JP2011048602A - Fingerprint collating device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accelerate collating speed by reducing the number of feature points to be collated by excluding feature points extracted approximately linearly at high concentration from collating targets as false feature points. <P>SOLUTION: The fingerprint collating device for collating an input fingerprint with a previously stored registered fingerprint includes: a fingerprint image acquisition means for acquiring a fingerprint image: a feature point extraction means for extracting feature points such as end points of a ridge line or a trough line or a branch line from the fingerprint image together with position information; a position feature point selection means which when the concentration of the feature points extracted by the feature point extraction part in a predetermined area exceeds a threshold, excludes the feature points existing in the area and selects a position feature point; a collating means for positioning the registered fingerprint and the input fingerprint by using the position feature point and collating identify between the registered fingerprint and the input fingerprint; and an output part for outputting a collating result of the collating means. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

The present invention relates to a fingerprint collation device that collates the identity of a fingerprint registered in advance with an input fingerprint, and in particular, extracts and extracts fingerprint ridge or valley end points and branch points as feature points from a fingerprint image. The present invention relates to a so-called feature point matching type fingerprint collation apparatus that uses the feature points for collation.

Conventionally, a fingerprint collation device is used for identity verification in an access management device that manages entering and exiting a room. In addition, it is also used for user confirmation of computers and identity verification in CD / ATM. In such a widely used fingerprint matching device, there is a feature point matching method as a method for matching fingerprint images. This feature point matching method extracts feature points useful for fingerprint identification, such as fingerprint peak (ridge), valley (valley) end points, and branch points, and the arrangement and feature of the extracted feature points. Matching is performed using feature quantities such as the direction of the points. Patent Document 1 is an example of a fingerprint collation apparatus using a feature point matching method. In the fingerprint collation device of Patent Document 1, a feature point is extracted after thinning processing of a fingerprint image, and feature points whose extracted feature point coordinates are adjacent within a predetermined range are deleted as pseudo feature points. Yes. Here, the pseudo feature points are feature points that do not contribute much to fingerprint collation due to various causes such as improper finger placement, dry fingers, and scratches.

JP-A-9-265528

The feature point matching fingerprint collation device uses many feature points to be extracted because it matches the feature quantity of the registered fingerprint and the input fingerprint feature points such as the arrangement and direction of the features. And the verification process takes time. On the other hand, if the number of extracted feature points is small, the verification processing time is shortened. For this reason, the collation time of the fingerprint collation apparatus varies depending on the fingerprint image registered or inputted. That is, in the feature point matching type fingerprint collation device, the time from when the fingerprint is read until the collation result is obtained varies. When used for user confirmation such as login to a computer or access control device, if the confirmation time differs greatly, psychological stress increases, which is not preferable. In particular, if a large number of feature points are extracted and it takes a long time to collate, the user's usability becomes worse.

  Further, as in the fingerprint matching device of Patent Document 1, a method of simply deleting pseudo feature points from extracted feature points, that is, removing all feature points adjacent in a predetermined range from the feature points, If there are few feature points to be extracted, the number of feature points to be extracted further decreases, which may adversely affect the matching accuracy. Further, even feature points that exist within a predetermined range are not necessarily limited to pseudo feature points, and may include feature points that are useful for matching. Furthermore, there is no conventional fingerprint collation device that considers the influence of collation time depending on the number of extracted feature points, and the main purpose is to delete pseudo feature points that adversely affect collation accuracy.

  By the way, the pseudo feature points include the outer peripheral portion of the fingerprint image and end points caused by scratches. These pseudo feature points are arranged in a substantially straight line and have a high density in a predetermined area. On the other hand, normal feature points are rarely arranged in a substantially straight line with high density. Therefore, the present invention aims at speeding up collation by excluding feature points extracted in a substantially straight line shape and with high density from collation targets and reducing the number of feature points to be collated. Furthermore, in the case of a fingerprint with a small number of extracted feature points, it is intended to prevent a reduction in matching accuracy by preventing feature points from being excluded from matching targets.

