a vein pattern is acquired by passing the finger 20 through that portion. At this time, it is not necessary to bring the finger into contact with the device itself. Moreover, the vein pattern of a plurality of fingers can be imaged by passing a plurality of fingers through the groove. Furthermore, it is also possible to capture a finger image by continuously passing a plurality of fingers 20 through the finger insertion slot 22 by swinging down a finger while drawing a circular trajectory around the shoulder, elbow, wrist and the like. When such a non-contact imaging method is taken, the captured finger image is rotated with respect to the imaging plane. Therefore, it is necessary to inevitably perform rotation correction on the captured finger image.
The direction of the groove may be any direction with respect to the ground, but in particular, when a groove perpendicular to the ground is provided, the user is a natural motion of swinging his hand down according to gravity. Since the pattern image can be acquired, there is an effect that it is easy to use.
In addition, such an interface structure in which a finger is passed through the groove is less likely to rotate around the central axis of the finger, so that the accuracy of authentication is improved.

FIG. 3 is another example of the finger vein pattern input interface 1. This is an example of an input interface having a hole-like finger insertion slot 22. The finger 20 is inserted into the insertion slot 22 of the finger, and the vein of the finger is acquired by one or a plurality of light sources / image sensors inside the finger 20. At this time, a vein pattern from multiple directions can be obtained by rotating the finger concentrically with respect to the central axis of the finger.

FIG. 4 shows an example of the non-contact type finger vein pattern input interface in consideration of safety during authentication and when passing near the interface installation location. (a) Chamfers the corners of the finger vein pattern input interface so that it is safe even when fingers, hands, arms, etc. come into contact. Also, depending on the installation location, it may be dangerous if the device protrudes. In that case, as shown in (b), the groove is provided sideways to narrow the width of the protrusion of the interface, or as shown in (c), a finger insertion opening is provided on the wall surface itself. However, in (c), there is a groove of sufficient width to swing the arm down and pass through. Further, in order to cope with imaging of the finger by swinging down the arm, as shown in (d), the back of the insertion slot of the finger is formed in an arc shape in accordance with the trajectory of the finger. By adopting such a structure, the finger is less likely to contact the device with respect to the circular movement of the arm. Moreover, the safety | security with respect to a finger contact can be improved by covering the surface of an apparatus and the inside of the insertion slot of a finger | toe with a soft material like a cushion. In this figure, a state in which the cushion 38 is affixed side by side to the arc-shaped portion is shown. The above-described finger vein pattern input interface is an example in which the groove is opened in the vertical direction and the vein pattern is obtained by moving the finger in the vertical direction. The moving direction can be arbitrary.

FIG. 5 shows an example of a configuration for imaging a vein pattern from multiple directions by a plurality of light sources / imaging elements inside the finger vein pattern input interface 1. A plurality of CCD cameras 4 having a light source 2 and an optical filter 3 are arranged in a concentric manner with respect to the central axis of the finger 20 at positions facing each other. When the finger 20 is inserted into the input interface 1, those image pickup devices pick up finger images from multiple directions. With this configuration, there is an effect that vein patterns from multiple directions can be imaged without rotating the finger. When the captured image is disturbed due to interference between the light sources, the images may be sequentially captured with the light source operating time shifted.

FIG. 6 shows an example of a system in which the non-contact type finger vein pattern input interface and the non-contact type automatic door are combined, and entrance / exit including authentication can be performed without contact. The finger vein pattern input interface 20 is installed on the wall surface next to the automatic door 42, and the finger 20 is inserted therein. When the vein pattern of the authentication requester 40 matches the vein pattern registered in the system, the automatic door 42 is automatically opened. At this time, it is a great feature that everything from authentication to opening and closing of the door can be performed without contact. In this case, various configurations shown in FIG. 4 are used as the input interface.

