JPH04313167A - Location confirmation system for characteristic point of picture - Google Patents

Abstract

PURPOSE:To confirm the relative location of the characteristic point of two objective pictures to collate the characteristic points of the picture. CONSTITUTION:First, in the extraction of the characteristic point being preprocessing, an analog picture (registration picture or inspection picture) picked up by an image pickup device 7 of an input device 6 i.s inputted to an A/D converter 2 of an electronic apparatus 1 for digitization, stored in a picture memory 3, processed in a CPU4 to obtain the characteristic point of the picture to be stored in a memory 5. On the other hand, in the collation of each characteristic point, the CPU 4 is divided by more than one circles around each characteristic point selected for the collation object to be obtained from the information on the characteristic point. By information of the surrounding characteristic point in the part area not including the center of the circle, the coincidence with the relative position of one registration picture characteristic point selected for collation object and one inspection picture characteristic point is decided.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] The present invention is applicable to pattern recognition of digitized images (fingerprints, figures, etc.) using electronic devices such as electronic computers, electronic exchanges, communication control devices, IC cards, image processing devices, recognition devices, etc. When using a device, etc., 2.
The present invention relates to a method for locating feature points of two images for determining matching of two images by matching based on matching or similarity of one or more feature points of the two images.

0002

2. Description of the Related Art As an example of an image to be subjected to pattern recognition of images such as digitized fingerprints and figures, a case will be explained in which the image is a fingerprint. This fingerprint is a pattern of ridges on a finger. Since the valley line is determined by the ridge line, a pattern drawn by the valley line may be used as the fingerprint instead of the ridge line.

[0003] Fingerprint input devices for personal identification are input from an imaging device (for example, a CCD (charge-coupled device) camera), and a prism method (for example, ``Fingerprint Information Detection Method Using a Prism'' by Akihiro Shimizu et al. ``Comparison of total internal reflection method and optical path separation method'', Transactions of the Institute of Electronics and Communication Engineers, Vol. J
68-D, No. 3, 1985) and holographic methods (for example, Seigo Igaki et al., “Personal Verification Device Using Holographic Fingerprint Sensor”, IEICE Technical Research Report, PRU87-31, pp.27-33, 198
7 years).

[0004] A fingerprint image of analog information input from an image pickup device of an input device is an A/D (analog/digital)
A converter converts it into a digitized grayscale image of the fingerprint. This gradation image of the fingerprint is represented by (x, y) coordinates, which are pixel addresses on the image memory, and the brightness of pixels, which are components of each pixel address in the image memory. The shading image of the fingerprint is then processed by the processing processor, CP.
Correction is performed by smoothing with U, correction based on the directionality of the ridge, and the like.

[0005] Here, the feature points representing the characteristics of a fingerprint include end points, branch points, and intersections. The minutiae of the digitized fingerprint gradation image is obtained by binarizing the fingerprint image and further thinning it to create a pixel range representing the minutiae, for example, the same pattern as a 3 x 3 pixel pattern centered on the minutiae. can be detected by its presence in a thinned image (for example, Sasakawa et al., "Fingerprint matching device for personal identification with improved ability to cope with low-quality images", Transactions of the Institute of Electronics, Information and Communication Engineers, Vol. J)
72-D-II, No. 5, pp. 707-714,1
990).

Next, in fingerprint verification, information for verification is registered in advance in a memory such as a file.The registered fingerprint (sometimes only the minutiae are registered; in that case, the registered fingerprint is (The image is not registered), and the fingerprint that is checked for consistency with this registered fingerprint is called a test fingerprint. These registered fingerprints and test fingerprints can be compared by comparing minutiae.

[0007] In this comparison of the minutiae of the registered fingerprint and the test fingerprint, since misalignment occurs when each fingerprint is input, it is necessary to align the registered fingerprint and the test fingerprint. Method a using the ridge direction
(the above-mentioned literature by Sasakawa et al.), method b using arbitrarily selected representative feature points and their neighboring feature points (Kazuharu Yamato et al.
"Fingerprint verification system using valley thinning", Transactions of the Institute of Electronics, Information and Communication Engineers, Vol. J71-D, No. 2, pp. 327
-335, 1988), method c using the center point of the fingerprint
(For example, Hiro Asai et al., "Automatic fingerprint verification using minutiae network characteristics - verification process", Transactions of the Institute of Electronics, Information and Communication Engineers, Vol. J72-D-D-II, No. 5,
pp. 733-740, 1989).

