JPH052635A - Individual identification device - Google Patents

Abstract

PURPOSE:To obtain superior identification precision in a small scale by executing pattern collation between the spatial spectrum of a fingerprint obtained from a feature extraction means and the registered spatial spectrum of a registrant, and judging whether the registrant and a person being the object of identification are the same or not. CONSTITUTION:A fingerprint picture input means 1 fetches the fingerprint of the specified finger of the person being the object of identification as picture data in terms of optoelectronics. The feature extraction means 2 two-dimensionally Fourier- transforms fetched picture data, and obtains the spatial spectrum which can simultaneously observe frequency components in the vertical direction/lateral direction of the fingerprint. A judgment means 3 executes pattern collation between the obtained spatial spectrum and the registered spatial spectrum of the registrant, which is previously and similarly processed, and judges whether the registrant is the same as the person being the object of identification. When the judgment means 3 judges that they are the same, an output means 4 transmits an unlock signal to a door lock device. Thus, a system can be made into the small scale since the data size of the spatial spectrum becomes considerably smaller than the data size of fetched picture data.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for identifying an individual by fingerprint collation.

[0002]

The following are known conventional techniques. A method for collating the positional relationship between fingerprint feature points (end points of fingerprint ridges, branch points, etc.) as individual feature data (prior art 1). A method in which the number of ridges between characteristic points is collated as individual characteristic data (prior art 2). A method for performing pattern matching between original images (prior art 3).

[0003]

However, the above-mentioned conventional techniques have the following drawbacks. The prior arts 1 and 2 require a large-scale system in terms of knowledge database search and processing time, which is costly. Prior art 3
Requires a small system, but the matching accuracy is poor. An object of the present invention is to provide a personal identification device that is small in scale and excellent in identification accuracy.

[0004]

[Means for Solving the Problems] In order to solve the above problems,
The present invention has the following configurations. (1) A fingerprint image input unit that captures a fingerprint of a specific finger of a person to be identified as image data by optoelectronics, a feature extraction unit that performs a two-dimensional Fourier transform of the captured image data to obtain a spatial spectrum, and the feature extraction. The spatial spectrum obtained by the means and the registered spatial spectrum of the registrant, which has been subjected to similar processing in advance, are pattern-matched to determine whether or not the registrant and the identification target person are the same. I will do it. (2) A fingerprint image input unit that captures the fingerprint of the specific finger of the identification target person as image data by optoelectronics, a feature extraction unit that performs a two-dimensional Fourier transform on the captured image data, and obtains a spatial spectrum, and the feature extraction. A masking means for removing a predetermined frequency region from the spatial spectrum obtained by the means, and a pattern matching between the residual spatial spectrum and the registered residual spatial spectrum of the registrant that has been subjected to similar processing in advance are performed to identify the registrant. And a determination means for determining whether or not the target person is the same.

[0005]

[Operation and effect of the invention]

(Claim 1) The fingerprint image input means captures the fingerprint of the specific finger of the person to be identified as image data in optoelectronics. The feature extracting means performs a two-dimensional Fourier transform on the captured image data to obtain a spatial spectrum in which the vertical and horizontal frequency components of the fingerprint can be simultaneously observed. In addition,
The data size is significantly reduced due to the two-dimensional Fourier transform. The determination means performs pattern matching between the obtained spatial spectrum and the registered spatial spectrum of the registrant that has been subjected to similar processing in advance, and determines whether the registrant and the identification target person are the same. Since the data size of the spatial spectrum is significantly smaller than the data size of the captured image data, the system can be downsized. It is configured to determine whether or not the registrant and the identification target person are the same by performing pattern matching between the spatial spectrum obtained by performing the two-dimensional Fourier transform on the image data and the registered spatial spectrum of the registrant that has been similarly processed in advance. Because of this, it has excellent identification accuracy.

(Claim 2) The fingerprint image input means captures the fingerprint of the specific finger of the person to be identified as image data in optoelectronics. The feature extracting means performs a two-dimensional Fourier transform on the captured image data to obtain a spatial spectrum in which the vertical and horizontal frequency components of the fingerprint can be simultaneously observed. Note that the data size is significantly reduced due to the two-dimensional Fourier transform. The masking means removes from the spatial spectrum a predetermined frequency region unnecessary for identifying fingerprint features. This removal further reduces the data size. The determination means performs pattern matching between the residual spatial spectrum and the registered residual spatial spectrum of the registrant that has been subjected to similar processing in advance, and determines whether the registrant and the identification target person are the same. Since the data size of the residual spatial spectrum is significantly smaller than the data size of the captured image data, the system can be downsized. The residual spatial spectrum obtained by performing two-dimensional Fourier transform on the image data and removing the predetermined frequency region and the registered residual spatial spectrum of the registrant, which has been previously processed in the same manner, are subjected to pattern matching to identify the registrant and the identification target person. Since it is configured to check whether or not it is excellent in identification accuracy.

[0007]

Embodiments of the present invention will be described with reference to FIGS. As shown in the figure, the personal identification device A includes a fingerprint image input unit 1 that captures a fingerprint 12 of a right index finger 11 as image data in optoelectronics, a feature extraction unit 2 that obtains a spatial spectrum from the captured image data, Determination means 3 for determining whether the registrant and the identification target person are the same;
When the determination means 3 determines that they are the same, the output means 4 is provided for sending an unlocking signal (neither is shown) to the door lock device.

