JPH0478975A - Device and method for collating pattern - Google Patents

Abstract

PURPOSE:To shorten time for collation by matching positions while collating the distribution of pattern non-coincidence degree between a window for position matching and a screen for registration with the distribution of a pattern non-coincidence degree between the window for position matching and a screen for collation. CONSTITUTION:The whole face of a prescribed window (window for position matching) 30 segmented from a screen 32 for registration is scanned in advance at prescribed intervals, pattern collation is executed at each position so as to calculate the degree of non-coincidence, and the distribution of the calculated non-coincidence degree is registered in a non-coincidence degree distribution storage part 9. In the case of collation, a position matching part 7 scans the prescribed window 30 at the same intervals as registering the prescribed area of a screen 31 for collation, pattern collation is executed at each position so as to calculate the degree of non-coincidence, and the distribution of the non-coincidence degree for collation is prepared. Then, this distribution of the non-coincidence degree is collated by being moved on the registered distribution of the non-coincidence degree, and the moving amount is calculated. Based on the calculated moving amount, the respective windows 30 are parallelly moved, and pattern collation is executed near each window respectively. Thus, time for collation is shortened.

Description

[Detailed Description of the Invention] [Summary] The present invention relates to a pattern matching device and a pattern matching method, and aims to speed up the matching process by selecting a predetermined window from among a plurality of windows cut out from the registration screen and registered. A pattern matching device that aligns a pattern between a screen and a matching screen, moves the position based on the obtained movement amount, and performs pattern matching with the matching screen, a mismatch distribution storage unit that stores a mismatch distribution of patterns obtained by moving the predetermined window at predetermined intervals on the registration screen; positioning, measuring the mismatch degree of the pattern at each position, comparing the obtained mismatch distribution with the stored mismatch distribution to determine the position, and calculating the amount of movement from the predetermined area. A discrepancy distribution of patterns between the alignment window and the registration screen is registered in advance, and the discrepancy distribution between the window and a predetermined area of the verification screen is registered in advance. The configuration is such that alignment is performed by measuring and comparing the mismatch distributions of both sides.

[Industrial Application Field] The present invention relates to improvements in a pattern matching device and a pattern matching method.

In a fingerprint verification device, an individual's fingerprint image is registered and compared with the fingerprint image input during identity verification. For reasons such as cross-checking, a plurality of windows are cut out from the registration screen and registered, and each of these windows is scanned over the cross-checking screen to perform pattern matching.

In this matching process, since the position of the fingerprint image on the screen generally differs between registration and verification, pattern alignment is performed by pattern matching using a certain window, but since there are multiple candidate points for alignment. , Performing well-intentioned pattern matching for each candidate point takes time.

In this way, if there is a positional deviation between the pattern to be compared when it is registered on the screen and when it is compared, it is necessary to quickly determine the positioning and shorten the matching time.

[Conventional technology]

A conventional verification method in a fingerprint verification device will be described below. For details, please refer to "Personal Verification Device Using Holographic Fingerprint Sensor" (IEICE PRU8 B-3
8) etc.

FIG. 8 is an explanatory diagram of the prior art, and FIG. 9 is a diagram showing changes in the degree of mismatch between alignment windows.

As a method for comparing fingerprints, the branch point 2 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 that compares the fingerprint image data input at the time of personal identification with the dictionary information of the individual.

At the time of registration, the fingerprint image data inputted from the fingerprint image input device is first binarized and stored in the memory as pattern data in which one screen is a contact permissible area of the finger from which the fingerprint image is taken. Next, a plurality of feature points are extracted from this pattern data by thinning the ridge portion and applying a feature extraction mask to the pattern data in eight neighborhoods surrounding the point of interest.

Next, as shown in FIG. 8(a), a region of a predetermined size including this feature point is cut out to form a window 30 (■ to ■ in FIG. 8), and the position coordinates of the window, pattern data in the window, etc. Register as dictionary information. At this time, one window (for example, ■) at the center of the plurality of windows 30-■ to ■ is defined as the alignment window.

