JP2006072554A - Fingerprint input image judgment processing device and fingerprint collation device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a fingerprint input image judgment processing device for judging whether or not inputted fingerprint image data are desired brightness and darkness-related image data regardless of the factor, and for converting the inputted fingerprint image data into the desired brightness and darkness-related image data based on the judgment result. <P>SOLUTION: The fingerprint input image judgment processing device 20 is provided with an input image judging part 21 and a brightness inversion processing part 22. The input image judging part 21 first generates binarized judgment image data based on the inputted fingerprint image data. Then, the input image judging part 21 detects an isolated region in the generated judgment image data, and counts the number of configuring elements in the isolated region constituted of a predetermined color. Then, the counted number of pixels is verified by using predetermined conditions, and the brightness and darkness relation of the inputted image data is judged. The brightness inversion processing part 22 inverts the brightness and darkness relation of the inputted fingerprint image data when the brightness and darkness relation of the input image is different from desired one. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to fingerprint collation processing, and in particular, determines the light / dark relationship between ridges and valleys of fingerprint image data input from a fingerprint reader or other fingerprint input device, and determines the light / dark relationship required by the fingerprint collation device. The present invention relates to a fingerprint input image determination processing device that can be changed to image data, and a fingerprint collation device.

  In recent years, there has been an interest in the importance of security based on identity verification, such as performing credential authentication when entering or leaving a building or room. There are various methods such as authentication using a password as a method for identity verification. As one of the methods, an apparatus that performs authentication using a fingerprint, which is unique biometric information, is often used.

  In the fingerprint authentication device of the feature point matching method, a finger is lightly touched on the fingerprint reading device to read the fingerprint of the finger, a plurality of feature points are extracted from the read fingerprint image, and the extracted feature points and the feature points of the registered fingerprint image Are compared and whether or not the input fingerprint is a registered fingerprint is checked. The feature points are extracted as various data of end points and branch points of the ridges by analyzing the ridge pattern of the fingerprint image. The fingerprint image data read by the fingerprint reading device and input to the collation device has the light / dark relationship between the ridges and valleys reversed depending on the reading method. For this reason, in the fingerprint authentication device, it is general to set in advance the relationship between the ridges and valleys of the fingerprint image data for extracting the fingerprint reading device to be used and the feature points. Here, the light-dark relationship between ridges and valleys is “the relationship in which ridges appear as high brightness (brighter) and valleys appear as low brightness (darker in fingerprint image data, or vice versa)” It shows that.

  For example, an optical type is often used as a reading method of a fingerprint reading device. As a typical optical fingerprint reader, a reflected light system that obtains image data by irradiating a finger from a light source and receiving reflected light from the finger, and receiving scattered light from the finger There is a scattered light method for acquiring image data.

  In the optical fingerprint reader using the reflected light method, the amount of light reflected from the ridge portion of the finger in contact with the fingerprint reading surface is less than the amount of light reflected from the valley portion not in contact with the reading surface. The image is low-intensity (dark) and the valley is high-intensity (bright).

  In the optical fingerprint reader using the scattered light method, the amount of light scattered from the ridge portion of the finger in contact with the fingerprint reading surface is larger than the amount of light scattered from the valley portion not in contact with the reading surface. An image having a high luminance (bright) in the line portion and a low luminance (dark) in the valley portion is obtained.

  Thus, if the reading method of the optical reading sensor is different, the contrast of the output image data is reversed. For this reason, if the fingerprint readers used for collation and registration differ, the image is reversed and the extracted feature data differs, so that fingerprint collation cannot be performed. Therefore, in the conventional fingerprint reader, the fingerprint reader of the same reading method is used at the time of registration and collation, or if different reading methods are used at the time of collation, the brightness and darkness of the input fingerprint image data is reversed. Thus, erroneous determination due to a difference in reading method was prevented. When the reading method is determined in advance, it is possible to prevent erroneous determination due to the difference in the reading method by matching the light and dark of the image data at the time of registration and collation in advance according to the reading method in this way. is there.

Furthermore, as a technology that enables fingerprint collation even when the reading method of the optical reading device used at the time of collation is not determined in advance, the reading method of the reading device is confirmed at the time of collation, and an input image is obtained if necessary. There has also been proposed a fingerprint reader that changes data to light-dark image data that is required (see Patent Document 1). This device compares the output image before placing the finger on the fingerprint reading device with the output image after placing the finger, and determines the reading method of the fingerprint reading device depending on which of the two image data has more bright pixels. Judgment.
JP 2001-84371 A

