JPH0424881A - Fingerprint picture registering system - Google Patents

Abstract

PURPOSE:To improve fingerprint collation rate by excluding the small part of low quality such as a bridge generated due to sweat in a fingerprint ridgeline part in a fingerprint picture from the selected object of dictionary data when a fingerprint dictionary is generated. CONSTITUTION:A fingerprint input part 1 receives the ridgeline part of a fingerprint by a CCD element 45 through a glass plate 43 and a lens 44, and outputs analog data. A picture input part 2 is provided with a binarization circuit and a binary memory, and binarizes and stores the analog data of the fingerprint in the memory. Then, a collation circuit part 3 is provided with a collation circuit and a dictionary memory, and stores the fingerprint registered before in the dictionary memory, and the collation circuit compares and collates inputted binary data and the contents of the dictionary memory with each other. At that time, the fingerprint picture detected by a fingerprint sensor is divided into plural small areas, and when the small area containing a feature point and brighter than a prescribed threshold is detected in the fingerprint picture, the picture data of the bright small area is excluded from the generation of the fingerprint dictionary. Thus, the collation rate can be improved.

Description

[Detailed Description of the Invention] [Summary] Regarding a fingerprint image registration method when registering a finger print, the purpose of this invention is to prevent erroneous matching in fingerprint registration and improve the matching rate. A fingerprint matching device having a feature extraction unit that extracts feature points of an image, an image input unit that binarizes analog data to the fingerprint sensor using a predetermined threshold, and a dictionary memory that stores a fingerprint dictionary of registered fingerprints. In the fingerprint registration and verification device, the fingerprint image detected by the fingerprint sensor is divided into a plurality of small areas, and when an image area having a certain area or more and brighter than a predetermined threshold is detected in the fingerprint image, The image data of the bright image area is excluded from the creation of the fingerprint dictionary to prevent pseudo-minutiae points that are erroneously regarded as minutiae points from being selected as minutiae points in the fingerprint dictionary. .

[Industrial application field]

The present invention is a fingerprint image registration method for registering finger prints. Fingerprint verification is widely used as one of the personal identification methods. Fingerprint verification systems are often used to control access to important facilities. Usually, a fingerprint matching system treats a fingerprint as an image, and therefore is comprised of an input device that converts a fingerprint into image data, and a fingerprint matching device that matches registered fingerprint data with input fingerprint data.

On the other hand, when registering a fingerprint image, continuous fingerprint ridges may be formed depending on the condition of the finger skin, but it is important to avoid erroneously regarding these as feature points of the fingerprint.

[Conventional technology]

The fingerprint registration and verification device has a dictionary memory that stores a fingerprint dictionary that records fingerprint minutiae data of each individual, and usually combines the stored minutiae data with minutiae points such as branching points and endpoints of the input fingerprint. The distribution (coordinate position) and its shape are compared, and the fingerprints are compared to see if they match.

FIG. 4.5 is a diagram showing the main part of the fingerprint sensor.

Normally, a fingerprint image human power device (fingerprint sensor) detects the uneven pattern of a fingerprint by placing a finger 41 on a glass plate 43, as shown in the figure.When the finger 41 is pressed against one surface of the glass plate 43, the protrusions are They touch, but the recesses do not.

When light is irradiated onto the plane against which the finger is pressed through the glass plate 43, the light is reflected and scattered on the surface and inside of the finger. Since the scattered light from the concave portion of the finger passes through the air once and enters the glass plate 43, there is no component that is totally reflected within the glass plate 43 and propagated.

On the other hand, the reflected and scattered light from the convex portion is transmitted directly from the finger to the glass plate 43.
is incident on the glass plate 4 as a spherical wave, reflected by the mirror 42, and a part of it satisfies the condition of total reflection in the glass plate 4.
3, propagates through repeated total reflections.

When this total reflection component is imaged by an optical system such as a suitable lens 44, a ridge pattern of convex portions can be obtained on the light receiving surface of the image pickup device (for example, a CCD device) 45. However, since the reflected and scattered light from this convex part is a spherical wave, if a single spherical lens is used to form an image with little aberration, the fifth
As shown in the figure, an aperture stop 46 placed at the center of curvature of the convex lens 46 is required.

