JPH0478967A - Fingerprint data registering device - Google Patents

Abstract

PURPOSE:To prevent the pseudo feature point from being registered due to sweat, drying, etc. by executing window image data with pseudo feature point image data and deciding whether an extracted feature point is true or not, and eliminating the window image data from the registration in the case it is false. CONSTITUTION:Pseudo feature point image data 31 containing a pseudo feature point is stored in advance in a memory 9. A discriminating part 4 executes pattern matching of fingerprint image data segmented like a window at the time of registration, that is, window image data 30 and pseudo feature point image data 31, respectively, and in the case the degree of noncoincidence is below a prescribed value, its feature point is discriminated to be a pseudo feature point, and eliminated from the registration. In such a way, since a true feature point can be extracted and registered without being influenced by being dry or wet of a finger, a drop of the collation rate and a rise of an erroneous collation rate can be prevented.

Description

[Detailed Description of the Invention] [Summary] The present invention relates to a fingerprint data registration device, and has the purpose of preventing the registration of false minutiae due to sweat, dryness, etc., and which extracts minutiae of a fingerprint from fingerprint image data. A fingerprint data registration device that cuts out fingerprint image data including the minutiae into a window shape and registers fingerprint minutiae information, the memory storing one or more pseudo minutiae image data including the pseudo minutiae; a discrimination unit that performs pattern matching on the window image data cut out into the window shape with the pseudo feature point image data, and discriminates the extracted feature point as a false feature point if the degree of mismatch is less than a predetermined value; The window image data is pattern-matched with the pseudo feature point image data to determine the authenticity of the extracted feature points, and if it is false, the window image data is excluded from registration. .

C Industrial application field] The present invention relates to improvements in fingerprint data registration devices.

A fingerprint verification device for identifying an individual is composed of a fingerprint image input device and a verification device, and verifies the identity of the individual by comparing the fingerprint image data registered in advance with the fingerprint image data input from the fingerprint image input device. .

This fingerprint image input device, which is also used during enrollment, generally uses optical contact between the finger and the glass, and if the finger is sweaty, bridges of sweat will form between the fingerprint ridges, making it appear as if the finger is sweaty. The fingerprint image is detected such that there are branch points in the ridge pattern. On the other hand, if the finger is dry, the contact between the finger and the glass is poor, the ridge pattern is cut off, and the fingerprint image is detected as if there were end points.

Therefore, if such fingerprint image data is registered, there is a risk that the matching rate will decrease and the false matching rate will increase, so a fingerprint data registration device that eliminates the registration of pseudo-fingerprint data due to sweat, dryness, etc. is required. be done.

[Conventional technology]

An overview of fingerprint verification will be explained using FIG. 7, which is an explanatory diagram of the fingerprint registration order method. For details, please refer to "Personal Verification Device Using Holographic Fingerprint Sensor" (IEICE P.
RUB8-38) etc.

As a method for comparing fingerprints, the branching point 9 dead-end point (referred to as an end point) of a ridge (a series of convex parts) is used as the minutiae of the fingerprint, and the distribution of the minutiae points of the fingerprint is registered in advance as dictionary information. A method is used in which input fingerprint image data is compared with the individual's dictionary information.

At the time of registration, first, the ridge part of the fingerprint image data inputted from the fingerprint image input device 2 (see Fig. 7) is thinned, and the binary data of 8 neighborhoods around the point of interest is applied to a feature extraction mask. Extract multiple feature points using a method.

Next, cut out the area that includes this feature point and open the window (Fig.
(2)) and register the window position coordinates, binary data in the window (window image data), etc. as dictionary information. At this time, one window (for example, ■) at the center of the plurality of windows (■ to ■) is defined as the alignment window.

At the time of verification, pattern matching is performed between the window image data and the input fingerprint image data while scanning the positioning window 2 in a spiral pattern, thereby performing positioning at the time of registration and input. When the alignment is completed, other windows are moved in parallel according to the amount of displacement of the alignment window (2), and pattern matching is performed in the vicinity of each window. If the degree of inconsistency of a predetermined number of windows above is less than or equal to a predetermined value, the identity of the person is determined.