  When the output data from the fingerprint reader is in a normal state, that is, in the case of the reflection type, the ridge is output as white (light) and the valley is black (dark). When the image data is output as black (dark) lines and white (bright) valleys, the light-dark relationship of the input fingerprint image data is determined and changed to the desired light-dark relationship according to the above-described prior art. It is possible. However, under certain conditions, the contrast between the ridges and valleys of the fingerprint image acquired by the fingerprint reader may appear in reverse to the output image data of the normal reading method. Such a light-dark reversal phenomenon may occur when an extremely dry finger or a finger to which alcohol or the like is attached is read. For example, in the case of using a scattered light type fingerprint reader, the valley portion of the finger that is white due to the effect of drying or alcohol may be scattered as the scattered light becomes strong and may be read as a bright portion. Due to the overlap of these various conditions, a light-dark reversal phenomenon occurs rarely. Hereinafter, in this specification, only reversal of light and darkness of an image that occurs under such a specific condition is referred to as “reversal”, and “inversion” refers to reversing light and darkness of an image consciously (artificially). Called.

  The reversal phenomenon of fingerprint reading data output from the fingerprint reading device as described above is due to a factor different from the original function of the device, and the ridges and valleys of the fingerprint are always constant according to the reading method of the input device. However, the above prior art that presumes that the output is correct in the light / dark relationship cannot be dealt with at all.

  The present invention has been made in view of the above-described problems. Regardless of the difference in the reading method of the fingerprint reading device and the reversal phenomenon of the image data, the light / dark relationship of the fingerprint image data input from the reading device or other input device is obtained. It is an object of the present invention to provide a fingerprint input image determination processing device and a fingerprint collation device that determine and convert and input light / dark relationship image data that matches the light / dark relationship set in the collation unit.

The fingerprint input image determination processing apparatus according to the first embodiment of the present invention includes a fingerprint input means for inputting fingerprint image data representing fingerprint ridges and valley line patterns, and all or one of the input fingerprint image data. A determination image generation unit that generates determination image data obtained by binarizing the luminance of each pixel, and an isolated region extraction unit that extracts an isolated region in the binary determination image data generated by the determination image generation unit. , One of the two luminance pixels among the pixels constituting the isolated region is counted, and based on the counted number of pixels, it is determined whether the pixel counted in the isolated region is a pixel constituting the sweat gland When determining that the counted sweat gland and the counted pixels are pixels constituting the sweat gland, the brightness-darkness relationship representing the ridges and valleys in the judgment image data is expressed by one luminance. Other lines Characterized in that it comprises the light and dark relationship determination means determines that brightness relationship expressed by luminance, a.

As a result, even when the brightness relationship between the ridges and valleys in the input fingerprint image data is unknown or reversed, the isolated region is observed to determine the brightness representing the sweat gland. Thus, it is possible to know the light / dark relationship between the ridges and valleys in the fingerprint image data. When this is used in a fingerprint collation device, if the feature information of the fingerprint is extracted from the ridge (or valley line) based on the determined light / dark relationship, the state of the light / dark relationship in the input fingerprint image data is obtained. Regardless, feature information can be appropriately extracted.

  In fingerprint image data, ridges and valleys are represented by opposite luminances. That is, when the luminance of the ridge is low (dark), the luminance of the valley is high (bright), and when the binarized value is expressed in black and white, the ridge is black (dark) and the valley Becomes white (bright). As described above, the light-dark relationship between the ridges and the valleys is expressed by high and low luminance (black and white).

  Also, the sweat glands are converted to the same brightness as that of the valley lines by binarization processing, and the brightness is opposite to that of the ridges. Therefore, a sweat gland distributed on the ridge can be detected as an isolated region. And by observing the isolated area and detecting the brightness of the sweat glands, the valley line is the same as the sweat gland, but the ridge line has the opposite brightness to the completed line, and the light-dark relationship between the ridge line and the valley line is known. Can do.

  The fingerprint input image processing apparatus according to another aspect of the present invention further includes a light / dark relationship storage unit for storing a predetermined light / dark relationship and a light / dark relationship in which the light / dark relationship determined by the light / dark relationship determining unit is stored. The reverse rotation detection means for determining whether the light / dark relationship is reversed, and when it is determined that the light / dark relationship is reverse, the brightness of each pixel of the fingerprint image data or the determination image data matches the stored light / dark relationship. And reversing processing means for correcting it.

  In this aspect, when the light-dark relationship between the ridges and valleys in the fingerprint image data reverses the desired light-dark relationship, the luminance of each pixel is corrected so that the desired light-dark relationship is obtained. Thereby, when used for a fingerprint collation device, feature information can be appropriately extracted from the input fingerprint image data.

  When correcting the binary determination image data when correcting the brightness, the brightness of each pixel may be simply replaced by light and dark (black and white) replacement. In the case of correction in fingerprint image data, correction is performed such that the lower the luminance, the higher the corrected luminance, and the higher the luminance, the lower the corrected luminance. For example, when the fingerprint face image data is expressed in 256 gradations, the luminance (brightness) of the pixels with 0 brightness may be corrected to 255 (brightness), and the brightness of 100 pixels should be corrected to 155. .

