JPS63236173A - Feature extraction device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to fingerprint matching, and is particularly used in fingerprint matching processing for personal identification by determining the identity of input fingerprints and registered fingerprints based on fingerprint pattern characteristics. , relates to a feature extraction device that creates a feature file that stores features through man-machine interaction.

(Conventional technology and its problems) Systems that use fingerprints to identify individuals are extremely reliable because fingerprints have the characteristics of being "unique for everyone" and "unchangeable throughout life." A device for automatically verifying the identity of a collected fingerprint and a pre-registered registered fingerprint has already been disclosed in Japanese Patent Application No. 54-39648.
Many proposals have been made, including Japanese Patent Application No. 57-111114.

On the other hand, as the information society has progressed in recent years, the need for personal identification when accessing computer terminals and databases and managing entrance to sensitive areas is increasing. Currently, many of these transactions are carried out using pin numbers or ID cards, but these can easily be forgotten or stolen.

If it is lost, it loses its effectiveness and its reliability must be said to be low.Therefore, great expectations have been placed on the realization of a personal identification device using fingerprint matching technology. This personal identification device requires a feature file that stores features used to compare input fingerprints with registered fingerprints. Feature files are conventionally used to perform shading correction on registered shading fingerprint images,
After performing various image processing such as shading enhancement and core line conversion, feature points are extracted, and their positions, directions, and the number of ridges existing between neighboring feature points are determined and created. However, conventional feature extraction devices that create feature files in this manner have been large-scale and expensive, comparable to mini-computer systems, due to the enormous amount of processing and complexity.

As mentioned above, existing automatic feature extraction devices have problems with the scale and price of the device itself. SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to eliminate the drawbacks of conventional large-scale and expensive feature extraction devices, and to provide a feature extraction device that is simpler and cheaper than conventional ones. Furthermore, in order to prevent illegal intrusion into sensitive areas using forged fingerprints and illegal access to important files, it is also necessary to prevent the registration of forged fingerprints other than human fingers.

(Means for Solving the Problems) In order to solve the above-mentioned problems, the present invention provides means for determining whether the input fingerprint image and the registered fingerprint image represent the same fingerprint. A feature extraction device that extracts and registers pattern features used for determining the pattern characteristics of the two fingerprint images through man-machine interaction and creates a feature file of the fingerprint, the device comprising: an optical image of the fingerprint; a fingerprint image input unit that inputs an optical image and converts the optical image into an electrical signal to generate an analog fingerprint image of the electrical signal; and A/A that converts the analog fingerprint image into a two-dimensional quantized image.
A D conversion unit detects and stores the gray scale value of a pixel at a predetermined position in the two-dimensional quantized image, checks that the fingerprint placement is completed based on a change in the gray value, and outputs the test result. Detects the time when the fingerprint placement inspection unit receives the placement completion detection output of the TL and W inspection unit, and the grayscale value of the pixel at the predetermined position in the image captured within a certain period of time after this time. and a forgery input determination section that compares the density values stored in the fingerprint placement inspection section and determines whether or not a normal fingerprint has been placed; an input image collection control unit that collects the image input by the fingerprint image input unit as an input image when the grayscale value of a pixel at the predetermined position or at a new position becomes equal to or higher than a predetermined threshold; an image storage section that stores the two-dimensional quantized image of the input image collected by the collection control section; an image output section that displays the two-dimensional quantized image stored in the image storage section; and the image storage section. a pointing device for inputting a point indicating the position and pattern direction of a feature point included in the fingerprint of the two-dimensional quantized image stored in the section; and determining four feature points nearest to the feature point;
a feature extraction unit that calculates the number of ridges existing between each feature point; and a feature point extraction unit that calculates the number of ridges existing between each feature point, and the position, pattern direction, and distance of the feature point extracted from the fingerprint of the two-dimensional quantized image stored in the image storage unit; a file storing a feature list in which features are registered including at least the number of ridges existing in the area and the distinction between end points and branch points; and an input section for inputting confirmation information for performing man-machine interaction when creating the feature list. , an output unit that outputs the feature extraction and feature file creation process, and a point indicating the position and direction of the feature point on the fingerprint image collected by the fingerprint image input unit is displayed on the image output unit by the user. When the position of the feature point on the binary image determined visually on the gray scale image and displayed on the output unit by the pointing device is input, the feature extraction unit determines the position of the feature point. and the direction thereof, and create a feature file of the fingerprint.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing one embodiment of the present invention. This embodiment is a feature extraction device that creates a feature file that stores features included in an input grayscale image through man-machine interaction. This embodiment includes a fingerprint image input unit 100 that collects a grayscale image of a placed fingerprint and photoelectrically converts it, an A/D converter 101 that converts the photoelectric conversion signal into two-dimensional quantized image data, and A fingerprint placement inspection unit 102 detects and stores the gray scale value of a pixel at a predetermined position of the image sensor, and detects and outputs that the fingerprint has been placed on the transparent body based on the gray scale change.
and the time after receiving the inspection output of the fingerprint placement inspection unit 102, and after this time, detect the grayscale value of the pixel at the predetermined position in the image captured within a certain period of time, and a forgery input determination unit 103 that compares the density value and determines whether a normal fingerprint and not a forgery is placed;
an image acquisition control unit 104 that acquires an input image when the gray value of a pixel at a predetermined position or a new position after the forged input determination unit 103 determines that the input is normal; An image storage unit 105 that stores quantized image data, an image output unit 115 that displays two-dimensional quantized image data, a feature extraction unit 110, and a file 11.
1, an output section 112, an input section 113, and a pointing device 114.

