JP2007079771A - Personal identification device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a personal identification device for easily and precisely collating the image of a biological pattern picked up in a non-contact status with registered image data without installing any additional device. <P>SOLUTION: This personal identification device is configured of a light source 1 for irradiating a finger 7 with the rays of light; an image pickup means 3 for picking up the image of a fingerprint as well as the finger 7; an optical system 2 installed so that the images of the fingerprint and the finger 7 can be picked up even when the image pickup object 7 is moved within a predetermined movement permissible range; a normalization parameter acquiring means 41 for calculating finger width as a normalization parameter from the pickup image picked up by the image pickup means 3; a storage means 61 for storing a feature parameter in a preliminarily pickup individual fingerprint image as registration data; a normalization processing means 42 for normalizing the pickup image by using the normalization parameter so that the coordinate system of the pickup image can be matched with the coordinate system of the registration data; a feature parameter extracting means 5 for extracting the feature parameter from the normalized pickup image; and a collating means 6 for collating the extracted feature parameter with the registration data. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a personal identification device that identifies a person by analyzing a biometric pattern captured in a non-contact manner.

In a conventional personal identification device, for example, in a fingerprint identification device, there is a method of collecting a fingerprint by pressing a finger against a flat sensor device, and in recent years, there is a method of imaging a fingerprint in a contactless manner with a camera from a remote location (for example, Patent Document 1).
Japanese Patent Application Laid-Open No. H10-228561 discloses a technique of obtaining a fingerprint image by imaging a finger while the finger is lifted without using a finger placement guide, and imaging light transmitted through the finger.

JP-A-2004-290664 (FIGS. 13 to 18)

In the above-described conventional device that captures a fingerprint pattern in a non-contact manner, since the position of the finger is not specified, the size of the fingerprint pattern varies depending on the distance between the finger and the image sensor. There is a problem that it is not possible to accurately collate the image with a previously registered image.
As a method of collating a fingerprint image whose size changes with a registered image, for example, using a collation method in a conventional contact-type fingerprint identification device, changing the size of the obtained image at various magnifications, Although a method of exhaustively searching for an enlargement rate that can be successfully verified can be applied, it takes time. There is also a method of measuring the distance to the finger using distance measurement hardware and normalizing the image size by multiplying the reciprocal of the distance, but an additional device is required, which is costly. There was a problem.
In addition, the same problem has been encountered not only for fingerprint patterns but also for devices that image other biological patterns used for personal identification such as blood flow patterns without contact.

  The present invention has been made in order to solve the above-described problems. The image data of a biological pattern captured in a non-contact manner can be easily registered without providing an additional device. It is an object to obtain a personal identification device that can be collated with high accuracy.

  The personal identification device according to the present invention includes a light source that irradiates light to an imaging target having an individual biological pattern, an imaging unit that images the biological pattern together with the imaging target, and the imaging target moves within a predetermined allowable movement range. However, the length of the specific portion of the imaging target is calculated as a normalization parameter from the optical system installed so that the biological pattern can be imaged together with the imaging target and the captured image captured by the imaging means. Normalization parameter acquisition means, storage means for storing characteristic parameters in an image of an individual biometric pattern captured in advance as registration data, and the normalization parameter so that the coordinate system of the captured image matches the coordinate system of the registration data Normalization processing means for normalizing using normalization parameters, feature parameter extraction for extracting feature parameters from captured images after normalization processing Stage, as well as those with matching means for matching the extracted the feature parameters and the registered data.

  The personal identification device according to the present invention includes a light source that irradiates light to an imaging target having an individual biological pattern, an imaging unit that images the biological pattern together with the imaging target, and the imaging target is within a predetermined movement allowable range. An optical system installed so that the biological pattern can be imaged together with the imaging target even if it moves, and a length of a specific portion of the imaging target as a normalization parameter from the captured image captured by the imaging means Normalization parameter acquisition means for calculating, storage means for storing feature parameters in an image of an individual biometric pattern captured in advance as registration data, feature parameter extraction means for extracting feature parameters from the captured image, and extraction of feature parameters The normalization using the normalization parameter is performed so that the coordinate system and the coordinate system of the registered data match. Processing means, as well as those with matching means for matching the normalized the feature parameters and the registered data.

