JP2005339230A - Fingerprint collation device and its relative position detection device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a fingerprint collation device for securing precision of positioning, and for reducing calculation processing quantity required for the selection of collation featured points. <P>SOLUTION: This fingerprint collation device is provided with a featured point arranging means 3 for arranging featured points being the endpoints and branch points of the ridges or valley lines of fingerprints by using the respective angles as standards, a featured point selecting means 5 for selecting the second featured points arranged by the featured point arranging means 3 having predetermined differences of angles or less to the angles of the respective points of the first featured points, a positional difference totaling means 6 for calculating and totaling the positional differences of the respective points of the first featured points and the second featured points, a relative position detecting means for rotating the second featured points within the range of the predetermined relative rotational amounts for every predetermined rotational amounts, and for performing the selection of the second featured points and the totaling of the positional differences, and for detecting the relative positional differences and relative rotational differences being the relative differences of positions and the relative differences of the rotational amounts between a fingerprint image being the base of the first featured point and a fingerprint image being the base of the second featured points, and a fingerprint collating means for performing fingerprint collation by using the relative positional differences and the relative rotational amounts as correction values. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a fingerprint collation device that collates captured fingerprint data with registered data to determine pass / fail and a relative position detection device thereof.

  In a conventional fingerprint collation apparatus, as one of the methods for performing alignment before fingerprint collation, a direction angle image showing a directional distribution of ridges that are peaks of a fingerprint pattern is registered in advance. If there is a relative position difference and a relative rotational shift between the registered image and the collation image input to the fingerprint collation apparatus, the relative rotation amount is calculated and the registration is performed. Then, after the registration is performed, the registered feature points including the end points and branch points of the ridges extracted from the pre-registered image, and the matching feature points extracted from the matching image input to the fingerprint matching device Fingerprint verification is performed. (For example, see Patent Document 1)

In addition, as another method for performing alignment before fingerprint matching, there is a method in which alignment is performed by calculating a position difference or an angle difference between a registered feature point and a matching feature point. (For example, see Patent Document 2)
In Patent Document 2, in calculating a relative position difference and a relative rotation amount from a registered feature point and a matching feature point, first, as shown in FIG. 2 and FIG. Extract only the matching feature points where the positional deviation and angular deviation in the X and Y directions are within a predetermined range from the angle and the angle, and create a histogram of the positional deviation amounts of the matching feature points in the X and Y directions. Yes. When the relative rotation amount is also detected, the above process creates a histogram after rotating a predetermined rotation angle within a predetermined range. This is repeated a predetermined number of times until a certain degree of accuracy is obtained, and the relative position difference and the relative rotation amount of the image are determined to perform alignment.

Japanese Patent Laid-Open No. 2001-243465 (pages 5-9, FIGS. 1-5) JP-A-4-357570 (pages 4-7, FIGS. 1-8)

  In the fingerprint collation device of Patent Document 1, when the registration image and the collation image are aligned using the direction angle image indicating the direction distribution of the ridges that are the peaks of the fingerprint pattern, the direction distribution of the direction angle image is loose. If it is close to a sector or parallel line, the direction distribution will not change even if it is shifted in parallel, and if the direction angle image is fan-shaped or concentric, the center of rotation of the direction angle image and the fan-shaped or concentric circle will match. Then, there is a problem that the alignment cannot be performed because the direction distribution does not change even if the direction angle image is rotated.

  Further, in Patent Document 2, in the case where registration is performed by calculating a position difference or a rotation difference from a registered feature point and a matching feature point, if the predetermined range for extracting the matching feature point is large, the position difference is large. However, although the accuracy of alignment can be secured to some extent, there is a problem that the processing time is prolonged against the purpose of Patent Document 2. Furthermore, especially when performing repeated calculation by rotating the matching feature point, the coordinates of the matching feature point change nonlinearly, so that each point of the registered feature point corresponds to each point each time the whole is rotated. There is a problem in that it is necessary to reselect all the matching feature points, which increases the amount of calculation processing.

  The present invention has been made in order to solve the above-described problems, and a fingerprint collation apparatus capable of reducing the amount of calculation processing required to select collation feature points while ensuring alignment accuracy. And it aims at obtaining the relative position detection apparatus.

  The present invention provides a feature point input means for outputting the position and angle of a feature point which is an end point and a branch point of a fingerprint ridge or valley line, and a feature point output from the feature point input means or input from the outside. A feature point organizing unit that organizes each angle as a reference, and a second feature point that is arranged by the feature point organizing unit that has an angle difference of a predetermined value or less with respect to the angle of each point of the first feature point is selected. Feature point selection means, position difference summing means for calculating and summing the position difference between each point of the first feature point and the second feature point, and the second feature point as a predetermined relative rotation amount In this range, the second feature point is selected at a predetermined rotation amount and the position difference is totaled to obtain a fingerprint image as a basis of the first feature point and the second feature point. Relative position difference and rotation amount difference with the base fingerprint image Relative position detection means for detecting a positional difference and relative rotation amount, and fingerprint verification means for performing fingerprint verification using the relative position difference and relative rotation amount detected from the relative position detection means as correction values It is. When the feature point of the fingerprint registered in advance is used as the first feature point, the feature point of the fingerprint collated with this is the second feature point.

  The present invention provides a feature point input means for outputting the position and angle of a feature point which is an end point and a branch point of a fingerprint ridge or valley line, and a feature point output from the feature point input means or input from the outside. A feature point organizing unit that organizes each angle as a reference, and a second feature point that is arranged by the feature point organizing unit that has an angle difference of a predetermined value or less with respect to the angle of each point of the first feature point is selected. Feature point selection means, position difference summing means for calculating and summing the position difference between each point of the first feature point and the second feature point, and the second feature point as a predetermined relative rotation amount In this range, the second feature point is selected at a predetermined rotation amount and the position difference is totaled to obtain a fingerprint image as a basis of the first feature point and the second feature point. Relative position difference and rotation amount difference with the base fingerprint image Since there is provided a relative position detecting means for detecting a positional difference and a relative rotation amount, and a fingerprint collating means for performing fingerprint collation using the relative position difference and the relative rotation amount detected from the relative position detecting means as correction values. Thus, it is possible to obtain a fingerprint collation apparatus that can reduce the amount of calculation processing required to select a collation feature point while ensuring alignment accuracy.