The feature extraction unit 110 connects to the external interface 106. Central processing unit 107. It consists of a program memory 108 and a working memory 109. Pointing device 1
The coordinate values on the output unit 112 that are pointed from 14 and various information input from the input unit 113 are received by the central processing unit 107 via the external interface 106. A program memory 108 stores a program, and the central processing unit 107 operates according to the program.A working memory 109 temporarily stores coordinate values of the minutiae and the like when creating a feature file. The feature file created here is assigned an ID code input from the input unit 113 and is stored in the file 111.

The fingerprint image input unit 100 uses an optical boundary condition caused by light from a light source to perform ITV (industrial telephony) on a finger placed on a transparent body.
A device that inputs a photoelectrically converted image of a fingerprint pattern stamp using an imaging device such as JP-A-54-6
9300 and Japanese Patent Laid-Open No. 54-85600 are available and are used.

Fingerprint placement inspection section 102. A section consisting of a forgery input determination unit 103 and an image acquisition control unit 104 inspects the completion of fingerprint imprinting based on changes in the density values of pixels at predetermined positions in images repeatedly captured from the ITV, and determines the time after receiving this inspection output. And after this time, the gray value of the pixel at the predetermined position in the image captured within a certain period of time is compared with the stored gray value to determine whether a normal fingerprint has been input, and after it is determined that the input is normal. When the density value of the pixel at the predetermined position yL or at a new position exceeds a predetermined threshold value, an input image is collected. be done. The multivalued two-dimensional quantized image thus obtained is stored in the image storage unit 105.
is memorized.

FIGS. 2(a), (b), and (c) show the fingerprint button, minutiae list, and direction including minutiae (hereinafter referred to as minutiae) in this embodiment.
) to (C), the characteristics of the fingerprint are the position (X, Y) of the minutiae M, the direction of the pattern, and the adhesion ratio with the neighboring minutiae rl+r21 r31 ! ``In units of 4,
-Multiple minutiae M+ (t=
1.2.・、・.

n) is expressed in a list form.Detailed definitions of these are described in Japanese Patent Application Laid-Open No. 138174/1982, and how to find the relations is described in Japanese Patent Application No. 259/1983.
989, detailed description thereof will be omitted here.

FIG. 3 shows a flowchart of creating a mechanical list in this embodiment. In FIG. 3, steps S1 and S2 are shown.
After inspecting the placement of the fingerprint using the method described above, it is determined whether or not it is a forged input, and in step S3, a re-stamping is requested in accordance with the determination result. Here, the number of times of re-stamping is continuous,
If the predetermined number of times is exceeded, the entire process is stopped. In step S4, fingerprint sampling is controlled by the method described above, and in step S5, a re-stamping is requested in accordance with the result. Here too, if the number of times of re-stamping continues and exceeds a predetermined number of times, the entire process is terminated. In step s6, a binary image based on the threshold value used at the time of image acquisition is displayed on the output unit 112. Here, the grayscale fingerprint image stored in the image storage section 105 is displayed on the image output section 115. In step s6, the user searches for a minutiae in the grayscale fingerprint image displayed on the image output section 115, and uses the binary image displayed on the output section 112 to search for the minutiae on the pointing device 11.
4, point two points indicating the minutiae and its direction, where the coordinates of minutiae M1 @
! (X+, Y+) is temporarily stored in working memory 109. The output unit 112 displays a binary image of the gray scale image in the image storage unit 105 using a predetermined threshold value, but in the binary image, breaks in ridges may occur due to sweat pores, etc., and the presence of minutiae may occur. becomes unstable, and the end point /
Since the distinction between branch points is also unstable, the image output unit 115
Minutiae extraction is more stable when the minutiae is selected on the grayscale image displayed. In step S7, a point representing the direction of the minutiae is pointed, and the coordinate value (X
,,y,) and the point (Xnt, Y+o) representing the direction, and the angle formed by the X-axis is calculated. In step S8, the feature extraction unit 110 determines whether the end point of the minutiae described above is a branching point based on the grayscale values of pixels of a plurality of neighboring points of the point pointed in step S6, and the output unit 112
The user checks the output results. The distinction between endpoints and branching points is made in the working memory 109 according to the format shown in Figure 2(b).
Memorized temporarily. In step S9, the user inputs an end signal from the input unit 113 or pointing device 114 indicating that the pointing of all minutiae has been completed, and the process proceeds to step S6. Otherwise, the same pointing is repeated. Here, for the minutia that has already been pointed, a line segment connecting the point indicating the direction and the center is displayed on the binary image displayed on the output unit 112, except for a circle centered at the center position. By doing this, you can know the progress of pointing. After the pointing of all minutiae is completed, each minutiae is set as the center coordinate, and the local a!
form a standard system, and the most recent example ration r in each quadrant of it
1 + r2 + r3 Hr4 in the above-mentioned patent application 1986-2
It is determined by the method described in No. 59989. Note that if the nearest neighbor minutiae does not exist, a predetermined constant (for example, 255
) as a relation.