  According to the fingerprint input image determination processing device according to another aspect of the present invention, the isolated region extraction unit scans each pixel of the determination image data with a matrix filter having a predetermined size, and surrounds the determination target pixel. When a predetermined number or more of pixels having different luminances exist in the region, the determination target pixel is determined to be a pixel constituting the isolated region.

Whether or not the fingerprint image data has a predetermined light / dark relationship does not need to accurately extract all the sweat glands, and it is sufficient to determine which luminance pixel constituting the isolated region is the sweat gland. Since many sweat glands exist on the fingerprint in isolation, if an approximate sweat gland is extracted as an isolated region and the number of pixels is counted, it is possible to determine which pixel has any luminance. Therefore, a matrix filter having a predetermined size is used, and an isolated region is roughly determined based on whether or not the pixel to be determined by the majority of pixels on the filter is isolated. This makes it possible to extract the pixels constituting the isolated region at high speed. Although the accuracy of sweat gland extraction decreases, it is possible to ensure sufficient accuracy for determining whether the light-dark relationship is reversed.

  According to the fingerprint input image determination processing device according to another aspect of the present invention, the sweat gland determination means counts one of the two luminance pixels among the pixels constituting the isolated region and includes the determination image data in the determination image data. When the ratio of the number of pixels counted in the determination image and the number of pixels counted in the isolated region is equal to or greater than a predetermined threshold, the pixels counted in the isolated region It is determined that the pixel is a constituent pixel.

  This mode prevents erroneous determination of sweat glands when an isolated region (called an “island”) having a brightness opposite to that of the sweat glands is formed in the valley line due to a strong pressing of the finger or the like. In this case, many islands may be formed depending on the degree of finger pressing, and the number of pixels becomes considerably large. Therefore, if only the number of pixels constituting the isolated area is determined, there is a risk of erroneous determination. Since the islands on the valley line increase when the finger is strongly pressed, in that case, the area of the ridge also increases. Therefore, the ratio of the area of the island on the valley line and the area of the ridge does not exceed a certain value, and the ratio between the number of isolated regions having the same luminance and the number of pixels having the same luminance in the entire image is constant. When the value is exceeded, it can be determined that the pixel counted in the isolated region is due to sweat glands.

  A fingerprint collation apparatus according to a first aspect of the present invention includes a fingerprint reading unit that reads a fingerprint from a placed finger and generates fingerprint image data representing fingerprint ridge and valley fingerprints, and an input fingerprint Determination image generation means for generating determination image data obtained by binarizing the luminance of each pixel for all or part of the image data, and extraction of isolated regions in the binary determination image data generated by the determination image generation means The isolated region extracting means, and counting the pixels having one of the two luminance values among the pixels constituting the isolated region, and the pixels counted in the isolated region are the pixels constituting the sweat gland based on the counted number of pixels. The sweat gland determination means for determining whether or not there is a pixel, and if the counted pixel is determined to be a pixel constituting the sweat gland, the brightness contrast representing the ridges and valleys in the determination image data is as follows. No shine The ridge is determined by the light / dark relationship determining means for determining that the ridge is the light / dark relationship expressed by the other luminance, the light / dark relationship storing means for storing the predetermined light / dark relationship, and the light / dark relationship determining means. The reverse detection means for determining whether the light / dark relationship is the stored light / dark relationship or the reverse light / dark relationship, and if it is determined that the light / dark relationship is the reverse light / dark relationship, it matches the stored light / dark relationship. Fingerprint image data or inversion processing means for correcting the brightness of each pixel of the determination image data, and identifying ridge and valley line patterns in the fingerprint image data based on a predetermined light / dark relationship, and fingerprint feature information A feature extraction unit that extracts the feature information, a registered fingerprint storage unit that stores feature information of a fingerprint registered in advance, feature information extracted from the input fingerprint image data, and feature information stored in the registered fingerprint storage unit Similarity Out, characterized in that it and a fingerprint verification unit for performing fingerprint collation by comparing the similarity with a predetermined threshold value.

  According to this aspect, even when the light / dark relationship of the input fingerprint image data is unknown or reversed, it is possible to appropriately extract the feature information from the fingerprint image data, thus enabling high brightness fingerprint matching. It becomes.

  According to the present invention, it is possible to detect how the brightness of the ridges and valleys of the fingerprint image data input from the fingerprint reader or other input device has a light-dark relationship. Moreover, it is possible to detect whether the detected light-dark relationship is what the fingerprint collation device requires, regardless of the reason. Therefore, fingerprint image data can be collated using a plurality of fingerprint readers with different reading formats by adopting a configuration in which the fingerprint image data is input to the collation device via the fingerprint input image determination processing device according to the present invention. It becomes.