FIG. 6 is an explanatory diagram of local region division of a fingerprint image. As shown, fingerprint image data is divided into several areas (e.g.
The average density value is determined for each divided area.

FIG. 7 is an explanatory diagram of determination of minutiae, and numerals 1 to 6 are windows of minutiae of the fingerprint. For these locations, the branches of the fingerprint are used as shown. This is because branch parts are easy to capture as feature points. The central window is the central window for matching.

When comparing the fingerprint registered in the dictionary memory with the input fingerprint, these windows and the registered pattern are compared, but first the central window is aligned and the minutiae patterns of multi-windows 1 to 6 are registered. Compare with the pattern of minutiae points.

FIG. 8 is an explanatory diagram of a low-quality portion of a fingerprint image, that is, a portion bridged by sweat. When oil, sweat, etc. adhere to a finger, the ridges of the fingerprint in that area are bridged, as shown by the arrow in the figure.

There is a risk that the feature extraction unit may consider such bridges between ridges to be minutiae points of the fingerprint.

FIG. 9 is a processing flowchart of a conventional fingerprint registration method. First, when fingerprint image data is manually input by the fingerprint sensor shown in FIG. 4.5, in step 1), the image is binarized by slicing it at a predetermined threshold value (step 2). Then, the feature points 1 to 6 and the central feature point as shown in FIG. 7 are extracted (step 3), and these feature points are registered in the dictionary memory (step 4). If some of the feature points in Figure 7 are unclear and cannot be registered, the number of registered feature points will be less than the predetermined number N (7 locations in this case), so a new location on the side will be extracted again. Register the feature points of the location (Step 5) [Problem to be solved by the invention] As mentioned above, since the fingerprint input device uses optical contact between the finger and the glass plate, sweat from the finger A fingerprint image from a portion optically bridged by the above method is a fingerprint image in which there are many bridges of fingerprint ridges.

Therefore, if such defective parts of a fingerprint are mistakenly regarded as features and feature extraction is performed, a fingerprint dictionary containing a large number of such pseudo minutiae points will be created. If verification is performed using the fingerprint dictionary created in this manner, the number of incorrect verifications will increase and the verification rate will decrease.

An object of the present invention is to prevent low-quality areas in a fingerprint image that have many consecutive fingerprint ridges from being mistakenly regarded as minutiae when creating a fingerprint dictionary, and to exclude them from being selected as minutiae dictionary data. By doing so, the purpose is to prevent false minutiae from being included in the fingerprint dictionary, thereby preventing an increase in erroneous verifications and improving the verification rate.

[Means to solve the problem]

The present invention includes a fingerprint sensor, a feature extraction unit that extracts feature points of a fingerprint image, an image input unit that binarizes analog data to the fingerprint sensor using a predetermined threshold, and a fingerprint dictionary of registered fingerprints. A fingerprint image registration method in a fingerprint registration and verification device equipped with a fingerprint verification unit having a dictionary memory for storing information, the fingerprint image being detected by the fingerprint sensor being divided into a plurality of small areas, When detecting an image area that is larger than the area and brighter than a predetermined threshold, the image data of the bright image area is excluded from the creation of the fingerprint dictionary, so that the image data of the bright image area is excluded from the creation of the fingerprint dictionary. This feature is characterized in that feature points are prevented from being selected as feature points.

[For production]

According to the present invention, when an area brighter than a predetermined threshold value is detected in a certain area or more in the fingerprint data during image processing of fingerprint data, this bright image area is excluded from the area for which a fingerprint dictionary is created. be.

〔Example〕

FIG. 1 is a system configuration diagram including a fingerprint matching circuit to which the present invention is applied. The fingerprint input unit 1 has a fingerprint sensor shown in FIG. 4.5, and as described above, the ridge portion of the fingerprint is detected by the CCD element 45 via the glass plate 43 and the lens 44, and outputs analog data. The image input unit 2 has a binarization circuit and a binarization memory, and binarizes the analog data of the fingerprint and stores it in the memory. The verification circuit unit 3 has a verification circuit and a dictionary memory, and stores already registered fingerprints in the dictionary memory.
The verification circuit compares and verifies the manually input binary data and the contents of the dictionary memory. The feature extraction unit 4 has a thinning circuit and a thinning memory, and performs a thinning process on the input binary data.