In step Sho, relations are found for all minutiae, and a minutiae network is constructed. In step S11, first, the relationship between a certain minutiae and the nearest neighboring minutiae in the first quadrant is displayed, and the user checks the relationship on the grayscale image displayed on the image output section 115 to calculate the correct value. If so, move on to the next relation check. if,
If an incorrect value has been calculated, enter the correct value in input section 1.
13, and after making corrections, move on to the next relation confirmation. Confirmed relations are stored in the minutiae list as needed. When the end of relation confirmation for all minutiae is detected in step S12,
After assigning the ID code input from the input unit 113 to the manual list created in step S13, the creation of the manual list is completed by storing it in the file 111.

Note that in this embodiment, the fingerprint image stored in the image storage unit 105 is a multivalued image, but by a known method,
Even if a binarized binary image is stored and processed, the essence of the processing remains unchanged.

(Effects of the Invention) As explained above, the present invention inspects whether the input grayscale fingerprint image is a forged fingerprint, and if the input is normal, the position and direction of the minutiae are displayed on the binary image. Provided is a feature extraction device that calculates the direction and relation of the minutiae without performing various image processing by indicating a point indicating the minutiae with a pointing device, and allows a user to sequentially check them and store them in a minutiae list. As described above, this apparatus does not require various complicated image processing, so it can be implemented on a personal computer system and can be implemented simply and at low cost.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing one embodiment of the present invention, and FIG. 2 (
a), (b), and (c) are diagrams showing the concepts of minutiae, minutiarist, and direction, respectively, in this embodiment, and FIG. 3 is a flowchart showing the processing procedure in this embodiment. 100...Fingerprint image input section, 101...A/D2f
Part i, 102... Fingerprint placement inspection unit, 103... Forgery input determination unit, 104... Image collection control unit, 105...
- Image storage unit, 106...external interface, 10
7... Central processing unit, 108... Program memory, 109... Working memory, 110... Feature extraction unit,
111...File, 112...Output section, 113.
... Input section, 114 ... Pointing device, 1
15... Image output section, 201... End point, 202...
- Branching point, 203...Relation, 204...Direction.

Claims (1)

[Scope of Claims] A pattern feature used to determine whether an input fingerprint image and a registered fingerprint image represent the same fingerprint based on the pattern features of the two fingerprint images, A feature extraction device that extracts and registers the fingerprint through man-machine interaction and creates a feature file of the fingerprint receives an optical image of the fingerprint, converts the optical image into an electrical signal, and generates an analog fingerprint image of the electrical signal. a fingerprint image input unit; and A converting the analog fingerprint image into a two-dimensional quantized image.
/D conversion unit; detects and stores the gray scale value of a pixel at a predetermined position in the two-dimensional quantized image, checks whether the fingerprint placement is completed based on a change in the gray value, and outputs the result of the test; A fingerprint placement inspection unit that outputs a fingerprint placement inspection unit, and detects the time when the placement completion detection output of this placement inspection unit is received and the grayscale value of the pixel at the predetermined position in the image captured within a certain period of time after this time. and a forgery input determination unit that compares the density values stored in the fingerprint placement inspection unit and determines whether or not a normal fingerprint has been placed; an input image collection control unit that collects the image input in the fingerprint image input unit as an input image when the grayscale value of a pixel at the subsequent predetermined position or a new position becomes equal to or greater than a predetermined threshold; and this input image collection. an image storage unit that stores the two-dimensional quantized image of the input image collected by the control unit; an image output unit that displays the two-dimensional quantized image stored in the image storage unit; and an image storage unit a pointing device for inputting points indicating the position and pattern direction of the minutiae included in the fingerprint of the two-dimensional quantized image stored in the fingerprint; a feature extraction unit that calculates the number of existing ridges, and the positions and pattern directions of minutiae extracted from the fingerprint of the two-dimensional quantized image stored in the image storage unit, and ridges existing between neighboring minutiae points. A file that stores a feature list in which features are registered, including at least a number and a distinction between an end point or a branch point;
It consists of an input section for inputting confirmation information for man-machine interaction when creating the feature list, and an output section for outputting the feature extraction and feature file creation process, The user visually determines a point indicating the position and direction of the feature point on the grayscale image displayed on the image output section, and uses the pointing device to determine the point indicating the position and direction of the feature point on the binary image displayed on the output section. A feature extraction device characterized in that when the position of the point is input, the feature extraction unit automatically calculates the position and direction of the feature point and creates a feature file of the fingerprint.