JP2002117405A - Palm shape authentication method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve the problems that much time and labor are required in a method for authenticating a person himself by a fixed palm shape, the method is unsuitable from the viewpoints of convenience and human psychology, and a large device is required therein. SOLUTION: Palm shape images (b), (c) detected by a CCD camera detect a deviation in rotation angle for a registered palm shape (a) to compensate an angle, the number of pixels is plotted as data per line in the horizontal direction of both images (d) to authenticate the person himself based on an error between both images.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention belongs to the field of biometrics identity verification in which the identity of a person is confirmed based on the biological characteristics of the individual (biometrics), and in particular, a palm shape detected from a non-fixed palm shape. The present invention relates to a palm-shape authentication method for authenticating a person based on features.

[0002]

2. Description of the Related Art Conventionally, in personal identification using a palm shape, a palm is fixed by sliding a finger on a table on which several pins are arranged, and the shape of the palm / finger, the width of the palm, the width of the palm, and the like. The authentication was performed using the area of the finger, the length of the finger, and the like.

[0003]

However, fixed palm shape authentication requires time and effort to fix the palm shape in addition to the time required to extract and authenticate features from the palm shape. It took a lot of trouble and time.

[0004] Accordingly, there has been a great demand for entry / exit management using a small number of people, but it is not suitable for application to a gate system or a fee collection system through which many people pass.

[0005] In addition, the fixed palm-shaped authentication involves a psychological resistance to sliding a finger into a table on which several pins are arranged, and in addition to an elderly person or a person with a handicapped person. It was an inconvenient authentication method.

[0006] Further, in the fixed palm shape authentication, since a general palm shape is targeted for authentication, it is incompatible with an unusual palm shape such as an infant's palm or an extremely large palm. . In addition, a fixed palm-shaped authentication device requires a platform on which pins for fixing the palm are required, which increases the size of the authentication device and is used for communication devices and network terminals that have become smaller due to the development of integrated technology. It was difficult to incorporate.

Furthermore, the left and right palm shapes of an individual generally have a line-symmetric shape, and in a fixed palm shape authentication device, when the reverse palm of the registered palm is fixed to the device, the registered palm shape is not registered. Since the palm cannot be fixed unless it is turned upside down, the left and right palm shapes cannot be distinguished. Due to reasons such as a change in the specifications of the palm shape detection device (such as a CCD camera), a change in the setting such as enlargement / reduction magnification, or a change in the distance between the palm shape detection device and the palm shape, the detected palm shape for the registered palm shape is changed. When the relative size of the shape was changed, the same palm shape could not be correctly identified.

[0008] An object of the present invention is to provide a palm-shape authentication method which solves the above problems.

[0009]

According to the present invention, a non-fixed palm shape placed in a free direction (rotational angle) in close contact with a table is provided.
By detecting using a non-contact detection device such as a camera or a three-dimensional shape measuring device or a contact detection device such as a pressure sensor, the time and labor for fixing the palm can be saved.

In the palm shape authentication, a rotation angle of the palm is detected, a position farthest from the body in the direction of the detected rotation angle is defined as a tip of the palm, and the palm ( Finger) By comparing the palm shape based on the change in width and the distance between other characteristic points of the palm shape (points at which the palm width change becomes steep), it is possible to deal with an unusual palm shape.

Further, focusing on the tip of the palm, the characteristic of the change of the palm (finger) width to another fingertip based on the tip position (usually the tip of the middle finger) and the distance to the other fingertip When,
By comparing the palm shape based on the length of the finger / palm, the width of the finger / palm, and the distance between feature points in other palm shapes, the image size handled in the matching process is reduced, and the matching time is reduced. By shortening the size of the palm shape detection device, it is possible to incorporate the device into a small communication device or a network terminal.

[0012] Further, a palm with aligned fingers is targeted for authentication, and a direction perpendicular to the direction in which the fingers of the palm are lined is defined and detected as a palm rotation angle, and the palm shape is collated, thereby obtaining a palm shape. To minimize the change of the shape, and improve the authentication accuracy using a non-fixed palm shape.