[0008]

[Problems to be Solved by the Invention] In the conventional image feature point matching technology, when the image is a fingerprint, for example, when the image is a fingerprint, it is difficult to There is a problem.

[0009] Method a, which uses the general ridge direction of the fingerprint of the prior art described above, has a problem in that the processing time increases when the input fingerprint has a large inclination or when the ridge direction fluctuates greatly. Method b, which uses an arbitrarily selected representative feature point and its neighboring feature points, has a problem in that the processing time increases when the input fingerprint has a large slope. Method c, which uses the center point of the fingerprint, requires time to find the center point, and has the problem that it cannot be applied to fingerprints with unclear center points.

[0010] It is an object of the present invention to be able to handle cases where an image, for example a fingerprint, has a large inclination, a large variation in the ridge direction, or a case where the center point of the fingerprint is unclear, and the position of the feature point of the image. It is an object of the present invention to provide a method for performing confirmation accurately in a short time.

[0011]

[Means for Solving the Problems] The present invention provides two digitized images, one of which is a registered image and the other image which is an inspection image, and each of the registered image and inspection image is stored in an image memory. The electronic device is equipped with an electronic device that performs matching on one or more feature points selected for matching, and each subregion is divided by one or more circles centered on each feature point selected for matching, and does not include the center of the circle. The method is characterized in that the matching regarding the relative positions of one registered image feature point and one inspection image feature point selected as a matching target is determined based on information on surrounding feature points that are existing feature points.

0012

[Operation] According to the present invention, information on minutiae located within a circle (here, the minutiae to be matched is the center of the circle) that essentially has the characteristic of not being affected by parallel movement or rotation is used. hand,
Since the relative position matching or similarity of two feature points in different images is determined, for example, if the image is a fingerprint, if the fingerprint is tilted greatly, the variation in ridge direction is large. This method has the advantage that the minutiae of the registered fingerprint and the verification fingerprint can be accurately compared even when the central point of the fingerprint is unclear.

[0013]

[Example] As an example, a case where the image is a fingerprint will be described. FIG. 1 is a block diagram of an example of a fingerprint image processing system implementing the method of the present invention. In the figure, 1 is an electronic device with an A/D converter 2, an image for storing a digitized fingerprint image, a binary image, and a thinned image (i.e., a thinned image) when necessary. memory 3,
It is equipped with a CPU 4 which is a processing processor, and a memory 5 for a work area and information storage. Reference numeral 6 denotes an input device, which includes an imaging device 7.

In this image processing system, a fingerprint inputted from the input device 6 is processed in the electronic device 1 and the minutiae points of the fingerprint are compared.

[0015] Here, a pixel address in the image memory 3 that stores a fingerprint image, which is a grayscale image of a digitized fingerprint, is expressed as (x, y) by x and y coordinates. The image memory 3 is composed of pixels, and the range of all pixel addresses is 0≦x≦xh, 0≦y≦yh.
Then, a further specified processing range within the range of this pixel address becomes the processing target. Note that pixels are expressed by brightness.

[0016] Also, which part of the luminance becomes the ridge line is
Depending on the method of the fingerprint input device 6 and the processing of the image after it has been digitized, in any case by presetting in the electronic device 1 processing the brightness characteristics corresponding to the ridges. Processing is possible.

In fingerprint verification, the fingerprint input from the input device 6 for registering minutiae points in the memory 5 of the electronic device 1 is referred to as a registered fingerprint, and the fingerprint input from the input device 6 for inspection is referred to as an inspection fingerprint. These are respectively referred to below.