  At that time, even if the fingerprint image data output by the reading device is reversed from the original light / dark relationship, the collation device is used as the correct light / dark fingerprint image data regardless of the cause of the reverse rotation. Can be entered. As described above, since the image data of the light and dark relationship to be obtained can be provided regardless of the reading method and regardless of the cause of the inversion, it is possible to construct a fingerprint collation system in which the reading methods are mixed. Even when the input image data is reversed, it is possible to obtain an effect that correct fingerprint collation is possible. Furthermore, it is possible to freely use a reading device of a different reading format as an input device of the fingerprint collation device, and no adjustment work is required for matching the type of the fingerprint reading device and the fingerprint collation device at the time of installation. There is also an effect of becoming.

Embodiments of the present invention will be described below in detail with reference to the drawings.
(Explanation of operation of fingerprint verification device)
The fingerprint input image processing apparatus according to the present invention can be used by being incorporated as a part of the fingerprint verification apparatus, or can be used as a fingerprint input image processing apparatus independent of the verification apparatus. Further, when a plurality of fingerprint readers are connected to a collation device via a communication line such as LAN or WAN and fingerprint collation is performed intensively, first, fingerprint image data input via the communication line is used as the present invention. After the determination process is performed by the fingerprint input data determination processing apparatus, the data may be input to the collation apparatus. Further, after the fingerprint image data read by the fingerprint reading device is subjected to the determination processing by the fingerprint input data determination processing device according to the present invention, the feature data is extracted, only the feature data is transmitted through a line such as a LAN, and only the collation processing is performed on the server. It is also possible to adopt a configuration in which centralized processing is performed, for example. In the following, an example will be described in which the fingerprint collation apparatus is integrally incorporated as a fingerprint input image determination processing unit.

  FIG. 1 is a functional block diagram illustrating a fingerprint collation apparatus including a fingerprint input image processing unit according to an embodiment of the present invention. The fingerprint collation apparatus 10 stores an operation display unit 11 having an input function and a display function, a fingerprint input unit 12 that acquires and inputs a fingerprint of a finger as image data, a data processing unit 13, and fingerprint data to be collated. And a registration data storage unit 14. The operation display unit 11 includes a display screen and a touch panel provided on the display screen, a touch pen, or a key input unit (not shown) independent of the screen, and sets various conditions, parameters, and the like of the fingerprint collation device 10. And an operation unit for performing an operation of fingerprint registration processing.

  The fingerprint input unit 12 is a device that reads a fingerprint of a finger placed at the time of fingerprint registration or verification and inputs the fingerprint image data to the data processing unit 13. In the case of fingerprint collation, generally, each pixel is output as 256 gray scale data. As described above, which luminance (brightness) of the ridge and the valley is higher depends on the reading method of the input device or the state of the finger to be read.

  The data processing unit 13 registers fingerprint data or collates input fingerprints. In the registration of fingerprint data, predetermined feature data is extracted from the inputted fingerprint image and stored in the registration data storage unit 14. The registration data storage unit 14 is a storage unit for registering and storing fingerprint data of a predetermined authorized person determined in advance. If the predetermined registration data can be read out at a relatively high speed, a semiconductor storage device However, it may be any of a disk drive system or a recording apparatus using various media such as CD-RW and DVD.

  When the input fingerprint matches the registered fingerprint data as a result of the verification of the data processing device 13, a verification confirmation signal is sent from the data processing device 13 to the output unit 19, and an external electronic lock or the like is monitored from the output device. A permission signal is output to a device (not shown). The monitoring device is a device that monitors a predetermined place, facility, room, object, or virtual space where access such as entry / exit, login, etc. is restricted, and access is performed based on a permission signal from the verification device. to approve.

  The fingerprint verification will be described in more detail. The data processing unit 13 includes a CPU, a ROM, a RAM, a register, a control program, and the like, and a PC or the like can also be used. The data processing unit 13 includes a control unit 15 that controls the operation of the entire data processing, a data storage unit 16 that primarily stores various data such as fingerprint images, a fingerprint input image determination processing unit 20, a feature extraction unit 17, and a matching unit 18. It has.

  The fingerprint image data input from the fingerprint input unit 12 is temporarily stored in the data storage unit 16 by the control unit 15. Next, the luminance (lightness) of the fingerprint image data input is reversed by the fingerprint input image determination processing unit 20 as necessary. The fingerprint input image determination processing unit 20 includes an input image determination unit 21 and a brightness inversion processing unit 22. The input image determination unit 21 detects the sweat glands of the input fingerprint image data, determines the light / dark relationship between the ridges and valleys from the luminance (lightness) of the pixels constituting the sweat gland, and the light / dark relationship is determined by the matching unit 18. Determine if it is the same as the setting. The predetermined light / dark relationship is stored in the collation unit 18 or a storage unit (not shown).

