JP2004171577A - Personal identification device and method for the same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide personal identification of high precision while using a blurred finger vein pattern image which has misregistration regarding the personal identification under circumstances non-contact is required. <P>SOLUTION: The device is provided with: a means for obtaining the finger vein pattern image with no contact; a means for performing rotation compensation by using the contour of the finger as a method of retrieving the vein pattern included in the obtained image; a means for normalizing the position of the finger image with a fingertip as a reference; a means for obtaining the overall vein pattern statistically by repeating tracing of parts where a luminance value is low just for an arbitrary length from an arbitrary position on the image; a matching means for comparing only parts where distinctive characteristics appear in the blood vessel patterns; and a means for dividing the image into small areas, performing matching for every small area independently and evaluating the amount of misregistration when they are matched. <P>COPYRIGHT: (C)2004,JPO

Description

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

  Personal authentication techniques include methods based on fingerprints, irises, voice, veins in the back of the hand, and the like. Personal identification devices based on fingerprints have already been commercialized and sold by a plurality of companies. These products read fingerprints by contacting them with a fingerprint sensor, recognize the end points and branch points of the fingerprints as feature points, and perform personal authentication by comparing them with feature points of registered fingerprints. I have.

Japanese Patent Application Laid-Open No. 10-295674 discloses a personal authentication device based on the vein of the back of the hand. According to this, the back of the hand is pointed at the imaging camera, and the reflected light of the light emitted therefrom is used to capture the blood vessel pattern to perform authentication. At this time, by holding the fixed rod-shaped guide, the position of the hand to be imaged for each authentication is not deviated.
Also, Japanese Patent Application Laid-Open No. 7-21373 discloses a personal authentication device using a finger vein. In particular, in order to reduce a loss of light amount during imaging, an optical fiber is brought into close contact with a finger to capture a finger image. It is disclosed.

JP-A-10-295674

JP 7-21373 A

  In the related art, a method that has a large psychological resistance, such as collecting a fingerprint at the time of authentication or putting light into the eyes, is adopted. Further, in the above-described conventional personal authentication device, it is necessary to bring a human body into contact with the authentication device, and it may not be appropriate to use the device in a medical site where hygiene is important. They can be counterfeited because they take advantage of features that are exposed outside the body.

An object of the present invention is to construct a security system in an environment where non-contact is required, such as a medical site. Therefore, the present invention provides an apparatus and a method for capturing 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, attention has been paid to a new problem that when a finger is photographed in a non-contact manner, rotation or uneven brightness is 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 highly accurate 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 comprises: a storage device for storing a vein pattern of a registered finger image; an interface including a light source and a camera for photographing finger transmitted light; Means for extracting a vein pattern included in the image, collating the extracted vein pattern with the vein pattern of the registered finger image and performing personal authentication, wherein the interface has a groove for inserting a finger in a non-contact manner. The light source and the camera are disposed so as to face each other with the groove interposed therebetween.

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

ADVANTAGE OF THE INVENTION According to this invention, the authentication using the inside of the living body which does not require contact with an apparatus, has low psychological resistance at the time of authentication, and is hard to forge. In addition, highly accurate personal authentication can be realized while using an unclear image with a positional shift peculiar to non-contact.

  Hereinafter, one 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. The vein pattern input interface 1 corresponding to a part into which a finger is inserted includes a light source 2, an optical filter 3, and a CCD camera 4. A vein pattern is obtained by inserting a finger between the light source 2 and the optical filter 3. Light transmitted through the finger is imaged by the CCD camera 4 through the optical filter 3. An image signal captured by the CCD camera 4 is captured by 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. The 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, the CPU 9 determines whether or not the captured image matches the registered image. Note that this program may be supplied to the device using an external storage medium. As the storage medium, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a nonvolatile memory card, a ROM, and the like can be used.