  According to the present invention, even if the size of a captured image changes due to changes in imaging conditions, changes in personal growth, body shape, etc., it can be easily verified at low cost against registered data and verified accurately. be able to.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a personal identification device according to Embodiment 1 of the present invention. When the light emitted from the light source 1 is irradiated from the back of the finger onto the finger 7 inserted in the allowable imaging range 21, the light is transmitted through the finger and the optical system 2 (shown as a single lens in the figure for simplicity). And the imaging unit 3 captures the biological pattern of the finger together with the finger. The captured biological pattern is a fingerprint pattern, a blood flow pattern, or an image in which they are superimposed depending on the wavelength of the light source 1 and the imaging method.
By taking a wide focusing range of the optical system 2, a biological pattern can be obtained regardless of where the finger 7 is in the allowable imaging range 21. Further, the optical system 2 and the image pickup means 3 are installed so as to be able to pick up an image of a range wider than the finger width in an allowable finger movement range.

As the allowable imaging range 21, the finger movement range of the finger at the upper end of the allowable imaging range 21 is set to the finger width in a state where the lower half (finger side of the finger) that may be imaged is at the upper end portion of the range 21. A value obtained by adding the required values becomes a lower limit value of the width of the upper end portion of the range 21. The upper limit of the width of the upper end portion of the range 21 is a width that can maintain the minimum required image resolution (that is, a resolution equal to or higher than the ridge interval) in the image pickup means 3 or the subsequent normalization means 4 or feature parameter extraction means 5. It becomes. As a specific calculation example, for example, when the user of the device can be limited to Japanese, the upper limit value of the width of the fingertip of the second finger, which is often used for authentication, is statistically about 24 mm. For example, 10 mm is added as the required degree of freedom value to set the lower limit value of the upper end width of the range 21 to 34 mm. Further, when the number of pixels in the width direction of the image sensor is, for example, 1400 pixels and the apparent minimum value of the ridge interval is, for example, 0.05 mm, the upper limit value of the upper end width of the range 21 is 1400 / √2 × 0.05. It can be estimated that ≈50 mm (÷ √2 reflects a decrease in resolution in a 45-degree oblique direction in the image sensor). Since it was confirmed that the upper limit value is larger than the lower limit value, the upper limit value is determined to be 50 mm.
In the above discussion, for the sake of simplicity, the degree of freedom in the front-rear direction of the finger and the degree of freedom of the lower end portion of the range 21 were not mentioned, but actually, the same calculation as described above is necessary for them.
Also, if the angle of view of the optical system 2 is wide, the device can be miniaturized because the focal length is short, but the difference in the movable range of the finger at the upper end and the lower end widens, and the limit of the image resolution tends to be reached especially at the upper end. Therefore, a longer focal length is advantageous for securing the degree of freedom. On the other hand, there is a negative aspect that if the focal length becomes long, the amount of light is insufficient and the image quality deteriorates. The optimal balance among the focal length, the angle of view, and the width of the allowable imaging range is the optical system, imaging device, image processing It also depends on the ratio of performance that can be secured at various points. As a guide for the angle of view, it is usually 30 degrees or less, and preferably 45 degrees or less at the maximum.
Further, if there are few problems related to insufficient light amount, the optical system 2 has a structure called a telecentric system that optically realizes a state corresponding to an infinite focal length, so that the allowable imaging range is suppressed while suppressing the focal length to some extent. It is also possible to minimize the difference between the top and bottom.

  FIG. 2A is a diagram illustrating a state in which a biological pattern (here, a fingerprint pattern) is captured. If the finger is away from the 7 allowable imaging range 21 and is not in focus, or if the movement of the finger 7 is faster than the imaging frame rate and a biological pattern cannot be obtained, a spatial differentiation process is applied to the image. Since it can be easily distinguished by a method such as examining the contrast of the edge portion, only the case where such a case does not occur will be described here.