  If it is determined that the image has been reversed, the lightness reversal processing unit 22 reverses the luminance of all the pixels of the input fingerprint image data (black and white reversal). For example, if each pixel of the input fingerprint image data is expressed by a luminance value (brightness) I (x, y) of 256 gradations, the luminance values of all the pixels are set to I ′ ( x, y) = 255-I (x, y). As a result, the image data of the input fingerprint image is inverted in black and white (light and dark) with 256 gradations. Alternatively, conversion may be performed such that I ′ (x, y) = 2A−I (x, y), where A is the average value of I (x, y).

  If it is determined that the image has not been reversed, the input fingerprint image data conforms to the setting of the collation unit, so that the input fingerprint image data is maintained as it is without doing anything. When the determination process by the fingerprint input image processing unit 20 is completed, the control unit 15 outputs the fingerprint image data subjected to the determination process to the feature extraction unit 17. The feature extraction unit 17 extracts feature data to be collated based on the fingerprint image data output from the inversion processing unit. It is also possible to perform image processing such as thinning processing before extracting feature data.

The feature data is represented by feature points that are the end points of fingerprint ridges and branch points that are the divergences of fingerprint ridges and valleys. Each feature point has its attribute data such as position, direction, reliability, etc. Has elements. The position of the feature point is represented by coordinates, and the direction is represented by an angle. Further, the reliability of the feature point can be obtained as a value proportional to the size of the direction vector obtained by adding the respective direction unit vectors of the neighboring feature points and averaging them. FIG. 2 illustrates fingerprint image data, and the feature points will be described. FIG. 2 shows the fingerprint image data after thinning processing. In the figure, the end points which are feature points are represented by numerals 40, 41 and 42, and the branch points are represented by numerals 43. Further, θ _T of the end point 40 is illustrated as the direction of the feature point. A plurality of these feature points are extracted from one fingerprint image, and attribute data including the coordinates of each of the plurality of feature points is output to the matching unit 18 as feature point list data.

  The collation unit 18 collates the feature point list data extracted by the feature extraction unit 17 with the feature point list data registered in the registered data storage unit 14 to calculate the similarity, and the calculated similarity is determined according to a predetermined value. By comparing with the threshold value, it is determined whether or not the registrant can be authenticated. The output unit 19 outputs a permission signal if the result of determination by the collation unit is that authentication is possible. The monitoring device that has received the permission signal unlocks the electric lock or permits login or the like, for example.

  An example of the overall processing procedure of the fingerprint collation apparatus 10 will be described using the flowchart of FIG. First, the fingerprint collator operates the operation display unit 11 to input his / her identification number (ID number) (S101). The ID number is used to increase the security level of fingerprint collation, and the ID number can be omitted.

  Next, the fingerprint collator places a predetermined finger on the fingerprint input device 12. The fingerprint input device reads the fingerprint image data of the placed finger and inputs it to the data processing device 13 (S102). The input fingerprint image data is first determined whether the input image data is image data related to light and darkness planned by the collation device, and the input image data is adapted as necessary to match the setting of the collation unit 18. Inversion processing is performed (S103). When the processing of the input image is completed, predetermined feature points are detected from the input fingerprint image data, and feature data is extracted (S104).

  When the feature data is extracted from the input fingerprint image data (S104), the registered fingerprint feature data is sequentially read from the registered data storage unit 14 and compared with the input fingerprint feature data (S104). S105). At the time of comparison, the similarity between the inputted fingerprint feature data and the feature data of each registered fingerprint is calculated (S106), and the feature data of both fingerprints match when the similarity exceeds a predetermined reference value. It is determined that it is to be performed (collation OK) (S107; YES). When a matching registered fingerprint is found (S107; YES), a permission signal is transmitted from the output unit 19 to the monitoring device (S108). At this time, a collation confirmation guidance such as “collation is OK” may be output by voice or display (S109).

  If a registered fingerprint that matches the input fingerprint is not found (S107; NO), no permission signal is output. At this time, it is possible to configure such that guidance such as “failure of collation” is played (S109).

(Fingerprint input image determination processing unit)
Next, the fingerprint input image determination processing unit (or fingerprint input image determination processing device, hereinafter referred to as “fingerprint input image determination processing unit”), which is a main part of the present invention, will be described in more detail.

  FIG. 4 is a functional block diagram of the fingerprint input image determination processing unit 20 according to the first embodiment of the present invention. The input image determination unit 21 of the fingerprint input image determination processing unit 20 includes a determination image generation unit 23 and a reverse rotation detection unit 26. The lightness determination processing unit 22 includes an inversion control unit 30, a lightness calculation unit 31, and an image reading unit. 32.