FIG. 2 shows an example of the structure of the finger vein pattern input interface 1 for acquiring a finger vein pattern image by a non-contact method. The interface has a finger insertion port 22 which is a groove-shaped gap, and a vein pattern is obtained by passing the finger 20 through that part. At this time, there is no need to bring a finger into contact with the device itself. Also, by passing a plurality of fingers through the groove, vein patterns of the plurality of fingers can be imaged. Furthermore, it is also possible to continuously draw the plurality of fingers 20 through the finger insertion opening 22 by swinging the fingers down in a circular orbit around the shoulders, elbows, wrists, and the like, thereby capturing a finger image. When such a non-contact imaging method is used, the captured finger image rotates with respect to the imaging plane. Therefore, it is necessary to inevitably perform rotation correction on the captured finger image.
In addition, the direction of the groove may be any direction with respect to the ground, especially when a groove is provided in a direction perpendicular to the ground, the user is a natural motion of waving his hand according to gravity, Since the pattern image can be obtained, there is an effect that the usability is good.
In addition, such an interface structure in which a finger is passed through the groove has the effect of increasing the accuracy of authentication because rotation around the central axis of the finger is less likely to occur.

  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 port 22. The finger 20 is inserted into the insertion slot 22 of the finger, and a vein of the finger is obtained by one or a plurality of light sources / image sensors inside the finger. At this time, a vein pattern can be obtained from multiple directions by rotating the finger in a concentric direction with respect to the center axis of the finger.

  FIG. 4 is an example of the above-mentioned non-contact type finger vein pattern input interface in consideration of security at the time of authentication and at the time of passing near the interface installation location. (a) is a shape in which a corner of the finger vein pattern input interface is chamfered so that a finger, a hand, an arm, or the like comes into contact with a safe shape. Also, depending on the installation location, there is a case where it is dangerous that the device protrudes. In this case, the width of the protrusion of the interface is reduced by providing a groove horizontally as shown in (b), or a finger insertion port is provided in the wall surface itself to be installed as shown in (c). However, in (c), the arm has a groove having a width sufficient to allow the arm to swing down. In addition, in order to cope with imaging of the finger by swinging the arm down, the depth of the insertion slot of the finger is formed in an arc shape in accordance with the trajectory of the finger as shown in (d). With such a structure, the finger is hardly brought into contact with the device with respect to the circular movement of the arm. In addition, by covering the surface of the apparatus and the inside of the finger insertion port with a soft material such as a cushion, safety against finger contact can be enhanced. This figure shows a state in which the cushions 38 are arranged and attached to the arc-shaped portions. Note that the above-described finger vein pattern input interface is an example in which a groove is opened in the vertical direction and a vein pattern is obtained by moving the finger in the vertical direction. The direction of movement can be arbitrary.

  FIG. 5 shows an example of a configuration for capturing a vein pattern from multiple directions by a plurality of light sources / image sensors in the finger vein pattern input interface 1. A plurality of CCD cameras 4 each having the light source 2 and the optical filter 3 are arranged concentrically with respect to the center axis of the finger 20 at positions facing each other. When the finger 20 is inserted into the input interface 1, those image sensors capture finger images from multiple directions. This configuration has the 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 while the operation time of the light sources is shifted.