When the picked-up image as shown in FIG. 2A is obtained by the image pickup means 3, the obtained picked-up image is sent to the normalization means 4 and subjected to the following processing.
First, the normalization parameter acquisition unit 41 obtains the finger width as a normalization parameter. As a procedure, the obtained captured image is subjected to image processing, the background portion 75 is removed from the captured image, and the finger region 70 is obtained. Unlike the finger region 70, the background portion 75 does not have an object in focus. For example, the background portion 75 is removed by processing such as removal of a spatial low-frequency component or the biological pattern is clear. The background portion 75 is removed using the fact that the optical wavelength characteristics are different from each other.

  As a background removal method by removing spatial low-frequency components, for example, a captured image is subjected to two-dimensional Fourier transform, etc., a component equal to or lower than a predetermined spatial frequency threshold value is converted to 0, and inverse conversion is performed. The region is filled and the noise region is erased by a method such as binarization with a threshold value and performing shrinkage / expansion several times. As the spatial frequency threshold, since the maximum value of the ridge interval is the state in which the finger is projected to the full width of the image, for example, the width of the captured image is set to the finger of the finger with statistically the least ridge. By dividing by a number sufficiently smaller than the number of ridges in the transverse direction (for example, 20 to 30), a sufficiently high frequency threshold can be set as compared with an object in which the background portion is not in focus.

  In addition, as a background removal method based on optical wavelength characteristics, a bandpass filter 20 that emits a wavelength of light emitted from the light source 1 and passes through a living body, and a background that prevents reflection and transmission of light having the above wavelength behind an imaging range. By providing the plate 10, a finger through which light having the above wavelength passes is imaged as a bright region, and a background region without light having the above wavelength is imaged as a dark region. Therefore, for example, if binarization processing is performed on a captured image obtained based on such a configuration, the finger region 70 can be easily obtained. In the apparatus of FIG. 1, a background plate 10 and a band pass filter 20 are provided.

When the finger area 70 is obtained, next, the principal axis 71 is obtained, and the finger width 72 in the direction perpendicular to the principal axis 71 is obtained.
For example, the principal axis 71 may use a technique such as principal component analysis. However, when the direction of the finger is substantially limited due to the structure of the apparatus, the principal axis 71 can be approximately obtained by a simple method as described below, for example. That is, as shown in FIG. 2A, if it is known that the finger is substantially in the vertical direction, the image is first scanned along the horizontal direction (x direction), and the finger for each horizontal line (position y) is scanned. The left end coordinate sequence xl (y) and the right end coordinate sequence xr (y) of the area 70 are detected.
Note that only the background portion is not subject to processing.
And the midpoint coordinate sequence xc (y) = (xl (y) + xr (y)) / 2
And the intercept and slope of the approximate line are determined by the least square method and set as the main axis 71.

Next, when all pairs orthogonal to the principal axis are selected from any combination of the points (xl (y1), y1) and the points (xr (y2), y2), the distance between the selected points is The finger width 72 along the main axis 71 is represented.
The following method is used to select all the sets orthogonal to the main axis. A reference point (xr (y2), y2) paired with a certain point of interest (xl (y1), y1)
tan θ = (yl−yr) / (xl (y1) −xr (y1))
Meet. Therefore, y = y2 satisfying the above equation can be obtained by searching by sequentially changing y from the point sequence xr (y). Therefore, when the above search operation is performed for each point of the point sequence xl (y1), all pairs of points orthogonal to the main axis are selected.
Therefore, the value Fw (y1) of the finger width 72 at y1 is Fw (y1) = √ ((xl (y1) −xr (y1)) ² + (yl−yr) ² ).
Or Fw (y1) = | cos θ × (xl (y1) −xr (y1)) |
Can be calculated.

  Among the obtained finger widths, for example, the width of the portion moved to the base side by a predetermined distance from the fingertip, the maximum width of the finger width that changes along the main axis 71 of the finger, or along the main axis 71 of the finger The representative value Fw of the finger width 72 is determined by a method such as an average value of the changing finger width.