In the non-fixed palm-shaped authentication, since the palm is not fixed with a pin, both right and left palms can be registered without turning over.

In addition, the collation of the palm shape is performed by using a collation algorithm which is not affected by the line symmetry of the target shape (a similar result is obtained for a line-symmetric shape), and an arbitrary palm rotation angle is obtained. By performing matching based on the distance between feature points in the rotation angle (measurement angle) direction component, the length of the finger / palm, and the width of the finger / palm, (a) the measurement angle becomes the same angle as the palm rotation angle By setting, the user can be authenticated by using either palm regardless of the registered palm. (Condition: The left and right palms of the individual have strong line symmetry, and the In the case of a person having a general palm shape in which the matching error between the other palm shapes is larger than the matching error). (B) By appropriately shifting the measurement angle from the palm rotation angle, it becomes possible to perform measurement while breaking the line symmetry of the left and right palm shapes of the same person, and thus the left and right palm shapes of the same person are identified and registered as different palms. It is possible to do.

By comparing and identifying the palm based on the ratio of the distance between the detected feature points of the palm, the feature data required for the comparison is reduced, and the capacity of the registered data is reduced.
In addition, due to changes in the specifications of the palm shape detection device (such as a CCD camera), changes in settings such as enlargement / reduction magnification, or changes in the distance between the palm shape detection device and the palm shape, the registered palm shape may not be used. Even if the relative size of the detected palm shape changes, it is possible to correctly identify the same palm shape.

As described above, the present invention is characterized by the following palm-shaped authentication method.

In a palm-shape authentication method for authenticating a person based on a personal palm-shape, palm-shape image data is detected from a non-fixed palm-shape placed in a free direction, and palm rotation is performed from the palm-shape image data. Detecting the angle, correcting the positional deviation between the rotation angle of the registered palm shape image obtained by processing the registered palm shape image registered in advance by detecting and registering the palm shape of the person in advance and the rotation angle of the detected palm shape image. According to another aspect of the present invention, the authentication of the person is performed by collation based on an error between the palm-shaped image whose position has been corrected and the registered palm-shaped image.

Further, the rotation angle is detected by performing a two-dimensional Fourier transform on the binarized palm-shaped image to obtain an intensity image centered on a DC component. An image in the form of a rectangular coordinate system with a vertical axis frequency is obtained, and the angle direction perpendicular to the rotation angle at which the maximum intensity cumulative value of the intensity value for each rotation angle of this image is obtained is defined as the palm-shaped rotation angle. And

In the above-mentioned collation, the palm end in the rotational angle direction of the detected palm-shaped image is set as the tip of the palm, and the change in the palm width or the change in the palm width based on the tip position of the palm is steep. It is characterized in that personal authentication is performed based on the ratio of the distance between points.

In the above-mentioned collation, the palm shape in which the fingers are aligned is to be authenticated, the direction perpendicular to the direction in which the fingers of the palm shape are arranged is detected as the palm rotation angle, and the forefront position of the fingertip portion is determined. Based on the characteristics of the change of the palm width as a reference, the distance to other fingertips, the length of the finger / palm, the width of the finger / palm, and the distance between the characteristic points predetermined in the palm shape, Is compared.

In the above collation, the collation of the palm shape uses a collation algorithm which is not affected by the line symmetry of the object, and the distance between the feature points and the finger in an arbitrary rotation angle direction component with respect to the palm shape rotation angle. -It is characterized in that the person is collated based on the length of the palm and the width of the finger / palm.

Further, the collation is performed based on the ratio of the distance between the detected feature points of the palm shape.

[0023]

DESCRIPTION OF THE PREFERRED EMBODIMENTS In an embodiment which will be described in detail below, a non-fixed palm shape freely extended with fingers aligned is used as a CCD.
This is a palm shape authentication method in which a palm shape image is detected by a camera or the like, and the detected non-fixed palm shape is compared with the registered palm shape to authenticate the user.