[0018] In the process of extracting the minutiae of a fingerprint, the analog image of the fingerprint input from the input device 6 is
The converter 2 digitizes the image, the CPU 4 corrects the image, separates the processing range of the image into a fingerprint area and other areas (valley line or background), and further binarizes the image, thins the image, and converts the image into a thin line. Feature point candidates that match the feature point pattern (for example, a bit pattern of 3 x 3 pixels) are extracted from the processed image, and pseudo feature points, that is, noise and thinning processing results, such as whiskers, cuts, and short lines, are extracted. This is done by extracting true feature points by removing pseudo feature points generated by such methods.

For these processes, conventional techniques or any improved or modified techniques thereof can be used as the pre-processing of the present invention.

The minutiae pattern in the thinned fingerprint image consists of 8 pixels adjacent to the minutiae (because it is a binary image, the pixel value is 0 or 1, where 1 is a ridge pixel and 0 is a valley pixel). or a background pixel), an arbitrary pixel in the thinned image is set as P0, and the surrounding eight pixels are set as P1.
, P2, . . . , P8 are used to represent bits (0 or 1).

In this case, each pixel address is P0=(x
m, ym), P1=(xm+1, ym), P2
= (xm+1, ym-1), P3=(xm, ym-1)
, P4=(xm-1, ym-1), P5=(xm-1,
ym), P6=(xm-1, ym+1), P7=(xm
, ym+1), P8=(xm+1, ym+1).

[0022] When 3×3 pixels centered on P0 are feature points (in this case, P0=1), feature point pattern = P8‖P7‖P6‖P5
Define by ‖P4‖P3‖P2‖P1. Here, ▼ represents a concatenation of bits (a raw combination). In this case, the image address of the feature point is P0
This is the address. P1 and P8 are considered to be adjacent consecutive bits. The feature point pattern can be determined by considering the feature point pattern generated by the thinning method,
An example is shown below (where, as mentioned above, P0=1).

[0023] The end points are, for example, ■ when only one of the eight bits of the feature point pattern is 1 (here, the other bits are 0), and ■ when the eight bits of the feature point pattern are consecutive. This is the case when only the two bits shown in the table are 1 (the other bits are 0).

[0024] The branch point is, for example, 8 of the feature point pattern.
only 3 of the bits are 1, and those 1
If does not appear consecutively (other bits are 0),
■If only four consecutive bits of the eight bits of the minutiae pattern are 1, and only two of those 1s appear consecutively in one place, and ■The eight bits of the minutiae pattern This is the case when five of the bits are 1, and only two of those 1's appear in two consecutive places (where the other bits are 0).

[0025] An intersection occurs, for example, when four bits out of eight bits of a feature point pattern are 1, and these 1s do not appear consecutively (here, the other bits are 0).

The present invention is a technique that is applied after feature points are extracted. Regarding the minutiae, the minutiae of the registered fingerprint will be referred to as the registered fingerprint minutiae, and the minutiae of the test fingerprint will be referred to as the test fingerprint minutiae.

The feature point is located at a position (x
coordinates and y-coordinates), minutia type (distinguishing between end points, branch points, intersections, etc.), and minutiae pattern (a pattern of bits represented by pixels around the minutiae position in a thinned fingerprint image. , the feature point type is also determined from the feature point pattern).

In this embodiment, among the minutiae of the registered fingerprint,
The minutiae position confirmation method of the present invention is applied to confirm the relative position of the minutiae selected as the representative minutiae with respect to the test fingerprint. In the following description, for convenience, it is defined that inside the circle includes the top of the circle, and outside the circle does not include the top of the circle, and inside the square includes the top of the side, and outside the square does not include the top of the side.

The radius of the circle, which is a parameter that defines the range of peripheral feature points, is defined as V1, V2, ..., Vh, (h≧1, V
1<V2<V3<...<Vh). The value of h is arbitrary (h≧1), but in this embodiment, h=2 is used. As the information on the peripheral feature points, information that is invariant to rotation and translation of the image is selected. In this embodiment, the number of feature points for each feature point type is used.

FIG. 2 shows that x, y on the image memory 3
FIG. 3 is an explanatory diagram of an example showing the situation of coordinates, a fingerprint area A, a background B, and two circles C1 and C2 centered on a selected minutiae (defining a partial area where peripheral minutiae exist).

Select a representative minutiae from the registered fingerprint minutiae,
Procedure A, which is a means for finding peripheral feature points, is as follows. FIG. 3 shows an outline of the processing flow of procedure A.