Embodiment 1 FIG.
In the first embodiment, the feature point organizing unit organizes the feature points by classifying them into a plurality of groups in advance according to the angles of the registered feature points and the matching feature points, and the feature point selecting unit registers the registered feature points. By selecting a group of matching feature points whose angle difference from the center angle of the group to which the group belongs is less than or equal to a predetermined angle, select all the matching feature points whose angle error with the registered feature point is less than or equal to a predetermined threshold The case where it does is demonstrated. Here, rotation is also detected.

FIG. 1 is a configuration diagram of an embodiment of a fingerprint collation apparatus according to the present invention.
FIG. 2 is a flowchart in one embodiment of the fingerprint collation apparatus according to the present invention.
First, in FIG. 1, the fingerprint image input means 1 inputs a fingerprint as an image in the feature point input means 10 (corresponding to s1 in FIG. 2). The fingerprint image input means 1 outputs the fingerprint image data to the feature point extraction means 2.
For example, a CCD camera or the like is used as the fingerprint image input means 1 to input a fingerprint image. Further, even if a sensor such as a camera is not necessarily provided, any means for inputting a fingerprint image from outside the fingerprint collation apparatus according to the present embodiment via a wired or wireless connection may be used.

The feature extracting means 2 in FIG. 1 uses the input fingerprint image data for registration or verification to detect the positions and angles of the feature points that are the end points and branch points of the ridges and valleys that are the peaks of the fingerprint pattern. And the feature point data is output to the feature point organizing means 3 (corresponding to s2 in FIG. 2).
The feature point extraction will be described using a specific example. FIG. 3 is an example of feature point extraction by the feature point extraction means 2 in the embodiment of the fingerprint collation apparatus according to the present invention. (A) is the data of the inputted fingerprint image. In the feature point extraction unit 2, for example, noise removal and image contrast enhancement are performed on the input fingerprint image data, and then binarization processing is performed to convert the fingerprint image data into black and white data. . Then, the image data as shown in FIG.

  Subsequently, when the feature point extraction unit 2 performs thinning processing on each or one of the white ridge portion and the black valley portion from the monochrome data of the fingerprint image, a line as shown in FIG. It becomes an image having. In the figure, for convenience, a solid line is indicated by a valley line and a broken line is indicated by a ridge line. Thereafter, from the thinned data, the feature extraction unit 2 searches for the end points and branch points of the thin lines shown in circles in FIG. 3C, and uses the searched points as feature points. Then, as shown in FIG. 3D, information on the coordinates (x, y) on the fingerprint image data and the angle θ of the ridge is extracted and stored. In FIG. 3D, the arrow indicating the feature point of the valley line is indicated by a bold line, and the arrow indicating the feature point of the ridge line is indicated by a thin line.

  The order of feature point data extracted by the feature point extraction means 2 often depends on the processing order of the feature point extraction method. Therefore, the feature points are not arranged, classified, or aligned based on the angle. Normally, the X and Y coordinates of image data are usually in descending or ascending order of X or Y, but in the case of a method in which the ridge trajectory is tracked for feature point extraction, the order of the ridge tracking is used. ing. Therefore, the feature point organizing means 3 organizes the feature point data in order of angle (corresponding to s3 in FIG. 2).

FIG. 4 is a diagram showing before and after the arrangement of collation feature points of the feature point organizing means 3 in the embodiment of the fingerprint collation apparatus according to the present invention. The table on the left is the coordinates X, Y and the angle θ before the arrangement of the matching feature points as the second feature points, and the table on the right is the group, the address of the arrangement, the angle θ and the coordinates X, Y after the arrangement of the matching feature points. Is shown. The same applies to the registered feature point, which is the first feature point, and the description thereof will be omitted.
The data order before the arrangement of the matching feature points is ascending order of the coordinates X as shown in the table on the left side of FIG. The feature point organizing means 3 classifies the data of the collation feature points before the collation feature points are organized into groups according to the magnitude of θ and stores them in the memory 4 of FIG. For example, when divided into N (N is a natural number) groups, the range of θ in the I (I is a natural number) group is (I−0.5) × when the angle is I × 360 / N. 360 / N ≦ θ <(I + 0.5) × 360 / N may be set. FIG. 4 shows a case where N = 90.
For the sake of explanation, it is assumed that the address range of the memory 4 assigned to each group is designated in advance. Then, the feature point organizing means 3 discriminates the group including the angle of each matching feature point and stores it in the address of the memory 4 assigned to the group. Then, as shown in the table on the right side of FIG. 4, the matching feature points are classified and arranged for each group corresponding to the angle θ.

Similarly, the feature point organizing means 3 is a case where the registered feature points, which are information on the positions and angles at the end points and branch points of the ridges, are not arranged in the order of angle from the personal fingerprint image registered in advance (FIG. 2). 2) (corresponding to s4 of FIG. 2), they are arranged in the order of angles and grouped according to the angle range (corresponding to s5 in FIG. 2). Assuming that the coordinates (xr, yr) and angle of an arbitrary registered feature point are θr, dividing into M (M is a natural number) as in the case of the matching feature point, θ of the Jth group (J is a natural number) The range is (J−0.5) × 360 / M <θr ≦ (J + 0.5) × 360 / M.
When the feature point organizing unit 3 determines the angle of each point of the data of each registered feature point and stores it in the address of the group including the angle, the registered feature point corresponds to the angle θr. It will be organized for each group.