For palm shape detection, the measurement angle direction is set to the column (vertical) direction of the image, thereby facilitating collation processing mainly by a computer. In addition, a matching algorithm that is not affected by the line symmetry of the shape to be compared is used for matching, and the palm rotation angle of the registered palm is measured to identify the left and right palms of the same person with line symmetry. Register by shifting from the angle.

(First Embodiment) FIG. 1 shows a specific authentication processing procedure of the palm-shaped authentication method according to the first embodiment. As shown in FIGS. 1A and 1B, the palm shape detected by the CCD camera (detected palm shape) involves a rotational angle shift with respect to a previously registered palm shape (registered palm shape). Therefore, in the present embodiment, before performing the image matching process, a rotational angle shift of the detected palm shape with respect to the registered palm shape is detected, and preprocessing of an image for correcting this is performed. In general, the pre-processing includes performing noise removal, smoothing, edge emphasis, thinning, position shift correction, and the like on the image as required in addition to the rotation angle shift correction. By performing pre-processing on the image, the accuracy of authentication can be improved, and the image size handled at the time of image collation and the calculation cost can be significantly reduced, which leads to a reduction in authentication time.

In this embodiment, first, a specific color component is extracted from a detected image of 256 gradations of RGB captured by a CCD camera, and is converted into a grayscale image based on grayscale values. This is for efficiently extracting only the finger portion from the background image.

Next, the gray scale image is binarized.
This is to avoid wrinkles of the hand and the influence of the background, and to emphasize only the direction component in which the fingers are arranged.

Next, a two-dimensional median filter is applied to the binarized image (FIG. 1B). This is to remove impulsive noise generated by binarization.

Next, a two-dimensional Fourier transform is performed on the binarized image to obtain an intensity (power spectrum) image centered on a DC component (without using phase information). This is to detect the rotation angle of the palm based on the spatial frequency of the image, regardless of the positional shift of the palm or a change in the lighting environment.

Next, the intensity image is subjected to polar coordinate conversion to obtain an image in an orthogonal coordinate system having a horizontal axis angle and a vertical axis frequency. This is to evaluate the spatial frequency of the image for each rotation angle.

In this embodiment, the intensity values are accumulated for each rotation angle (power spectrum is frequency-integrated), and the angle direction perpendicular to the rotation angle at which the maximum intensity accumulation value is obtained is the palm rotation angle ( Angle at which the finger is extended). Similarly, by detecting the palm shape rotation angle of the registered palm shape ((a) in FIG. 1), and calculating the difference between the detected palm shape and the palm shape rotation angle,
A rotational angle shift of the detected palm shape with respect to the registered palm shape is detected, and the detected palm shape is corrected (FIG. 1C).

Up to this point, the rotation angle deviation correction processing procedure has been described. In addition to the procedure described in the embodiment, by performing necessary processing from the standpoint of signal theory, such as applying a window function at the time of Fourier transform, it is possible to improve the correction accuracy of the rotational angle deviation. Using the detected palm shape (FIG. 1C) subjected to the pre-processing and the registered palm shape (FIG. 1A), a palm shape matching process is performed.

In the collation processing in this embodiment, palm shape matching is performed by detecting a change in palm width in the column (vertical) direction of the palm shape image. The change in palm width includes information on the width and length of each finger. Since the palm portion is represented as white pixels by binarizing the image, the number of white pixels is counted for each row in the horizontal direction of the image in order to detect a change in palm width in the column (vertical) direction. The number of rows counted is plotted on the horizontal axis, and the number of white pixels in each row is plotted on the vertical axis, and the detected change in palm width is plotted. FIG. 1 (d) plots the change in palm width for the first time, with the first row of white pixels being counted in order to match the registered palm shape with the detected palm shape based on the tip of the palm. The result. The number of white pixels indicates the width of the finger, and it can be seen that the point where the pixel increase is sharp is the point where the finger newly appears.