(Procedure A) Step A1: Set the processing range of the thinned fingerprint image to xs≦
x≦xs+xL-1, and ys≦y≦ys+yL-1,
If the center of the processing range is (x0, y0),

0033
]

[Math 1]

[0034] [ ] indicates truncation after the decimal point.

It is assumed that the positions of the reference points for selecting the representative minutiae from the registered fingerprint minutiae are the following five pixel addresses. Here, K is a parameter, K≧1, and K
The value of is determined in advance. The position and number of reference points can be set arbitrarily.

Note that within a circle, for any radius V, |
x-x0|/V≦cosθ, and |y-y0|/V≦
Since sin θ, cos α,si for any α
If nα is made into a table and the existence of θ can be confirmed from the table, it may be determined that the shape is circular.

(x0+[xL/K], y0+[yL/K]), (x0
−[xL/K], y0+[yL/K]), (x0−[x
L/K], y0-[yL/K]), (x0+[xL/K
], y0-[yL/K]), and (x0, y0) Based on these reference points, the first found feature point that satisfies all of the following three conditions is set to that reference point. Let this be the representative minutiae of the registered fingerprint minutiae.

Condition 1: Among the registered fingerprint minutiae, the minutiae closest to the position of the reference point is selected preferentially. Here, the search range is searched by sequentially widening the square range for pixel addresses centered on the reference point. The maximum side length of this square is MAX (xL/2, yL/2). MAX (A, B) is the larger value of arbitrary numbers A and B, and when A=B, it is one of the values.

■Condition 2: The fingerprint area is within the circle with radius Vh (in this example, since h=2, radius V2), which is the largest circle centered on the pixel address of the representative minutiae of the registered fingerprint minutiae. Must be a feature point within.

Condition 3: When a plurality of representative minutiae of the registered fingerprint minutiae are selected, the pixel addresses must be different from each other.

The positions of the effective representative minutiae of the registered fingerprint minutiae obtained from the above are expressed as (xi, yi), (i=1, 2,
..., n; n≦standard score). Here, when the number of effective representative feature points≦G, the process is terminated (G is a parameter determined in advance, and G≦n).

As described above, the representative minutiae {position, minutiae type, minutiae pattern} is found from among the registered fingerprint minutiae points.

Step A2: The partial area that is the search range for peripheral feature points is divided into a radius V centered on each representative feature point.
1, as well as outside the circle with radius V1 and inside the circle with radius V2 (
Assuming that V1<V2), peripheral minutiae are found for each representative minutiae of each registered fingerprint minutiae. The fact that the pixel address (x, y) is within a circle with radius V1 centered at (x0, y0) means that

[0044]

[Math 2]

It can be determined by: Pixel address (x, y
) is outside the circle with radius V1 centered on (x0, y0) and inside the circle with radius V2,

[0046]

[Math 3]

It can be determined by:

[0048] Given a center pixel (x0, y0) and a radius V, when determining whether the pixel is within the (x, y) coordinate range within the circle, |x-x0|/V≦1 Katsu｜y
Unless -y0|/V≦1, the processing time can be shortened by taking advantage of the fact that it is clearly outside the circle with radius V.

[0049] As described above, as peripheral minutiae for each representative minutiae of each registered fingerprint minutiae, the position within the circle with radius V1, the minutiae type, the number of end points, the number of branching points, the number of intersections, and the circle with radius V1 The position outside and inside the circle of radius V2, the feature point type, the number of end points, the number of branch points, and the number of intersections are determined.

Step A3: As a representative minutiae of the registered fingerprint minutiae, the position, minutiae type, and minutiae pattern are stored in the memory 5, and information on surrounding minutiae regarding the representative minutiae is stored within a circle with a radius of V1. The number of end points, the number of branch points, the number of intersections, and the number of end points, the number of branch points, and the number of intersections outside the circle with radius V1 and inside the circle with radius V2 are stored in the memory 5 (Step A
end).

An embodiment of the means for performing the initial comparison between the registered fingerprint minutiae and the test fingerprint minutiae is shown in procedure B. Figure 4 shows step B.
An outline of the processing flow is shown below.