The control section 6 has a CPU and a work memory, and controls the operations of the image human power section 2, the matching circuit section 3, and the feature extraction section 4 via the internal interface 5 based on a predetermined program.

FIG. 2 is a processing flowchart of one embodiment of the present invention.

When fingerprint image data is manually input from the fingerprint sensor shown in FIG. 4.5 (step 1), the fingerprint image data is divided into a plurality of local regions as shown in FIG. 6 (step 2). Then, the average value of image brightness is determined for each region (step 3). Next, a threshold value (second threshold value) for binarizing this image data is determined for each region (step 4). Here, in order to judge the brightness of the image, a first threshold value is determined separately in advance.

Next, compare the magnitude of the predetermined first threshold value and the second threshold value obtained from the average value (step 5), and the first threshold value is larger than the second threshold value. When I was small (Y
ES), the image data is binarized using the first threshold (step 7), but if the first threshold is larger than the second threshold (NO), the L threshold is small. The address of the area is stored in memory (step 6). By storing the address of the target area in memory in this way, if it matches this address of the cropped image,
This minutiae is not included in the fingerprint dictionary, and another minutiae is newly selected.

When the fingerprint image is binarized in step 7, thinning processing and feature point extraction are performed on the fingerprint image data (
In step 8), it is determined whether or not the feature points extracted in step 8 are in a relationship where the first threshold is larger than the second threshold as in step 5.Step 9) Yes (YES) ), the process returns to step 8, and if this relationship is not satisfied (N○), feature points are registered (step 10), and if the number of registered feature points is less than the predetermined value N (YES). ), return to step 8 and perform thinning and feature point extraction for other locations (step 11).
After going through steps 9 and 10, the number of registrations is determined again, and if it is larger than the predetermined number N, the registration process is ended.

FIG. 3 is a processing flow chart of another embodiment of the present invention. In this process, the average value of the brightness of the entire fingerprint image data is calculated (step 2), and in the example shown in FIG. Then, determine the first threshold value from this average value (step 3
). Then, a second threshold value for binarizing the divided area is determined (step 4), and the magnitude of the first threshold value and the second threshold value are compared (step 5).

The subsequent steps are similar to those in the embodiment shown in FIG. 2, so their explanation will be omitted.

〔Effect of the invention〕

As explained above, according to the present invention, when creating a fingerprint dictionary, each time a low-quality image part of a fingerprint image with many continuous fingerprint ridges is removed from the dictionary data selection target, the fingerprint dictionary is created. It is possible to prevent false minutiae from being included in the dictionary, thereby preventing an increase in erroneous verifications and improving the fingerprint verification rate.

[Brief explanation of the drawing]

Fig. 1 is a system configuration diagram to which the present invention is applied, Fig. 2 is a processing flowchart of one embodiment of the present invention, Fig. 3 is a processing flowchart of another embodiment of the invention, and Fig. 4 is the main part configuration of a fingerprint sensor. Figure 5 is a configuration diagram of the main parts of a fingerprint sensor equipped with an aperture diaphragm, Figure 6 is an explanatory diagram of local region segmentation of a fingerprint image, Figure 7 is an explanatory diagram of a minutiae window, and Figure 8 is a fingerprint image. FIG. 9 is an explanatory diagram of a low-quality portion of FIG. 9, and FIG. 9 is an explanatory diagram of a conventional fingerprint registration method. (Explanation of symbols) Two fingerprint human power section: Image input section, : Verification circuit section, : Feature extraction section, : Internal interface, Two control sections.

Claims (1)

[Scope of Claims] A fingerprint sensor (1), a feature extraction unit (4) that extracts feature points of a fingerprint image, and an image input unit (4) that binarizes analog data to the fingerprint sensor using a predetermined threshold. 2) and a fingerprint matching unit (3) having a dictionary memory for storing a fingerprint dictionary of registered fingerprints, the fingerprint image detected by the fingerprint sensor (1) is divided into a plurality of small areas. fingerprint image registration, wherein when an image area having a certain area or more and brighter than a predetermined threshold is detected in the fingerprint image, image data of the bright image area is excluded from creation of the fingerprint dictionary. method.