During verification, one method is to use the coordinates of the alignment window ■ as a starting point and scan in a spiral pattern, for example every other pixel, while performing pattern matching between the window ■ and the verification screen.
Performs alignment processing to match positional deviations during registration and input. Figure 8 [Yes]) Figure 9 shows the transition of the degree of pattern mismatch.The position where the degree of mismatch is smaller than the threshold value is taken as a candidate point, and each time a candidate point is detected, the amount of movement from the starting point is changed. In order to correct the distortion of the fingerprint image during registration and verification, the fingerprint image is scanned in the vicinity of the fingerprint image and pattern matching is performed. Then, if the degree of inconsistency in the predetermined counting window is less than or equal to a predetermined value, the identity of the person is confirmed. Figure 8 (C
) If the result of condition pattern matching is unsuccessful, window (2) is scanned again to search for the next candidate point, and condition pattern matching is performed in the same manner as described above.

(Problems to be Solved by the Invention) In conventional positioning using a positioning window, pattern matching is performed by scanning every other pixel in a spiral pattern, and when a candidate point is detected, the state is moved and the The matching process of scanning the neighborhood, performing pattern matching, and searching for the next candidate point if it is determined that the patterns do not match is repeated.

In condition pattern matching, even if alignment is performed using the alignment window to correct distortion, the vicinity of each condition is scanned as described above, so if pattern matching is performed for each candidate point, matching will be difficult. it takes time.

In view of the above problems, it is an object of the present invention to provide a pattern matching device and a pattern matching method that quickly determine positioning and shorten matching time.

[Means for Solving the Problems] In Fig. 1, a diagram of the principle of the present invention, 9 is a mismatch degree distribution storage unit, which moves a predetermined window at predetermined intervals over the entire registration screen and records the mismatch of patterns at each position. The discrepancy distribution obtained by measuring the discrepancy is stored.

Reference numeral 7 denotes a positioning unit, which moves a predetermined window over a predetermined area of the matching screen at the predetermined intervals, measures the degree of mismatch between patterns at each position, and converts the obtained mismatch distribution into the stored mismatch distribution. In addition to determining the position by comparing the position with the specified area, the amount of movement from the predetermined area is calculated.

[For production]

In advance, a predetermined window cut out from the registration screen (alignment window) is scanned over the entire surface of the registration screen at a predetermined interval, pattern matching is performed at each position to determine the degree of mismatch, and the obtained degree of mismatch distribution is calculated. It is registered in the mismatch degree distribution storage section 9.

During verification, the alignment unit 7 scans a predetermined window within a predetermined area of the verification screen, for example, within a predetermined range up, down, left and right around the coordinate position of the registration screen of the predetermined window at the same intervals as during registration; Pattern matching is performed at each position to calculate the degree of mismatch, and a degree of mismatch distribution for matching is created.

Then, this mismatch degree distribution is moved on the registered mismatch degree distribution and compared, and, for example, a position where the sum of the differences in the corresponding mismatch degrees is less than or equal to a predetermined value is searched and positioned, and the amount of movement is calculated.

Then, the condition is translated in parallel based on the obtained movement amount, and pattern matching is performed in each vicinity.

Compared to the conventional method of detecting candidate points by matching patterns between a predetermined window and the matching screen, this method targets an area larger than the window area, so the mismatch distribution is uniquely Therefore, the positioning can be quickly determined, and the verification time for the condition can be significantly shortened.

〔Example〕

Embodiments of the present invention will be described in detail using FIGS. 1 to 7.

Figure 2 is a configuration diagram of one embodiment, Figure 3 is an explanatory diagram for creating a discrepancy distribution table for registration, Figure 4 is an explanatory diagram for creating a discrepancy distribution table for verification, and Figure 5 is a diagram for matching discrepancy distribution tables. FIG. 6 is a registration processing flowchart, and FIG. 7 is a matching processing flowchart.

FIG. 2 shows an example of a fingerprint registration and verification device, which includes a numeric keypad 1 for inputting a registration number, etc., a fingerprint image input device 2 for detecting a fingerprint image from a finger 20 touched during registration and verification, and a registration verification section 3. It consists of In the registration verification unit 3, 4 is a registration unit that cuts out a plurality of windows 30 from the registration screen 32 (Figure 1) as shown in Figure 8(a), and records pattern data in the windows and window coordinate data. etc. are registered in the window information storage unit 10 as window information, and the inconsistency distribution table 32a for registration created by the inconsistency distribution creation unit 5 is registered in the inconsistency distribution storage unit 9.