In this embodiment, the registered feature points of the personal fingerprint are stored in the memory 4 or the hard disk in the fingerprint collation device.
In general, the registered feature points of personal fingerprints are not necessarily provided in the fingerprint collation device, but are recorded in a memory 4 or hard disk existing in another computer, a server or via a network. It suffices if the information can be transmitted to the feature point organizing means 3 by communication or the like. Further, it may be provided with a function of inserting a card into the fingerprint verification device and a function of storing registered feature points in the card so that information of registered feature points can be transmitted to the feature point organizing means 3.

After organizing the feature points, the feature point selection unit 5 in FIG. 1 obtains each point of the registered feature points from the data of the collation feature points rotated by an angle α degrees (including α = 0) by the relative position detection unit 7. A matching feature point having an angle difference equal to or smaller than a predetermined angle is selected. Hereinafter, a process at a certain relative rotation angle α will be described.
FIG. 5 is a diagram showing a matching feature point selection method of the feature point selection means 5 in the embodiment of the fingerprint matching apparatus according to the present invention. The table on the left side shows each group of registered feature points, the addresses of registered feature points belonging to each group, and the angle θr. The table on the right side shows each group of matching feature points, the address of the matching feature point belonging to each group, an angle θα rotated by α degrees, and coordinates (xα, yα). That is, equation (1) is obtained.

  Further, the central angle of the group J to which the angle θr of the registered feature point belongs is set to θrj, and the central angle of the group I to which the angle θα of an arbitrary point of the matching feature point belongs is set to θi.

  FIG. 7 is a diagram showing an example before and after the collation feature point rotation processing performed by the relative position detection means 7 in the embodiment of the fingerprint collation apparatus according to the present invention. (A) shows before rotation processing and (b) shows after rotation processing. In both (a) and (b), the horizontal axis is the X axis, the vertical axis is the Y axis, the point is the position of the matching feature point, and the arrow indicated by the point indicates the angle of the matching feature point. The collation feature points shown at both ends of the rotation angle α in FIG. 7B indicate before and after rotating one collation feature point counterclockwise by the rotation angle α. Although the rotation center is shown in FIG. 7 when the origin is rotated to the center (Xc = Yc = 0), it may be an arbitrary fixed point. The range in which α is changed is determined by the fingerprint image input means 1 as an allowable range with reference to a relative rotation angle that may occur when a fingerprint image is actually input.

In order to select a matching feature point that has an angle difference of a predetermined value or less with respect to the angle of each registered feature point, first, for each registered feature point group, a matching in which the difference between θi and θrj is a predetermined angle or less Select a group of feature points and associate them.
For example, in FIG. 5, the feature point selection means 5 selects and associates groups 1, 2, and 3 as indicated by arrows as groups of matching feature points corresponding to group 2 on the registered feature point side. Specifically, all of the matching feature points that may include the matching feature point satisfying the angle difference | θα−θr | ≦ th (th = 2 degrees in FIG. 5) with the points in the group of J = 2. Arrow up the group. The method of pasting the arrow can be determined offline for the combination of θα and θr in the actual machine, so the processing time is very short. Similarly, groups 2, 3, and 4 are selected as groups of matching feature points corresponding to group 3, and arrows are applied. (Equivalent to s6 in FIG. 2)
By adopting such a selection method, most of the matching feature point groups that are known to contain no matching feature point satisfying the angle difference condition | θα−θr | ≦ th for each registered feature point are selected. It is possible to eliminate the necessity of selecting all the matching feature points.

  After the selection of the matching feature points, the position difference counting means 6 in FIG. 1 compares the matching feature points included in the group of matching feature points selected by the feature point selection means 5 for each point of the registered feature points. The angle difference | θα−θr | is calculated. If the angle difference is equal to or smaller than the predetermined angle error, the position difference (xα−xr, yα−yr) is calculated (corresponding to s7 in FIG. 2). A two-dimensional histogram of the X axis and the Y axis is created with the calculated values (corresponding to s8 in FIG. 2). However, since only feature points with originally small angular errors have already been selected by the above-described group selection, even if | θα−θr | exceeds the error, it is a slight value. Therefore, it is possible to further speed up the processing by omitting the determination of the angle error. Conversely, as in Patent Document 2, it is possible to constrain the range for the position difference, but there is a merit of saving memory and processing time in the summation of position differences and a demerit that fingerprints with large relative position differences cannot be collated. is there.

The above processes S7 to S8 are performed for all J and sequentially added to the histogram 9 in the position difference counting means 6 (S81).
FIG. 6 is a diagram showing an example of a two-dimensional histogram 9 created by the position difference counting means 6 in the embodiment of the fingerprint collation apparatus according to the present invention. The horizontal axis represents the X axis, the vertical axis represents the Y axis, the circle in the figure represents the calculated position difference, and the size of the circle represents the magnitude of the value of the histogram 9 at that position. In this example, there is no relative rotation of the image between the entire registered feature point and the entire matching feature point. If there is no rotation, the image is registered in the group selected by the feature point selection means 5 in this way. Since the matching feature point that should originally correspond to the feature point is included, coordinates showing a large value as shown in the coordinates (1, 5) in FIG. 6 appear. On the other hand, normally, since the rotation is often relatively, the correct group including the matching feature point that should originally correspond to the registered feature point is not selected by the feature point selection unit 5 when there is such a rotation. The coordinates indicating a large value as shown in the coordinates (1.5) of the histogram 9 in FIG. 6 do not appear. Therefore, it is necessary to find a relative rotation amount α that maximizes the number of registered feature points having a small positional difference and angular difference while rotating the entire matching feature point.