The present invention relates to a fingerprint collation device that collates a registered fingerprint stored in advance with an input fingerprint, and includes a fingerprint image obtaining unit that obtains a fingerprint image, an end point or a branch point of a ridge or valley line from the fingerprint image, and the like A feature point extracting unit that extracts the feature points together with position information, and feature points existing in the region when the density of the feature points extracted by the feature point extracting unit exceeds a threshold value A position feature point selecting means for selecting a position feature point by removing the position feature, a registration means for aligning the registered fingerprint and the input fingerprint using the position feature point, and a matching means for matching the identity of the registered fingerprint and the input fingerprint; Provided is a fingerprint collation apparatus including an output unit that outputs a collation result by collation means.

With this configuration, when there is a region having a high density of feature points in a predetermined region from the extracted feature points, the present invention increases the reliability when the feature points in the region are excluded from matching targets or in matching. It is set low. For this reason, since the number of feature points to be collated does not increase, the collation time is increased, and when the number of feature points to be extracted is small, the feature points are not deleted, and the collation accuracy is not lowered.

As a preferred embodiment, the position feature point selection means scans a fingerprint image using a rectangular area window for a predetermined area, counts the number of feature points in the area window, and sets the density as the density. Thereby, it is possible to target feature points whose density standards are arranged in a rectangular shape, that is, in a substantially straight line.

As a preferred mode, the threshold value of the density is set with a correlation with the number of ridges or valleys that can enter the region, and is set to be lower as the number is smaller.

Further, as a preferred aspect, the position feature point selection means calculates the density excluding the branch point among the feature points extracted by the feature point extraction means, and also when the density exceeds the threshold value. The branch point is not excluded and is selected as a position feature point. As a result, it is possible to prevent a branch point having a low possibility of a pseudo feature point from being inadvertently excluded from the position feature point, and to prevent a decrease in accuracy of position matching.

It is possible to maintain collation accuracy without deleting feature points from a fingerprint with few feature points, while realizing a faster collation time of the fingerprint collation device.

Functional block diagram of fingerprint collation device 1 Overall processing flow of fingerprint verification device 1 Processing flow of position feature point selection means The figure which shows the feature point extracted by the feature point extraction means The figure which shows the position feature point selected by the feature point extraction means Figure showing the image of scanning a vertically long rectangular window Diagram showing the image of scanning a horizontally long rectangular window

Hereinafter, a fingerprint collation apparatus 1 to which the present invention is applied will be described with reference to the drawings. FIG. 1 shows functional blocks of the fingerprint collation device 1. The fingerprint collation apparatus 1 includes an operation / display unit 11, a fingerprint photographing unit 12, a fingerprint collation processing unit 13, a storage unit 14, and an output unit 15, and includes fingerprint data of an input fingerprint image photographed by the fingerprint photographing unit. The fingerprint collation processing unit 13 collates the fingerprint data of the registered fingerprint registered in advance in the storage unit 14, and outputs the collation OK or collation NG as the result from the output unit 15.

The operation / display unit 11 is composed of a touch panel type liquid crystal panel and its peripheral circuits, and a user displays a guidance display / collation result display for a fingerprint collation processing start operation or a fingerprint image capture by the fingerprint photographing unit 12. It is an interface that performs etc.

The fingerprint imaging unit 12 includes a prism having a mounting table on which a user places a finger and a camera that captures a fingerprint image of the user's finger mounted on the mounting table via the prism. The fingerprint image, which is an image, is output to the fingerprint verification processing unit 13. More specifically, the fingerprint photographing unit 12 is an LED (Light) for photographing a finger fingerprint clearly.
(Emitting Diode) Illumination and a guide member for stably holding the finger on the mounting table are further provided. A sharp and stable fingerprint image is photographed with a predetermined resolution by the image sensor, and the captured fingerprint image is captured. The digital image signal is output to the fingerprint verification processing unit 13. As such a fingerprint photographing unit, various commercially available ones may be used as appropriate, and will not be described in detail here. In this embodiment, an optical image sensor is used, but other image sensors such as a capacitance type, a pressure sensitive type, and a thermal type may be used.