  FIG. 6 shows an example of a system in which the above-mentioned non-contact type finger vein pattern input interface and a non-contact type automatic door are combined to enable non-contact entry and exit including authentication. The finger vein pattern input interface 20 is installed on a wall surface next to the automatic door 42, and the finger 20 is inserted therein. When the vein pattern of the authentication claimant 40 matches the vein pattern registered in the system, the automatic door 42 is automatically opened. At this time, a major feature is 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 is an example of a personal authentication device in which a plurality of vein patterns and fingerprints, irises, voices, handwritings, faces, and the like are combined. The authentication requester 40 is authenticated by an authentication device installed at a place to be authenticated. A vein is imaged by the input interface 1 for imaging a vein pattern, and before and after that, other personal characteristics are input by various input interfaces. For example, the personal identification number is input by the personal identification number input key 43, and then the finger is inserted into the vein pattern input interface 1 to be authenticated. In order to further improve the accuracy of authentication, a fingerprint is input by the fingerprint input interface 44, and an iris and a face image of the person to be authenticated are captured by the iris image capturing camera 46 and the face image capturing camera 48. Then, a character is written on the handwriting input tablet 52 with the handwriting input pen 50 for handwriting confirmation, and a voice is emitted, so that the microphone 54 picks up the voice. The characteristics of these various individuals are analyzed by the PC 6, and it is finally determined whether or not to be authenticated. The method of combining personal characteristic information used in combination with vein pattern authentication is arbitrary, and it is not necessary to use all of them, but in this example, a lot of characteristic information is used to further increase the 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 a finger once in a predetermined place. This saves time and effort for the user and enables more accurate personal authentication.

  FIG. 8 shows an example of a system for providing a personal identification number and personal characteristic information to an authentication device using an IC card. The authentication claimant 40 holds an IC card 60 in which personal characteristic information such as a personal identification number, a person's vein pattern, fingerprint, voice, iris, handwriting, and face are recorded. Although this drawing shows an example using a contact type IC card, a contact type IC card can also be used. The information recorded on 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 characteristic information is sent to the personal authentication device, but only the number for identifying the individual is sent to the device, and the personal characteristic information corresponding to the number is read out of the personal characteristic information stored in the external storage device 10 in advance. By doing so, personal characteristic information may be obtained. In the example of this figure, a vein pattern is given as personal characteristic information. Thereafter, a finger is inserted into the finger vein pattern input interface 1 to acquire a vein pattern, and the authentication is performed by matching the vein pattern with the vein pattern read from the IC card 60 or the external storage device 10. In this figure, only the combination of the vein pattern and the IC card is shown, but it is also possible to use various personal characteristic information shown in FIG. 7 together.

In the following, a detailed description will be given of a software flow executed by the above-mentioned hardware, in particular, the CPU 9 for solving the above-mentioned problem. The software program for realizing this flow may be supplied to the device using an external storage medium.
As the storage medium, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a nonvolatile memory card, a ROM, and the like can be used. FIG. 9 is a schematic flowchart from when a finger image is captured to when 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 steps. First, after performing various initializations (200), the outline of the finger is detected (202) to extract only the finger portion in the image. At this point, it is possible to detect at what angle and position the finger is being imaged. After performing a rotation correction (204) to make the inclination of the finger horizontal so that the image can be correctly recognized even if the image is captured at any angle or position, cut out the finger image (206) to extract only a part necessary for recognition. Do. The image obtained at this time has not only vein patterns but also shadows and uneven brightness which are unnecessary for authentication. Then, in order to extract only the vein pattern, blood vessel tracking (208) is performed. Using this result, matching (210) of the vein pattern between the registered image and the captured finger image is performed, and the correlation between the images is calculated as an evaluation value. It is determined that the subject is the person or the other person according to the magnitude of the evaluation value. However, if the evaluation value is indeterminate for both (212), the image is further divided into small areas and matching is performed. A division matching (214) to be evaluated is performed. Then, it is finally determined whether or not the user is himself (216).

Hereinafter, each item of the flowchart shown in FIG. 9 will be described in detail.
FIG. 10 shows an example of a finger outline detection process. (a) is a schematic diagram of a captured image. Here, a case where the finger is imaged horizontally from the left side of the image and the fingertip is located on the right side will be described. First, the contrast is adjusted to emphasize the boundary between the finger 20 and the background. However, it is not necessary to adjust the contrast of the whole image. For example, when the lower contour is unclear, the contrast of only the lower part of the image may be adjusted. This process emphasizes the edge of the finger. (b) shows a state in which the edge of the finger 20 in (a) is sharp.