  The relative position detecting means 7 changes α, and the coordinate having the largest value of the histogram 9 from the value of the two-dimensional histogram 9 obtained by the position difference counting means 6 is set as a temporary position difference with respect to α at that time. And the maximum value of the histogram 9 (corresponding to s9 in FIG. 2).

Next, the relative position detecting means 7 checks whether or not the processing of the histogram 9 has been performed for all α in the predetermined rotation angle range. (Corresponding to s10 in FIG. 2).
If the processing of all the histograms 9 has not been completed, the relative position detection means 7 rotates the matching feature points extracted by the matching feature point detection means 2 by changing the rotation angle α in the equation (1) ( This corresponds to s11 in FIG. The change width of α is preferably at least equal to or less than the determination value th of the angle difference. On the other hand, from the viewpoint of processing speed, for example, if a multiple of the group angle range 360 / N is used, the processing becomes simpler and faster. In the rotation, it is only necessary to rewrite the coordinates and angles without changing the address of the matching feature point in the memory 4, and it is not necessary to change the group division. If the rotation amount is set to α = k × 360 / N (k is an integer), the group number is shifted by k and the arrow is also shifted by k. Needless to say, if k exceeds the value range of I, it is included in the value range by the remainder calculation.

  Thereafter, steps S6 to S10 are repeated, and the process proceeds to s12 when the temporary position difference and the histogram value at that time are obtained for all α in the predetermined rotation angle range.

If the true rotation angle to be obtained matches the rotation angle α, the matching feature point group I selected corresponding to the registered feature point group J is always the matching feature point corresponding to the registered feature point. If the position counting means 5 calculates the difference (xα−xr, yα−yr) between the registered feature points and the matching feature points according to the α and creates the two-dimensional histogram 9, the original position is obtained. The value of the histogram 9 at the coordinates of the position difference to be matched should be the maximum.
Therefore, the relative position detecting means 7 calculates the difference (xα−xr, yα−yr) on the coordinates when the value of the histogram 9 is the largest among the histograms 9 created for k with respect to α, as a whole relative position. The relative position is detected by determining the difference and determining the rotation angle α at that time as the overall relative rotation angle (corresponding to s12 in FIG. 2).

  According to the conventional invention, when the loop of s6 to s11 in FIG. 2 is performed, it is determined whether or not a condition for performing addition to the histogram 9 for all the matching feature points is satisfied for one registered feature point. There was a need to do. Specifically, all the addresses in the memory 4 of the matching feature points are accessed, and a large amount of calculation is required. The collation feature is also included when the relative position detection unit 7 performs a command to rotate the rotation angle α of the collation feature point k times and then reselects the collation feature point. If the coordinates of the points change, it is necessary to determine all the matching feature points for one registered feature point. That is, the number of accesses to the memory 4 = the number of registered feature points nr × the number of matching feature points nv × k times (both nr and nv are natural numbers).

  On the other hand, in the case of the embodiment of the present invention, for one registered feature point, the registered feature points that need to determine the conditions for addition to the histogram 9 are not all points, but the group to which the registered feature points belong. Therefore, the amount of calculation is greatly reduced. In FIG. 5, for example, the threshold range th of the angle range of the group and the angle difference condition of the feature points is 4 degrees, so the group of matching feature points selected for the registered feature point group 2 is 1, 2, 3 When the feature point organizing unit 3 organizes the feature points, the total number of accesses to the memory 4 when storing the feature points in the addresses of the memory 4 corresponding to each group is 90 per registered feature point. Since three of the matching feature point groups are extracted, assuming that one group of matching feature points includes nv / 90 points on average, the number of accesses is nr × (nv / 90) × 3 × k = nr. Since it is * nv * k / 30, it becomes 1/30. Accordingly, the number of accesses to the memory 4 is reduced, and the speed can be greatly increased.

  Using the relative position difference and the relative rotation angle detected by the relative position detecting means 7 in this way, the fingerprint collating means 8 is used to register the personal fingerprint image registered in advance and the fingerprint image input from the fingerprint image input means 1. Are aligned (corresponding to s13 in FIG. 2). Then, the matching ratio between the matching feature point and the registered feature point or the difference in the gray value of the image is used as the target of collation, and the score when the fingerprint image is collated is determined. The fingerprint is collated by a method such as (corresponding to s14 in FIG. 2). If it is determined only by the matching ratio between the matching feature point and the registered feature point, it is determined that the user is the person if only the matching feature point and the registered feature point are equal to or greater than a predetermined threshold without using a fingerprint image. It is also possible to do.

  As described above, the present invention has the feature point input means 10 for outputting the position and angle of the feature points which are the end points and branch points of the ridge or valley line of the fingerprint, and the feature point input means 10 which outputs the feature points. The feature points are arranged by the feature point organizing means 3 that arranges the input feature points based on the angles, and the feature point organizing means 3 that makes an angle difference of a predetermined value or less with respect to the angles of the registered feature points registered in advance. Feature point selecting means 5 for selecting the collated feature points, position difference summing means 6 for calculating and summing the position difference between each registered feature point and the matching feature points, and a predetermined relative rotation of the collated feature points. Relative fingerprint image that is the basis of registered feature points and fingerprint image that is the basis of matching feature points Difference in position and rotation amount Relative position detection means 7 for detecting the position difference and relative rotation amount, and fingerprint verification means 8 for performing fingerprint verification using the relative position difference and relative rotation amount detected from the relative position detection means 7 as correction values. Since it is provided, it is possible to obtain a fingerprint collation apparatus that can reduce the amount of calculation processing required to select collation feature points while ensuring alignment accuracy.

  In the present embodiment, the accuracy may be increased by selecting a matching feature point or a group of registered feature points and then performing an angle check again. However, in this case, the second embodiment described below is more suitable. In addition, in S12 of FIG. 2, the relative position difference and the relative rotation angle are not uniquely obtained, and a plurality of candidates are left and collation is attempted for each candidate, thereby improving resistance to unexpected noise. Can do.