[Problem to be solved by the invention]

As shown in FIG. 7, the fingerprint image input device 2 includes a glass plate 2.
This device detects a fingerprint image using light reflected from the contact surface 21 of a finger 20 brought into contact with the glass plate 2.
Fingerprint ridges that contact 2 and valley lines that do not contact are optically separated, and the light is received by a sensor (not shown) and binarized.

Therefore, when the fingers become sweaty and sweat bridges form between the fingerprint ridges, the fingerprint image appears as if there are branch points in the ridge pattern, as shown in Figure 9, which shows examples of true/false branch points. is detected, and conversely, if the finger is dry, the contact between the finger and the glass will be poor, and the ridge pattern will be cut off, as shown in Figure 8, which shows an example of true/false end points. The endpoints are detected as if they existed.

If feature points are extracted from fingerprint image data that includes such pseudo feature points, there is a risk that pseudo feature points will be extracted by mistake, resulting in a decrease in the matching rate or an increase in the false matching rate.

Therefore, by taking advantage of the fact that there are not many true feature points in the vicinity, for example, if the extracted feature point is an endpoint, the number of endpoints in the vicinity is counted, and if it is more than a predetermined number, it is not adopted as a feature point. If the feature point is a branch point, there are methods to determine whether it is a pseudo branch point or not based on whether there is a paired branch point nearby and remove it. If the number of points is large, the true endpoint is excluded from the feature points, and in the latter case, if a pseudo branching point exists nearby, the true branching point is determined to be a pseudo branching point and is removed. In any case, there is a problem in that true feature points are excluded and it is not possible to obtain the required number of feature points for matching.

SUMMARY OF THE INVENTION In view of the above problems, an object of the present invention is to provide a fingerprint data registration device that identifies pseudo minutiae.

[Means to solve the problem]

In FIG. 1, a diagram showing the principle of the present invention, 9 is a memory in which one or more pseudo feature point image data 31 including pseudo feature points are stored.

4 is a discrimination unit that performs pattern matching on window image data 30 cut out into a window shape including the extracted feature points with the pseudo feature point image data 31, and if the degree of mismatch is less than a predetermined value, the feature point is is determined to be a false feature point.

[For production]

(1) Pseudo feature point image data 3 including pseudo feature points in advance
1 is stored in the memory 9. The discrimination unit 4 uses fingerprint image data cut out into a window shape at the time of registration, that is, window image data 3.
0 and the pseudo feature point image data 31, and if the degree of mismatch is less than a predetermined value, the feature point is determined to be a pseudo feature point and excluded from registration.

(2) As another method for the above-mentioned determination, if the degree of matching in pattern matching is greater than or equal to a predetermined value, the point is determined to be a pseudo feature point.

(3) When there is a plurality of pseudo feature point image data 31, the direction information of each ridge or valley line is recorded, and the direction of the ridge or valley line in the window image data is measured to correspond to the corresponding pseudo feature point. Only image data 31 is subjected to pattern matching.

(4) Similarly, the width of the ridge or valley line in the pseudo feature point image data 31 is recorded, and the line width of the ridge or valley line in the window image data is measured, and the corresponding pseudo feature point is measured. Only image data 31 is subjected to pattern matching.

(5) If the area of the pseudo feature point image data 31 is larger than the window area and includes multiple pseudo feature points, pattern matching is performed while scanning in a spiral pattern, and the degree of mismatch (matching degree) is less than or equal to a predetermined value (or more). ) is detected, it is determined to be a false feature point.

Note that the processes (3) to (5) above are for improving the processing speed of pattern matching.

As described above, by pattern matching with the pseudo feature point image data 31, it is possible to determine the authenticity of feature points, obtain the necessary number of feature points, and prevent false feature points from being registered in error. The reduction in the matching rate and the increase in the false matching rate are improved.

〔Example〕

Embodiments of the present invention will be described in detail with reference to the drawings.

Fig. 2 is a configuration diagram of one embodiment, Fig. 3 is a registration processing flowchart, Fig. 4 is a diagram showing the case where the pseudo feature point image data area is larger than the window area, and Fig. 5 is a pseudo feature based on the ridge direction. FIG. 6 is a diagram illustrating the selection of pseudo feature point image data based on the ridge width.