(Decision image generator)
The determination image generation unit 23 generates determination image data for determining the light / dark relationship of the image data. The determination image generation unit 23 includes a correction processing unit 24 and a binarization processing unit 25, and performs correction and binarization processing of input image data. As correction of the image by the correction processing unit 24, for example, the density unevenness of the fingerprint image caused by the difference in the pressing force between the central part and the peripheral part of the finger placed on the fingerprint input part is eliminated, and the entire density is made uniform. Perform such correction. More specifically, processing such as dividing a fingerprint image into partial areas and converting the luminance values so that the average and variance of the luminance values for each partial area become the average and variance of the luminance values of the entire fingerprint image. Thus, density unevenness correction can be performed. Correcting the density unevenness in this way can improve the accuracy of the binarization process.

  The binarization processing unit 25 converts gray scale image data in which density unevenness and the like are corrected into binary image data of white (bright) or black (dark) based on a predetermined threshold (threshold). Instead of the binarization processing unit, ternarization and other multi-value processing are also possible. However, there is a problem that the processing load increases when the number of values is increased. On the other hand, the input image determination process requires high-speed processing in order to shorten the verification time. Also, the presence / absence of reversal of the image can be determined with high accuracy also from the binarized image. From the above viewpoint, the determination image can be a multi-valued image, but is preferably a binary image.

  Note that, in order to generate determination image data (binary image data), if there is a large amount of data (number of pixels) to be processed, both the determination image data generation process and the light / dark determination process of the image data take time. As a result, the fingerprint collation process takes a lot of time. On the other hand, the purpose of generating the determination image data is to determine how the input fingerprint image data has the light / dark relationship (the light / dark relationship between the ridge and valley pixels). The color of the ridges and valleys can be determined by the color of the sweat gland pixels, and the color of the sweat glands and the pixels can be sufficiently detected even if they are part of the input fingerprint image data. .

  Therefore, from the viewpoint of high-speed processing, it is desirable to generate determination image data using only a part of the input fingerprint image data. Since there are individual differences, gender differences, and age differences in finger sizes, it is desirable to generate determination image data from fingerprint image data in the central portion of the fingerprint reader as much as possible. For example, it is possible to generate determination image data from image data having an area of about ¼ of the entire area that occupies the central portion of the image acquisition region of the fingerprint reader. Of course, when the processing speed is not a problem and more accurate determination is necessary, determination image data can be generated by processing all of the input fingerprint image data.

(Reverse rotation detection unit)
The reverse rotation detection unit 26 includes an isolated region detection unit 27, a counting unit 28, and a determination unit 29. First, the detection principle of the reverse rotation detection unit 26 will be described. In the present invention, the light / dark relationship of the input image data is determined by paying attention to the following properties of the sweat glands in the fingerprint.
1) The sweat glands in the fingerprint are present only on the ridges, and are isolated and scattered in large numbers.
2) Since the sweat glands are hollow, their brightness (luminance) in gray scale is closer to that of valleys than ridges.

  That is, the sweat gland exists at least as an isolated region in the image data. Further, since the sweat gland is a depression, it can be seen that the image data binarized to white or black has the same color as the valley pixel. When the sweat gland is detected based on the above conditions, the sweat gland existing region portion can be recognized as the ridge portion, and the light / dark relationship of the input image data can be determined based on the brightness of the sweat gland pixels.

  On the other hand, since sweat glands are present in about 10% of fingerprints, even if the color of the sweat gland pixels is white or black, there are a certain number or more than a certain percentage of isolated regions independent of the surrounding colors. In this case, it is possible to estimate the isolated region as a sweat gland with high accuracy. Note that an isolated region having a color opposite to the sweat gland may be formed in the fingerprint image data due to noise or the like. However, since the number of isolated regions due to noise is much smaller than that of sweat glands, the isolated regions due to noise can be ignored from the viewpoint of detection of sweat glands. By detecting the color of the sweat glands in the binarized determination image data in this way, it is possible to determine the light / dark relationship between the ridges and valleys.

  In the first embodiment of the present invention shown in FIG. 4, in order to detect sweat glands, an isolated region is detected by an isolated region detector 27, and a pixel of an isolated region of a predetermined color of white or black is detected by a counter 28. The number is counted. Based on the counting result, the determination unit 29 determines whether the input image data is not reversed from the predetermined light / dark relationship set by the collation unit.

  This will be specifically described with reference to FIG. FIG. 5 is an example of determination image data after the input fingerprint image data is binarized by the binarization processing unit 25. FIG. 5A illustrates an example of image data in which the ridge 50 of the input fingerprint image data is black (dark: low luminance) and the valley 51 is white (bright: high luminance). FIG. 5B shows an example of image data in which the ridge is white and the valley is black, contrary to FIG. In FIG. 5A, sweat glands 52 and valley pixels are white, and in FIG. 5B, black. In FIGS. 5A and 5B, 53 and 58 also appear as stripe-shaped “isolated regions” having colors opposite to those of the sweat glands 52 and 57. 53 and 58 and FIGS. 5C and 5D will be described later.