Embodiment 2. FIG.
In the first embodiment, the registered feature point side is also organized by angle, but another embodiment is described in which only the collation side is organized and the collation side group number is calculated directly from θr. That is, a standard of I corresponding to a certain θr is calculated by, for example, I = Int ((θr + 0.5 + 360) mod 360) / N (Int () is an integer by rounding, mod () is a remainder calculation) Are processed in the same manner as in the first embodiment. Further, as another processing method, the processing of the registered feature point and the collated feature point after rotation may be exchanged to arrange only the registration side. As described above, various variations are possible, but which one is faster depends on the number of registered fingerprints and collation fingerprints and the nature of hardware and applications.

Embodiment 3 FIG.
In the third embodiment, the feature point organizing unit arranges the feature points by arranging them in ascending order according to the angles of the registered feature points and the matching feature points, and the feature point selecting unit arranges each of the registered feature points. The case where the collation feature point whose angle difference with the angle of a point is below a predetermined angle is selected will be described.

The configuration of the present embodiment is the same as that of the first embodiment.
A difference in processing from the first embodiment will be described.
When the feature point organizing unit 3 sorts the collation feature point data in the order of angles, in the first embodiment, the feature point organizing unit 3 uses the angle information of the registered feature points and the collation feature points according to the magnitude of θ. In contrast to the classification into groups, in the present embodiment, the registered feature points and the matching feature points are arranged in order of angle. The direction of the arrangement is, for example, ascending order.
The feature point selection unit 5 selects the matching feature point group corresponding to each group of registered feature points in the first embodiment when extracting the matching feature point. In this embodiment, when an allowable angle error is set to th, a matching feature point satisfying | θa−θr | ≦ th is selected for an arbitrary registered feature point, and the matching feature point is arranged in ascending order of the angle of the registered feature point. Do it.

The operation will be described.
FIG. 8 is a flowchart in the third embodiment of the fingerprint collation apparatus according to the present invention.
From the fingerprint image input by the fingerprint image input means 1 in the feature point input means 10 (s21 in FIG. 8), a matching feature point is extracted by the feature point extraction means 2 (s22 in FIG. 8), and the position and angle The data is output to the feature point organizing unit 3 as data including a large number of matching feature points including information.

The feature point organizing means 3 rearranges each point of the matching feature points and arranges them in ascending order based on the angle (s23 in FIG. 8).
FIG. 9 is a table showing before and after the arrangement of collation feature points of the feature point organizing means 3 in the third embodiment of the fingerprint collation apparatus according to the present invention. The table on the left shows the coordinates X, Y and the angle θ before arrangement of the matching feature points, and the table on the right shows the address, the angle θ and the coordinates X, Y of the arrangement after the arrangement of the matching feature points.
Specifically, since the address range of the memory 4 in which the matching feature points are stored is designated in advance, the feature point organizing means 3 sorts the data by the order of the angles of the data of each point of the matching feature points. Are stored in the addresses of the memory 4 in the order of angles.

  Similarly, with respect to the registered feature points, when the registered feature points are not arranged in the order of angles (s24 in FIG. 8), the registered feature points are arranged in the order of angles, and the respective points of the registered feature points are rearranged to obtain the angle as a reference. Are arranged in ascending order (s25 in FIG. 8). Similarly, since the address range of the memory 4 where the registered feature points are stored is also designated in advance, the feature point organizing means 3 performs rearrangement by rearranging the order of the angles of the data of each point of the registered feature points, It memorize | stores in the address of the memory 4 in angle order.

  The feature point selection means 5 selects, for each registered feature point, a matching feature point that satisfies | θα−θr | ≦ th (s26 in FIG. 8). Since they are arranged in ascending order, it is not necessary to access the memory 4 of all the matching feature points for one registered feature point. Details will be described below.

FIG. 10 is a diagram showing a selection example of the matching feature point selection means 5 in the second embodiment of the fingerprint matching device according to the present invention. The left column shows registered feature point data, the address on the left and the angle on the right. The upper column shows the data of the matching feature points, the address on the second level and the angle on the third level. The lower right part shows the value of θr−θα and the determination result of −th ≦ θr−θα ≦ th when the matching feature point selection means 5 selects from the registered feature points and the matching feature points.
Assume that the angles of registered feature points are arranged in ascending order from 0 to 360 degrees, and the matching feature points are arranged in ascending order from 0 to 360 degrees as well. Suppose that th is set to 2 degrees.

  First, when the angle of the registered feature point is 0 degree, it is determined by accessing in the ascending order of the address of the memory 4 of the matching feature point. Since each of the addresses in the memory 4 is stored in ascending order of the angles of the matching feature points, accessing in ascending order of the addresses in the memory 4 is the same as determining in ascending order of the angles of the matching feature points.

  Of the matching feature points, as shown in FIG. 10, 0 degree is selected as applicable because −2 ≦ θr−θα ≦ 2. Next, 4 degrees is not selected because 0-4 (θr−θα) <− 2 (−th).

  This time, the registered feature point is set to twice, and the address of the memory 4 of 0 degree which is the first selected feature point when the previous access is performed is stored, and the determination is made from there. Then, 0 degree is selected because 2 (th) ≦ 2-0 (θr−θα). 4 degrees is selected because −2 ≦ 2-4 (θr−θα) ≦ 2, and 5 degrees is selected because 2−5 <−2 (−th).

  Next, the registered feature point is set to 4 degrees, and discrimination is performed from 0 degree which is the collation feature point first selected at the previous access. Then, 0 degree is not selected because 2 <4-0, and 4, 5, and 6 degrees are selected because −2 ≦ θr−θα ≦ 2, and 8 degrees are not selected because 4-8 <−2. And so on.

