JPH10334237A - Fingerprint collation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain sure and stable finger placing discrimination without mis- discrimination by extracting a characteristic only when a finger placing discrimination means discriminates a finger placed state and collating finger data with registration data. SOLUTION: An image input means 11 receives an image, a finger placing discrimination means 12 applies binarization processing to the image to divide the input image into a block consisting of 16×16 pixels, binarizes a contrast image by using a mean density of each block as a threshold level so as to discriminate whether or not a skin edge line in each block is made stable. In the case that many blocks are discriminated for finger placing, it is discriminated that a finger is placed and a characteristic discrimination means 13 is used to correct intermission or connection of a skin edge line by skin edge direction calculation processing and contrast image correction processing. A collation means 14 is used to calculate a degree of skin edge directions and a degree of coincidence of characteristic points thereby calculating how much the characteristic point is coincident with a characteristic point of a concerned person registered in advance. When the coincidence of the skin edge direction is smaller than a preset threshold level and when the coincidence of the characteristic point is larger than other threshold level, the concerned person is authenticated.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fingerprint matching device for personal identification.

[0002]

2. Description of the Related Art Conventionally, as this kind of technology, for example, IEICE Technical Report PRU87-52, 1987,
There is a “high-precision personal identification device using fingerprints” published on pp. 49-57.

FIG. 17 is a block diagram showing a fingerprint collating apparatus according to this conventional example. In the figure, 1 is an image input means, 2
Denotes a finger placement determination unit, 3 denotes a feature extraction unit, and 4 denotes a collation unit. As shown in FIG. 18, the image input means 1 is composed of a prism 1a, a light source 1b, and a camera 1c, and their positional relationship is as shown in FIG. When the finger is not placed, all the light emitted from the light source 1b is totally reflected by the prism surface 1a. On the other hand, when a finger is placed on the prism surface 1a, the light emitted from the light source 1b is totally reflected at the valley portion of the fingerprint (portion not in close contact with the prism), and the ridge portion of the fingerprint (contact with the prism). Scattered light). The light reflected by the prism surface 1a is transmitted to the camera 1
By receiving light at c, the entire image becomes a high-luminance image when the finger is not placed, and when the finger is placed, the luminance is high at the valley portion of the fingerprint and the luminance is high at the ridge portion. A low density fingerprint image is obtained. The image of the camera 1c is A / D converted and taken into the processing unit.

The feature extracting means 3 performs a binarizing process and a ridge direction detecting process. The ridge direction detecting process includes a block dividing process, a mask pattern scanning process, and a ridge direction calculating process. It is. The feature extracting means 3 performs a grayscale image correction process, a re-binarization process, a thinning process, and a feature point detection process. Further, the matching means 4
The ridge direction matching degree calculation processing, the feature point matching degree calculation processing, and the processing for determining whether or not the user is the principal are performed.

Next, the operation will be described with reference to FIGS. First, in step S1, an image is captured by the image input unit 1, and in step S2, the digital video input by the finger placement determination unit 2 is subjected to image processing to determine whether or not a finger is placed.
When the finger is not placed, the process returns to step S1 to capture the image again by the image input unit 1. When the finger is placed, the process proceeds to step S3, and the process by the feature extracting unit 3 is executed.

The specific processing contents of the finger placement determining means 2 are disclosed, for example, in Japanese Patent Publication No. 7-74747.
That is, some N × N pixel small areas are cut out from the gray-scale fingerprint image input by the image input unit 1, and a density histogram of each small area is created. FIG. 2 uses the average value μ of the entire density histogram created in each small area as a threshold value.
It is divided into a black class and a white class like 0, and the average values μ1 and μ2 of both classes are calculated. The calculated average value of both classes is substituted into the following equation (1) to calculate the degree of separation R. If the degree of separation R is larger than a predetermined threshold value, it is determined that the finger is placed in the small area. When it is determined that the finger is placed on all of the cut out small areas, the processing of the finger placement determining means 2 is terminated, and the processing of the next-stage feature extracting means 3 is executed. On the other hand, when it is determined that the finger is not placed even in one of all the small areas, the finger placement determining means 2 ends, and the process returns to the image input means 1 again.

R = (μ2-μ1) ÷ μ2 (1)

Next, in the feature extracting means 3, first, in step S3, a binarization process is performed to binarize the gray-scale fingerprint image. This is because the gray-scale fingerprint image is divided into 16 × 16 pixel blocks, the average density of each block is obtained, and the average density is set as a threshold, and “1” is given to a pixel having a higher density than the average density. Is given "0". Next, the binary image is divided into blocks of 16 × 16 pixels (step S4), and each block has 3 × 3 pixels corresponding to four directions of 0 °, 45 °, 90 °, and 135 ° as shown in FIG. (Step S)
5). By scanning the mask pattern, the frequency of the pixels that match each mask pattern (histogram for each direction) can be determined, and the ridge direction in each block can be set in the range of 0 to 180 degrees using the frequency of each mask pattern. It is determined (step S6). Steps S4 to S6 are ridge direction detection processing.

Next, in step S7, a gradation image correction process is performed to correct breaks or adhesion of ridges. First, the ridge direction of each block is quantized in eight directions as shown in FIG. 22 (results are, for example, as shown in FIG. 23), and the ridge directions quantized in these eight directions are along the ridge directions. K ×
This is realized by applying an average filter of one pixel to the original image. Next, in step S8, the grayscale image correction result is binarized again by the same method as the binarization processing in step S3.
Then, in step S9, the binarization result is thinned to obtain a line figure having a line width of one pixel. Then, step S1
At 0, a feature point detection process is performed, and a mask pattern of 3 × 3 pixels is scanned over this thinned image, and the point of interest is a value “1” and only one of the eight peripheral pixels has a value “1”.
Is extracted as the end point, and the coordinates of the point of interest such that the point of interest has the value “1” and only three of the eight surrounding pixels have the value “1” are extracted as the branch points. I do. At the same time, the ridge line extending from the feature point is traced to calculate the vector angle of the feature point (θ in FIG. 24).

After the processing of the feature extraction means 3 in steps S3 to S10 is completed in this way, the collation means 4 in the next stage first calculates the ridge direction coincidence in step S11, Calculate how well the overall flow directions of the lines match. This is because the ridge direction Vd (Bx, By) of each block detected by the ridge direction detection processing in steps S4 to S6 and the ridge direction Ed (Bx) of each block registered in advance by the subject.
x, By) are overlapped while translating in the horizontal direction in the range from -Px to + Px and in the vertical direction in the range from -Py to + Py, and searching for a position where the ridge direction is the best overall, The degree of coincidence is calculated.
Specifically, the matching degree Sd (px, py) at a certain shift amount (px, py) is calculated by the following equation (2).

Sd (px, py) = (1 / Bn) × {Vd (Bx + px, By + py) −Ed (Bx, By)} ² ByBx (2)

Here, Bn represents the number of overlapping blocks. Then, the smallest coincidence Sdmin (px, p
y) is used in the determination process in step S13 to be described later to determine whether the user is a person or another person.

Next, at step S12, a feature point coincidence calculation is performed, and the coordinates of the feature points such as the end points and the branch points detected by the feature point detection processing at step S10 and the vector angles thereof are registered in advance. Calculate the degree of coincidence with the coordinates and vector angles of the person's feature points. This is because when the distance between certain feature point coordinates detected in advance from the registered fingerprint and certain feature point coordinates detected from the collation fingerprint is short and the difference between the vector angles of these feature points is small, these feature points Is considered as a pair, the number of feature points of the pair is calculated with respect to the total number of feature points by the following equation (3), and this is defined as the degree of coincidence of feature points.

Sm = (2 × Cp) / (Cv + Ce) (3)

Here, Cv is the number of feature points detected from the collation fingerprint, Ce is the number of feature points detected from the registered fingerprint, Cp
Is the number of feature points determined as a pair between the registered fingerprint and the verification fingerprint. Finally, in the determination processing in step S13,
If the ridge direction coincidence Sdmin (px, py) is smaller than a preset threshold α and the feature point coincidence Sm is larger than another preset threshold β, it is determined that the person is the subject,
Otherwise, it is determined to be another person.