  In order to make the description easy to understand, in the following description, an image (ridge line) showing the light / dark relationship in FIG. 5A as a predetermined light / dark relationship in the fingerprint image data from the feature extracting unit 17 and the collating unit 18. The description will be made assuming that 50 is black and the valley line 51 is white).

(Isolated area detection)
The isolated area detection unit 27 detects all isolated areas in the generated binary image data. The detection of the isolated region can be performed using a majority filter composed of a matrix of a predetermined size. Specifically, for example, each pixel of the fingerprint image data is placed at the center of the majority filter, and how many pixels of the central pixel color are in the entire matrix by comparing the color of the central pixel with the color of the other pixels of the filter. It is possible to determine whether or not the central pixel is isolated depending on whether it is present at the ratio of.

  For example, when the resolution is such that the sweat gland appears in almost one pixel, using a 3 × 3 pixel matrix filter and judging by the majority rule, five or more of nine pixels in the matrix are different from the color of the central portion. In this case, the central pixel is determined as a part of the isolated region. The size of the 3 × 3 matrix filter described above is an example, and the size can be arbitrarily determined based on the resolution of the fingerprint image data and the size of the sweat glands. When the resolution is large, a matrix of 5 × 5, 7 × 7, etc. can be used. As the size of the matrix increases, the processing load increases, and when the processing capability is low, it takes a lot of time, and it is possible to select a matrix filter optimal for the use environment. However, at least the size of the sweat gland is one pixel or more, and the matrix filter is desirably a filter sufficiently larger than the sweat gland. In addition, since the size of the sweat gland varies depending on the biological characteristics such as race, it is possible to select the input condition and the matrix filter according to the use environment. Alternatively, the center pixel may be automatically replaced according to the majority filter described above, and the original fingerprint image data and the processed fingerprint image data may be compared to detect a pixel having a color change as an isolated region.

(Judgment of reverse rotation detection)
As described above, the isolated region detection unit 27 detects all isolated regions in the image data. The counting unit 28 counts the total number of pixels in the isolated area composed of pixels of brightness (luminance) of either white (bright) or black (dark) among the detected isolated areas. Now, since image data in which the pixels of the ridges 50 as shown in FIG. 5A are black are required, the sweat gland is white, so the counting unit 28 counts the number N _W of isolated white pixels. Will be described as counting. The determination unit 28 determines that the white isolated area is a sweat gland when the number N _W of pixels of the white isolated area exceeds the number indicated by the predetermined threshold T ₁ . Therefore, when the sweat gland 52 is composed of white pixels as shown in FIG. 5A, it is determined that this condition is satisfied and the input image data shown in FIG. Then, the determination result is output to the brightness inversion processing unit 22. In this case, since the input fingerprint image data is in the light-dark relationship that is required, the inversion control unit does not perform any special processing, and the input image data is input as it is to the feature detection unit 17 (see FIG. 1). In this way, out of the pixels constituting the isolated region, the pixel of lightness (luminance) of one of black and white binary values is counted, and it is determined whether or not the counted pixel is a pixel constituting the sweat gland. Is calculated. Then, it can be seen that the brightness same as the brightness of the sweat gland is a valley line, and the opposite brightness is a ridge.

When the input data is fingerprint image data as shown in FIG. 5B, the sweat glands 57 are composed of black pixels, and the isolated area 58 of white pixels is slight. Therefore, in such fingerprint image data, since the isolated region of the white pixel does not reach a predetermined threshold value T _1, the matching unit 18 determines that there is no light and dark relationship for obtaining the fingerprint image data, the brightness inversion processing unit 22 Tell. The reversal control unit 30 of the lightness reversal processing unit 22 reads the input fingerprint image data, the lightness calculation unit 31 calculates the reversal lightness (luminance) of each pixel and writes it back, and collates the input fingerprint image data. The image data is converted into light and dark fingerprint image data required by the unit 18.

In the description of FIG. 5, when the ridges of the fingerprint image data to determine the black is to count the pixels of white isolated area was described as judged by whether the number exceeds the threshold value T _1. However, by counting the pixels in the isolated area of the black, it is also possible to image data if the number is a predetermined threshold value T ₂ above is configured to determine to be reversed.

  The fingerprint input image determination processing unit 20 can be configured by a CPU and a control program. An example of the processing procedure of the fingerprint input image determination processing unit 20 will be described with reference to the flowchart of FIG. The flowchart in FIG. 6 corresponds to the process in step S103 in FIG. When the input image is input to the determination image generation unit 23 by the process of S102 of FIG. 3, first, a process (block 60) for generating determination image data for determining the light / dark relationship of the input image is executed. . First, density unevenness correction or the like is performed (S201), and when the predetermined correction is completed, the input image after correction is binarized (S202), and binarized determination image data is generated.