When the finger width 72 is calculated as a normalization parameter by the processing as described above, the normalization processing unit 42 uses the information on the finger width 72 to adjust the size of the captured fingerprint pattern to a predetermined size. At the same time, normalization is performed so that the orientation of the fingerprint pattern is aligned with a predetermined direction.
The apparent enlargement ratio of the fingerprint pattern existing in the finger region 70 is not only a component inversely proportional to the distance from the focal position of the optical system 2, but also the original finger generated by the growth of the individual to be identified, the body shape change, etc. Another component of individual difference in size is included. Therefore, the two cannot be separated without the measured distance value. Further, even if there is no image expansion / contraction due to the distance, image processing corresponding to the expansion / contraction of the biological pattern due to growth or the like is required.
In this embodiment, when normalizing the fingerprint pattern to a predetermined size, the finger width of the captured image is used as a normalization parameter, and the coordinate system of the captured image matches the coordinate system of the registered data. Normalization is performed so that As a result, it is not necessary to separate a component that is inversely proportional to the distance from the focal position of the optical system 2 and a component due to individual differences in the original size of the finger in the apparent magnification of the fingerprint pattern. Further, distance measurement is not necessary. In other words, since biometric patterns such as fingerprints are life-long, if the size of the finger changes due to growth or obesity, the biometric pattern only expands and contracts in proportion to the expansion and contraction of the skin, and the number of fingerprint ridges increases. Such a thing cannot happen. Therefore, the above component due to individual differences may be considered as a component that expands or contracts according to the apparent finger width. In addition, regarding the expansion / contraction of the image depending on the distance, it can be considered as a component that expands and contracts according to the apparent finger width, so by applying a normalization process that aligns the finger width to the predetermined finger width, the influence of each component Can be canceled together.

Specifically, the captured image is normalized by the following conversion. For example, the reference finger width (predetermined fixed value) is W, the principal axis 71 of the finger image of the finger width is the Y axis, the axis perpendicular to the Y axis is the X axis, and the origin is set to a predetermined position to be the reference coordinates System. The origin of the coordinate system is, for example, simply the intersection of the Y axis and the lower end of the image, or the tip position of the finger is further detected in the (+) direction along the Y axis, and W × k ( k is determined as a point moved in the Y-axis (−) direction by a predetermined magnification.
If the finger width in the captured image obtained by the normalization parameter acquisition unit 41 is Fw, the coordinates of each point of the captured image after conversion by the normalization process are
(X, Y) = (W / Fw) R (Xo, Yo) + (Xc, Yc) (1)
It is represented by However, (Xo, Yo) is the coordinates of each pixel in the coordinate system of the original captured image (in this case, the main axis 71 is the Yo axis, the axis perpendicular to it is the Xo axis, and the origin is determined so that the finger image is in the center. ) R is a reverse rotation matrix corresponding to the tilt angle between the main axis 71 in the coordinate system of the captured image and the main axis 71 of the finger image in the reference coordinate system, and (Xc, Yc) is the origin of the (X, Y) plane and ( Xo, Yo) is the amount of translation of the origin of the plane. The inverse rotation matrix R uses an angle θ obtained as an arc tangent of the inclination of the main axis 71, and

Represented as:
A captured image converted by the normalization process is shown in FIG.

In this way, the captured image that has been converted so as to coincide with the reference coordinate system is then processed by the feature parameter extraction means 5 to extract the feature parameter.
On the other hand, a fingerprint pattern is captured separately with a finger in advance, and this finger image is also normalized so that it matches the reference coordinate system, and an image having a finger width of a predetermined value is obtained. . Further, feature parameters are extracted from this image, and the extracted feature parameters are individually registered in the database 61.
The collating unit 6 collates the characteristic parameters of the captured image extracted by the characteristic parameter extracting unit 5 with the characteristic parameters (registered data) registered in the data pace 61.
As a result, the positional shift of the feature parameter due to enlargement / reduction is greatly reduced, and the matching can be performed with high accuracy.