FIG. 7 shows an example of a personal authentication device that combines a plurality of vein patterns, fingerprints, irises, voices, handwriting, faces, and the like. The authentication requester 40 is authenticated by an authentication device installed at a place where authentication is to be performed. The vein is imaged by the input interface 1 that images the vein pattern, but before and after that, other personal features are input by various input interfaces. For example, a personal identification number is input by the personal identification number input key 43, and then a finger is inserted into the vein pattern input interface 1 for authentication. In order to further improve the accuracy of authentication, a fingerprint is input through the fingerprint input interface 44, and the iris and face images of the person to be authenticated are captured by the iris imaging camera 46 and the face image capturing camera 48. Then, the handwriting input pen 50 writes characters on the handwriting input tablet 52 for handwriting confirmation, and the microphone 54 picks up the sound by further producing sound. These various individual characteristics are analyzed by the PC 6 and it is finally determined whether or not to be authenticated. The combination of personal feature information used in combination with vein pattern authentication is arbitrary. Of course, it is not necessary to use all of them, but in this example, a lot of feature information is used to improve authentication accuracy. . Alternatively, the fingerprint input device and the vein pattern imaging device may be integrated so that both inputs can be performed simultaneously by placing the finger once at a predetermined location. This saves the user trouble and enables more accurate personal authentication.

FIG. 8 shows an example of a system that uses a smart card to give a personal identification number and personal characteristic information to an authentication apparatus. The authentication requester 40 holds an IC card 60 in which personal feature information such as a personal identification number, a vein pattern of the person, a fingerprint, voice, iris, handwriting, and face is recorded. This figure shows an example using a contact type IC card, but a contact type IC card can also be used. The information recorded in the IC card 60 is automatically read by the IC card reader 62 when approaching the IC card reader 62. At this time, the personal feature information is sent to the personal authentication device, but only a number for identifying the individual is sent to the device, and the personal feature information stored in advance in the external storage device 10 is read according to the number. Thus, personal feature information may be acquired. In the example of this figure, a vein pattern is given as individual feature information. Thereafter, a finger is inserted into the finger vein pattern input interface 1 to obtain a vein pattern, and matching with the vein pattern read from the IC card 60 or the external storage device 10 is performed for authentication. In this figure, only the combination of the vein pattern and the IC card is shown, but various individual feature information shown in FIG. 7 can be used together.

Hereinafter, a software software flow executed by the hardware, particularly the CPU 9, for solving the above-described problem will be described in detail. Note that the software program for realizing this flow may be supplied to the apparatus using an external storage medium. As the storage medium, for example, a flexible disk, hard disk, optical disk, magneto-optical disk, CD-ROM, CD-R, magnetic tape, nonvolatile memory card, ROM, or the like can be used. FIG. 9 is a schematic flowchart from when a finger image is captured until authentication is performed. In order to perform personal authentication using a vein pattern, it is necessary to extract a vein pattern from the captured finger image and compare it with a registered vein pattern. Therefore, the captured finger image must be converted into a vein pattern that can be compared with the registered image through several stages. First, after performing various initializations (200), finger contour detection (202) is performed in order to extract only the finger portion in the image. At this time, it can be detected at what angle and position the finger is imaged. After performing the rotation correction (204) for making the finger tilt horizontal so that it can be correctly recognized no matter what angle or position the image is picked up, the finger image clipping (206) for extracting only the part necessary for recognition is performed. Do. The image obtained at this time includes not only the vein pattern but also shadows and uneven brightness that are unnecessary for authentication. Therefore, blood vessel tracking (208) is performed to extract only the vein pattern. Using this result, the vein pattern matching (210) between the registered image and the captured finger image is performed, and the correlation between the images is calculated as an evaluation value. Depending on the size of the evaluation value, it is determined that the person is the person or the other person, but if both are evaluation values that are difficult to determine (212), the image is further divided into small areas and matching is performed. Split matching (214) to be evaluated is performed. Then, it is finally determined whether or not the person is himself (216).

Below, each item of the flowchart shown in FIG. 9 is demonstrated in detail.
FIG. 10 shows an example of finger contour detection processing. (a) is a schematic diagram of a captured image. Here, a case where the finger is imaged in the horizontal direction from the left side of the image and the fingertip is positioned on the right side will be described. First, the contrast is adjusted to emphasize the boundary between the finger 20 and the background portion. However, it is not necessary to adjust the contrast of the entire image. For example, when the lower outline is unclear, it is only necessary to adjust the contrast only on the lower side of the image. This process enhances the edge of the finger. A state in which the edge of the finger 20 in (a) is clear is shown in (b).