Next, the registered feature point is set to 6 degrees, and discrimination is performed from 4 degrees which is the matching feature point selected first when the previous access was made. Then, 4, 5, 6, and 8 degrees are selected because −2 ≦ θr−θα ≦ 2, and 8 degrees are selected because 4-8 <−2. And so on. Therefore, it continues when the registered feature points are set to 8 degrees and 10 degrees.
That is, conventionally, it has been necessary to determine whether or not to select a matching feature point for all the columns in the combination table of FIG. 10, but in the present embodiment, substantially along the diagonal line in the figure. Only the determination processing relating to the remaining columns is necessary, and determination calculation does not have to be performed for blank columns that occupy most of the other columns.

  Although omitted in the description, it goes without saying that it is necessary to solve the boundary conditions in θα <0 degrees and θα ≧ 360 degrees. For example, the matching feature points included in 360−β ≦ θα <360 are required. Copy the matching feature points included in 0 ≦ θα ≦ β to the region of −β ≦ θα <0 or the region of 360 ≦ θα <360 + β, or the portion of θα = 0 and the portion of 360 are continuous regions There are some ingenuity in implementation such as realization of accessible circular reference memory in hardware or software.

Thereafter, after selecting the matching feature point in s26 of FIG. 8, the loop of s26 to 31 is exactly the same as s6 to 11 of FIG. 2 of the first embodiment except that J is changed to a loop of θr.
Also in this embodiment, by arranging the registered feature points and the matching feature points in ascending order based on the angle, only one memory 4 near the selected matching feature point is accessed for one registered feature point. The number of accesses to the memory 4 can be greatly reduced, and the speed can be greatly increased.
In addition, when the matching feature points are rotated again by α degrees and reselected, the feature points are arranged once in ascending order by the feature point organizing means 3, so that the order of the feature points is not changed by the rotation, so it is necessary to sort again. In addition to the rotation calculation of coordinates, the calculation start amount and the end point of the angle at which selection of the matching feature point is started can be changed by changing the address. There is no need to reselect the matching feature points.

  Also, the processing of s32 to 34 in FIG. 8 is exactly the same as s12 to 14 of the first embodiment.

  As described above, in this embodiment, the feature point organizing unit 3 arranges the pre-registered matching feature points and the registered feature points in ascending order based on the angles, so that the accuracy of the alignment is increased. It is possible to obtain a fingerprint collation apparatus that can reduce the amount of calculation processing required to select a collation feature point while ensuring the above.

  In this embodiment, the registered feature points and the matching feature points are arranged in ascending order to arrange them in the order of angles. However, even when all of them are performed in descending order, the same effect as in this embodiment is obtained. I can do it.

Embodiment 4 FIG.
In the fourth embodiment, in contrast to the first embodiment, the feature point selecting unit selects a group of registered feature points whose angle difference from the central angle of the group to which the matching feature point belongs is equal to or less than a predetermined angle. A case will be described in which all of the registered feature points whose angle error with the matching feature point is equal to or less than a predetermined threshold value are selected.

The configuration of this embodiment is the same as that of the first embodiment shown in FIG.
FIG. 11 is a flowchart in one embodiment of the fingerprint collation apparatus according to the present invention. s41 to s45 are the same as s1 to s5 of the first embodiment.
After the arrangement of the feature points, the feature point selection unit 5 in FIG. 1 uses the registered feature point data rotated by an angle α degrees (including α = 0) by the relative position detection unit 7 to check each point of the matching feature points. A registered feature point having an angle difference of a predetermined value or less with respect to the angle is selected (corresponding to s46 in FIG. 11). This is only performed by exchanging the registered feature points and the matching feature points in the case of the first embodiment.

  After completing the selection of the registered feature points, the position difference counting unit 6 calculates the angular difference between the registered feature points included in the group of matching feature points selected by the feature point selecting unit 5 for each point of the registered feature points. If it is less than a predetermined angle error, the position difference is calculated (corresponding to s47 in FIG. 11). A two-dimensional histogram of the X axis and the Y axis is created with the calculated values (corresponding to s48 in FIG. 11). This is also only performed by replacing the registered feature points and the matching feature points with respect to the case of the first embodiment.

  Similar to the first embodiment, the above processes S47 to S48 are performed for all the Js and sequentially added to the histogram 9 in the position difference counting means 6 (S83). Then, the relative position detecting means 7 changes α, and from the value of the two-dimensional histogram 9 obtained by the position difference counting means 6, the coordinate having the largest value of the histogram 9 is determined as a temporary position difference with respect to α at that time. And stored together with the maximum value of the histogram 9 (corresponding to s49 in FIG. 11).

Next, the relative position detecting means 7 checks whether or not the processing of the histogram 9 has been performed for all α in the predetermined rotation angle range. (Corresponding to s50 in FIG. 11).
If the processing of all the histograms 9 has not been completed, the relative position detecting means 7 rotates the registered feature points by changing the rotation angle α (corresponding to s51 in FIG. 11).
Thereafter, Steps S46 to S50 are repeated, and the process proceeds to s52 when the temporary position difference and the histogram value at that time are obtained for all α in the predetermined rotation angle range. S52 to s54 in FIG. 11 are the same as s12 to s14 in the first embodiment.

  As described above, the present invention performs the processing performed by the feature point selection unit 5, the position difference tabulation unit 6, and the relative position detection unit 7 with respect to the first embodiment with the registered feature points and the matching feature points reversed. However, as in the first embodiment, it is possible to obtain a fingerprint collation apparatus that can reduce the amount of calculation processing required to select collation feature points while ensuring alignment accuracy.