Subsequently, the outline of the finger is actually detected. First, the tracking point 104 is placed at the center of the image 100. From this position, the starting position at which the outline of the finger is detected is obtained by moving one pixel upward and downward separately. Since the finger portion is shown in the center of the image, all pixels above and below that position have relatively bright luminance values. As it moves up and down, it eventually reaches the boundary between the finger and the background. At this time, the luminance value of several pixels inside the image (finger side) should be relatively high, but the luminance value of several pixels outside the image (background side) should be low. Therefore, the difference is obtained by subtracting the sum of the luminance values of the outer n pixels from the sum of the luminance values of the inner n pixels at the current position of the tracking point 104, and the difference is obtained by moving the tracking point 104 to the upper or lower end of the image. It can be determined that the position showing the high difference value is the boundary between the background and the finger.
(c) indicates that the tracking point 104 has moved up and down and has reached the boundary between the finger and the background. Next, the contour of the finger is traced from this position. In this case, it is necessary to follow two directions, rightward and leftward. When following 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, left, upper left, and lower left of the current position, and the position having the largest difference value is set as the next position. When the image is traced to the left end, the locus becomes an outline. In the right direction, since it is necessary to follow the fingertip portion, the upper contour tracking is a few pixels below and to the right of the current position, and the search range is directly below the current position. A few pixels above and to the upper right of the position and directly above the current position were also used as the search range. As a result, it is possible to detect even a curve having a high curvature at the fingertip.
(d) finally shows the result of the tracking point 104 following the outline of the finger. In the above description, a simple tracking method is adopted because the captured image is fixed, but it goes without saying that the accuracy can be improved by using various other contour tracking methods used in image processing. .

FIG. 11 shows an example of a finger rotation correction process using the finger contour information obtained by the above-described method. (a) shows an image before finger rotation correction. The inclination of the finger when the image is captured can be determined by examining the shape of the contour. Correction is performed so that the inclination is constant in all the finger images, thereby normalizing the two-dimensional rotation with respect to the photographing plane of the camera.
The inclination of the finger can be considered as an angle between a line when the finger is approximated by a straight line and a horizontal line. Here, as an example, a rotation correction method using an upper contour line will be described. In the case of the same finger, the contours of the fingers have the same shape even when the insertion inclination is different. Here, an approximate straight line of the outline is obtained. Generally, the relative positional relationship between a curve and its approximate straight line is always constant. Therefore, contour lines having different slopes and the same shape can be normalized by an approximate straight line. At this time, the accuracy is more improved when an approximate straight line is obtained from a curve that is as close to a straight line as possible. Therefore, an approximate straight line is obtained from a straight upper contour line. Specifically, for example, only a contour located between a position advanced by 128 pixels in the same direction from a position advanced by 16 pixels from the fingertip in the root direction of the finger is used. The reason for avoiding fingertips is to avoid high curvature portions.

  Next, several pixels of the contour portion to be used are extracted at equal intervals, and an approximate straight line 106 of the upper contour is obtained by the least squares method. Finally, the whole is rotated so that this straight line is horizontal to the image. However, the center of rotation is assumed to be the intersection of this straight line and the left end of the image. The position normalization 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 of cutting out a portion required for authentication from a finger image whose rotation has been corrected. Normally, the position of the finger in the horizontal and vertical directions at which the image is taken differs every time the finger is taken. Therefore, it is necessary to normalize the position of the finger in order to easily use the finger image for matching. On the other hand, in order to perform recognition, it is sufficient that a vein pattern is captured in an image, and it is not necessary to hold other unnecessary parts. Therefore, an image smaller in size than the original image is cut out from the original image. At this time, if the cutout position is always set to the same part of the finger, the finger position is normalized. The outline information of the finger is used for positioning the cutout portion. First, using the fingertip 110 obtained from the outline information, the position in the left-right direction is determined to be, for example, a position corresponding to the right end of an image cut out by the fingertip 110. Next, the position in the vertical direction is determined such that the center axis 108 of the finger is obtained by using the upper and lower contours of the finger, and the center axis 108 passes through the center of the image to be cut out. With such positioning, a cut-out image 114 is obtained. Regardless of the position where the finger 20 is picked up in the image 100, the same finger portion is cut out from the cutout image 114 every time, and the finger position is normalized.