Subsequently, finger contour detection is actually performed. First, the tracking point 104 is placed at the center of the image 100. The starting position for detecting the contour portion of the finger is obtained by moving one pixel upward and downward from this position. Since the finger portion is shown in the center of the image, all the pixels above and below that position have relatively bright luminance values. As you move up and down, you will eventually reach the border between your finger and background. At this time, the luminance values of several pixels inside the image (finger side) should be relatively high, but the luminance values of several pixels outside the image (background side) should be low. Accordingly, the sum of the luminance values of the inner n pixels at the current position of the tracking point 104 is subtracted from the sum of the luminance values of the outer n pixels to obtain a difference, and the tracking point 104 is moved to the upper or lower end of the image. It can be determined that the position showing a high difference value is the boundary between the background and the finger. (c) shows the tracking point 104 moving up and down to reach the boundary between the finger and the background. Next, the outline of the finger is traced from this position. In this case, it is necessary to follow the right direction and the left direction. When tracing in the left direction, the difference between the luminance values of the upper n pixels and the lower n pixels is calculated for the three points on the left, upper left, and lower left of the current position, and the position having the largest difference value is set as the next position. In this way, when the image is traced to the left end, the locus becomes an outline. For the right direction, it is necessary to follow the fingertip part, so the upper contour tracking is set to the lower several pixels below the current position and the search range is also directly below the current position, and the lower contour tracking is The search range was also set for several pixels in the upper right of the position and just above the current position. As a result, even a curve with a high curvature of the fingertip can be detected. (d) represents the result of the tracking point 104 finally following the contour of the finger. In the above, since a captured image is fixedly determined, a simple tracking method is adopted, but it goes without saying that various other contour tracking methods used in image processing can be used to improve accuracy. .

FIG. 11 shows an example of a finger rotation correction process using the finger contour information obtained by the above-described method. (a) represents an image before finger rotation correction. The inclination of the finger when the image is taken can be found by examining the shape of the outline. By correcting so that the tilt is constant in all finger images, normalization with respect to two-dimensional rotation with respect to the imaging plane of the camera is performed. The tilt of the finger is considered to be an angle formed by a line obtained by approximating the finger with a straight line and a horizontal line. Here, as an example, a rotation correction method using the upper contour line will be described. In the case of the same finger, the shape of the finger outline is the same even if the inclination of insertion is different. Here, an approximate straight line of the contour line is obtained. In general, the relative positional relationship between a curve and its approximate straight line is always constant. Therefore, the contour lines having different inclinations and the same shape can be normalized by the approximate straight line. At this time, since it is more accurate to obtain an approximate straight line from a curve that is as close to a straight line as possible, an approximate straight line is obtained from a straight upper contour line. Specifically, for example, only the contour between the position advanced 16 pixels from the fingertip and the position advanced 128 pixels in the same direction is used. The purpose of avoiding the fingertip is to avoid a portion with a high curvature.

Next, several pixels of the contour portion to be used are extracted at equal intervals, and the approximate straight line 106 of the upper contour is obtained by the least square method. Finally, the whole is rotated so that this straight line is horizontal to the image. However, the rotation center is assumed to be the intersection of this straight line and the left end of the image. The normalization of the position will be described later. (b) is the result of horizontal rotation correction of the approximate straight line 106 of the upper contour. FIG. 12 shows an example of a process for cutting out a part necessary for authentication from a rotation-corrected finger image. Usually, the horizontal and vertical positions where the finger is imaged differ each time the finger is imaged. Therefore, in order to make it easy to use the finger image for matching, it is necessary to normalize the finger position. On the other hand, in order to perform recognition, it is only necessary to capture a vein pattern in the image, and it is not necessary to hold other unnecessary portions. Therefore, an image having a smaller size than the original image is cut out from the original image. At this time, if the cutout position is always matched with the same part of the finger, the finger position is normalized. The finger outline information is used for positioning the cutout portion. First, the position in the left-right direction is determined as a position that matches, for example, the right edge of the image cut out by the fingertip 110 using the fingertip 110 obtained from the contour information. Next, the position in the vertical direction is determined such that the central axis 108 of the finger is obtained using the vertical contours of the finger and the central axis 108 passes through the center of the image to be cut out. A clipped image 114 is obtained by such positioning. This cutout image 114 cuts out the same finger portion every time the finger 20 is imaged at any position in the image 100, and the finger position is normalized.

