JPS5932064A - Extractor for feature of fingerprint - Google Patents

Abstract

PURPOSE:To improve accuracy for collation of fingerprints with an extractor for fingerprint features which uses an optical system and to realize the miniaturization of said extractor as well as an automatic collation, by providing a mechanism which extracts the features of finger prints by means of a Fourier conversion at a minor region of a fingerprint image obtained from a scan of a laser beam. CONSTITUTION:The laser light is irradiated to an optional minor region of the surface where a finger 5 is pressed to a prism 4 from a laser light source 1 through a laser beam control circuit 8. The laser irradiated part is sent to a Fourier conversion lens 6 in the form of a fingerprint image. The result of Fourier conversion is fetched into a detector 7 and then converted into a digital signal by a detector input circuit 9 to be sent to a processing system. The directivity of the minor region of the fingerprint image is extracted by means of the 1st and 2nd peaks of the Fourier conversion surface. Then the out-of-focus degree of the 2nd peak is evaluated to extract the periodic property. The feature quantity is obtained for the input exponent of the finger 5 by integrating two features of the directivity of the minor region of the fingerprint image and the periodic property.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a fingerprint extraction device using an optical system.

Conventionally, devices that perform a Fourier transform similar to the present invention have been put into practical use as devices that perform fingerprint verification using an optical system. This device uses lasers, lenses, etc. to first perform full-screen Fourier transform on an input fingerprint image to create a hologram, and then superimposes the hologram of the fingerprint image in the file obtained using the same procedure and performs inverse Fourier transform. It performs a transformation, determines the optical correlation, and performs fingerprint verification.

This conventional device performs fingerprint verification using full-screen Fourier transform, and although it is capable of roughly classifying the shape of fingerprints, it is not suitable for detailed identification. In other words, is it suitable for extracting fingerprints that are similar in shape to the input fingerprint from among a huge number of files? :The required application pressure is not suitable for the car. Additionally, the need to process the input sample into a film is also considered to be a major impediment to automated fingerprint verification.

This invention has been made in view of the above points, and is characterized by having a mechanism for extracting features of a small region of a fingerprint image obtained by scanning a laser beam. The purpose of the present invention is to realize a Bouet fingerprint feature extraction device suitable for increasing the precision of fingerprint matching, miniaturizing the device, and automating matching. Hereinafter, this invention will be explained with reference to the drawings.

FIG. 1 shows an embodiment of the present invention, in which 1 is a laser light source, 2 is a laser beam focusing lens, 3 is a laser beam control mechanism such as a galvano mirror, 4 is a prism, 5 is a finger, 6 is a Fourier transform lens, γ is a detector (CCJ
), 8 is a laser beam control circuit, 9 is a detector input circuit, 10 is a memory, 11 is an input control circuit, 12 is an output control circuit, 13 is a CPU, and 14 is a common bus.

The apparatus according to the ``%'' embodiment shown in FIG. The laser is irradiated onto an arbitrary small area of the surface of the prism 4 that is pressed against the prism 4.The laser irradiated area is divided into ridges ((Is is dark) based on the principle of total reflection of the prism 4.
The resulting fingerprint image is sent to the Fourier transform lens 6. The result of the Fourier transform performed by the Fourier transform lens 6 is taken into a detector 7 such as a COD image sensor, converted into a digital signal by a detector input circuit 9, and sent to a processing system.

Fourier transform has a shift imparance property in which the result of the transform does not change even if the region to be Fourier transformed is moved in parallel, so there is no need to move the Fourier transform lens 6, detector γ, etc. with the laser beam scan. .

In this way, the Fourier transform results of a small region of a fingerprint irradiated with a laser can be loaded onto the processing system, but in the process of processing, more detailed information about the regions near a certain small region is needed. If the laser beam control mechanism 3. By controlling the laser beam from the laser light source 1 by the laser beam control circuit 8, it is also possible to intensively scan the vicinity. In addition, when the laser beam is irradiated onto the 50 fingerprint images, the appropriate beam diameter is about 0.5n to 5nm, but this varies depending on the type of processing system.
It can be set either circular or square.

The operating principle of this embodiment has been described above. Next, an example of processing for the Fourier transform plane taken into the processing system by the detector T and the detector input circuit 9 will be described.

FIG. 2 is a gradation representation of the data on the Fourier transform plane that is taken into the processing system by the detector 7 and the detector input circuit 9. 15 in Figure 2 is the first peak, which is the largest peak at wave number 0, and 16 in Figure 2.
is the second peak, which is the next largest peak. The original fingerprint image from which the Fourier transform plane of FIG. 2 is obtained has a shape as shown in FIG. 3, but due to the nature of Fourier transform, the first peak 15 and the second peak 16 are connected in The diagonal line 1T is approximately normal to the ridge line 18 of the fingerprint image shown in FIG. That is, by having a mechanism & for recognizing the positions of the first beak 15 and the second beak 16 on the Fourier transform surface, the directionality of the original fingerprint image small area can be calculated. The positions of the first beak 15 and the second beak 16 can be easily realized as a digital circuit by using a COD image sensor as the detector 7.