[0016]

In the conventional fingerprint collating apparatus, whether or not a finger is placed is determined based on the density histogram of a small area as described above. When only the light and dark fingerprint images are obtained, the degree of separation R shown in the above equation (1) becomes very small. On the other hand, when the light source has uneven brightness, even when the finger is not placed, the density histogram of the small area has a certain spread, so that the degree of separation R is relatively large. As a result, depending on the determination threshold for performing the finger placement determination, it is not possible to detect that a blurred finger is placed, or it is erroneously determined that the finger is placed due to unevenness of the light source even though the finger is not placed. There was a problem that judgment was sometimes made.

Further, in the conventional fingerprint collation device, even if the position where the finger is placed at the time of registration and the position where the finger is collated are shifted, the collation is performed as it is. Has been reduced, and the user is erroneously determined to be another person in spite of his / her own identity.

Further, the conventional fingerprint collating apparatus does not judge whether or not a finger is a bow-shaped pattern or whether only a fingertip is placed. In the case of only a bow-shaped crest or a fingertip of a finger, the degree of coincidence obtained with a correct positional deviation amount when calculating the ridge direction coincidence degree and the degree of coincidence deviating from the correct positional deviation amount in a direction orthogonal to the ridge line There is no significant difference in the degree of coincidence. For this reason, there has been a problem that an erroneous displacement amount is calculated due to some kind of noise, and the degree of coincidence is deteriorated even in the case of the persons themselves, and the person is erroneously determined to be another person.

Further, the conventional fingerprint matching apparatus treats all feature point pairs with the same weight when calculating the degree of coincidence of feature points, so that pseudo feature points generated by, for example, scratches, wrinkles, fading, etc. There is a problem in that it has an adverse effect and causes the person to be incorrectly distinguished from another person.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems. Even when a blurred finger is placed, it can be reliably detected and the finger is placed. It is an object of the present invention to provide a fingerprint matching device that can prevent erroneous determination that a finger is placed due to unevenness of a light source in spite of absence. In addition, it is possible to prevent erroneous determination when the position where the finger is placed at the time of registration and the position at the time of collation are misaligned, and even if there is an erroneous misalignment due to some noise, it can be reliably detected. It is an object of the present invention to provide a fingerprint collation device capable of performing reliable discrimination without being adversely affected by pseudo feature points generated by scratches, wrinkles, fading, and the like.

[0021]

According to a first aspect of the present invention, there is provided a fingerprint matching apparatus which processes a fingerprint image obtained by a sensor to extract feature data, and stores feature data unique to a user registered in advance. In the fingerprint matching device that determines whether or not the user is the user by comparing, a binarization processing unit that binarizes the fingerprint image, a plurality of fingerprint images binarized by the binarization processing unit A block division processing unit for dividing into blocks, a mask pattern scanning processing unit for creating a histogram for each direction by scanning a mask pattern corresponding to a predetermined angle in each block divided by the block division processing unit, and the mask A finger placement determination unit including a stability determination processing unit that determines the stability of the directional angle of each block from the directional histogram of each block based on the output of the pattern scanning processing unit; Means in which performs collation with the observed feature extraction and registration data upon detecting a state of being placed with the finger.

According to a second aspect of the present invention, in the first aspect of the present invention, the fingerprint collating device is connected to the finger placing judging means, and based on the judgment result of the stability judging processing section of the finger placing judging means, each of the blocks is stored in each block. A ridge direction calculation processing unit that calculates the ridge direction of the fingerprint of the fingerprint, a gray image correction processing unit that corrects the gray image along the ridge direction of the fingerprint calculated by the ridge direction calculation processing unit, and the gray image correction A re-binarization processing section for re-binarizing the grayscale image correction result in the processing section, a thinning section for thinning the result binarized again in the re-binarization processing section, and a thinning in the thinning section And a feature extracting unit including a feature point detecting unit that scans the obtained image and detects feature points of the image.

The fingerprint matching device according to the third aspect of the present invention processes the fingerprint image obtained by the sensor to extract characteristic data, and compares the extracted characteristic data with user-registered characteristic data registered in advance. In a fingerprint matching device that determines whether or not the person is the person, a binarization processing unit that binarizes the fingerprint image,
A block division unit for dividing the fingerprint image binarized by the binarization processing unit into a plurality of blocks, and a ridge direction calculation process for calculating a ridge direction of a fingerprint in each block divided by the block division unit Unit, a ridge direction storage unit that stores a calculation result obtained by the ridge direction calculation processing unit input a minute time ago, and a block-by-block calculation unit that calculates the ridge direction calculation processing unit from an image input at the current time. A ridge direction change detection processing unit that compares the ridge direction of the fingerprint with the ridge direction of the fingerprint for each block before the short time stored in the ridge direction storage unit and detects a change in the ridge direction; A finger placement determination unit including a spatial stability determination processing unit that detects a group of blocks having the same direction angle and that are adjacent to each other with respect to the blocks determined to have no change in the direction change detection processing unit and determines the degree of clumping; Equipped, said finger rest Means in which performs collation with the observed feature extraction and registration data upon detecting a state of being placed with the finger.

According to a fourth aspect of the present invention, in the third aspect of the present invention, the fingerprint collating device is connected to the finger placement judging means, and outputs the fingerprint based on the judgment result of the spatial stability judgment processing section of the finger placing judging means. A gray-scale image correction processing unit that corrects a gray-scale image along the ridge direction, a re-binarization processing unit that binarizes the gray-scale image correction result in the gray-scale image correction processing unit again, and the re-binarization processing unit. It includes a thinning unit for thinning the result of binarization again, and a feature point detecting unit for scanning an image thinned by the thinning unit and detecting a feature point of the image.

The fingerprint matching device according to the fifth aspect of the present invention processes the fingerprint image obtained by the sensor to extract characteristic data, and compares the extracted characteristic data with user-registered characteristic data registered in advance. In a fingerprint matching device that determines whether or not the person is the person, a binarization processing unit that binarizes the fingerprint image,
A block division processing unit that divides the image binarized by the binarization processing unit into a plurality of blocks, and scans each of the blocks divided by the block division processing unit with a mask pattern corresponding to a predetermined angle. A first feature extraction including a mask pattern scanning processing unit for creating a histogram for each direction, and a ridge direction calculation processing unit for calculating a ridge direction of a fingerprint in each block based on an output of the mask pattern scanning processing unit Means, and a position shift amount calculation unit for calculating a position shift amount of the finger at the time of registration and at the time of collation based on the output of the first feature extracting means,
A displacement presenter that displays the displacement calculated by the displacement calculator, and determines whether the displacement is within a predetermined range based on the output of the displacement calculator. A position shift amount detection unit including a position shift amount determination unit, and instructs the user to re-place a finger when the position shift amount exceeds a predetermined range. The collation with the registered feature data is performed to determine whether or not the user is the principal.

According to a sixth aspect of the present invention, in the fingerprint matching device of the fifth aspect, the fingerprint matching device is connected to the displacement amount detecting means, and the fingerprint collating device determines the fingerprint based on the result of the decision by the displacement amount determining section of the displacement amount detecting means. A gray-scale image correction processing unit that corrects a gray-scale image along the ridge direction, a re-binarization processing unit that binarizes the gray-scale image correction result in the gray-scale image correction processing unit again, and the re-binarization processing unit. A second feature extracting unit including a thinning unit for thinning the result of the binarization again, and a feature point detecting unit for scanning an image thinned by the thinning unit and detecting a feature point of the image; It is provided.

According to a seventh aspect of the present invention, there is provided a fingerprint collating apparatus which processes a fingerprint image obtained by a sensor to extract characteristic data, and compares the extracted characteristic data with user-registered characteristic data registered in advance. In a fingerprint matching device that determines whether or not the person is the person, a binarization processing unit that binarizes the fingerprint image,
A block division processing unit that divides the image binarized by the binarization processing unit into a plurality of blocks, and scans each of the blocks divided by the block division processing unit with a mask pattern corresponding to a predetermined angle. And a ridge direction calculation processing unit for calculating a ridge direction of a fingerprint in each block based on an output of the mask pattern scanning processing unit. Extracting a ridge direction value of a fingerprint for each block obtained from the same image of the fingerprint image divided into a plurality of blocks based on an output of the extracting means and the first feature extracting means, An autocorrelation calculating unit for calculating a difference between ridge direction values of fingerprints for each fingerprint to generate an autocorrelation map, and processing the autocorrelation map obtained by the autocorrelation calculating unit to place a fingertip or an arcuate pattern of Is determined by the fingertip / bow print determining unit that determines whether the finger is placed, and when the fingertip / bow print fingerprint determination unit determines that the finger is a finger of a finger print or an arc print, the result is displayed and the registration process is continued as it is. Fingertip / bow shape including the fingertip / bow print presenting part that asks the user outside or outside, that is, asks the user or the administrator to determine whether to perform the registration process again after temporarily interrupting and repositioning the finger or changing the finger. The fingerprint data is registered only in the case of not being determined as a fingertip / bow-shaped fingerprint in the registration mode, and only when there is an instruction to continue the registration from the outside, that is, the user or the administrator. Is to judge only the fingertip or not and request an external user or administrator to reposition the finger.