  The arithmetic processing executed in the normalizing means 4, the characteristic parameter extracting means 5 and the collating means 6 is executed by inputting the obtained captured image to a computer and by a program in the computer. The

  For example, a method as described in Japanese Patent Application Laid-Open No. 2005-10842 can be used for extracting characteristic parameters, and a method as described in Japanese Patent Application Laid-Open No. 4-357570 can be used for collation.

  In the above embodiment, the image used for the registration data is also normalized to a finger width of a predetermined value (fixed value), and the finger width of the captured image is set to the predetermined value (fixed value). The normalization process that fits the finger width of the image is described. However, the image used for the registration data is not normalized, and when the captured image is normalized, the finger width of the image used for the registration data is reduced. It is good also as a treatment to match. In this case, data relating to the finger width is registered in the database 61 together with the registration data.

  As described above, in this embodiment, the finger width of the captured image is used as a normalization parameter, and normalization is performed using the normalization parameter so that the coordinate system of the captured image matches the coordinate system of the registered data. At the same time, feature parameters are extracted from the captured image after normalization, and the above-mentioned feature parameters are registered with the registered data. Even if the size changes, the captured image can be compared with the registered data easily and at low cost, and the matching can be performed with high accuracy without shifting the characteristic parameter.

  In image processing such as fingerprints, a spatial pattern filter limited to a certain spatial frequency is used as described in, for example, Japanese Patent Application Laid-Open No. 2005-10842, and only a portion of a ridge interval close to that frequency is captured image. The method of extracting from is used. Since the spatial frequency of the biological pattern is basically proportional to the apparent finger width except for individual differences, the normalization of the present invention allows both the component due to finger distance and the component due to growth, etc., from fluctuations in the spatial frequency on the image. Is removed, and only the fluctuation component due to the original individual difference is obtained. Therefore, since the spatial frequency value distribution is more limited, there is an advantage that there is no need for additional processing such as preparing filters corresponding to many spatial frequencies in order to cope with a wide spatial frequency range. It is done.

  In the above embodiment, the light from the light source is incident from the back of the finger and the fingerprint pattern is obtained by the transmitted light transmitted through the finger. However, the fingerprint pattern is directly imaged by irradiating light on the finger abdomen. The same effect can be obtained by providing the normalizing means 4 of the present embodiment.

The biometric pattern of the finger is not limited to the fingerprint pattern, and the same effect can be obtained even if the blood flow pattern is imaged.
Further, the same effect can be obtained by providing the same normalizing means for an apparatus that captures a biological pattern of a palm, for example, a palmprint pattern or a blood flow pattern of the palm. However, the normalization parameter in this case is not the finger width, but the palm width, for example, the width at the third joint of the little finger and the third joint of the index finger, and the like, where the palm is relatively less fluctuating with respect to movement, and It is preferable to use as a parameter the value at which the biological pattern expands and contracts in proportion to the change in body shape.

  In the above embodiment, the captured image is normalized so that the coordinate system of the captured image matches the coordinate system of the registration data. However, the registration data and the captured image are interchanged, and the captured image is normalized. Instead, the registration data may be normalized so that the coordinate system of the registration data matches the coordinate system of the captured image.

Embodiment 2. FIG.
FIG. 3 is a block diagram showing a personal identification device according to Embodiment 2 of the present invention. In the first embodiment, the feature parameter is extracted after normalizing the captured image itself. However, in the present embodiment, the feature parameter is first obtained from the captured image, and then the coordinate system of the feature parameter and the registration data are obtained. Normalize to match the coordinate system of. That is, in FIG. 3, the feature parameter extraction unit 5 is arranged between the normalization parameter acquisition unit 41 and the normalization process 42. Other configurations are the same as those of the first embodiment.