As mentioned above, by using the first peak 15 and the second peak 16 of the Fourier transform plane, it is possible to extract the directionality of a fingerprint image small region, but some fingerprint image small regions have clear directionality. If so, there are some that are unclear, so the first
A means is required to assist in extracting the most dominant direction using only the beak 15 and the second peak 16.

An example of this is given below.

As shown in FIG. 2, the Fourier transform surface is symmetrical about the first peak point. Therefore, as shown in FIG. 4, for a region 20 on one side of the Fourier transform plane divided by a straight line 19 passing through the first beak 15 and perpendicular to the straight line 1T connecting the first beak 15 and the second beak 16, the second
The degree of periodicity is evaluated by the sum of the square of the vector from the beak 16 to each point in the region 20 multiplied by the size of that point. That is, P=Σrl(rl−r.)2・・・・・・・・・・・・
...(1) ro is from the first beak 15 to the second beak 16
The vector leading to , 1 line is from the first beak 15 to the area 20
is a vector that leads to each point within. The lid is a rosette that defines each point within the region 20, and defines only the points within the region shown at 21 in FIG. Although this area 21 is determined empirically, it is effective to make it approximately a rectangular area in order to make it suitable for digital processing. fm (The larger the value of P determined by Equation 11, the more non-periodic the small region of the fingerprint image becomes, and the smaller the value, the more pronounced the monoperiodicity becomes.The function to evaluate the degree of blur of this second peak 16 is Alternatively, by implementing it with simple hardware and adding it, it can be used as a feature amount of a fingerprint image small area together with the directionality described above.

The two features of the directionality and the degree of periodicity of the fingerprint image small area described above are integrated as the feature amount of the 50-finger input fingerprint,
It performs matching of the fingerprint characteristics written in the card information input from the input control circuit 11, accesses the output control circuit 12, and performs operations depending on whether or not matching is possible. For example, in an access control system, this is the 01 of the door.
1 - Corresponds to the closing operation.

The above input control, matching, and output control processing is performed by the CP
This is realized by U13 communicating with each control circuit via the common bus 14 and communicating with the memory 10.

As for the prism 4, as shown in FIG. 5, if a coating 22 is applied to one surface of the prism 4 to block the incidence of light and the detector 7 and the laser beam are placed on the same side, the fingerprint prism can be used. Due to the difference in the paths of light from the contact part 24 and the non-contact part 23, dark fingerprint ridges are obtained due to the total internal reflection principle shown in FIG. 1, whereas medium-pressure fingerprint ridges on a dark background are bright. You get what you get. In this case, an image with a higher signal-to-noise ratio can be obtained, but the control of the laser beam control mechanism 3 becomes somewhat more complicated. Other principles are similar.

In addition, a function is added to control the laser beam diameter by varying the position of the laser beam converging lens 2. In addition, the directionality of the ridge 18 as well as the end point of the ridge 18 can be adjusted.

It is also possible to find feature points such as branch points.

As described above, the fingerprint feature extraction device of the present invention uses an optical system to collect a fingerprint image and extract features from a small area of the fingerprint image, so that verification can be speeded up and automated. Furthermore, since the Fourier transform of a small region of the fingerprint image is used as a feature, the K<< preprocessing is reduced due to the influence of noise. In addition, by changing the laser beam scanning method and laser beam diameter, it is possible to extract a wider variety of features and achieve higher accuracy in matching.Also, it is possible to flexibly respond to changes in matching software. In addition, it has the advantage that it can be implemented on a small device scale.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of the present invention, Fig. 2 is a diagram showing the shading of the Fourier transform surface data taken in from the detector, and Fig. 3 shows the original image that produces the Fourier transform surface shown in Fig. 1. FIG. 4 is a diagram showing the applicable range of equation (1), and FIG. 5 is a diagram of a prism input system that provides a high S/N ratio. In the figure, 1 is a laser light source, 2 is a lens for focusing one section of laser light,
3 is a laser beam control mechanism, 4 and 1 prism, 5 are fingers,
6 is a Fourier transform lens, T is 1 day tech, 8 is a laser beam control circuit, 9 is a detector input circuit, 10 is a memory, 11 is an input control circuit, 12 is an output control circuit, 1
3 is CPU, 14)! Bud (pass bus, 15 is the first peak,
16 is the second peak, 17° 19 is a straight line, 18 is a ridge, 2
0,21)-! , the territory area 22 is a court, 23 is a non-contact part, 241'! , is the contact part. Figure 1 5 Figure 4 Port 0 Figure 5 3

Claims (2)

[Claims] (1) A mechanism for scanning a laser beam, a fingerprint image extraction optical system for extracting a fingerprint image for each small area irradiated by the laser beam, and an optical fingerprint image obtained by the fingerprint image extraction optical system. A patent characterized in that it is equipped with a Fourier transform optical system that performs a Fourier transform of the image, and a conversion mechanism that converts the Fourier image obtained by the Fourier transform optical system into an image. (2) The converter M has a mechanism that recognizes the positional relationship between the first peak, which is the largest peak, and the second peak, which is the next largest peak, based on the wave number O, from the Fourier image, and the same mechanism as the second peak. Claim (1) comprises a mechanism for calculating the degree of dispersion of Fourier components having a wave number of a magnitude, and a mechanism for extracting features of a fingerprint image from the positional relationship and the degree of dispersion. fingerprint feature extraction device.