The fingerprint collating device according to the invention of claim 8 is the invention according to claim 7, wherein the fingerprint collating device is connected to the fingertip / bow print detection means, A gray-scale image correction processing unit that corrects a gray-scale image along the ridge direction of the fingerprint based on the determination result; a re-binarization processing unit that binarizes the gray-scale image correction result in the gray-scale image correction processing unit again; A thinning section for thinning the result of binarization again by the binarization processing section, and a feature point detecting section for scanning an image thinned by the thinning section and detecting a feature point of the image. 2 is provided with feature extraction means.

The fingerprint matching device according to the ninth aspect of the present invention processes the fingerprint image obtained by the sensor to extract characteristic data, and compares the extracted characteristic data with user-registered characteristic data registered in advance. A fingerprint matching device for determining whether or not the user is an identity person; a binarization processing unit that binarizes at least the fingerprint image; and a block that divides the fingerprint image binarized by the binarization processing unit into a plurality of blocks. A feature extracting unit including a dividing unit, a mask pattern scanning processing unit that scans a mask pattern corresponding to each predetermined angle in each of the blocks divided by the block dividing unit, and creates a histogram for each direction; Directional angle stability calculating means for calculating the directional angle stability of each block from the directional histogram of each block extracted by the extracting means; It is intended to reflect the feature point matching degree calculated using the calculation result as a weight of the reliability of the feature points.

[0030] According to a tenth aspect of the present invention, there is provided a fingerprint collating apparatus.
10. The ridge direction coincidence calculating unit, which is connected to the feature extracting unit and the direction angle stability calculating unit, and calculates the ridge direction coincidence of the fingerprint based on the output of the feature extracting unit. A feature point coincidence that calculates a feature point coincidence using the ridge direction coincidence calculated by the ridge direction coincidence calculator and the directional angle stability calculated by the directional angle stability calculator. A matching unit including a calculation unit and a determination processing unit that determines whether or not the user is a principal based on the matching degree of the feature points calculated by the feature point matching degree calculation unit.

According to the eleventh aspect of the present invention, there is provided a fingerprint collating apparatus comprising:
In a fingerprint matching device that determines whether or not the user is the user by processing the fingerprint image obtained by the sensor to extract feature data and comparing the feature data with pre-registered user-specific feature data, A binarization processing unit for binarizing an image, and a ridge for calculating a ridge direction of a fingerprint in each block of the fingerprint image which is divided into a plurality of blocks after being binarized by the binarization processing unit A direction calculation processing unit, a gray image correction processing unit that corrects a gray image along the ridge direction of the fingerprint based on the output of the ridge direction calculation processing unit, and a gray image correction result in the gray image correction processing unit. A feature extraction unit including re-binarization processing for binarization, and after dividing the binarized image before and after the grayscale image correction into a plurality of blocks, comparing the binarized image for each block, and calculating the number of points of change. Change point counting section for counting, the change point counting section Weighting means including a weight calculation processing unit for calculating a weight for each block in accordance with the number of change points calculated by the unit, wherein the weight of the block in which the feature point exists is used as the reliability of the feature point. This is reflected in the degree calculation.

[0032] According to a twelfth aspect of the present invention, there is provided a fingerprint collating apparatus.
12. The ridge direction coincidence calculating unit, which is connected to a feature extracting unit and a weight determining unit, and calculates a ridge direction coincidence of a fingerprint based on an output of the feature extracting unit. A feature point coincidence calculation that calculates a feature point coincidence using the coincidence in the ridge direction of the fingerprint calculated by the coincidence calculator and the weight for each block according to the number of change points calculated by the weight determining means And a collation unit including a judgment processing unit for judging the identity of the user based on the coincidence of the feature points calculated by the feature point coincidence calculation unit.

[0033]

BEST MODE FOR CARRYING OUT THE INVENTION

Embodiment 1 FIG. Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a configuration diagram showing Embodiment 1 of the present invention. In the figure, 11 is an image input means,
As the image input means 11, the same as the above-mentioned image input means 1 is used. Reference numeral 12 denotes a finger placement determination unit connected to the image input unit 11, 13 denotes a feature extraction unit connected to the finger placement determination unit 12, and 14 denotes a matching unit connected to the feature extraction unit 13. The finger placement determining means 12 performs a binarizing process and a ridge direction stability determining process. The ridge direction stability determining process includes a block dividing process, a mask pattern scanning process, and a stability determining process. Processing.

The feature extracting means 13 includes a ridge direction calculating process, a grayscale image correcting process, a re-binarizing process, a thinning process,
And feature point detection processing. Collation means 14
The same as the above-mentioned matching means 4 is used.

Next, the operation will be described with reference to FIGS. First, in step S21, an image is captured by the image input unit 11, and then the binarization process is first performed by the finger placement determination unit 12 (step S21).
22) As in the conventional case, the input image is divided into blocks of 16 × 16 pixels, and the grayscale image is binarized using the average density of each block as a threshold. Next, the binary image is divided again into blocks of 16 × 16 pixels (step S23), and a mask pattern of 3 × 3 pixels corresponding to four directions of 0 °, 45 °, 90 ° and 135 ° in each block (step S23). (See FIG. 21)
Is scanned (step S24). By scanning the mask pattern, the frequency of the pixels that match each mask pattern (histogram for each direction) can be known. Therefore, the stability determination process, that is, the ridge direction in the block is stably performed using the frequency of each mask pattern. Is determined (step S2).
5).

The stability determination process in step S25 is realized by, for example, calculating the following equation (4). As a result of the equation (4), the stability F in the ridge direction is smaller than a preset threshold value γ. In this case, the block is determined to be a background area (area where no finger is placed), and if the stability F in the ridge direction is larger than the threshold γ, it is determined that a finger is placed in the block. If there are many blocks in which it is determined that the finger is placed, it is determined that the finger is placed, and the processing (steps S26 to S30) in the next stage feature extracting means 13 is executed. If the number of blocks determined to be present is small, it is determined that the finger is not placed or the finger placement is insufficient, and the process returns to step S21.

F = | H0−H90 | + | H45−H135 | (4)

In the equation (4), H0, H45,
H90 and H135 are the frequencies in each direction in the direction-specific histogram created in the mask pattern scanning process in step S24.

As an example, in the case of the background area, as shown in FIG. 3, for example, as shown in FIG. 3, the number of pixels that match the pattern as a whole is small (see FIG. 3A), or the frequency in the orthogonal direction is almost the same. (See FIG. 3 (b)),
This is strange as a ridge. On the other hand, in the block on which the finger is placed, for example, as shown in FIG. 4, the frequency in one direction becomes dominant (see FIG. 4A), or the frequency in the adjacent direction becomes the same (see FIG. 4A). 4 (b)). Therefore, the stability F in the ridge direction of the above equation (4) decreases in the background region, and the stability F in the ridge direction increases in the region where the finger is placed. Thus, here, step S
Steps S23 to S25 are the ridge direction detection processing in steps S23 to S25.

Next, in step S26, the ridge direction calculation processing is performed by the feature extraction means 13 in the same manner as described above.
The ridge direction for each block is set to 0 using the histogram for each direction.
From 180 degrees. Next, in step S27, a gradation image correction process is performed to correct breaks or adhesion of ridges. First, the ridge direction of each block is quantized into eight directions (see FIGS. 22 and 23).
This is realized by applying an average filter of K × 1 pixels along the ridge direction quantized in the direction to the original image.