The storage unit 14 is configured by a semiconductor storage element such as a ROM / RAM, and stores in advance fingerprint data of a user who is an authentication target in the fingerprint collation apparatus 1. The storage unit 14 also stores an operation program and various parameters of the fingerprint collation apparatus 1, and is read out and used at an appropriate timing in the collation processing in the fingerprint collation processing unit 13. Examples of the various parameters stored include a window size (W / H) serving as a reference for the density of feature points to be described later, a threshold value for the density, and the like.

The output unit 15 is an interface that outputs a collation result in the fingerprint collation processing unit 13 to an external device (not shown) such as an electric lock control device that performs opening / closing control of the electric lock installed on the door. For example, when the electric lock control device receives an output of verification OK, the electric lock installed on the corresponding door is unlocked. On the other hand, in the case of output with collation NG, the electric lock is not unlocked.

The fingerprint collation processing unit 13 includes an arithmetic processing chip such as a CPU / MPU and peripheral circuits, and reads various programs and various parameters stored in the storage unit 14. Then, the fingerprint collation processing unit 13 determines the identity of the input fingerprint image captured by the fingerprint imaging unit 12 with a registered fingerprint image registered in advance in the storage unit 14. Thus, a process for authenticating whether or not the user who has input the fingerprint image is a person registered in advance is performed.

Specifically, the fingerprint collation processing unit 13 is a fingerprint image acquisition unit 131 that acquires an input fingerprint image input from the fingerprint photographing unit 12, and a feature point extraction that processes the acquired input fingerprint image and extracts a feature point. Using the means 132, the position feature point selecting means 133 for selecting the position feature point used for position matching from the feature points extracted by the feature point extracting means 132, and the position matching feature point selected by the position feature point selecting means 133 After performing position matching with the feature points of the registered fingerprint image registered in advance in the storage unit 14, the feature values at all the feature points are used for a plurality of positional relationships having high similarity in the position matching. The collation unit 134 is configured to collate and determine the identity of both fingerprint images.

The feature point extraction unit 132 performs a contrast correction process on the input fingerprint image acquired by the fingerprint image acquisition unit 131 in order to reduce the influence of the image contrast variation caused by the change of the shooting environment. In order to clearly distinguish the ridges and valleys in the fingerprint, a binarization process is performed to distinguish the ridges into black and the valleys into white. Thereafter, a thinning process is performed on the ridge that is a set of black pixels. The ridge end points and branch points of the input fingerprint image thus obtained are extracted as feature points. At this time, each feature point is extracted by associating individual feature amounts such as position information on the input fingerprint image and direction information in which the ridges flow. And it outputs to the position feature point selection means 133. In the present invention, a general method may be adopted as a feature point or feature amount extraction method, and therefore, detailed description thereof is omitted to simplify the description.

FIG. 4 shows the feature points extracted by the feature point extraction unit 132 superimposed on the fingerprint image. In the figure, the black linear portion indicates the ridge, and the white portion indicates the valley line. The feature points are indicated by Δ, and position information and ridge direction information are also provided. It can be seen that a large number of feature points are extracted in a substantially straight line with high density in the outline portion and the flaw portion of the fingerprint image. Such feature points are extracted because the ridges are interrupted in the contour portion of the fingerprint image.

The position feature point selection unit 133 selects a position feature point that is a feature point obtained by excluding those that do not contribute much to the alignment process between the input fingerprint image and the registered fingerprint image from the feature points obtained by the feature point extraction unit 132. To do. Here, the registration processing refers to the registration of the registered fingerprint image and the input fingerprint image by obtaining the degree of similarity in which the positional relationship between the extracted feature points of the registered fingerprint image and the input fingerprint image is similar to each other. To do. Then, the position feature point and other feature points are distinguished and output to the matching unit 134. Detailed processing of the position feature point selection unit 133 will be described in detail with reference to FIG. 3 when the operation flow of the fingerprint collation apparatus 1 is described.