In this embodiment, a plurality of pseudo feature point image data 31 are provided,
An example will be shown in which pseudo feature point image data 31 is selected by measuring the ridge direction and ridge width.

FIG. 2 shows an example of the configuration of a fingerprint verification device having a registration function. In the figure, 1 is a numeric keypad for inputting registration/verification instructions, registration numbers, etc.

2 is a fingerprint image input device which inputs fingerprint image data to the verification device 3 in real time.

Reference numeral 3 denotes a verification device, which has a dictionary information registration function and a fingerprint verification function.

Further, in the verification device 3, an image input unit 8 binarizes the fingerprint image data inputted from the fingerprint image input device 2 and stores it in the image memory 11.

Reference numeral 6 denotes a registration unit that searches the fingerprint image data stored in the image memory 11 to extract a plurality of feature points, cuts out a region of a predetermined size including the feature points as a window, and extracts a region of a predetermined size including the feature points as a window.
The value data (window image data 30) is stored in the window image memory 5, and only the window image data 30 determined to be a true feature point is registered in the dictionary memory 10 as dictionary information.

Reference numeral 9 denotes a memory in which a plurality of pseudo feature point image data 31 are stored in advance together with direction information and line width information.

4 is a determination unit that performs pattern matching on the extracted window image data 30 and the pseudo feature point image data 31, and if the degree of mismatch is less than a predetermined value, it is determined that it is a pseudo feature point and notifies the registration unit 6.

Reference numeral 7 denotes a collation unit which reads out corresponding dictionary information from the dictionary memory 10 based on the registration number inputted from the numeric keypad 1;
Compare with input fingerprint image data.

With the above configuration, the registration process described below is performed.

(1) Set the registration mode using the numeric keypad 1, input the registration number, and place the finger 20 on the contact surface of the fingerprint image input device 2. As a result, the fingerprint image is read, converted into binary data, and stored in the image memory 11.

(2) The registration unit 6 searches the fingerprint image data stored in the image memory 11 and extracts feature points. Then, data of a predetermined area including the extracted feature points is cut out and stored in the window image memory 5 as window image data 30.

(3) The discrimination unit 4 performs the following feature point discrimination processing,
The authenticity is notified to the registration unit 6.

First, line thinning processing is performed to erase the periphery of the ridge in the window image data 30 one dot at a time, the line width is measured by counting the number of thin lines, etc., and based on the thinned final image data, Measure direction.

(4) Select the corresponding virtual feature point image data 31 based on the measured direction and line width.

Figure 5 shows pseudo feature point image data classified in two directions.
This shows the case where ``■'' is selected, and the ridge direction in the window image data 30 is recognized by measuring the gradient of the thinned image data, etc., and the virtual feature point image data ``■'' and ``■'' are selected. Indicates that pattern matching is to be performed.

Since a fingerprint has a spiral shape, it can be classified into multiple ridge directions, but when classifying into multiple directions such as the horizontal direction and the vertical direction, pseudo minutiae image data 31 that is close to the measurement result is selected.

Figure 6 shows an example of selecting pseudo feature point image data ■ to ■ with different line widths.The line widths are measured during the line thinning process, and the pseudo feature point image data ■ and ■ with similar line widths are selected. select.

In this embodiment, since both the direction and the line width are measured and selected, if the pseudo feature point image data 31 is from ■ to ■, only ■ will be selected. Time is significantly reduced.

(5) Perform pattern matching to detect matching/mismatching of corresponding pixels between the window image data 30 and the selected pseudo feature point image data 31, and count the number of mismatches to determine the degree of mismatch.

(6) If the degree of mismatch is less than or equal to the threshold value, it is determined that it matches the virtual feature point image data 31, and the registration unit 6 is notified.

(7) The registration unit 6 stores the window image data 30 determined as a true feature point in the dictionary memory 1 as dictionary information together with position information.
0 and extract the next feature point.

Similar determination and registration processing is performed thereafter, and the registration processing is ended after the number of true feature points required for verification has been extracted.