Next, in step S28, the grayscale image correction result is binarized again in the same manner as in the binarization processing in step S22, and in step S29, the binarization result is thinned to a line width of one pixel. Get a line figure. Then, in step S30, a feature point detection process is performed,
A mask pattern of 3 × 3 pixels is scanned over this thinned image, and the point of interest is a value “1” and one of eight peripheral pixels is one.
The coordinates of the point of interest such that only the pixel has a value of “1” are extracted as an end point. Extract coordinates as branch points. At the same time, the ridge line extending from the feature point is traced to calculate the vector angle of the feature point (θ in FIG. 24).

Thus, steps S26 to S30
When the processing of the feature extracting means 13 in the above is completed, the collating means 14 in the next stage firstly performs the processing in step S3 as described above.
In 1, the ridge direction matching degree calculation is performed to calculate how much the overall flow direction of the ridges matches. This is because the ridge direction Vd (Bx, Bx) for each block detected by the ridge direction detection process in step S26.
y) and the ridge direction Ed (Bx, By) of each block registered in advance by the subject in the horizontal direction from -Px to + P
The overlapping is performed while moving in parallel in the range of x and the range of -Py to + Py in the vertical direction, a position where the ridge direction is the best overall matches, and the degree of matching at that time is calculated. Specifically, a certain shift amount (px, py)
The degree of coincidence Sd (px, py) at the time of
Is calculated.

Next, in step S32, the feature point coincidence calculation is performed, and the coordinates of the feature points such as the end points and the branch points detected in the feature point detection processing in step S30 and their vector angles are registered in advance. Calculate the degree of coincidence with the coordinates and vector angles of the person's feature points. This is because when the distance between certain feature point coordinates detected in advance from the registered fingerprint and certain feature point coordinates detected from the collation fingerprint is short and the difference between the vector angles of these feature points is small, these feature points Is considered as a pair, the number of feature points of the pair is calculated with respect to the total number of feature points by the above equation (3), and this is defined as the coincidence of feature points. Finally, in the determination processing in step S33, the ridge direction coincidence Sdmin (px, py) is smaller than the preset threshold α and the feature point coincidence Sm is larger than another preset threshold β. The person is determined to be the other person, and the others are determined to be others.

As described above, in this embodiment, it is determined whether or not the finger is placed based on whether or not the detected ridge direction is sufficiently stable, that is, when detecting the ridge direction. The ridge direction stability can be evaluated using the directional component ratios of 0 °, 45 °, 90 °, and 135 ° obtained for each block in the executed mask pattern scanning.

Embodiment 2 FIG. 5 is a configuration diagram showing a second embodiment of the present invention. In FIG. 5, portions corresponding to those in FIG. 1 are denoted by the same reference numerals, and detailed description thereof will be omitted. In the figure, reference numeral 12A denotes a finger placement determination unit connected to the image input unit 11, and 13A denotes a feature extraction unit connected to the finger placement determination unit 12A. Finger placement determination means 12A
Performs a binarization process and a ridge direction stability determination process. The ridge direction stability determination process includes a block division process, a mask pattern scanning process, a ridge direction calculation process, and a ridge direction calculation process. It includes a line direction change detection process and a spatial stability determination process. Further, the finger placement determining means 12A has a ridge direction storage function. The feature extracting unit 13A performs a grayscale image correction process, a re-binarization process, a thinning process, and a feature point detection process.

Next, the operation will be described with reference to FIG. In step S21, the image input unit 11
The image is captured by the
A, steps S21 to S24 are the same as those in FIG.
Steps S21 to S24 in the finger placement determining means 12
By scanning the mask pattern, the frequency (pixel-by-direction histogram) of pixels that match each mask pattern can be determined, and the ridge direction of each block can be changed from 0 using the frequency of each mask pattern. It is determined within a range of 180 degrees (step S34). This step S34
In this example, the ridge direction for each block obtained by the same method from the image input a minute time ago is stored.

Next, in step S35, ridge direction change detection processing is performed, and the ridge direction of each block obtained from the image at the current time and the ridge direction of each block obtained from the image in step S34 are determined. After quantization in eight directions (see FIG. 22) (the one quantized in eight directions is called a direction code), the two are compared for each block to determine whether the direction has changed. Then, the direction code is held as it is for a block in which the ridge direction has not changed, and a special code (for example, -1) indicating the change is given to the changed block.

Further, in step S36, a spatial stability determination process is performed, and the same code group having the same direction code and adjacent to each other is detected for the block determined to have no change in step S35. The average degree of lump (lump area) of the same code group is obtained. If the degree of clumping is larger than the predetermined value, it means that the ridge direction is stable both temporally and spatially, so it is determined that the finger is placed and step S
The process proceeds to 27, where the processing in the feature extracting means 13A is executed. On the other hand, if the degree of clumping is smaller than the predetermined value or if there are many blocks in which the direction code has changed, it is determined that the finger is not placed in a sufficient state, and the process returns to step S21.

This is because when the finger is placed in a sufficient state, a stable ridge direction can be detected, and it does not change for a very short time. It is based on the idea that it will be unstable with time. Thus, here, steps S22 to S24, step S34
Steps S36 to S36 are processes in the finger placement determination unit 12A,
Each step other than step S22 is the ridge direction stability determination processing.

The subsequent processing of steps S27 to S30 in the feature extracting means 13A is the same as the processing of steps S27 to S30 in the above-described feature extracting means 13 of FIG. The processing in steps S31 to S33 is also the same as the processing in steps S31 to S33 in the matching unit 14 in FIG. 1, and a description thereof will be omitted.

As described above, in this embodiment, it is determined whether or not the finger is placed based on whether or not the detected ridge direction is sufficiently stable, that is, the block detected at a certain time t. The ridge direction of each block is compared with the ridge direction of each block detected at the time t-Δ before the short time Δ, and the direction is stable temporally and spatially stable. Can be evaluated.

Embodiment 3 FIG. FIG. 7 is a configuration diagram showing a third embodiment of the present invention. In FIG. 7, parts corresponding to those in FIGS. 1 and 5 are denoted by the same reference numerals, and detailed description thereof will be omitted. In the figure, 12B is the image input means 1
1 is a finger placement determining means connected to 1; 13B is a feature extracting means (first feature extracting means) connected to the finger placing determining means 12B; 15 is a displacement detecting means connected to the feature extracting means 13B; The feature extracting means 13A (second feature extracting means) and the collating means 1 are provided after the displacement detecting means 15.
4 are cascaded.

As the finger placement judging means 12B, the same one as shown in FIG. 17 may be used here. The feature extracting unit 13B performs a binarization process and a ridge direction detection process. The ridge direction detection process includes a block division process,
A mask pattern scanning process and a ridge direction calculation process are included. Further, the displacement detecting means 15 performs a displacement calculation process, a displacement presentation process, and a displacement determination process.

Next, the operation will be described with reference to FIGS. In step S21, an image is captured by the image input unit 11, and in step S37, as in step S2 in FIG.
Image processing is performed on the digital video input by 2B to determine whether a finger is placed. When the finger is not placed, the process returns to step S21 to take in the image again by the image input unit 11, and when the finger is placed, the process proceeds to step S22 to execute the processing by the feature extracting unit 13B.

In the feature extracting means 13B, step S22
In steps S21 to S24, the same processing as steps S21 to S24 in the finger placement determination unit 12 in FIG. 1 described above is performed, and in step S26, the frequency of pixels that match each mask pattern by scanning the mask pattern (in each direction) Histogram), the ridge direction of each block is obtained in the range of 0 to 180 degrees using the frequency of each mask pattern. Thus, here, steps S22 to S2
6 is processing in the feature extraction means 13B, and steps S23 to S26 in the processing are ridge direction detection processing.

Next, in step S38, the ridge direction Vd (Bx, By) for each block calculated in the ridge direction detection processing in step S26 is calculated in the same manner as the ridge direction coincidence calculation processing in step S31. , Ridge direction Ed (Bx, B
y) are overlapped while translating in the horizontal direction in the range from -Px to + Px and in the vertical direction in the range from -Py to + Py, and searching for a position where the ridge direction is the best overall. Specifically, the matching degree Sd (px, py) at a certain shift amount (px, py) is calculated by the following equation (5).