(Procedure B) Step B1: In the pixel address of the thinned image in the image memory 3, the range that takes into account the deviation at the time of input of the test fingerprint minutiae and the registered fingerprint minutiae, for example, the representative feature of the registered fingerprint minutiae. For the pixel address (xi, yi) of a point, the square range xi-(2q+1)≦x≦xi+2q+1
, yi-(2q+1)≦y≦yi+2q+1, (q is a parameter and the value is set in advance),
For the reference points corresponding to the representative minutiae of the registered fingerprint minutiae, a representative minutiae that satisfies all of the following conditions is searched from the test fingerprint minutiae, and the obtained minutiae is tested to match the registered fingerprint minutiae. It is regarded as a fingerprint minutiae and the process goes to step B2.

[0053] If there are a plurality of test fingerprint minutiae satisfying the following conditions 1 to 3 for one registered fingerprint minutiae, one of them is arbitrarily selected. This will occur less frequently if each parameter of the present invention is appropriately set.

[0054] Condition 1: The representative minutiae of the registered fingerprint minutiae and the minutiae type match.

Condition 2: The representative minutiae of the registered fingerprint minutiae and the minutiae pattern match or are similar.

Here, the consistency of the minutiae patterns can be determined by checking whether the two minutiae patterns are the same, and if they do not match, the similarity can be determined, for example, by checking whether the two minutiae patterns are the same. When two or more ridge pixels (that is, the pixel value is 1) are compared, the amount of change m bits of the corresponding bits is within the range of -w≦m≦+w (w is a parameter, w≧1.w The feature point pattern located at the bit position (the value of which is determined in advance) can be regarded as a similar feature point pattern.

■Condition 3: Consistency regarding the relative positions of the registered fingerprint minutiae and the test fingerprint minutiae is determined based on the range of the circle of the peripheral minutiae (in this example, as described in step A3, the radius V1 and The number of feature points for each feature point type (range defined by the circle of V2) must match or be similar. Here, the range on the feature points for determination is a parameter, and the value is set in advance. For example, information on peripheral minutiae points of a registered fingerprint minutiae is as follows: R11 = number of end points within a circle with radius V1, R12 = radius V
1, R13 = number of intersections within a circle with radius V1, R21 = number of end points outside the circle with radius V1 and within the circle with radius V2, R22 = outside the circle with radius V1 and radius V2
The number of branching points within the circle of R23 = the number of intersections outside the circle with radius V1 and within the circle with radius V2, and the information on the peripheral minutiae of the test fingerprint minutiae is Q11 = radius V
1, Q12 = number of branching points in a circle with radius V1, Q13 = number of intersections in a circle with radius V1, Q
21 = Number of end points outside the circle with radius V1 and inside the circle with radius V2, Q22 = Number of branch points outside the circle with radius V1 and inside the circle with radius V2, Q23 = Outside the circle with radius V1 and inside the circle with radius V2 Suppose that there is a certain number of intersections, then all i, j, (i=1, 2
;j=1,2),

[0058]

[Math 4]

The following is a condition for considering that the numbers of peripheral minutiae of each minutiae type match between the registered fingerprint minutiae and the test fingerprint minutiae. Mij is a parameter for the range of the number of feature points determined to be a match.

Step B2: Regarding the results of step B1, the range of reference points where the representative feature point is empty (that is, the range of reference points where no representative feature point exists) is excluded from comparison.
When the registered fingerprint minutiae and the test fingerprint minutiae match L or more representative minutiae points, which is the threshold value for initial matching, the initial matching (that is, matching using the representative minutiae points) is determined to be a match.

[0061] Here, L is a parameter, and the value of L is set in advance (L≧1). If the initial comparison is a match and the number of representative feature points matched by the initial match is two or more, the process proceeds to step B3. When the initial verification results in a mismatch, the registered fingerprint minutiae and the test fingerprint minutiae are determined to be a mismatch in the initial verification, and the process of step B is terminated, and the verification process is terminated.

[0062] If there is only one representative feature point among the registered fingerprint feature points, or if only one representative feature point matches in the initial verification, the verification process ends in step B1.