FIG. 13 is a flowchart showing the flow of blood vessel tracking processing for a cut finger image. The finger image captured by the CCD camera does not clearly show the vein pattern necessary for authentication. The obtained image contains a lot of information unnecessary for authentication, such as background noise, irregular shadows due to uneven thickness of finger bones and flesh, and uneven brightness. Therefore, in order to use such an image for authentication, it is necessary to extract only the vein pattern from the image or to emphasize the vein pattern. The finger image is an image of light transmitted through the finger. Since the transmitted light has a wavelength that is absorbed by hemoglobin in blood, the blood vessel portion has a dark luminance value. Bright light leaks from the joints. Therefore, the background luminance varies greatly in space, and it is not possible to emphasize only blood vessels by simply enhancing the edges. However, the blood vessel portion is darker than the surroundings if limited to a spatially local range. Therefore, there is a high possibility that the locus when moving from a certain position to a darker portion is a blood vessel. Usually there are multiple blood vessels instead of one, and the number and length of them are not known in advance. Therefore, in this embodiment, a large number of trajectories of various lengths are obtained at various positions, and the blood vessel pattern is statistically emerged by superimposing them.
Based on this idea, blood vessels were tracked by the following method. First, a score table having the same size as the image is prepared to hold the history of tracking blood vessels, and all of them are initialized to 0 (300). In the blood vessel tracking loop (302), which is repeatedly executed as many times as necessary j to reveal the entire blood vessel pattern, first, the starting position of the tracking point for one blood vessel tracking is determined by a random number (304). To do. However, since the blood vessel cannot be traced correctly if the background, the fingertip, the base of the finger, or the vicinity of the contour of the finger is used as a starting point, the contour information of the finger is used so that they do not become the starting point.

It is easier to follow the blood vessel if the starting point is placed on the blood vessel. Therefore, a plurality of starting point candidates are determined, and the pixel having the darkest luminance value among them is set as the starting point. However, if the starting point is always determined under this condition, it is difficult to track the blood vessels in the bright part.Therefore, the pixel with the darkest luminance value among the multiple starting point candidates is not always set as the starting point, but with a certain probability. The pixel with the brightest luminance value is taken as the starting point. This probability is determined by a random number. Next, the direction in which the tracking point is easy to move is determined (306). This property is used to determine the movable point described later. As an example of this determination method, it is determined by random numbers that it has a property of easily moving right or left and upward or downward. Subsequently, “life” is determined (308), and at the same time, the value is given to the tracking point as an initial value. The tracking is stopped when the tracking is performed for a distance determined by the lifetime. In other words, the tracking length represents the tracking point life, and the tracking point tracking length is represented, and the tracking is terminated when the lifetime is exhausted every time one pixel is traced. This lifetime is determined using a random number.