FIG. 13 is a flowchart showing the flow of the blood vessel tracking process for the clipped finger image. The finger image captured by the CCD camera does not clearly show a vein pattern required 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 enhance the vein pattern. The finger image is obtained by capturing light transmitted through the finger. Since the transmitted light has a wavelength that is absorbed by blood hemoglobin, the blood vessel portion has a dark luminance value. Bright light is leaking from the joints. Therefore, the luminance of the background changes greatly spatially, and it is not possible to emphasize only the blood vessel by simply enhancing the edge. However, the blood vessel portion is darker than the surroundings only in a spatially local range. Therefore, there is a high possibility that the trajectory when moving from a certain position to a darker part is a blood vessel. Usually, not one blood vessel exists but a plurality of blood vessels, and the number and length of the blood vessels cannot be known in advance. Therefore, in the present embodiment, a large number of trajectories having various lengths are obtained at various positions, and by superimposing the trajectories, a blood vessel pattern is statistically made to emerge.
Based on this idea, blood vessels were tracked by the following method. First, a score table of the same size as an image is prepared to hold a history of tracking blood vessels, and all of them are initialized to 0 (300). In a blood vessel tracking loop (302) that is repeatedly executed by the number of times j necessary to make the entire blood vessel pattern emerge, first, a starting position of a tracking point for performing one blood vessel tracking is determined by a random number (304). I do. However, if the starting point is the background, the fingertip, the root of the finger, or the vicinity of the contour of the finger, the blood vessel cannot be correctly traced. Therefore, the information is used as the starting point by using the contour information of the finger.

  In addition, it is easier to follow the blood vessel when the starting point is arranged on the blood vessel. Therefore, a plurality of starting point candidates are determined, and a pixel having the darkest luminance value among them is set as a starting point. However, if the starting point is always determined under these conditions, it will be difficult to track the blood vessels present in the bright part, so the pixel with the darkest luminance value among the multiple starting point candidates will not always be the starting point, but with a certain probability. The pixel having the brightest luminance value is 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 for determining a movable point described later. As an example of this determination method, a random number is determined to have a property of easily moving right or left and up or down. 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 terminated when the tracking is performed for the distance determined by the life. In other words, the tracking length is defined as the life of the tracking point, and the tracking length of the tracking point is represented. The life is reduced each time one pixel is traced, and the tracking is terminated when the life has expired. This lifetime is determined using random numbers.

Subsequently, the tracking point is moved. First, a point to which the tracking point can move next is determined (310). Although many blood vessels run in the long axis direction of the finger, the blood vessels are emphasized more by adjusting the tendency of the tracking point to move in accordance with the tendency of the blood vessels in the running direction. Therefore, the tendency of the movement of the moving point is controlled by giving a tendency to the candidate for the next moveable point.
As an example of the tendency of movement, 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 set as movable points, and for the remaining 50%, 30% of the finger short axis direction. In order to make it easy to move, the upper and lower three neighborhoods are set as movable points.
In other cases, the vicinity of 8 is a movable point. However, in either case, the user 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 ends.
Subsequently, among the movable points, the pixel is moved to the pixel having the darkest luminance value (314). Then, the current position is registered and updated as trajectory information (316) so that the current tracking point does not follow the trajectory traced in the past again. At this time, the score is added to the position of the score table corresponding to the coordinates of the pixel (318). Here, five points are added as an example. Further, the life, which is the tracking length of the tracking point, is reduced by one (320). It is determined whether or not the tracking point has a lifetime (322), and if it has, 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 life has expired, the information of the traced path is initialized (324), and the tracking at the current tracking point ends. Such a blood vessel tracking process is repeatedly executed. When all the repetitions are completed, the score of the score table corresponding to the position is higher as the pixel has been traversed a larger number, that is, the portion having a higher probability of being a blood vessel. Conversely, a position with a low score has a high probability of not being 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 obtained in this manner 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 examples (328). First, it is assumed that no blood vessel exists at a pixel position having a low score. It is also assumed that a pixel position with a high score is highly likely to be a blood vessel. Then, it is assumed that the pixel position having an intermediate score is an ambiguous region that may be a blood vessel but cannot be definitely called a blood vessel. Further, a pixel located outside the outline of the finger is set as a background. By associating these four types with luminance values, an image of a vein pattern can be obtained.
Finally, dilation (330) was performed on the blood vessel portion and the ambiguous portion to fill in the holes of the pixels that were not accidentally tracked. The dilation process examines the 8 neighborhoods of the pixels in the blood vessel portion or the ambiguous portion for all the pixels present in the image, and if the number of pixels in the non-blood vessel portion is four or less, the non-vascular portion is unclear. This is done by converting to parts.