FIG. 5 shows the position feature points extracted by the position feature point selection means 133 superimposed on the fingerprint image. As is clear from the comparison with FIG. 4, it can be seen that a large number of feature points extracted in the outline and scratches of the fingerprint image are excluded.

The matching unit 134 reads the feature points of the registered fingerprint image registered in advance in the storage unit 14. Then, the collation unit 134 collates the registered fingerprint image with the input fingerprint image using the feature points of the registered fingerprint image read out, the position feature points from the position feature point selection unit 133 and other feature points. Execute. Specifically, the collation unit 134 first executes the alignment process using the position feature point selected by the position feature point selection unit 133 and the feature point of the registered fingerprint. At this time, the similarity indicating the degree of matching is obtained. Next, a plurality of upper positional relationships having a high degree of similarity, in other words, a high degree of matching of the positions of feature points, are obtained.

Then, the collation unit 134 considers the feature amount such as the ridge direction using the feature points of the registered fingerprint image and the feature points extracted by the feature point extraction unit 132 of the input fingerprint image for the upper positional relationship. To obtain the similarity. If the similarity is such that identity can be authenticated, it is determined that the verification is OK. On the other hand, if the degree of similarity is not enough to authenticate the identity, it is determined as collation NG. If at least one of the upper positional relationships is determined to be verified, it is determined that the registered fingerprint image and the input fingerprint image have the same identity, and verification is successful. In the present embodiment, all feature points extracted by the feature point extraction unit 132 are used for collation. However, if further speedup is desired, collation is performed using only the position feature points. Also good.

Generally, in fingerprint collation processing, the processing for obtaining the similarity of the positional relationship between feature points has the highest load and requires a long collation time. The similarity of the positional relationship between feature points is to identify all combinations of many feature points in one fingerprint image, and compare the positional relationship between the extracted feature points between the registered fingerprint image and the input fingerprint image. become. For this reason, as the number of feature points increases, the number of combinations increases exponentially. Therefore, if the number of feature points used for the alignment process is reduced, the speed of the matching process is greatly improved.

Next, the operation of the fingerprint collation apparatus 1 configured as described above will be described with reference to FIGS. FIG. 2 is an overall operation flow of the fingerprint collation apparatus 1, and FIG. 3 is a processing flow in the position feature point selection unit 133.

First, it is assumed that the fingerprint collation apparatus 1 has registered fingerprint images acquired from legitimate users, feature points of the fingerprint images, and the like registered in the storage unit 14 and is ready for collation processing. In step S <b> 20, the user stands by until an operation for starting collation is performed from the touch panel of the operation / display unit 11. Here, in the present embodiment, the operation signal in the operation / display unit 11 is used as a trigger. However, the trigger may be that the user has approached the fingerprint collation apparatus 1 using an infrared sensor. The trigger may be that the finger is placed on the fingerprint photographing unit 12.

When the user operates the operation / display unit 11, guidance for verification from the guidance means (not shown) of the fingerprint verification processing unit 13 to the user is displayed on the operation / display unit (S21). Specifically, the user is prompted to place a finger on the mounting table of the fingerprint imaging unit 12.

Next, the fingerprint imaging unit 12 outputs a fingerprint image obtained by photographing the fingerprint image of the placed finger to the fingerprint collation processing unit 13 (S22). When the fingerprint image is acquired by the fingerprint image acquisition unit 131, the feature points of the fingerprint image acquired by the feature point extraction unit 132 are extracted (S23). Each feature point is
It has three feature quantities: the position of the feature point that is the end point or branch point of the ridge or valley line, the direction of the ridge at that position, and the reliability of the feature point.