FIG. 4 shows an example in which the pseudo feature point image data area is made larger than the window area and includes a plurality of pseudo feature points, which has the effect of reducing the number of pseudo feature point image data. In this case, as shown in the figure, pattern matching is performed while scanning the window image data 30 in a spiral pattern from approximately the center of the pseudo feature point image data 31, and the degree of mismatch at each position is calculated, and whether there is a threshold value or less. The determination is made by detecting whether or not the

In addition, all of the above were determined by calculating the degree of discrepancy, but -
The same result is obtained even if we distinguish based on political change.

As described above, the authenticity of the extracted feature points can be determined by pattern matching with the pseudo feature point image data 31, and the number of feature points necessary for matching can be determined from the collected fingerprint image data regardless of whether the finger is wet or dry. can be extracted, and there is no possibility of erroneously registering pseudo feature points, thereby reducing the matching rate or increasing the false matching rate.

〔Effect of the invention〕

As explained above, the present invention distinguishes and registers the authenticity of extracted feature points by pattern matching with pseudo feature point image data, and extracts true feature points without being affected by the dryness or humidity of the finger. Since the matching rate can be registered using the finger, it is possible to prevent a decrease in the matching rate and an increase in the false matching rate, and it is also possible to extract the required number of feature points regardless of whether the finger is wet or dry.

[Brief explanation of the drawing]

Fig. 1 is a principle diagram of the present invention, Fig. 2 is a configuration diagram of an embodiment,
3 is a registration process flowchart, FIG. 4 is a diagram showing the case where the pseudo feature point image data area is larger than the window area, FIG. 5 is an explanatory diagram for selecting pseudo feature point image data according to the ridge direction, and The figure is an explanatory diagram for selecting pseudo minutiae image data based on ridge width, Fig. 7 is an explanatory diagram for fingerprint registration verification method, Fig. 8 is an illustration showing an example of authenticity of end points, and Fig. 9 is an example of authenticity of branch points. FIG. In the figure, 1 is a numeric keypad, 2 is a fingerprint image input device, 3 is a verification device, 4 is a discrimination unit, 5 is a window image memory, 6 is a registration unit, 7 is a verification unit, 8 is an image input unit, 9 is a memory, 10 is a dictionary memory, 11 is an image memory, 20 is a finger, 21 is a contact surface, 22
is a glass plate, 30 is window image data, and 31 is pseudo feature point image data. Principle diagram of the present invention Fig. 1 1. FIG.

Claims (5)

[Claims] (1) A fingerprint data registration device that extracts fingerprint minutiae from fingerprint image data, cuts out the fingerprint image data including the minutiae into a window shape, and registers fingerprint minutiae information, including pseudo minutiae. Performing pattern matching on a memory (9) storing one or more pseudo feature point image data (31) and the window image data (30) cut out into the window shape with the pseudo feature point image data (31), respectively; A discriminating unit (4) is provided which discriminates the extracted feature point as a false feature point if the degree of mismatch is less than a predetermined value, and the window image data is extracted by performing pattern matching with the pseudo feature point image data. A fingerprint data registration device characterized by determining whether the minutiae point is true or not, and if it is false, excluding the window image data from registration. (2) The fingerprint data registration device according to claim (1), further comprising a discriminating unit that discriminates the extracted minutiae as a false minutiae if the degree of matching is equal to or higher than a predetermined value as a result of pattern matching. . (3) Detect the direction of the fingerprint ridge or valley line in the window image data, select the pseudo minutiae image data corresponding to the detected direction from among the plurality of pseudo minutiae image data, and perform pattern matching. The fingerprint data registration device according to claim 1 or 2, characterized in that the fingerprint data registration device is a discrimination unit that performs the following operations. (4) Measure the line width of the fingerprint ridge or valley line in the window image data, select the pseudo minutiae image data corresponding to the line width from among the plurality of pseudo minutiae image data, and perform pattern matching. Claim (1) characterized in that the discriminating unit performs
), (2) or (3). (5) If the pseudo feature point image data has an area larger than the window area and includes multiple pseudo feature points, pattern matching is performed while scanning the window image data within the area, and the degree of mismatch or match is determined. Claims (1), (2), (3), or (4), characterized in that the discriminating unit determines that the point is a pseudo feature point when the degree is less than or equal to a predetermined value or more than a predetermined value.
The fingerprint data registration device described.