Reference numeral 5 denotes a mismatch distribution creation unit, which compares the registration screen 32 with the alignment window 30-■ (hereinafter referred to as window ■) to create a mismatch distribution table 32a for registration, as will be described later.

Reference numeral 6 denotes a matching unit that reads out a plurality of pieces of window information corresponding to the registration number and the discrepancy distribution table 32a, performs pattern matching processing for the condition 30 based on the amount of movement calculated by the alignment unit 7, and identifies the person. confirm.

7 is the positioning section, which displays the input collation screen 31 and the window ■
A matching inconsistency distribution table 31a is created by pattern matching within a predetermined area, and the read out inconsistency distribution table 32
The position is determined by comparing with a, and the amount of movement is calculated.

An image input unit 8 binarizes the fingerprint image data input from the fingerprint image input device 2 and stores it in the screen information storage unit 12 as screen information.

In the above configuration, registration and verification processing as shown below is performed.

[Registration Process] See FIG. 6 (1) When a fingerprint image for registration is input, the image input unit 8 binarizes this fingerprint image data and stores it in the screen information storage unit 12 as the registration screen 32.

(2) The registration unit 4 extracts feature points from the registration screen 32, cuts out a plurality of windows 30 of a predetermined size including the feature points, and registers them in the window information storage unit 10.

At this time, the window ■ near the center of the screen is defined as the alignment window.

(3) The inconsistency distribution creation unit 5 displays the window ■ and the registration screen 3.
2 is read out, the window 3 is scanned over the entire surface of the registration screen 32 by d pixels to perform pattern matching, and the degree of mismatch is calculated at each position, edited and registered in the degree of mismatch distribution table 32a.

FIG. 3 shows an example of creating a discrepancy distribution table 32a from the window ■ and the registration screen 32.
At the top, the window ■ is scanned every 6 pixels in the X direction and every d pixels in the Y direction, pattern matching is performed at each position, and the degree of mismatch (
%) or the number of mismatched pixels. This allows XY
Assuming N pixels in both directions, (N/d)'' mismatch degrees are obtained, and as shown in the figure, a mismatch degree distribution table is created by mapping in accordance with the screen.

[Verification Process] See FIG. 7 (1) When a verification instruction is given and a registration number is input, the verification section 6 reads out the corresponding window information and the discrepancy distribution table 32a.

(2) When the verification fingerprint image is input, the image input section 8 edits it into the verification screen 31 and stores it in the screen information storage section 12.

(3) The positioning unit 7 moves the window ■ on the verification screen 31 within a predetermined range every d pixels vertically and horizontally,
The degree of inconsistency is calculated by pattern matching, edited into the inconsistency distribution table 31a for comparison, and stored in the inconsistency distribution storage unit 11 for comparison.

FIG. 4 shows an example of creating the inconsistency distribution table 31a for matching, in which the window ■ is moved n times from the coordinate position on the registration screen 32 every d pixels up, down, left and right on the screen, and pattern matching is performed. and calculate the degree of discrepancy. This results in
(2n)'' discrepancy degrees are obtained, and mapped as shown in the figure to create a matching discrepancy distribution table 31a.

(4) Once the matching mismatch degree distribution table 31a is created, the alignment unit 7 scans the matching mismatch degree distribution table 31a over the mismatch degree distribution table 32a section by section to check the mismatch degree. Refer to Figure 5. This verification does not necessarily result in a match if the structure of the mismatch distribution table is rough, so for example, if the difference in the corresponding mismatch degrees is calculated and the number of sections for which the difference is less than the threshold value is greater than or equal to a predetermined number, then Or, if the sum of the differences of each degree of mismatch is less than a threshold value, or the position where the sum of the differences is the minimum is determined as having been matched by the degree of mismatch distribution tables 31a and 32a, the position is determined, and the movement amount A (vector, or ΔX, Δy).