Embodiment 5 FIG.
In the fifth embodiment, the feature point organizing unit arranges by arranging in advance in ascending order according to the angle of each point of the registered feature point and the matching feature point, and the feature point selecting unit sorts each of the matching feature points. The case where the collation feature point whose angle difference with the angle of a point is below a predetermined angle is selected will be described.

The configuration of the present embodiment is the same as that of the second embodiment.
FIG. 12 is a flowchart in the fifth embodiment of the fingerprint collation device according to the present invention.
A difference in processing from the second embodiment will be described.
The feature point selection means 5 selects a registered feature point that satisfies | θ−θrα | ≦ th (θrα is an angle θr rotated by α degrees) for each point of the matching feature point (s106 in FIG. 12). Since the registered feature points and the matching feature points are arranged in ascending order in the angular order, it is not necessary to access the memory 4 of all the matching feature points for one matching feature point. This is the same as described in the second embodiment.

Then, after selecting the matching feature point of s106 in FIG. 12, with respect to the loop of s106 to 111, the functions of the registered feature point and the matching feature point are opposite to those of s26 to 31 of FIG. Exactly the same.
Also, the processing of s112 to 114 in FIG. 12 is exactly the same as s32 to 34 of the second embodiment.

  As described above, in the present embodiment, even when the roles of the registered feature points and the matching feature points in the second embodiment are switched, as in the second embodiment, the alignment accuracy is ensured. In addition, it is possible to obtain a fingerprint matching device that can reduce the amount of calculation processing required to select matching feature points.

  In this embodiment, the registered feature points and the matching feature points are arranged in ascending order to arrange them in the order of angles. However, even when all of them are performed in descending order, the same effect as in this embodiment is obtained. I can do it.

  In the first to fifth embodiments, the histogram 9 is described as two-dimensional. For example, regarding histogram processing, X and Y are considered as independent variables, and two one-dimensional histogram memories are independently set for X and Y, respectively. By using a general speed-up method such as reducing the processing memory and processing time by using it, it is possible to further speed up, or simply adding 1 to one feature point when adding to the histogram Instead, it is desirable to improve the accuracy by adding reliability information calculated in advance by some method to the feature points and adding a score corresponding to the reliability.

  Furthermore, since the feature point organizing means 3 can be executed regardless of the combination of the first fingerprint and the second fingerprint, there is no need to wait for the feature points of both fingerprints to be matched in online matching. Therefore, it is possible to execute offline for registration fingerprints at the time of registration, and it only needs to be executed once at the time of input for verification fingerprints that are input online. When the collation process is repeated, the execution time of the feature point organizing means is equivalent to 1 / (the number of collation processes) per collation. Therefore, if the number of collation processes increases, the execution time is negligibly small. It will become.

Embodiment 6 FIG.
Among the fingerprint collation devices described in the first to fifth embodiments, the relative configuration comprising the feature point selection means 3, the feature point selection means 5, the position difference counting means 6, and the relative position detection means 7 described in FIG. The configuration of the position detection device may be used not only for detecting the relative position between the registered feature point and the matching feature point but also when registering the feature point. This case will be described below.

FIG. 13 is a block diagram of a sixth embodiment of a relative position detection apparatus according to the present invention, and FIG. 14 is a flowchart of a sixth embodiment of the relative position detection apparatus according to the present invention.
When registering feature points, when registering fingerprint data, it is common to collect images several times from the same finger continuously or at different times, and finally only feature points that can be extracted relatively stably in all of those images. Registered as registered feature point data. In this process, registration is required between a plurality of registered images. To that end, one of the two registered images to be registered is regarded as a collation image, and Embodiments 1 to 5 are used. The relative position detection device may be applied in the same manner as described above.
First, it is determined whether or not the feature point that has come to the feature point organizing unit 3 is a temporary registered feature point (s121 in FIG. 14), and the feature points are grouped according to the angle (corresponding to s122 in FIG. 14). The next feature point is determined as a matching feature point, NO is determined in s121, and the matching feature point is classified into groups according to the angle in s123.
The subsequent steps s124 to s130 are exactly the same as the steps s6 to s12 in FIG. 2 of the first embodiment.

  After that, although not shown in the figure, the provisional registered feature point and the matching feature point are collated using the determined relative position difference and relative rotation amount, and several times continuously from the same finger as shown above. Alignment is performed using a plurality of relative position differences and relative rotation amounts obtained by performing the processing from s121 to s130 in the same manner using the matching feature points processed from the collected images. For example, the position does not change. Only feature points that are successfully matched with high probability are registered as true registered feature point data.

  Therefore, the registered feature point is determined even when the relative position detecting device including the feature point selecting unit 3, the feature point selecting unit 5, the position difference counting unit 6, and the relative position detecting unit 7 is used for determining the registered feature point. The amount of calculation processing required at the time can be reduced.

  Needless to say, the same effect can be obtained when the relative position detection device is applied as a part of the fingerprint collation device. Moreover, even if the process of selecting the feature points used in the second to fifth embodiments and determining the relative position difference and the relative rotation amount is applied to the present embodiment, the same effect can be obtained.