  According to the above procedure, the score table was converted into an image of a vein pattern, and at the same time, was converted into a form easily usable 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-described method matches a registered vein pattern. As an algorithm for comparing two images, SSDA (sequential similarity detection algorithm) was adopted. This is a method of stopping the calculation when a certain threshold is exceeded, using the property that the mismatch monotonically increases. First, various initializations (400) are performed. Subsequently, n pixels on the outer periphery of one 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 overlapping pixels are compared (404). At this time, if a pixel having a high possibility of being a blood vessel and a pixel having 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, pixels of the larger image that did not overlap with 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 several pixels in a range in which the image having the reduced size does not protrude from the large image (n pixels at the top, bottom, left, and right with respect to the center of the image), and the number of mismatches at each position is counted. At this time, mismatches in the current overlap are counted in units of one pixel in the entire image (410). However, if the current minimum mismatch number is exceeded during the counting of mismatches, a smaller mismatch number cannot be obtained. Is stopped (412). If the current mismatch number does not exceed the minimum mismatch number, the past minimum mismatch number is rewritten to the current mismatch number (416). The images are superimposed over the entire range, and the minimum number of mismatches finally obtained is the number of mismatches between these two images.

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

  FIG. 15 is a flowchart showing an example of a method of performing matching by another method for a case where it is difficult to determine whether the person is an individual or another person with a mismatch rate in the above-described matching method. In many cases, the method of obtaining the mismatch of the entire image can stably distinguish the subject from others. However, some data is ambiguous near the threshold. Therefore, if data near the threshold can be authenticated by another matching method, the recognition rate can be further improved as a whole.

  The method of division matching is as follows. For one of the two images, the image is divided into two or more m small areas (502). Each of the small areas is matched with the other image (506). Here, the number of pixels whose brightness values match is simply counted. The position with the largest number of matched pixels is set 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 most closely matches is obtained (508). That is, in each of the small areas, how much the position is shifted from the initial position is stored in the form of a two-dimensional vector. When the matching has been completed for all the small areas, the information of 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. Conversely, when these points are sparse, there is almost no correlation.

  For the evaluation of the density, a weight p is added to the plotted point, and a value smaller by Δp is added every time a pixel is separated from the point (514). If the plotted points are dense, their weight values are repeatedly added, so a larger value appears on the plane. If the match positions of the m small areas are all the same, the maximum value of the added weight values is m * p. Conversely, if the match positions are sparse, the maximum evaluation value is p. Since the match positions may be accidentally overlapped, the maximum value of the evaluation value is larger than p even in an uncorrelated image. The plane is scored in this way, and the largest score is searched (516). This value is the correlation between the two vein patterns. If this value is large, the probability of being the person is high, and if it is small, the probability of being another person is high. However, there may be a case where there is a large correlation due to the coincidence of the match positions. In this case, the probability that the other person is mistaken for the person is increased. Therefore, if there are few points plotted within a circle of a certain radius from the position where the maximum value of the evaluation value occurs, it is determined that the evaluation value has accidentally increased (518), and it is determined that the user 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 method from the acquisition of the finger image to the end of the authentication is different from that of the present invention. The method is to apply a uniform filter to the acquired image to emphasize the vein pattern, perform a two-dimensional convolution operation between the registered template and the image, and evaluate the sharpness of the peak in the short axis direction of the finger. It is. In the performance evaluation, four fingers of each of 678 subjects were imaged, and one of them was regarded as a registered template, and collation was performed with finger images of the subject and others. The matching method matches the vein patterns of the subject and another person by matching all combinations of the registered templates of all subjects. In addition, a finger image that is not a registered template is compared with a registered template of the same person.