Next, the error of the palm width change data obtained from the registered palm shape and the detected palm shape is calculated. In the present embodiment, a mean square error (MSE: Mean Square) between data is set.
dError), the absolute value of the difference for each row, and the maximum value of the absolute value of the difference for each row. FIGS. 1E and 1F show the results of plotting the calculated absolute value of the difference for each row with “+” on the drawing obtained in FIG. 1D. FIG. 1E shows an example of an error detected in the palm of the subject when the palm shape authentication is performed according to the present embodiment, and FIG. 1F is similarly detected in the palm of another person. It is an example of an error. In the palm of another person, a large error is detected because the width of the finger, the timing from the longest finger to the next finger, and the like are different.

The variance of the error obtained between the same palm of the person,
Based on the variance of errors obtained between different palms of another person or between other persons' palms, etc., an appropriate allowable value of the error for identifying the person is determined and used for personal authentication. For example, if the allowable error value in the present embodiment is defined as MSE <30 and the maximum absolute value of the difference for each row <15, (e) in FIG. 1 is “person”, and (f) in FIG. Can be identified as "others".

When actually performing palm-shape authentication using this method, by appropriately setting the allowable error range,
False acceptance rate required for palm-shape authentication system (FAR: F
alseAcceptRate) and rejection rate (FR
R: FalseRejectRate).
FRR is an index relating to the ease of use of the personal identification system, and FAR is an index relating to the security and reliability of the system. Since FRR and FAR have a trade-off relationship with each other, they are usually set to appropriate values according to the nature of the application. For example, in a gate system where many people pass, the FRR is set small so that it can be passed by one authentication, and conversely, in a financial system, F
Set the AR to be small and allow multiple authentications.

(Second Embodiment) In the present embodiment, the first
Similarly to the procedure of the embodiment, after binarizing the detected image, correcting the rotational angle shift, and counting and plotting the number of white pixels for each row in the horizontal direction of the image, as shown in FIG. A sharply increasing point is detected by calculating a differential coefficient of the plot data, and palm shape authentication is performed based on a ratio of the distance (the number of rows) between the detected points.

The point at which the number of white pixels sharply increases is a point where a new finger appears. Therefore, in the second embodiment, the palm shape is determined based on the positional relationship (the ratio of the distance to the measurement angle direction) of the tip of each finger. Authentication.
Since at least three points are required to calculate the ratio between the points, when the palm shape is registered, the derivative is calculated while counting the white pixels for each row until three points are detected.

In the second embodiment, since the data to be registered in advance for palm-shape authentication only needs to be at least the ratio of the distance between each point, it is possible to greatly reduce the amount of registered data. it can. In addition, as shown in FIG.
When the relative size of the detected palm shape changes with respect to the registered palm shape due to a change in settings such as enlargement / reduction magnification or a change in the distance between the palm shape detection device and the palm shape. However, since the ratio between the detected points is always constant in the same palm shape, correct identification becomes possible.

If this method is used, the palm shape detection device prepared for authentication can be freely selected within a range in which the relative size of the palm shape varies, so that the palm shape authentication device is used. There are many advantages, such as being able to change specifications in consideration of the environment, convenience, cost, etc. of the site where the device is installed, and facilitating the entry of device development manufacturers.

When the palm shape authentication is actually performed by using the present method, similarly to the first embodiment, by setting an allowable error range appropriately, a third party acceptance required for the palm shape authentication system can be obtained. The rate (FAR) and the false reject rate (FRR) are adjusted.

[0042]

As described above, the following effects can be obtained by using the palm-shaped authentication method of the present invention.

(1) The palm shape authentication using the non-fixed palm shape can save time and labor for fixing the palm, so that a gate system or a fee collection system that many people pass through can be used. It can be applied.

(2) It is an authentication method that can reduce the psychological burden on the user at the time of personal authentication, and is highly convenient for the elderly and those who have disabilities in their palms.