Step B3: In this step, if the number of representative minutiae points matched by the initial verification is two or more, coordinate transformation (translation and rotation) is performed for the inspection fingerprint minutiae to match the position with the registered fingerprint minutiae point. conduct.

Let the coordinate system of the tested fingerprint minutiae be (X, Y) and the coordinate system of the registered fingerprint minutiae be (u, v). Let the valid representative minutiae of the test fingerprint minutiae be (Xi, Yi), (i=1,
2,...,n). The representative minutiae of the registered fingerprint minutiae corresponding to these are (Ui, Vi), (i=1, 2,..., n
), n is the number of valid representative minutiae points of one fingerprint. The number of empty representative feature points is not included in n.

Next, the coordinate system (u, v) and the coordinate system (X,
Find the relationship Y). Observed values of each coordinate system (ui, vi)
, (Xi, Yi), (i=1, 2,..., n) to a, b
, g, h are approximated as follows.

[0066]

[Math 5]

[0067]

[Math 6]

[0068]

[Math 7]

[0069]

[Math. 8]

[0070]

[Math. 9]

[0071] In these equations, a, b, g,
h is calculated as follows.

First, m is calculated, and then a and b are calculated. Next, g and h are calculated using a and b. Next, the coordinates (
For (X, Y), the coordinates are transformed using the following equation, and the new coordinates (U, V) after the transformation are determined and used as the pixel address of the new inspection fingerprint minutiae.

[0073]

[Math. 10]

[0074] In step B3 of procedure B, only the rotation and parallel movement of the coordinates are considered, so

[0075]

[Math. 11]

It can be expressed as: m, a, b, g, h are calculated using the least squares method,

[0077]

[Math. 12]

From [0078], ∂F/∂a=0, ∂F/∂b=0,
This is the formula obtained by setting ∂F/∂g=0 and ∂F/∂h=0.

Next, all feature points are compared using procedure C. This is a process in which all minutiae points are compared with respect to the information on the newly inspected fingerprint minutiae points and the information on the registered fingerprint minutiae points, and the number of matches is output. Representative feature points that match have already been tested and therefore do not need to be retested, but are included in the number of feature points and the number of matched feature points.

[0080] Procedure C is shown below. FIG. 5 shows an outline of the processing flow of procedure C.

(Procedure C) Step C1: Registered fingerprint minutiae (position,
Select the feature point type, feature point pattern).

Step C2: Regarding the minutiae of the new test fingerprint, place it within a square range centered on the position corresponding to the pixel address of the minutiae of the registered fingerprint (the size of the square is a parameter and is determined in advance). For all new test fingerprint minutiae, the position, minutiae type, and minutia pattern of the minutiae of the same minutiae type as the registered fingerprint minutiae are calculated.
Find candidates corresponding to the minutiae points of the registered fingerprint.

Step C3: New test fingerprint features whose minutia type matches the minutiae type of the registered fingerprint minutiae and whose minutiae pattern matches or is similar to the registered fingerprint minutiae points from among the selected new test fingerprint minutiae candidates. Select points as candidates. Here, the matching or similarity of the feature point patterns is checked by the means described in the condition 2 of step B1.

[0084] If there are multiple new test fingerprint minutiae that have a similar minutiae pattern and are close to the registered fingerprint minutiae, among them, the distance (the type of distance can be selected arbitrarily, The new test fingerprint minutiae that is closest to the new test fingerprint (selected from 4-neighborhood distance, 8-neighborhood distance, Euclidean distance, etc.) is stored as a minutiae that matches the registered fingerprint minutiae.

When there are a plurality of new inspection fingerprint minutiae having the same distance from the target registered fingerprint minutiae, the verification of the registered fingerprint minutiae is postponed. If a determination cannot be made even after the postponed comparison, priority is given to the newly inspected fingerprint minutiae whose x-coordinates are close. When the x-coordinates are also the same, any one is selected as the new test fingerprint minutiae that matches the registered fingerprint minutiae. Here, the new inspection fingerprint minutiae that has once been deemed to match the registered fingerprint minutiae is not determined to match the other registered fingerprint minutiae again.