Subsequently, the tracking point is moved. First, a point at which the tracking point can be moved next is determined (310). Many of the blood vessels run in the long axis direction of the finger, but the blood vessels are more emphasized by matching the tendency of the tracking point to move in the direction of the blood vessel running direction. Therefore, the tendency of movement of the moving point is controlled by giving a tendency to the candidate point that can be moved next. As an example of the movement tendency, in order to make it easy to move in the left and right major axis directions with a probability of 50%, the left or right three neighborhoods are made movable points, and 30% of the remaining 50% are in the minor axis direction of the finger. In order to make it easier to move to, the upper or lower three neighborhoods are set as movable points. Otherwise, the vicinity of 8 is a movable point. However, in any case, it cannot move to the outside of the trajectory or finger that has been followed. In this way, the movable point is obtained. If there is no movable point (312), the tracking at the current tracking point is terminated. Subsequently, the movable point is moved to the pixel having the darkest luminance value (314). Then, the current position is registered and updated (316) as the trajectory information so that the trajectory traced by the current tracking point does not follow again. At this time, the score is added to the position of the score table corresponding to the coordinates of the pixel (318). Here, 5 points are added as an example. Further, the lifetime which is the length of tracking of the tracking point is reduced by one (320). The presence / absence of the life of the tracking point is determined (322). If there is a life, the process returns to the determination of the movable point (310) again, and the movement, the addition of the score, and the update of the trajectory information are repeated. When the lifetime is exhausted, the trace information traced is initialized (324), and the tracking at the current tracking point ends. Such a blood vessel tracking process is repeated many times. When all of the repetitions are completed, the score of the score table corresponding to the position becomes higher as the pixel that has been traced more frequently, that is, the portion having a higher probability of being a blood vessel. Conversely, a position with a low score has a high probability that it is not a blood vessel. Therefore, the vein pattern itself appears in this score table. Therefore, by capturing this score table as an image, an image obtained by extracting only the vein pattern can be obtained.

In order to make the vein pattern thus obtained easy to use for matching, each column of the score table as a vein pattern is classified according to the score. Here, it is classified into four types as an example (328). First, it is assumed that there is no blood vessel at a pixel position with a low score. Further, it is assumed that a pixel position having a high score is likely to be a blood vessel. A pixel position having an intermediate score is an ambiguous region that may be a blood vessel but is not surely a blood vessel. Furthermore, the pixel located outside the outline of the finger is used as the background. By associating these four types with luminance values, a vein pattern image can be obtained. Finally, in order to fill the hole of the pixel that was not tracked accidentally, the blood vessel part and the ambiguous part were subjected to dilation (330). The dilation process examines all pixels in the image for 8 neighborhoods of vascular or ambiguous pixels, and if the number of non-vascular parts is 4 or less, those non-vascular parts are ambiguous. This is done by converting into parts.

The score table was converted into a vein pattern image by the above procedure, and at the same time converted into a form that could be easily used for matching. FIG. 14 is a flowchart showing an example of a method for determining whether or not the vein pattern obtained by the above method matches the registered vein pattern. SSDA (sequential similarity detection algorithm) was adopted as an algorithm for comparing two images. This is a technique that uses the property that the mismatch increases monotonically and terminates the calculation when a certain threshold is exceeded. First, various initializations (400) are performed. Subsequently, the outer peripheral n pixels of the two vein pattern images to be compared are cut off (402) to reduce the image size. Next, the central portions of the two images are matched and overlapped, and the luminance values of the overlapped pixels are compared (404). At this time, if a pixel that has a high possibility of being a blood vessel and a pixel that has a high possibility of not being a blood vessel overlap, this pixel is said to be mismatched. The number of mismatched pixels is counted for the entire image. However, for the larger image, the pixels that did not overlap the smaller image are ignored. The number of mismatches at this time is set as the initial value of the minimum number of mismatches. Subsequently, the image is shifted by one pixel or in units of several pixels within a range where the smaller image does not protrude from the larger image (up and down, left and right n pixels from the center of the image), and the number of mismatches is counted for each position. At this time, the mismatch in the current overlap is counted in units of one pixel in the entire image (410). However, if the current minimum mismatch number is exceeded even during the counting of mismatches, a smaller mismatch number cannot be obtained. Counting is stopped (412). If the current number of mismatches does not exceed the minimum number of mismatches, the past minimum number of mismatches is rewritten to the current number of mismatches (416). The minimum number of mismatches finally obtained by superimposing images over the entire range is the number of mismatches between the two images.