It is a block diagram in one Embodiment of the fingerprint collation apparatus concerning this invention. It is a flowchart in one Embodiment of the fingerprint collation apparatus concerning this invention. It is a figure which shows the example of extraction of the collation feature point by the feature point extraction means 2 in one Embodiment of the fingerprint collation apparatus concerning this invention. It is a figure which shows before and after arrangement | positioning of the collation feature point of the feature point arrangement | positioning means 3 in one Embodiment of the fingerprint collation apparatus concerning this invention. It is a figure which shows the selection method of the collation feature point of the feature point selection means 5 in one Embodiment of the fingerprint collation apparatus concerning this invention. It is a figure which shows an example of the two-dimensional histogram which the position difference totaling means 6 in one Embodiment of the fingerprint collation apparatus concerning this invention produced. It is a figure which shows an example before and after the rotation process of the collation feature point which the relative position detection means 7 in one Embodiment of the fingerprint collation apparatus concerning this invention performed. It is a flowchart in 3rd Embodiment of the fingerprint collation apparatus concerning this invention. It is a figure which shows before and after arrangement | positioning of the collation feature point of the feature point arrangement | positioning means 3 in 3rd Embodiment of the fingerprint collation apparatus concerning this invention. It is a figure which shows the selection example of the feature point selection means 5 in 3rd Embodiment of the fingerprint collation apparatus concerning this invention. It is a flowchart in 4th Embodiment of the fingerprint collation apparatus concerning this invention. It is a flowchart in 5th Embodiment of the fingerprint collation apparatus concerning this invention. It is a block diagram in 6th Embodiment of the relative position detection apparatus concerning this invention. It is a flowchart in 6th Embodiment of the relative position detection apparatus concerning this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Fingerprint image input means, 2 Feature point extraction means, 3 Feature point arrangement means, 4 Memory, 5 Feature point selection means, 6 Position difference totaling means, 7 Relative position detection means, 8 Fingerprint collation means, 9 Histogram, 10 Feature points Input means.

Claims (8)

Feature point extraction means for outputting the positions and angles of feature points that are the end points and branch points of fingerprint ridges or valleys from the inputted fingerprint image, and features outputted from the feature point extraction means or inputted from the outside Feature point organizing means for organizing points on the basis of each angle, and collation feature points arranged by the feature point organizing means having an angle difference of a predetermined value or less with respect to the angle of each registered feature point registered in advance A feature point selecting means for selecting a position difference, a position difference counting means for calculating and summing up the position difference between each point of the registered feature points and the matching feature point, and the matching feature point within a predetermined relative rotation amount range. By selecting the matching feature point and counting the position difference by rotating each predetermined amount of rotation, the fingerprint image that is the basis of the registered feature point and the fingerprint image that is the basis of the matching feature point are relative to each other. Position difference and times Relative position detection means for detecting a relative position difference and a relative rotation amount, which is a difference in amount, and fingerprint verification for performing fingerprint verification using the relative position difference and the relative rotation amount detected from the relative position detection means as correction values And a fingerprint collating apparatus. The feature point organizing unit classifies in advance into a plurality of groups according to the angles of the respective points of the matching feature points, and the feature point selecting unit is configured to perform the above processing with respect to the center angle of each group of the registered feature points. The fingerprint collation apparatus according to claim 1, wherein a group in which an angle difference from a central angle of the collation feature point is a predetermined angle or less is selected. The fingerprint collation apparatus according to claim 1, wherein the feature point organizing unit arranges the registered feature points and the collation feature points in ascending or descending order of angles. Feature point extraction means for outputting the position and angle of feature points that are the end points and branch points of fingerprint ridges or valleys from the inputted fingerprint image, and features outputted from the feature point extraction means or inputted from the outside Feature point organizing means for organizing each point based on an angle, and registered feature points registered in advance that have an angle difference of a predetermined value or less with respect to the angle of each point of the matching feature points arranged by the feature point organizing means A feature point selecting means for selecting a position difference, a position difference aggregating means for calculating and aggregating a position difference between each point of the registered feature points and the matching feature point, and the registered feature points within a predetermined relative rotation amount range. By selecting the registered feature points and summing up the positional differences by rotating each predetermined amount of rotation, the fingerprint image that is the basis of the registered feature points and the fingerprint image that is the basis of the matching feature points are relative to each other. Position difference and times Relative position detection means for detecting a relative position difference and a relative rotation amount, which is a difference in amount, and fingerprint verification for performing fingerprint verification using the relative position difference and the relative rotation amount detected from the relative position detection means as correction values And a fingerprint collating apparatus. The feature point organizing unit classifies in advance into a plurality of groups according to the angles of the respective points of the registered feature points, and the feature point selecting unit is configured to perform the above processing with respect to the center angle of each group of the matching feature points. The fingerprint collation apparatus according to claim 4, wherein a group whose angle difference from the center angle of the registered feature point is a predetermined angle or less is selected. 5. The fingerprint collation apparatus according to claim 4, wherein the feature point organizing unit arranges the registered feature points and collation feature points in ascending or descending order of angles. Feature point organizing means for organizing information on the position and angle of feature points that are the end points and branch points of fingerprint ridges or valleys from fingerprint images, and each point of registered feature points registered in advance A feature point selecting means for selecting matching feature points arranged by the feature point arranging means having an angle difference equal to or less than a predetermined angle, and a position difference between each point of the registered feature points and the matching feature point The position difference counting means for calculating and counting, and by rotating the matching feature point for each predetermined rotation amount within a range of a predetermined relative rotation amount, selecting the matching feature point and totaling the position difference, Relative position detection means for detecting a relative position difference and a relative rotation amount that are a difference in relative position and a rotation amount between a fingerprint image that is a basis of a registered feature point and a fingerprint image that is a basis of the matching feature point It is characterized by having Versus position detecting device. Feature point organizing means for organizing information on positions and angles of feature points which are end points and branch points of fingerprint ridges or valley lines from a fingerprint image, respectively, and collation features arranged by the feature point organizing means A feature point selection means for selecting a registered feature point registered in advance with an angle difference equal to or less than a predetermined angle with respect to the angle of each point; and a positional difference between each point of the registered feature point and the matching feature point The position difference counting means for calculating and totaling, and rotating the registered feature points for each predetermined rotation amount within a range of a predetermined relative rotation amount to select the registered feature points and totalize the position differences, A relative position detecting means for detecting a relative position difference and a relative rotation amount, which are a relative position difference and a rotation amount difference between a fingerprint image which is a basis of a registered feature point and a fingerprint image which is a basis of the matching feature point; Phases characterized by having Position detector.

Images