(3) An unusual palm shape, such as a palm that is too small, a palm that is too large, a palm that has been injured or deformed, or a palm of an animal other than a human, is used. Authentication and a wide range of individuals can be authenticated regardless of the person.

(4) By performing authentication focusing on the tip of the palm, it is possible to reduce the collation time and to incorporate the device into a small communication device or network terminal.

(5) The palm with aligned fingers is targeted for authentication, so that the accuracy of authentication using a non-fixed palm shape can be improved.

(6) Both the right and left palms can be registered as the same palm without turning over and used for personal authentication.

(7) By changing the registration information, the left and right palm shapes of the same person can be identified and registered as different palms.

(8) By using the ratio of the distance between the points at which the finger appears for collation, the amount of feature data to be registered can be reduced, or the palm-shaped photographing device (such as a CCD camera) and the object to be detected (palm) can be used. ) And expansion of palm-shape photography equipment
A palm shape authentication can be realized using a palm shape image in which the detected palm shape size changes relatively with a change in the reduction magnification.

[Brief description of the drawings]

FIG. 1 is a flowchart showing a palm-shaped authentication process according to a first embodiment of the present invention.

FIG. 2 is a flowchart showing a palm-shaped authentication process according to a second embodiment of the present invention.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Yoshimitsu Otani Inventor, Nippon Telegraph and Telephone Corporation 2-3-1, Otemachi, Chiyoda-ku, Tokyo (72) Inventor Yasuhiro Nagai 2--3, Otemachi, Chiyoda-ku, Tokyo No. 1 Nippon Telegraph and Telephone Corporation F term (reference) 5B043 AA01 AA09 BA03 CA10 DA05 EA07 EA08 EA10 EA12 EA13 FA03 FA07 GA05 5J104 AA07 KA01 KA16

Claims (6)

[Claims] 1. A palm shape authentication method for authenticating an individual based on a palm shape of an individual, wherein palm shape image data is detected from a non-fixed palm shape placed in a free orientation, and palm shape image data is detected from the palm shape image data. The rotational angle of the registered palm and the rotational angle of the detected palm shape image obtained by processing the registered palm shape image registered by detecting the palm shape of the person in advance in the same manner are detected. A palm shape authentication method comprising: performing authentication based on an error between the palm shape image that has been corrected for positional deviation and a registered palm shape image based on an error. 2. The rotation angle is detected by performing a two-dimensional Fourier transform on a binarized palm-shaped image to obtain an intensity image centered on a DC component. An image in the form of a rectangular coordinate system with a vertical axis frequency is obtained, and the angle direction perpendicular to the rotation angle at which the maximum intensity cumulative value of the intensity value for each rotation angle of this image is obtained is defined as the palm-shaped rotation angle. The palm shape authentication method according to claim 1, wherein 3. The method according to claim 1, wherein the collation is performed such that the palm end in the rotational angle direction of the detected palm-shaped image is set as the tip of the palm, and the palm width change with respect to the palm tip position or the palm width change is steep. 3. The palm-shaped authentication method according to claim 1, wherein the personal authentication is performed based on a ratio of a distance between points. 4. The method according to claim 1, wherein the palm shape in which the fingers are aligned is a target of authentication, a direction perpendicular to the direction in which the fingers of the palm shape are arranged is detected as a palm rotation angle, and the foremost position of the fingertip portion is determined. The characteristics of the palm width change as a reference, the distance to other fingertips,
The method according to any one of claims 1 to 3, wherein the verification of the person is performed based on the length of the finger / palm, the width of the finger / palm, and the distance between the feature points predetermined in the palm shape. The palm shape authentication method described. 5. The collation using a collation algorithm that is not affected by the line symmetry of the object in the palm shape collation, the distance between the feature points in an arbitrary rotation angle direction component and the finger, The palm shape authentication method according to any one of claims 1 to 4, wherein the personal identification is performed based on the length of the palm and the width of the finger / palm. 6. The palm shape according to claim 1, wherein the collation is performed based on a ratio of a distance between the feature points of the detected palm shape. Authentication method.