Finally, the mismatch rate is obtained from this result. First, the sum of the number of pixels that are highly likely to be blood vessels in two images is obtained (420). However, for the larger image, the outer peripheral n pixel portion is ignored. Using this result and the number of mismatches, the mismatch rate of the two vein patterns can be obtained (422). Here, the mismatch rate is defined as (mismatch number) / (total number of pixels that are highly likely to be blood vessels in two images). When the two vein patterns are the same, the mismatch rate is 0 or a very small value. However, the mismatch rate becomes very large when the vein patterns are different. When this numerical value is smaller than a certain threshold, it is determined that the person is the person, and when the numerical value is larger, the person is the other person.

FIG. 15 is a flowchart showing an example of a technique for performing matching using another technique for a mismatch rate in the above-described matching technique that cannot be determined by the person himself / herself or others. In many cases, the method for obtaining the mismatch of the entire image can stably distinguish the person from the other person. However, there are some ambiguous data near the threshold. Therefore, if the data in the vicinity of the threshold can be authenticated by another matching method, the recognition rate as a whole can be further improved.

The method of division matching is as follows. For one of the two images, the image is divided into two or more m small regions (502). Each small area is matched with the other image (506). Here, the number of pixels whose luminance values match is simply counted. The position with the largest number of matched pixels is taken as the matched position. At this time, the moving range is limited so as not to accidentally match a portion that cannot be clearly matched in position. Then, information on the position where the small area is most matched is obtained (508). That is, how much each small area matches from the initial position is held in the form of a two-dimensional vector. When matching is completed for all the small regions, the information on the m matching positions is plotted on a plane (512), and the density of each point is evaluated. When these points are dense, there is a large correlation between the two images, and conversely, when they are sparse, there is almost no correlation.

In order to evaluate the density, a weighted value p is added to the plotted points, and a value smaller by Δp is added (514) every time one pixel moves away from the point. If the plotted points are dense, the weight values are repeatedly added, so that a larger value appears on the plane. When the match positions of the m small regions are all the same, the maximum value of the added weight value is m * p. Conversely, when the match positions are sparse, the maximum evaluation value is p. Since coincidence match positions may overlap, the maximum evaluation value is larger than p even in an uncorrelated image. In this way, the plane is scored and the largest score is searched (516). This value is the correlation between the two vein patterns. When this value is large, the probability of being a person is high, and when the value is small, the probability of being a stranger is high. However, there may be a large correlation due to coincidence of the match position. At this time, the probability of mistaking another person as the person is increased. Therefore, when there are few points plotted in a circle of a certain radius from the position where the maximum evaluation value is generated, it is determined that the evaluation value has increased accidentally (518), and it is determined that the person is another person (520). .

FIG. 16 is a comparison result of performance evaluation when the present invention is implemented and performance evaluation by another method. The different method is different from the present invention in the method from the acquisition of the finger image to the end of the authentication. The method emphasizes the vein pattern by uniformly filtering the acquired image, performs a two-dimensional convolution operation between the registered template and the image, and evaluates the sharpness of the peak in the short axis direction of the finger It is. In the performance evaluation, four little fingers of each of 678 subjects were imaged, and one of them was regarded as a registered template and collated with the finger images of the person and others. The collation method collates the vein patterns of the person and the other person by collating all combinations of registered templates of all subjects. Also, a finger image that is not a registered template is compared with a registered template of the same person.

As a result of the verification, a false rejection rate (FRR: False Reject Rate) in which the person is mistaken for the other person and a false acceptance rate (FAR: false acceptance rate) in which the other person is mistaken for the person are required. Here, the relationship between FRR and FAR is used for performance evaluation. (a) is a comparison result of performance evaluation when only one image is used in the collation of the registered template and another image of the same person. Both FRR and FAR are better as they are smaller, but it can be seen that FRR and FAR in the present invention are already about one-tenth smaller than the result of another method when the whole matching is performed. Furthermore, even better results are obtained when split matching is performed. (b) is a comparison result of performance evaluation when an image that produces the best result among the three images of different images is selected in the verification of the registered template and another image of the same person. As a result of another method, there is still data that cannot correctly distinguish the principal from the other, but the method according to the present invention can completely distinguish the principal from the other. From these results, it can be concluded that the effect of the present invention is great.