Sd (px, py) = (1 / Bn) × {Vd (Bx + px, By + py) −Ed (Bx, By)} ² ByBx (5)

Here, Bn represents the number of overlapping blocks. Then, px and py at which Sd is minimized are determined, and are presented to the user as the amount of displacement of the finger placement. As a method of presenting the user, for example, as shown in FIG. 9, a position deviation allowable range 15b is drawn on a monitor or an LCD screen 15a as a position deviation amount presenting means, and the current finger position 15c is determined from the finger position at the time of registration. Display in relative blocks. In the case of the figure, a display example is shown in the case where a positional shift of px = 3 blocks and py = 3 blocks is detected.

Next, in step S39, it is determined whether or not the amount of positional deviation calculated in step S38 falls within the allowable range of positional deviation 16b shown in FIG. , The process proceeds to step S27, and the process in the next stage feature extraction unit 13A is executed.
If not, the process returns to step S21. On the other hand, the user shifts the finger position while viewing the position shift amount presentation screen 15a such that the finger placement position 15c falls within the position shift allowable range 15b.

The subsequent processing of steps S27 to S30 in the feature extracting means 13A is the same as the above-described processing of steps S27 to S30 in the feature extracting means 13 in FIG. 1 or 13A in FIG. The processing of steps S31 to S33 in the next stage collating means 14 is the same as the processing of steps S31 to S33 in the collating means 14 of FIG. 1 or FIG.

As described above, in this embodiment, by presenting to the user how much the current finger placement position is different from the position at the time of registration, the user can place the finger at the same position as at the time of registration. I can do it. That is, the ridge direction matching degree calculation is executed immediately after the ridge direction detection processing, and the resulting positional deviation amount can be presented to the user via the user interface device. In the above description, a visual presentation means such as a monitor or an LCD screen is used. However, the same applies to a case where the sound is presented in accordance with the amount of misalignment, or an LED is regularly arranged.

Embodiment 4 FIG. 10 is a configuration diagram showing a fourth embodiment of the present invention. In FIG. 10, parts corresponding to those in FIGS. 1, 5, and 7 are denoted by the same reference numerals, and detailed description thereof will be omitted. In the figure, reference numeral 16 denotes a fingertip / fingerprint detection means connected to the feature extraction means 13B (first feature extraction means 1), which performs an autocorrelation calculation process, a fingertip / bow print determination process, and a fingertip / bow print process. This is to perform a pattern printing process. At the subsequent stage of the fingertip / fingerprint detecting means 16, a feature extracting means 13A (second feature extracting means) and a collating means 14 are connected in cascade.

Next, the operation will be described with reference to FIG. Steps S21, S37 and S22 to S
Since the processing of 26 is the same as that of FIG. 8,
The description is omitted. Now, fingertip / fingerprint detection means 16
Calculates the autocorrelation in step S40 using the direction angle of each block detected in step S26 as an input. That is, a copy is created in the ridge direction Vd (Bx, By) for each block detected in step S26, and the blocks are translated in the horizontal direction within a range from -Px to + Px and vertically in the range from -Py to + Py. Then, the degree of coincidence in the ridge direction at each shift amount is calculated by the following equation (6).

Sd (px, py) = (1 / Bn) × {Vd (Bx + px, By + py) −Vd (Bx, By)} ² ByBx (6) where Bn represents the number of overlapping blocks.

Next, in step S41, the fingertip / bow print detection means 16 determines whether the current time is the time of collation or the time of registration. If the time is collation, in step S42, the result of the autocorrelation calculated in step S40 is determined. Use to judge only that the fingertip is placed. Then, when it is determined that the fingertip is placed, in step S43, the fact is presented to the user or the administrator by a fingertip / bow print presenting means (not shown) such as a monitor or an LCD screen. When presented, the user or the like repositions the finger so that the entire surface of the finger touches the prism. On the other hand, as processing
The process returns to step S21 in order to input and process the replaced fingerprint image again.

On the other hand, if it is not at the time of collation at step S41, that is, at the time of registration, at step S44,
Similarly, it is determined whether the fingertip is placed or the finger with the bow is placed using the calculation result of the autocorrelation in step S40. Specifically, since all ridges of a fingertip or bow-shaped crest are nearly straight and flow in parallel,
Even if the degree of coincidence in the ridge direction is calculated while being shifted in the direction orthogonal to the ridge, the value hardly changes. This means that among the results of the autocorrelation calculation, the result Sd when shifted by one block in the direction orthogonal to the ridge and the result Sd when shifted by several blocks output almost the same value. Therefore, the determination by the fingertip / bow-shaped pattern detecting means 16 is based on the S when shifting by one block in the direction orthogonal to the ridge.
d is compared with Sd when it is shifted by several blocks, and when the difference is small, it is determined that the fingertip or the bow is placed.

When it is determined in step S44 that the fingertip or the bow is placed, the monitor and the L
This is presented to the user or the like by a fingertip / bow print presenting means such as a CD screen. Next, when the user or the like is presented, it is determined in step S46 whether there is an instruction to change the process, that is, whether the user or the like replaces the finger with the finger as it is, changes the finger, or Determine whether to continue registration as it is, and if there is an instruction to change the process,
That is, when the user or the like replaces the finger with the finger as it is or changes the finger, the process returns to step S21 to repeat the above operation, and when there is no instruction to change the process,
If the user or the like continues registration with the finger as it is, the process proceeds to step S27.

If it is not determined in step S42 that the finger is a fingertip, and if it is not determined in step S44 that the finger is a fingertip or an arcuate pattern, the process proceeds to step S27. Run. Steps S27 to S3 in the subsequent feature extraction means 13A
0 is performed in steps S27 to S3 in the feature extraction unit 13 in FIG. 1 or the feature extraction unit 13A in FIG.
0, and the processing in steps S31 to S33 in the next stage collating means 14 is the same as steps S31 to S33 in the collating means 14 in FIG. 1 or FIG.

As described above, in this embodiment, the autocorrelation is calculated using the ridge direction obtained by the ridge direction detection processing, so that the finger placed at the time of registration is an arched pattern or not. To determine whether only the fingertip is placed, to determine whether the fingertip is placed at the time of collation, and to request the user to change or reposition the finger in the case of a bow or a fingertip Thus, it is possible to avoid a state where misjudgment is likely to occur.

The feature extracting means 13A and the collating means 14 operate in the same manner as the feature extracting means 13A and the collating means 14 shown in FIGS. 5 and 7, except that the collating means 14 is not executed at the time of registration but is extracted. Needless to say, the ridge direction and the feature point are stored.

Embodiment 5 FIG. FIG. 12 is a configuration diagram showing a fifth embodiment of the present invention. 12, parts corresponding to those in FIGS. 1, 5, and 10 are denoted by the same reference numerals, and detailed description thereof will be omitted. In the figure, 13C is a feature extraction unit connected to the finger placement determination unit 12B, 17 is a direction angle stability calculation unit connected to the feature extraction unit 13C, and 14A is a feature extraction unit 13C and the direction angle stability calculation unit 17 This is the connected collation means.

As the feature extracting means 13C, the same one as shown in FIG. 17 may be used here. The direction angle stability calculation means 17 calculates how stable the ridge direction of each block is, using the histogram for each block obtained by the above-described mask pattern scanning process. The matching means 14A uses the directional angle stability calculated by the directional angle stability calculating means 17 in addition to the ridge direction coincidence calculating processing and the processing for determining whether or not the user is the same as described above. Perform calculation processing.

Next, the operation will be described with reference to FIG. Steps S21, S37 and S22-
The processing in S26 is the same as that in the case of FIG. Also, the processing in steps S27 to S30 after the ridge direction calculation processing in step S26 is the same as that in the case of FIG. By the way, the direction angle stability calculation means 17 determines in step S4
In step 7, the degree of ridge direction of each block is calculated using the histogram for each block obtained in the mask pattern scanning process in step S24.

The direction angle stability calculation means 17 uses the direction-specific histogram for each block obtained in the mask pattern scanning process in step S24 to determine how stable the ridge direction of each block is. Is calculated. Specifically, the number of pixels matching the mask pattern indicating the 0-degree direction is H0, the number of pixels matching the mask pattern indicating the 45-degree direction is H45, and the number of pixels matching the mask patterns indicating the 90-degree and 135-degree directions. Are H90 and H135, the ridge direction stability F for each block is calculated by the following equation (7).