In addition, this movement amount A is the number of movement sections with respect to the coordinates of the window ■
It can be found as d.

(5) Once the amount of movement A is determined, the matching unit 6 moves the coordinates of the condition by the amount of movement, scans the vicinity of that position, and performs pattern matching.

(6) If it is determined by pattern matching that the fingerprint images match, an identity verification message is output.

It goes without saying that after positioning using non-components, detailed positioning may be performed by pattern matching.

As described above, if the degree of pattern mismatch between the registration screen and the alignment window is measured at the time of registration, it is sufficient to create a mismatch degree distribution table for a relatively small area and perform the mismatch degree comparison at the time of matching. Compared to the pattern matching of a small area using (2), it is possible to narrow down the number of candidate points to a smaller number, thereby reducing the number of well-intentioned pattern matchings and shortening the matching time.

〔Effect of the invention〕

As explained above, the present invention performs alignment by comparing the pattern mismatch distribution between the alignment window and the registration screen and the pattern mismatch distribution between the alignment window and the verification screen. Therefore, compared to the conventional method of performing well-intentioned pattern matching for each of a plurality of candidate points, the matching time is significantly shortened, and the operational effects are great.

[Brief explanation of the drawing]

Fig. 1 is a principle diagram of the present invention, Fig. 2 is a configuration diagram of an embodiment,
Figure 3 is an illustration of creating a discrepancy distribution table for registration, Figure 4 is an illustration of creating a discrepancy distribution table for verification, Figure 5 is an illustration of positioning by matching the discrepancy distribution table, and Figure 6 is registration. FIG. 7 is a flowchart of the matching process, FIG. 8 is an explanatory diagram of the prior art, and FIG. 9 is a diagram showing the transition of the mismatch degree of the alignment window. In the figure, 1 is a numeric keypad, 2 is a fingerprint image input device, 3 is a registration verification unit, 4 is a registration unit, 5 is a mismatch distribution creation unit, 6 is a verification unit, 7 is an alignment unit, 8 is an image input unit, 9 is a discrepancy distribution storage unit, 10 is a window information storage unit, 11 is a matching discrepancy distribution storage unit, 12 is a screen information storage unit, 20 is a finger, 30
is a window, 31 is a verification screen, 31a is a discrepancy distribution table for verification, 32 is a registration screen, -JP, Jiku-e! ! ρi■ Figure 2 Figure 1 Figure 3 Figure (C) Diagram representing good faith pattern matching Illustration of conventional technology Figure Discrepancy diagram

Claims (2)

[Claims] (1) Align the pattern between a predetermined window among the multiple windows cut out and registered from the registration screen and the verification screen, and move the position of each window based on the obtained movement amount. A pattern matching device that performs pattern matching with each of the matching screens, and a mismatch degree that stores a mismatch distribution of patterns obtained by moving the predetermined window at predetermined intervals on the registration screen. a distribution storage unit (9); moving the predetermined window at the intervals on the predetermined area of the matching screen to measure the degree of pattern mismatch at each position; a positioning unit (7) that performs positioning by comparing with the degree distribution and calculates the amount of movement from the predetermined area, and has a mismatch degree distribution of patterns between the positioning window and the registration screen A pattern matching device characterized in that the pattern matching device registers the window, measures the discrepancy distribution between the window and a predetermined area of the verification screen at the time of verification, and performs alignment by comparing the discrepancy distributions of both sides. . (2) When multiple windows are cut out from the registration screen and registered, pattern matching is performed between the windows and the matching screen, and a predetermined pattern in the registration screen and the matching screen is matched. Move a predetermined window of the windows at a predetermined interval on the registration screen, calculate the mismatch degree of the pattern at each position, register the mismatch degree distribution, and move the window on the predetermined area of the verification screen at the predetermined interval. move with
Pattern matching is performed at each position to calculate a mismatch distribution, and the window is determined by comparing it with the registered mismatch distribution to determine the position of the window, and the amount of movement from the predetermined area is calculated to determine the position of each window. A pattern matching method characterized in that after moving a pattern, pattern matching is performed in the vicinity thereof with a matching screen.