The present invention is useful as a technique for performing personal authentication using a vein pattern obtained by imaging light transmitted through a human finger.

DESCRIPTION OF SYMBOLS 1 ... Finger vein pattern input interface, 2 ... Light source, 3 ... Optical filter, 4 ... CCD camera, 5 ... Image capture board, 6 ... Computer, 7 ... Interface, 8 ... Memory, 9 ... CPU, 10 ... External storage device, DESCRIPTION OF SYMBOLS 20 ... Finger, 22 ... Finger insertion port, 30 ... Finger vein pattern input interface with chamfered corners, 32 ... Finger vein pattern input interface with finger insertion port open sideways, 33 ... Wall, 34 ... Finger insertion port on wall surface Finger vein pattern input interface 36: finger vein pattern input interface having a circular arc at the back of the finger insertion slot, 38 ... cushion, 40 ... authentication requester, 42 ... automatic door, 43 ... PIN code input key, 44 ... Fingerprint input interface, 46 ... Iris imaging camera, 48 ... Face imaging camera, 50 ... Handwriting input pen, 52 ... Handwriting input Tablet, 54 ... Microphone, 60 ... IC card, 62 ... IC card reader, 100 ... Finger image, 104 ... Tracking point, 106 ... Approximate straight line of the upper contour of the finger, 108 ... Finger central axis, 110 ... Fingertip, 112 ... The center of the finger image to be cut out, 114...

Claims (11)

A housing having an area for placing a finger;
A light source that emits light to the finger;
Imaging means capable of imaging feature information of the finger by irradiating light from the light source;
The feature information includes a finger blood vessel and a fingerprint,
The feature image photographing apparatus characterized in that the feature information can be imaged with one finger placement for the finger placement region. In the characteristic image photographing device according to claim 1,
It further includes an arithmetic unit,
The said calculating part extracts the 1st biological information based on the imaged finger blood vessel, and the 2nd biological information based on the said fingerprint, The characteristic image imaging device characterized by the above-mentioned. In the characteristic image photographing device according to claim 2,
The said calculating part can perform said personal authentication by comparing said extracted 1st biometric information, said 2nd biometric information, and the information memorize | stored previously, The characteristic image imaging device characterized by the above-mentioned. In the characteristic image photographing device according to claim 1,
A characteristic image photographing apparatus, wherein the light source and the imaging means are provided inside the housing. In the characteristic image photographing device according to claim 1,
The image pickup device includes an image pickup device that can pick up an image including blood vessels and an input device that can pick up an image including a fingerprint. A finger placement section for guiding the finger to a predetermined position;
A light source that illuminates the finger;
Imaging means capable of imaging light that has been irradiated to the finger and passed through the finger and reflected light, and from the image captured by the imaging means by one finger placement on the finger placement unit, A personal authentication device comprising a calculation unit capable of calculating a biometric feature based on a blood vessel and a biometric feature based on a fingerprint. The personal authentication device according to claim 6,
The said calculating part performs personal authentication using the biometric feature based on the blood vessel of the said finger | toe, and the biometric feature based on the said fingerprint, The personal authentication apparatus characterized by the above-mentioned. A light source;
Imaging means;
A finger placement unit,
An arithmetic operation capable of acquiring a biometric feature based on a blood vessel and a biometric feature based on a fingerprint captured by the imaging unit from an image captured by the imaging unit while presenting a finger to the finger placement unit. A personal authentication device characterized by comprising a unit. A common finger presentation unit on which fingers are presented in order to acquire feature information including fingerprints and blood vessels of the fingers,
The body housing,
A light source provided inside the main body housing and an imaging means;
The personal authentication apparatus characterized in that the feature information can be acquired by one finger presentation to the finger presentation unit. The personal authentication device according to any one of claims 6, 8, and 9,
The personal authentication device, wherein the imaging means includes an imaging device capable of capturing an image including a blood vessel and an input device capable of capturing an image including a fingerprint. A light source and an imaging unit;
A personal authentication device that extracts a blood vessel pattern based on an image captured by the imaging unit and performs personal authentication based on the blood vessel pattern.

Images