If there is no new test fingerprint minutiae that matches the registered fingerprint minutiae, there is no new test fingerprint minutiae corresponding to the registered fingerprint minutiae.

Step C4: Step C1 to Step C
3 is repeated for all registered fingerprint minutiae.

Step C5: As an evaluation standard for identity verification, calculate S1=number of matched feature points between registered fingerprint minutiae and new test fingerprint minutiae/number of registered fingerprint minutiae points,

[0089]

[Math. 13]

The registered fingerprint and the test fingerprint are considered to match when [0090]. T1 is a threshold for identity verification, and is determined based on actual measurement values (end of procedure C).

The present invention is not limited to the embodiments described above, but can also be applied to the following expansions or modifications.

Although the image feature point position confirmation method according to the present invention is applied to a feature point selected as a representative feature point among feature points, it can also be applied to any feature point. Although surrounding feature points within two circles centered on the feature point were used to match the representative feature point, the number of circles can be set freely.

[0093] Information on peripheral minutiae can be used as long as it is unaffected by parallel movement or rotation of the circle. In this embodiment, as information on peripheral minutiae, information for each minutiae type can be used. Although the number of feature points was used, for example, the Euclidean distance from the feature point at the center of the circle to the surrounding feature points for each feature point type (Euclidean distance D of arbitrary two points (xa, ya), (xb, yb) is

[0094]

[Math. 14]

It is defined as: ) can also be used, and furthermore, it is also possible to use these pieces of information together. Further, an embodiment of the present invention includes the use of a figure close to a circle (for example, a polygon) in place of the circle.

The format of the evaluation standard for identity verification in step C is arbitrary; for example, instead of S1 in step C5, S2=2.(number of matched feature points between registered fingerprint minutiae and new test fingerprint minutiae). It is also possible to use S2 based on /(number of registered fingerprint minutiae points+number of inspected fingerprint minutiae points).

In this example, the method of the present invention was applied to the minutiae points of the thinned fingerprint image, but it is not limited to the minutiae points of the thinned image, and the method of the present invention is applied to the minutiae points of the thinned fingerprint image. It is applicable even when matching is performed by finding feature points.

[0098]

[Effects of the Invention] As explained above, the method of the present invention enables the identification of feature points using information on surrounding feature points (for example, the total number of feature points for each feature point type) within a circle centered on the target feature point. Since matching is performed, it is possible to confirm the position of the feature points without being affected by image rotation or translation. It has the advantage of being able to confirm location.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an example of a fingerprint image processing system implementing the method of the present invention.

FIG. 2 is an explanatory diagram of a circle defining a partial area where peripheral feature points exist on an image memory.

FIG. 3 is a diagram showing an outline of the processing flow of procedure A.

FIG. 4 is a diagram showing an outline of the processing flow of procedure B.

FIG. 5 is a diagram showing an outline of the processing flow of procedure C.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1...Electronic device, 2...A/D converter, 3...Image memory, 4...CPU, 5...Memory, 6...Input device, 7...Imaging device.

Claims (4)

[Claims] Claim 1: Regarding two digitized images, one image is a registered image and the other image is a test image, and one or more feature points each of the registered image and the test image has on an image memory. Peripheral features, which are feature points that exist in partial areas that are divided by one or more circles centered on each feature point selected for verification and do not include the center of the circle; A method for confirming the position of a feature point in an image, the method comprising determining the relative positional consistency between one registered image feature point and one test image feature point selected as a matching target based on point information. 2. The image feature point position confirmation method according to claim 1, wherein the information on the surrounding feature points is the number of feature points for each feature point type of the surrounding feature points in each partial region. 3. Claim 1, wherein the information on the peripheral feature points is the sum of Euclidean distances from the feature point at the center of the circle to each peripheral feature point for each feature point type in each partial region. A method for confirming the location of feature points in the described image. [Claim 4] The electronic device is the one selected for verification.
One registered image feature point selected as the matching target and one inspection image feature point are determined by the means for determining the match of the feature point type and the match and similarity regarding the feature point pattern between the two registered image feature points and the one inspection image feature point. 2. The method for confirming the position of feature points in an image according to claim 1, further comprising determining the consistency of feature points in the test image.