F = | H0−H90 | + | H45−H135 | (7)

Next, in the same manner as described above, first, in step S31, the matching means 14A calculates the degree of coincidence of the ridge direction, and calculates the degree of coincidence of the overall flow direction of the ridge. That is, the ridge direction of each block is compared with the ridge direction of the person registered in advance, and the degree of coincidence is calculated. In step S48, the degree of coincidence of the feature points is calculated using the ridge direction stability F for each block calculated by the directional angle stability calculation means 17. That is, as described above, a paired feature point that matches the registered feature point and the matching feature point is found. On the other hand, the stability F in the ridge direction of the block to which each feature point belongs is given to each feature point as a weight related to reliability, and the feature point coincidence is calculated by the following equation (8) using the weight given to the feature point. Calculate the degree.

Sm = Σ (Fpe + Fpv) / (ΣFe + ΣFv) (8)

Here, Fe is the weight of each registered feature point, and Fv is the weight of each feature point detected from the collation fingerprint. Fpe is the weight on the registered feature point side, and Fpv is the weight on the collated feature point side, out of the feature points paired between the registered feature point and the matching feature point. It is assumed that the weights of the registered feature points are determined by the same processing at the time of registration, and are stored together with other feature data. Finally, in the determination processing in step S33, the user is determined using the result of the ridge direction coincidence calculation processing (step S31) similar to that described above and the result of the characteristic point coincidence calculation processing (step S48) unique to the present embodiment. Determine whether or not. In other words, when the ridge direction coincidence Sdmin (px, py) is smaller than a preset threshold α and the feature point coincidence Sm is larger than another preset threshold β, it is determined that the user is a person,
Otherwise, it is determined to be another person.

As described above, in this embodiment, the reliability of a feature point is evaluated using the stability of the ridge direction around the feature point, and the degree of matching of the feature points is calculated based on the evaluation result. It is possible to improve the ability of discriminating a person or another person.

Embodiment 6 FIG. FIG. 14 is a configuration diagram showing Embodiment 6 of the present invention. In FIG. 14, the same reference numerals are given to portions corresponding to FIGS. 1, 5, 10, and 12, and the detailed description thereof will be omitted. In the figure, 1
8 is a weight determining means connected to the feature extracting means 13C,
4B is a matching unit connected to the feature extracting unit 13C and the weight determining unit 18. The weight determining means 18 performs a change point counting process and a weight calculating process, compares the binary images before and after the grayscale image correction process, evaluates the stability of the image for each block, and determines the weight of the feature point. Also, the collating means 14
B calculates the degree of coincidence of the feature points using the weight for each block calculated by the weight determining means 18 in addition to the ridge direction coincidence degree calculation processing and the identity determination processing as described above.

Next, the operation will be described with reference to FIG. Steps S21, S37 and S22-
The processing in S26 is the same as that in the case of FIG. Also, the processing in steps S27 to S30 after the ridge direction calculation processing in step S26 is the same as that in the case of FIG. Now, in steps S49 and S50, the weight determination means 18 compares the binary images before and after the gray-scale image correction processing, evaluates the stability of the image for each block, and determines the weight of the feature point.

Specifically, for example, the gray scale image of a certain block before the gray scale image correction is as shown by reference numeral 100 in FIG.
Suppose that it has been modified to the one shown in 2. The respective binarization results at that time are as shown by reference numerals 104 and 106. Then, the change point counting process (step S4)
In 9), the difference between these binary images is obtained to obtain a result indicated by reference numeral 108, and the number of pixels corresponding to the change is counted.
Also, odd reference numbers 101, 103, 105, 107
The same applies to the cases 109 and 109.

Next, a weight calculation process (step S50)
Then, the weight W representing the reliability of each block is calculated using the number of changed pixels for each block, which is the result of the change point counting process. In this case, for example, as shown in the following equation (9), the weight is large when the number of changed pixels is small, and the weight is small when the number of changed pixels is large.

W = 1− (G / Q) (9) Here, G is the number of changed pixels for each block as a result of the change point counting process, and Q is the area of the block.

Next, in the same manner as described above, first, in step S31, the collating means 14B calculates the degree of coincidence of the ridge direction, and calculates the degree of coincidence of the entire flow direction of the ridge. That is, the ridge direction of each block is compared with the ridge direction of the person registered in advance, and the degree of coincidence is calculated. Then, in step S51, the degree of coincidence of the feature points is calculated using the weight W for each block determined by the weight determining means 18. That is, a paired feature point that matches the registered feature point and the matching feature point is found by the same method as described above. On the other hand, the weight W of the block to which each feature point belongs is given to each feature point as a weight relating to reliability, and the feature point coincidence is calculated by the following equation (10) using the weight given to the feature point.

Sm = Σ (Wpe + Wpv) / (ΣWe + ΣWv) (10)

Here, We is the weight of each registered feature point, and Wv is the weight of each feature point detected from the collation fingerprint. Wpe is the weight on the registered feature point side, and Wpv is the weight on the collated feature point side, of the feature points paired between the registered feature point and the matching feature point. It is assumed that the weights of the registered feature points are determined by the same processing at the time of registration, and are stored together with other feature data.

Finally, in the determination processing in step S33, the ridge direction coincidence degree calculation processing (step S3
Using the result of 1) and the result of the characteristic point matching degree calculation processing (step S51) unique to the present embodiment, it is determined whether or not the user is the principal. That is, the ridge direction coincidence Sdmin (px, py) is smaller than the preset threshold α, and the feature point coincidence S
If m is larger than another preset threshold value β, it is determined that the person is the person, otherwise, it is determined that the person is another person.

As described above, in this embodiment, the reliability of a feature point is evaluated using the stability of the ridge direction around the feature point and the stability of the binary image, and the evaluation result is used as a weight. By calculating the degree of point matching, it is possible to improve the ability of discriminating a person or another person.

[0090]

As described above, according to the present invention, a binarization processing unit for binarizing a fingerprint image, and the fingerprint image binarized by the binarization processing unit is divided into a plurality of blocks. A block division processing unit, a mask pattern scanning processing unit that scans a mask pattern corresponding to each predetermined angle in each block divided by the external block division processing unit to create a histogram for each direction, and the mask pattern scanning processing unit And a finger placement determination unit including a stability determination processing unit that determines the stability of the directional angle of each block from the directional histogram of each block based on the output of the block. Only when the state is detected, feature extraction and collation with the registered data are performed, so that the finger is not fixed with a blurred finger or erroneously determined that the finger has not been placed, and is stably and reliably placed. There is an effect that it is judgment.

Further, according to the present invention, a binarization processing section for binarizing a fingerprint image, a block division section for dividing the fingerprint image binarized by the binarization processing section into a plurality of blocks, A ridge direction calculation processing unit that calculates a ridge direction of a fingerprint in each block divided by the block division unit, and a ridge direction storage that stores a calculation result obtained by the ridge direction calculation processing unit input a short time ago. Unit, the ridge direction of the fingerprint for each block calculated by the ridge direction calculation processing unit from the image input at the current time and the fingerprint of the block for a minute time ago stored in the ridge direction storage unit. A ridge direction change detection processing unit that compares the ridge direction with the ridge direction to detect a change in the ridge direction, and blocks that have the same direction angle and are adjacent to each other with respect to the blocks determined to be unchanged by the ridge direction change detection processing unit Detecting groups and how much they are A finger placement determination unit including a spatial stability determination processing unit for determination is provided, and only when the finger placement determination unit detects a state where the finger is placed, the feature extraction and the comparison with the registered data are performed. There is an effect that a stable and reliable finger placement determination can be performed without erroneously determining that the finger has not been set or the finger has not been placed.

Further, according to the present invention, a binarization processing section for binarizing a fingerprint image, a block division processing section for dividing an image binarized by the binarization processing section into a plurality of blocks, A mask pattern scanning processing unit that scans a mask pattern corresponding to each predetermined angle in each block divided by the block division processing unit and creates a histogram for each direction, based on an output of the mask pattern scanning processing unit A first feature extraction unit including a ridge direction calculation processing unit for calculating a ridge direction of a fingerprint in each block; and a positional shift of the finger between registration and collation based on an output of the first feature extraction unit. A position shift amount calculation unit for calculating the amount, a position shift amount presentation unit for displaying the position shift amount calculated by the position shift amount calculation unit, and a position shift amount based on an output of the position shift amount calculation unit. Whether it is within the specified range A position shift amount detection unit including a position shift amount determination unit for determining, and when the position shift amount exceeds a predetermined range, an external user or an administrator is instructed to reposition a finger, and the Only when it is within the range, it is compared with the previously registered feature data and it is determined whether or not the user is the subject. Has the effect of increasing the probability that feature points will be paired when calculating, and improving the ability to discriminate between the subject and another person.