In the normalization parameter acquisition unit 41, the finger width Fw is calculated by the same method as that performed by the normalization unit 4 in the first embodiment, and the mathematical expression such as the image scaling factor W / Fw and the rotation matrix R is calculated. The parameter (1) is calculated.
The feature parameter extraction unit 5 extracts feature parameters from the captured image. At this time, parameters such as the enlargement / reduction ratio calculated above are also used depending on the nature of the extraction process. For example, in the aforementioned Japanese Patent Application Laid-Open No. 2005-10842, a ridge interval is used for processing that uses a spatial pattern filter limited to a certain spatial frequency and extracts only a portion of the ridge interval close to that frequency from the captured image. Is scaled, the filter frequency also needs to be adjusted. In this case, what is necessary is just to make it the frequency which multiplied the reciprocal number of the expansion / contraction rate to the frequency used normally.

  The normalization processing unit 42 performs coordinate transformation using the same formula as the formula (1) so that the coordinate system of the extracted feature parameter matches the coordinate system of the registered data, and normalizes the extracted feature parameter. . However, (Xo, Yo) is the coordinate of the feature parameter in the coordinate system of the captured image.

  The collating means 6 collates the normalized characteristic parameter with the registered data (characteristic parameter) registered in the data pace 61. As a result, the positional shift of the feature parameter due to enlargement / reduction is greatly reduced, and the matching can be performed with high accuracy.

  Since the information amount of the characteristic parameter is generally smaller than that of the image itself, this configuration can obtain the same effect as that of the first embodiment with a smaller processing amount.

In this embodiment as well, as in the first embodiment, the coordinate system of registered data registered in advance may be normalized to a reference coordinate system, or extracted from the captured image without normalization. When normalizing the feature parameters that have been performed, a procedure may be adopted in which the feature parameters are adjusted to the finger width of the image used for the registration data.
Also, in normalization, the registered data is normalized so that the coordinate system of the registered data matches the coordinate system of the feature parameter extracted from the captured image without normalizing the feature parameter extracted from the captured image. Also good.

It is a block diagram which shows the personal identification device by Embodiment 1 of this invention. It is a figure explaining operation | movement of the personal identification device by Embodiment 1 of this invention. It is a block diagram which shows the personal identification device by Embodiment 2 of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Light source, 2 Lens system, 21 Permissible imaging range, 3 Imaging means, 4 Normalization means, 41 Normalization parameter acquisition means, 42 Normalization processing means, 5 Feature parameter extraction means, 6 Collation means, 61 Database, 7 fingers, 70 finger area, 71 main axis, 72 finger width, 75 background area.

Claims (4)

A light source that irradiates light to an imaging target having an individual biological pattern, an imaging unit that images the biological pattern together with the imaging target, and the biological pattern is captured even if the imaging target moves within a predetermined allowable movement range. Normalized parameter acquisition means for calculating the length of a specific part of the imaging target as a normalization parameter from an optical system installed so as to be capable of imaging with the target, and a captured image captured by the imaging means. Storage means for storing feature parameters in individual biometric pattern images as registration data, and normalization using the normalization parameters so that the coordinate system of the captured image matches the coordinate system of the registration data Processing means, feature parameter extraction means for extracting feature parameters from the captured image after normalization processing, and the extracted features Personal identification device characterized by comprising a verification means for collating the parameters and the registered data. A light source that irradiates light to an imaging target having an individual biological pattern, an imaging unit that images the biological pattern together with the imaging target, and the biological pattern is captured even if the imaging target moves within a predetermined allowable movement range. Normalized parameter acquisition means for calculating the length of a specific part of the imaging target as a normalization parameter from an optical system installed so as to be capable of imaging with the target, and a captured image captured by the imaging means. Storage means for storing feature parameters in individual biometric pattern images as registration data, feature parameter extraction means for extracting feature parameters from the captured image, a coordinate system of the extracted feature parameters, and a coordinate system of the registration data Normalization processing means for normalizing using the normalization parameters so as to match, and the normalized above Personal identification device characterized by comprising a verification means for collating the symptoms parameter and the registration data. The personal identification device according to claim 1, wherein the imaging unit images the fingerprint together with the finger, and the normalization parameter acquisition unit calculates the finger width as a normalization parameter from the captured image. The personal identification device according to any one of claims 1 to 3, wherein a coordinate system of registration data is normalized to a reference coordinate system.

Images