  As a result of the collation, a false reject rate (FRR: False Reject Rate) in which the person is mistaken for another person and a false accept rate (FAR: False Accept Rate) in which the false person is mistaken for the person are obtained. Here, the relationship between FRR and FAR is used as performance evaluation. (a) is a comparison result of the performance evaluation when only one different image is used in matching the registered template with another image of the same person. It is clear that the smaller the FRR and the FAR, the better, but it is understood that the FRR and the FAR in the present invention are already about 10 times smaller than the result obtained by another method when the overall matching is performed. Further, even better results are obtained when the division matching is performed. (b) is a comparison result of the performance evaluation when the image that gives the best result among the three different images is selected in the comparison between the registered template and another image of the same person. As a result of another method, there is still data that cannot correctly distinguish a person from another, but the method according to the present invention can completely correctly distinguish a person from another. From these results, it is concluded that the effect of the present invention is great.

1 is an example of a system configuration for realizing the present invention. It is an example of an input interface configuration for acquiring a finger vein. It is an example of an input interface configuration for acquiring a finger vein. It is an example of a finger vein pattern input interface configuration in consideration of safety. 5 is an example of an arrangement of a light source and a CCD camera in an input interface for imaging a vein pattern from multiple directions. This is an example of a system configuration in which entry and exit including authentication can be performed without contact. This is an example of a system configuration for performing authentication by combining personal characteristic information such as a password, a fingerprint, an iris, a voice, a handwriting, and a face, in addition to a vein pattern. 1 is an example of a system configuration for acquiring a template image of a vein pattern using an IC card. 5 is a flowchart showing an outline of processing by software for realizing the present invention. FIG. 7 is an image diagram showing a contour tracking method of a finger image. FIG. 9 is an image diagram showing a method of rotationally correcting the inclination of a finger image. FIG. 7 is an image diagram showing a method of normalizing a cut-out portion of a finger image. It is an example of a flowchart for extracting a vein pattern from a finger image. It is an example of a flowchart for calculating the mismatch rate of two vein patterns. 9 is an example of a flowchart for calculating a correlation between vein patterns using partial images of two vein patterns. 5 is a performance comparison between the method according to the present invention and another method in the relationship between the false acceptance rate (FAR) and the false rejection rate (FRR) of the authentication system.

Explanation of reference numerals

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, 20 finger, 22 finger insertion port, 30
... A finger vein pattern input interface with chamfered corners, 32. A finger vein pattern input interface with a finger insertion opening sideways, 33... Wall, 34. Finger vein pattern input interface with an arc-shaped finger insertion port inside, 38 ... Cushion, 40 ... Authenticator, 42 ... Automatic door, 43 ... Personal number input key, 44 ... Fingerprint input interface, 46 ... Iris imaging camera, 48 ... Face image capturing 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 ... Finger upper contour Approximate straight line, 108: center axis of finger, 110: fingertip, 112: center of finger image to be cut out, 114: finger image cut out.