Further, according to the present invention, a binarization processing section for binarizing a fingerprint image, a block division processing section for dividing an image binarized by the binarization processing section into a plurality of blocks, A mask pattern scanning processing unit that scans a mask pattern corresponding to each predetermined angle in each block divided by the block division processing unit and creates a histogram for each direction, based on an output of the mask pattern scanning processing unit A first feature extraction unit including a ridge direction calculation processing unit for calculating a ridge direction of a fingerprint in each block, and a fingerprint image divided into a plurality of blocks based on an output of the first feature extraction unit An autocorrelation calculation unit that calculates the difference between the ridge direction values of the fingerprint of each block while shifting the positions of the ridge direction values of the fingerprint of each block obtained from the same image to generate an autocorrelation map, Autocorrelation A fingertip / bow print determining unit that processes the autocorrelation map obtained by the arithmetic unit to determine whether a fingertip is placed or a finger with an arch print is placed, When it is determined that the finger is an arched finger, whether to display the result and continue the registration processing as it is, or to suspend the registration once and then re-register the finger and then perform the registration processing again A fingertip / bow print detection unit including a fingertip / bow print presentation unit for requesting a determination from the outside, that is, a user or an administrator, and in the registration mode, when the fingertip / bow print is not determined, The feature data is registered only when the user or the administrator instructs to continue the registration, and in the verification mode, only the determination of the fingertip is performed, and the user or the administrator is asked to replace the finger. Judgment of misregistration amount when calculating line direction coincidence Eliminates Rukoto, than Te, can be expected is higher matching degree in the case between the person in the calculation of the subsequent characteristic points matching degree, the effect of improving the person and others discrimination performance.

According to the present invention, at least a binarization processing section for binarizing a fingerprint image, a block division section for dividing the fingerprint image binarized by the binarization processing section into a plurality of blocks, A feature extraction unit including a mask pattern scanning processing unit that scans a mask pattern corresponding to each predetermined angle in each block divided by the block division unit and creates a histogram for each direction; Directional angle stability calculating means for calculating the stability of the directional angle of each block from the extracted histogram for each direction of the block; Since it is used as the weight of the gender and reflected in the feature point coincidence calculation, for example, the weight of the pseudo feature point generated due to scratches, wrinkles, blurring, etc. can be reduced,
This has the effect of improving the ability to discriminate between the person and another person.

Further, according to the present invention, a binarization processing section for binarizing a fingerprint image, and each block of the fingerprint image divided into a plurality of blocks after being binarized by the binarization processing section A ridge direction calculation processing unit that calculates the ridge direction of the fingerprint in the
A gray-scale image correction processing unit that corrects a gray-scale image along the ridge direction of a fingerprint based on the output of the ridge direction calculation processing unit, and a binarization unit that binarizes the gray-scale image correction result in the gray-scale image correction processing unit again A feature extraction unit including a binarization process; and a change point counting process of dividing the binarized image before and after the grayscale image correction into a plurality of blocks, comparing the binarized images for each block, and counting the number of change points. And a weight determination unit including a weight calculation processing unit that calculates a weight for each block in accordance with the number of change points calculated by the change point counting processing unit. Is reflected in the feature point coincidence calculation, so that the weight of pseudo feature points caused by, for example, scratches, wrinkles, blurring, etc. can be reduced, thereby improving the discrimination ability between the subject and others. is there.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a first embodiment of the present invention.

FIG. 2 is a flowchart for explaining the operation of the first embodiment of the present invention;

FIG. 3 is a diagram showing an example of a mask pattern frequency in a background region according to the first embodiment of the present invention.

FIG. 4 is a diagram illustrating an example of a mask pattern frequency in a region where a finger is placed according to the first embodiment of the present invention;

FIG. 5 is a configuration diagram showing a second embodiment of the present invention.

FIG. 6 is a flowchart for explaining the operation of the second embodiment of the present invention;

FIG. 7 is a configuration diagram showing a third embodiment of the present invention.

FIG. 8 is a flowchart for explaining the operation of the third embodiment of the present invention;

FIG. 9 is a diagram illustrating a display state of a displacement amount according to the third embodiment of the present invention;

FIG. 10 is a configuration diagram showing a fourth embodiment of the present invention.

FIG. 11 is a flowchart for explaining the operation of the fourth embodiment of the present invention;

FIG. 12 is a configuration diagram showing a fifth embodiment of the present invention.

FIG. 13 is a flowchart for explaining the operation of the fifth embodiment of the present invention;

FIG. 14 is a configuration diagram showing a sixth embodiment of the present invention.

FIG. 15 is a flowchart for describing an operation according to the sixth embodiment of the present invention;

FIG. 16 is a diagram illustrating a change in a binarized image according to the sixth embodiment of the present invention.

FIG. 17 is a configuration diagram showing a conventional fingerprint matching device.

FIG. 18 is a diagram illustrating an example of a fingerprint image input unit.

FIG. 19 is a flowchart for explaining the operation of a conventional fingerprint matching device.

FIG. 20 is a diagram exemplifying a density histogram of a small area in a finger placement determination unit of a conventional fingerprint matching device.

FIG. 21 is a diagram illustrating eight directions when a ridge direction is quantized into eight directions.

FIG. 22 is a diagram illustrating an example of a mask pattern used when calculating a ridge direction.

FIG. 23 is a diagram illustrating a ridge direction detection result.

FIG. 24 is a diagram illustrating end points / branch points and vector angles.

[Explanation of symbols]

11 image input means, 12, 12A, 12B finger placement determination means, 13, 13A, 13B, 13C feature extraction means, 14, 14A, 14B collation means, 15 displacement amount detection means, 16 fingertip / bow print detection means, 17 directional angle stability calculating means, 19 weight determining means.

Claims (12)