Step S24 is a process of the position feature point selection unit 133, and executes the process shown in FIG. In FIG. 3, in step S30, either a vertically long rectangle or a horizontally long rectangle is selected as a search window to be used as an area serving as a reference for the density of position feature points used for alignment. In the present embodiment, two types of vertically long rectangles and horizontally long rectangles are prepared, but either one may be provided, and three or more windows having different rectangular widths and heights may be prepared. In this case, the aspect ratio of the window size is preferably 4.0 or more. This is to use the property that the pseudo feature points are linearly arranged.

Here, the window defines a predetermined area for determining the density of feature points, and is defined by the window size width and height. In the present embodiment, since the resolution of the fingerprint image captured by the fingerprint photographing unit 12 is 500 dpi, two types of windows, a vertically long rectangle having a width of 10 pixels × 240 pixels in height and a horizontally long rectangle having a width of 240 pixels × 10 pixels in height, are used. Is used. These two types of windows are stored in the storage unit 14. The reason why the vertically long rectangle and the horizontally long rectangle are used is that the pseudo feature points have a property when arranged in a straight line. In step S30, a vertically long rectangular window is first selected.

Then, the selected vertically long rectangular window 61 is set at the upper left position of the fingerprint image (S31), and the number of feature points located in the window at the set position is counted (S32). Here, the setting position of the vertically long rectangular window will be described with reference to FIG. FIG. 6 shows a state in which the vertically long rectangular window 61 is set to the upper left, and scans in the right direction sequentially from the left along the route indicated by the reference numeral 62, and when it reaches the right end, shifts downward and moves from right to left. Scan to. When it reaches the left end, it shifts further downward and scans from right to left. In this way, the entire fingerprint image is scanned. Similarly, in FIG. 7, the horizontally long rectangular window 71 is scanned along the route 72. Since the same scanning as in FIG. 6 is performed, repeated description is omitted.

It is determined whether or not the counted number of feature points exceeds a predetermined threshold value (S33), and if it exceeds the predetermined threshold value, a feature point located in the window is set as a deletion candidate (S34). On the other hand, if the predetermined threshold is not exceeded, the process proceeds to the window position setting in step S31 without setting the deletion candidate.

Here, the predetermined threshold will be described. The predetermined threshold value is used for determining the density of feature points located in the window. For example, in a fingerprint image acquired at a resolution of 500 dpi, if the window is 10 pixels wide by 120 pixels high, the number of ridges is about 10, so it is preferable that the number is less than about 5 that is half of that. This is because the number of feature points that can be extracted from the fingerprint image is affected by the resolution of the fingerprint image, that is, when using a window of the same size, the higher the resolution, the smaller the number of ridges that enter the window. As a result, a small number of feature points are extracted, and when the resolution is low, the number of ridges entering the window increases, resulting in an increase in the number of feature points extracted. Also, the number of ridges entering the window differs depending on the window size, and the number of feature points extracted also differs.

This time, we would like to reduce the number of position feature points used for alignment as much as possible without affecting the accuracy of alignment. On the other hand, if the number of position feature points is reduced too much, the alignment accuracy will be reduced. Therefore, in consideration of these, a predetermined threshold value is determined, but it is abnormally that one feature point is extracted per ridge. As a rule of thumb, it has been found from experience that if there are about one feature point in two ridges, that is, about 50% of the number of ridges, the possibility of a pseudo feature point is extremely high. ing. Therefore, the predetermined threshold is at least smaller than about 50% and can be appropriately determined to such an extent that the alignment accuracy can be maintained.

In the window position setting process in step S31, the window is shifted from the first upper left position to the right by about half the window width. FIG. 5 shows the order of scanning the fingerprint image by the window. The left figure shows scanning when scanning a vertically long rectangular window. The processes from S31 to S34 are repeated until the entire fingerprint image is scanned through the vertically long rectangular window.