As shown in FIG. 2, the fingerprint image input means 1 includes a triangular prism 13 in which the fingerprint surface of the right index finger 11 is pressed against the slope 131, and the illumination light 14 perpendicular to the end surface 132.
Light source 15 for projecting light and reflected light 1 reflected on the slope 131
A known CCD device 17 is arranged so that 6 can be captured vertically, and the reflected light 16 corresponding to the ridges of the fingerprint ridge is captured as a dark image and the reflected light 16 corresponding to the valleys of the fingerprint ridge is captured as a bright image.
Composed of and.

As shown in FIG. 3, the feature extracting means 2 is C
Thinning process 2 for erasing the ridge thickness information of the fingerprint from the binary two-dimensional fingerprint data (image data) transferred by the CD element 16.
1. Fingerprint center position extraction 22 for finding the fingerprint center point M of the thinned binary two-dimensional fingerprint data 20. This fingerprint center point M
From the determination data acquisition 23 that captures a fixed distance range 201 (for example, 8 mm × 8 mm) as the determination data 202, and the two-dimensional Fourier transform 24 that obtains the two-dimensional Fourier transform image 241 (spatial spectrum) from the determination data 202. It has a process.

In the present embodiment, the thinning process 21 uses a 3 × 3 mesh mask (not shown) as a starting point from the upper left end point of the binary two-dimensional fingerprint data 20 from left to right and from top to bottom. The parallel processing method of eight concatenations is used, in which whether or not the pixel in the center of the mask is to be deleted is determined from the values of eight neighboring pixels of the pixel. A method using differentiation may be used.

The fingerprint center position extraction 22 is a thinning process 21.
A histogram (not shown) of black dots in the horizontal direction and the vertical direction of the binary two-dimensional fingerprint data 20 obtained by performing the above is created, and the coordinate point forming the apex of the histogram in each direction is defined as the fingerprint center point M. In addition, in order to obtain the fingerprint center point more accurately, the fingerprint center point M ′ is extracted by the same method from the data obtained by polar coordinate conversion (for example, 45 °) of the thinned two-dimensional two-dimensional fingerprint data 20 and the polar coordinate conversion is not performed. The center point of the fingerprint center point M obtained from the data may be used as the fingerprint center point.

The two-dimensional Fourier transform 24 obtains a two-dimensional Fourier transform image 241 (spatial spectrum) shown in FIG. 5 by performing a two-dimensional Fourier transform on the judgment data 202. The data size of the two-dimensional Fourier transform image 241 is ¼ of the data size of the binary two-dimensional fingerprint data 20.
Becomes After obtaining the two-dimensional Fourier transform image 241, the masking means 25 removes the high-frequency component 242 other than the frequency component of the fingerprint ridge to obtain the remaining two-dimensional Fourier transform image 243 (residual spatial spectrum). Is also good. By doing so, the data size can be reduced to about ⅕ of the binary two-dimensional fingerprint data 20.

The judging means 3 is a two-dimensional Fourier transform image 24.
1 (or the remaining two-dimensional Fourier transform image 243) and the registered two-dimensional Fourier transform image (or the registered remaining two-dimensional Fourier transform image) of the registrant, which has been subjected to the same process in advance, are pattern-matched, and the matching result is predetermined. If it exceeds the threshold value (for example, 95%), it is determined that the identification target person is the same as the registrant. Since the pattern matching of the two-dimensional Fourier transform image 241 and the registered two-dimensional Fourier transform image is performed, it is not affected by the positional deviation or the rotational deviation of the finger 11. Also, the masking means 2
5, the high frequency component 242 other than the frequency component of the fingerprint ridge
When is removed, the noise component can be remarkably reduced, so that the determination accuracy is further improved.

[Brief description of drawings]

FIG. 1 is a block diagram of a personal identification device according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of fingerprint image input means in the apparatus.

FIG. 3 is a block diagram of feature extraction means in the apparatus.

FIG. 4 is an explanatory diagram illustrating binary two-dimensional fingerprint data, a fingerprint center point, and determination data.

FIG. 5 is an explanatory diagram illustrating a two-dimensional Fourier transform image and a high frequency component.

[Explanation of symbols]

A personal identification device 1 fingerprint image input means 2 feature extraction means 3 determination means 4 output means 11 right index finger (specific finger) 12 fingerprint 25 masking means 241 two-dimensional Fourier transform image (spatial spectrum) 242 high frequency component (predetermined frequency domain) 243 Remaining two-dimensional Fourier transform image (remaining spatial spectrum)

Claims (2)

[Claims] 1. A fingerprint image input means for optoelectronically capturing a fingerprint of a specific finger of a person to be identified as image data, and a feature extracting means for performing a two-dimensional Fourier transform on the captured image data to obtain a spatial spectrum. Performs pattern matching between the spatial spectrum obtained by the feature extraction means and the registered spatial spectrum of the registrant that has been subjected to similar processing in advance,
A personal identification device comprising: a determination unit that determines whether or not the registrant and the identification target person are the same. 2. A fingerprint image input means for optoelectronically capturing a fingerprint of a specific finger of a person to be identified as image data, a feature extracting means for performing a two-dimensional Fourier transform on the captured image data to obtain a spatial spectrum, Masking means for removing a predetermined frequency region from the spatial spectrum obtained by the feature extraction means, and pattern matching between the residual spatial spectrum and the registered residual spatial spectrum of the registrant that has been subjected to similar processing in advance, and the registrant And a determination means for determining whether or not the identification target person is the same.