[Claims] 1. A fingerprint which determines whether or not a user is a user by extracting a feature data by processing a fingerprint image obtained by a sensor and comparing the feature data with feature data unique to a user registered in advance. In the verification device, a binarization processing unit that binarizes the fingerprint image, a block division processing unit that divides the fingerprint image binarized by the binarization processing unit into a plurality of blocks, and a division unit that divides the fingerprint image by the block division processing unit A mask pattern scanning processing unit that scans a mask pattern corresponding to each predetermined angle in each of the blocks to generate a histogram for each direction, and the histogram for each direction of each block based on an output of the mask pattern scanning processing unit. And a finger placement determination unit including a stability determination processing unit that determines the stability of the directional angle of each block from the above, and is characterized only when the finger placement determination unit detects a state where the finger is placed. Fingerprint collation apparatus and performs collation with output and registration data. 2. A ridge direction calculation processing unit which is connected to a finger placement determination unit and calculates a ridge direction of a fingerprint in each block based on a determination result of a stability determination processing unit of the finger placement determination unit. A gray-scale image correction processing unit that corrects a gray-scale image along the ridge direction of the fingerprint calculated by the ridge direction calculation processing unit, and re-binarization that binarizes the gray-scale image correction result in the gray-scale image correction processing unit again A processing unit, a thinning unit for thinning the result of re-binarization by the re-binarization processing unit, and a feature check for scanning a thinned image by the thinning unit to detect a feature point of the image. 2. The fingerprint matching device according to claim 1, further comprising a feature extracting unit including an output unit. 3. A fingerprint which determines whether or not the user is the user by processing a fingerprint image obtained by the sensor to extract feature data and comparing the feature data with feature data unique to the user registered in advance. In the matching device, a binarization processing unit that binarizes the fingerprint image, a block division unit that divides the fingerprint image binarized by the binarization processing unit into a plurality of blocks, and a block that is divided by the block division unit A ridge direction calculation processing unit that calculates the ridge direction of the fingerprint in each block, a ridge direction storage unit that stores the calculation result of the ridge direction calculation processing unit input shortly before, and a ridge direction storage unit that is input at the current time. The ridge direction of the fingerprint of each block calculated from the image obtained by the ridge direction calculation processing unit is compared with the ridge direction of the fingerprint of each block before the short time stored in the ridge direction storage unit. Ridge direction change to detect the change Output processing unit, a spatial stability determination processing unit that detects blocks that have the same direction angle and that are adjacent to each other with respect to the blocks determined to have no change by the ridge direction change detection processing unit and determine the degree of clumping. A fingerprint matching device comprising: a finger placement determining unit that includes: a feature extraction unit that performs feature extraction and comparison with registered data only when the finger placement determining unit detects a state in which a finger is placed. 4. A gray-scale image correction processing unit connected to the finger placement determination unit and correcting a gray-scale image along the ridge direction of the fingerprint based on the determination result of the spatial stability determination processing unit of the finger placement determination unit. A re-binarization processing unit for re-binarizing the gray-scale image correction result in the gray-scale image correction processing unit; a thinning unit for thinning the result binarized again by the re-binarization processing unit; 4. The fingerprint matching device according to claim 3, further comprising a feature extraction unit including a feature point detection unit that scans the image thinned by the conversion unit and detects feature points of the image. 5. A fingerprint for determining whether or not a user is himself / herself by processing a fingerprint image obtained by a sensor to extract characteristic data and comparing the characteristic data with characteristic data unique to the user registered in advance. In the verification device, a binarization processing unit that binarizes the fingerprint image, a block division processing unit that divides the image binarized by the binarization processing unit into a plurality of blocks, and a block that is divided by the block division processing unit A mask pattern scanning processor for scanning a mask pattern corresponding to each predetermined angle in each block to generate a histogram for each direction, and a fingerprint ridge in each block based on an output of the mask pattern scanning processor. A first feature extraction unit including a ridge direction calculation processing unit for calculating a line direction; and a displacement amount for calculating a displacement amount of a finger at the time of registration and at the time of collation based on an output of the first feature extraction unit. Calculation unit, A position shift amount presenting unit for displaying the position shift amount calculated by the position shift amount calculation unit, and a position for determining whether the position shift amount is within a predetermined range based on an output of the position shift amount calculation unit. A displacement amount detecting means including a displacement amount determination unit is provided, and when the displacement amount exceeds a predetermined range, the user is instructed to put the finger again, and only when the position is within the specified range, the registration is performed in advance. A fingerprint collating apparatus comprising: performing collation with determined feature data to determine whether or not the user is the principal. 6. A gray-scale image correction processing unit connected to a positional deviation amount detecting unit, and correcting a gray-scale image along a ridge direction of a fingerprint based on a determination result of a positional deviation amount determining unit of the positional deviation amount detecting unit. A re-binarization processing unit for re-binarizing the gray-scale image correction result in the gray-scale image correction processing unit; a thinning unit for thinning the result binarized again by the re-binarization processing unit; 6. The fingerprint matching device according to claim 5, further comprising a second feature extraction unit including a feature point detection unit that scans the image thinned by the conversion unit and detects feature points of the image. 7. A fingerprint for determining whether or not the user is the user by processing fingerprint images obtained by the sensor to extract characteristic data and comparing the extracted characteristic data with characteristic data unique to the user registered in advance. In the verification device, a binarization processing unit that binarizes the fingerprint image, a block division processing unit that divides the image binarized by the binarization processing unit into a plurality of blocks, and a block that is divided by the block division processing unit A mask pattern scanning processor for scanning a mask pattern corresponding to each predetermined angle in each block to generate a histogram for each direction, and a fingerprint ridge in each block based on an output of the mask pattern scanning processor. A first feature extraction unit including a ridge direction calculation processing unit for calculating a line direction; and a fingerprint image obtained by dividing the fingerprint image divided into a plurality of blocks based on an output of the first feature extraction unit. Was An autocorrelation operation unit that calculates an autocorrelation map by calculating a difference between fingerprint ridge direction values for each block while shifting the ridge direction values of the fingerprints for each lock, and obtained by the autocorrelation operation unit A fingertip / bow print determining unit that processes the auto-correlation map to determine whether the fingertip is placed or the finger with the bow print is placed, When judged, this result is displayed, and the fingertip for which the registration process is to be continued as it is, or a request to the outside to determine whether the registration process is to be performed again after temporarily suspending and repositioning the finger or changing the finger, is performed. A fingertip / bow print detecting means including an bow print presentation unit, and in the registration mode, only when the fingertip / bow print is not determined and when there is an external instruction to continue registration, the feature data is provided. Register and judge whether it is a fingertip in collation mode Fingerprint matching device, characterized in that it seek outside to re-place the finger go himself. 8. A grayscale image is connected to a fingertip / bow print detecting means, and a grayscale image is corrected along a ridge direction of a fingerprint based on a determination result of a fingertip / bow print determination unit of the finger tip / bow print detection means. A gray-scale image correction processing unit, a re-binarization processing unit for binarizing the gray-scale image correction result in the gray-scale image correction processing unit again, and thinning the result binarized again by the re-binarization processing unit 8. The fingerprint matching apparatus according to claim 7, further comprising a second feature extracting unit including a thinning unit and a feature point detecting unit that scans the image thinned by the thinning unit and detects a feature point of the image. 9. A fingerprint which determines whether or not a user is himself / herself by processing a fingerprint image obtained by a sensor to extract feature data and comparing the feature data with pre-registered user-specific feature data. In the matching device, at least a binarization processing unit that binarizes the fingerprint image, a block division unit that divides the fingerprint image binarized by the binarization processing unit into a plurality of blocks, Feature extracting means including a mask pattern scanning processing unit for creating a histogram for each direction by scanning a mask pattern corresponding to each predetermined angle in each block, and a direction of each block extracted by the feature extracting means. Directional angle stability calculating means for calculating the directional angle stability of each block from another histogram, and calculating the stability calculated by the directional angle stability calculating means with respect to a feature point. Fingerprint verification apparatus characterized by reflecting the feature point matching degree calculated using as weights. 10. A ridge direction coincidence calculating unit which is connected to a feature extracting unit and a direction angle stability calculating unit, and calculates a ridge direction coincidence of a fingerprint based on an output of the feature extracting unit. A feature point coincidence calculating unit that calculates a feature point coincidence using the fingerprint ridge direction coincidence calculated by the direction coincidence calculator and the directional angle stability calculated by the directional angle stability calculator; 10. The fingerprint matching apparatus according to claim 9, further comprising a matching unit including a determination processing unit for determining whether the user is a person based on the matching degree of the feature points calculated by the feature point matching degree calculation unit. 11. A fingerprint which determines whether or not a user is himself / herself by processing a fingerprint image obtained by a sensor to extract feature data and comparing the feature data with feature data unique to the user registered in advance. In the matching device, a binarization processing unit for binarizing the fingerprint image, a ridge direction of a fingerprint in each block of the fingerprint image divided into a plurality of blocks after being binarized by the binarization processing unit Ridge direction calculation processing section for calculating the ridge direction calculation processing section, a gray image correction processing section for correcting a gray image along the ridge direction of the fingerprint based on the output of the ridge direction calculation processing section, and a shading in the gray image correction processing section. A feature extraction unit including a re-binarization process for re-binarizing the image correction result, and dividing the binarized image before and after the gray-scale image correction into a plurality of blocks, and comparing the binarized image for each block. , A change point counting unit for counting the number of change points Weight determining means including a weight calculation processing unit for calculating a weight for each block in accordance with the number of change points calculated by the change point counting processing unit; A fingerprint matching apparatus characterized in that the fingerprint degree is reflected in the calculation of the feature point coincidence degree. 12. A ridge direction coincidence calculating unit connected to a feature extracting unit and a weight determining unit, and calculating a ridge direction coincidence of a fingerprint based on an output of the feature extracting unit. A feature point coincidence calculation unit that calculates a feature point coincidence using the degree of coincidence of the ridge direction of the fingerprint calculated by the calculation unit and the weight of each block according to the number of change points calculated by the weight determination unit; 12. The fingerprint matching apparatus according to claim 11, further comprising a matching unit including a determination processing unit that determines whether the user is a person based on the matching degree of the feature points calculated by the feature point matching degree calculation unit.