Claims (18)

A storage device for storing the vein pattern of the registered finger image,
An interface with a light source and a camera that captures light transmitted through the finger,
Means for extracting a vein pattern included in the photographed finger transmitted light image, performing personal authentication by comparing the extracted vein pattern with the vein pattern of the registered finger image,
The personal identification device, wherein the interface has a groove into which a finger is inserted, and the light source and the camera are arranged to face each other with the groove interposed therebetween.   2. The personal authentication according to claim 1, wherein the groove of the interface is formed so that a finger of the person to be authenticated can be inserted while drawing an arc-shaped trajectory centered on a shoulder, elbow, or wrist. apparatus.   The groove of the interface according to claim 2, wherein the groove of the interface is formed in an arc shape corresponding to a trajectory for inserting a hand.   The personal identification device according to claim 2, wherein the groove of the interface is provided in a direction perpendicular to the ground.   5. The device according to claim 1, wherein the means for performing personal authentication corrects rotation of the captured finger image on an imaging plane that occurs when the finger is inserted into the interface, and performs the rotation correction. A personal authentication device for extracting a vein pattern included in a finger image and performing personal authentication. A storage device for storing the vein pattern of the registered finger image,
An interface including a light source and a camera that captures light transmitted through the finger, an interface into which a finger is inserted, and for a captured finger image, a rotation on an imaging plane that occurs when the finger is inserted into the interface is corrected, and the rotation is corrected. Means for extracting a vein pattern included in the finger image and performing personal authentication by comparing the extracted vein pattern with the vein pattern of the registered finger image.   7. The personal authentication device according to claim 5, wherein the means for performing personal authentication cuts a finger image of the rotation-corrected finger image based on a fingertip and a center axis of the finger, and extracts a vein pattern included in the cut finger image. A personal authentication device for performing personal authentication.   7. The method according to claim 5, wherein the means for performing personal authentication repeats at least once from an arbitrary position of the finger image to a portion having the lowest luminance value around the current position by an arbitrary length. A personal authentication apparatus for extracting a vein pattern.   9. The personal authentication apparatus according to claim 8, wherein a position at which the portion having a low luminance value starts to be traced is determined by a random number.   9. The personal authentication device according to claim 8, wherein the length of the portion following the low luminance value is determined by a random number. 9. The method according to claim 8, wherein the means for performing personal authentication classifies the finger image into a blood vessel part, a non-blood vessel part, or a part that cannot be said when comparing the vein pattern, and superimposes the finger image on a registered finger image. A personal authentication apparatus that evaluates an overlap between the blood vessel part and the non-blood vessel part, and does not evaluate an overlap between a part that cannot be said to be neither.   9. The method according to claim 8, wherein the means for performing personal authentication divides the finger image and the registered finger image into small regions when performing matching on the extracted vein pattern, and performs matching independently for each small region. Personal authentication device.   7. The personal authentication device according to claim 1, wherein the vein pattern of the registered finger image is registered in an IC card held by a person to be authenticated and is input to the storage device at the time of authentication.   7. The identification device according to claim 1, wherein identification information for identifying the user is read from an IC card held by the user, and a vein pattern of a registered finger image corresponding to the identification information is read from the storage device and used for comparison. A personal authentication device, characterized in that:   7. The interface according to claim 1, wherein the interface has means for rotating the camera or a plurality of cameras when photographing a finger image, and performs personal authentication using the finger images photographed from multiple directions. Characteristic personal identification device.   7. The personal authentication device according to claim 1, wherein personal authentication is performed using a plurality of finger vein patterns. A storage device for storing the vein pattern of the registered finger image,
An interface with a light source and a camera that captures the inserted finger transmitted light without contact,
Means for extracting a vein pattern included in the photographed finger transmitted light image, performing personal authentication by comparing the extracted vein pattern with the vein pattern of the registered finger image,
With a door,
When the means for performing personal authentication determines that the extracted vein pattern matches the vein pattern of the registered finger image, the security system opens the door. Take a transmitted light image of your finger,
Correct the rotation on the imaging plane that occurs when inserting the finger into the imaging device for the captured finger image,
A personal authentication method, wherein personal authentication is performed by matching a vein pattern included in the finger image with a vein pattern of a registered finger image.

Images