Next, when scanning with a vertically long rectangular window is completed, it is determined in step S35 whether or not the processing for all windows has been completed. In the case of the present embodiment, since scanning with a horizontally long rectangular window remains, the process returns to step S30 and the same processing is performed. The right figure of FIG. 5 shows the situation of scanning a horizontally long rectangular window. When all the windows have been scanned, feature points excluding feature points that are candidates for deletion in step S36 are selected as position feature points. Then, the position feature points and other feature points that are not position feature points are made identifiable and output to the matching unit 134.

Returning to FIG. 2, in step S <b> 25, the collating unit 134 executes collation processing. That is, the matching unit 134 reads the feature points of the registered fingerprint registered in advance in the storage unit 14. Then, the collation unit 134 collates the registered fingerprint image with the input fingerprint image using the read feature points of the registered fingerprint, the position feature points from the position feature point selection unit 133, and other feature points. To do.

In step S26, if the collating unit 134 determines that the collation is OK (S26 YES), a collation OK signal is output from the output unit 15 (S27). If the collating unit 134 determines that the collation is NG (S26 NO), the output unit 15 outputs a verification NG signal to an external device (not shown) (S28). Then, the process returns to step S20 and waits until the next collation operation is performed.

In the present embodiment, in step S31, the window is linearly scanned in an upright state. However, the window may be scanned while being rotated to about ± 30 degrees at each window position.

In the present embodiment, both the position feature points and other feature points are used for collation, but only the position feature points may be used. In this case, the processing speed of the collation process is further improved.

In this embodiment, the feature point selection unit 133 is operated in all cases regardless of the number of feature points extracted by the feature point extraction unit 132. However, the number of extracted feature points affects the matching time. The feature point selection means 133 may be operated only when a large number of effects are extracted. As a result, the pseudo feature point can be deleted only when the verification time is likely to be long, so that the processing time can be stabilized.

In the present embodiment, the feature points have been described without distinction between the end points and the branch points. However, the pseudo feature points generated due to the outer peripheral portion of the fingerprint image or scratches usually do not include the branch points. For this reason, if the inner branch point of the feature point does not become a count and deletion candidate in step S32 and step S34, the feature point can be prevented from being deleted unnecessarily.

DESCRIPTION OF SYMBOLS 1 ... Fingerprint collation apparatus 11 ... Operation part / display part 12 ... Fingerprint imaging | photography part 13 ... Fingerprint collation process part 131 ... Fingerprint image acquisition means 132 ... Feature point extraction means 133 ... Position feature point selection means 134 ... collation means 14 ... storage unit 15 ... output unit

Claims (4)

A fingerprint collation device for collating a registered fingerprint stored in advance with an input fingerprint,
Fingerprint image acquisition means for acquiring a fingerprint image;
A feature point extracting means for extracting feature points such as end points or branch points of ridges or valleys together with position information from a fingerprint image;
Position feature point selecting means for selecting a position feature point excluding the feature points existing in the area when the density of the feature points extracted by the feature point extracting means exceeds a threshold value; ,
A registration unit that performs registration of the registered fingerprint and the input fingerprint using the position feature points, and collates the identity of the registered fingerprint and the input fingerprint;
An output unit for outputting a result of collation by the collating unit;
A fingerprint collation apparatus characterized by comprising:   2. The fingerprint collation apparatus according to claim 1, wherein the position feature point selection unit scans a fingerprint image using a rectangular area window in a predetermined area, counts the number of feature points in the area window, and obtains a density. .   The said position feature point selection means sets the threshold value of the said density | concentration with the correlation with the number of the ridge or valley line which can enter in the said area | region, and it makes it so low that the said number is small. Fingerprint verification device.   The position feature point selection unit calculates a density by excluding a branch point from the feature points extracted by the feature point extraction unit, and also excludes a branch point even when the density exceeds a threshold value 4. The fingerprint collation apparatus according to claim 1, wherein the fingerprint collation device is selected as a position feature point.