  Next, in block 61, image data is determined. First, an isolated area is detected from the binarized image data (S203), and pixels of a predetermined color are counted from the detected isolated area (S204). Based on the number of pixels of a predetermined color constituting the isolated region, it is determined whether the input image data is in the light-dark relationship that is desired (S205). If the contrast of the input image data is opposite (reversed) from the desired contrast (S206; YES), the input image data is inverted (S207). If not reversed (S206; NO), the process proceeds to step S104 (FIG. 3) as it is.

(Second embodiment of reverse rotation detection unit)
In the fingerprint image determination processing unit 20 according to the first embodiment, after the input image determination unit 21 generates a binarized image for determination, the number of pixels in an isolated region that is thought to constitute the sweat gland is counted, and the pixel depending on whether the number exceeds a predetermined threshold value T ₁ or T _2, to determine the brightness relationship between the image data. However, isolated regions appearing in the image data include the following properties.
1) Sweat glands in reversed image (see sweat glands 52 and 57 in FIG. 5)
2) A streak-shaped “island” formed in a valley portion in a normal image (see islands 53 and 58 in FIG. 5).
3) Noise (not shown) formed in valleys in normal images

  In the present embodiment, only 1) is the target to be used for reverse rotation detection, but 2) and 3) are also included in the isolated region. Of these, the number of isolated regions generated by noise in 3) is extremely small compared to the number of sweat glands (about 10% of the total pixels: about 20% of the ridge or valley pixels). Is hardly a problem. The problem is streaky “islands” 53 and 58, which are isolated regions appearing in the valley line of 2).

The streak-like “islands” 53 and 58 appearing on the valley lines 51 and 56 cause the ridges to be crushed by pressing too hard on the glass surface of the fingerprint input area, and the color of the sweat glands near the center of the valley lines. It is thought that stripe-shaped “islands” of the opposite color occur. Such a streak-like “island” may appear greatly depending on how the finger is pressed. Accordingly, the number of pixels of the stripe-like “island” in 2) above increases, and the number of pixels in the isolated region of the color opposite to the sweat gland may exceed the threshold value T ₁ or T ₂ , so that the inversion of the image data There is a risk of misjudgment.

Therefore, in the second reverse rotation detection unit,
1) Line-shaped “islands” 53 and 58 are generated when the ridges 50 and 55 are crushed by strongly pressing a finger.
2) If the ridge is pressed and crushed, the areas of the ridges 50 and 55 should be large and the areas of the valleys 51 and 56 should be small.
This problem was solved by comparing the ratio of the number of pixels constituting the isolated region to the number of pixels of the entire image.

  That is, the color of the stripe-shaped “islands” is the same as the color of the ridges, so when the stripe-shaped “islands” appear, the area of the ridges also increases, and the area of the ridges The ratio of the “island” area does not increase significantly. Therefore, by measuring the ratio of the number of pixels of the isolated region to the number of pixels of the same color as the isolated region in the entire image, the fingerprint image data in the fingerprint image data reversed with the stripe-like “island” in the fingerprint image data not reversed It is possible to distinguish the sweat glands from each other, and the light / dark relationship of the image data can be determined.

  This will be specifically described with reference to the flowchart of FIG. FIG. 7 is a flowchart showing a reverse rotation detection processing procedure according to the second embodiment. FIG. 5A will be described as correct image data. The color of the isolated area to be counted may be either white or black, but in FIG. 7, the description will be made assuming that the black isolated area is counted.

First, the number _{N B} of the pixels constituting the black in the binary image is counted (S301). Next, the number of pixels _NG in the isolated region composed of black is counted (S302), and the ratio N _G / N _B to the whole is calculated (S303). When the ratio N _G / N _B to the whole exceeds a predetermined ratio T _G1 and when the number of _NG exceeds a predetermined number T _G2 , it is determined that the image data is reversed (S304). This condition is satisfied when the sweat gland 57 is composed of black pixels as shown in FIG. 5B, and this condition is satisfied when the stripe-like “island” is black as shown in FIG. 5A. There is no satisfaction.

Here, not only the composition ratio N _G / N _B > T _G1 but also the number of pixels N _G > T _G2 is based on the condition that when the fingerprint image data that has been blurred is input, the isolated area is in the blurred area. This is because there is a possibility that a partial determination error may occur, and N _G / N _B > T _G1 may be satisfied by mistake.

  As described above, after the determination, when the light and dark are reversed (S305; YES), the fingerprint image data is reversed as described above (S306), and when the light is not reversed, the input is performed. Based on the fingerprint image data as it is, processing by the feature extraction unit is performed (S104).

(Another embodiment of fingerprint collation device)
In the above embodiment, the detected fingerprint detects whether the color of the ridge or valley of the insertion data is the color planned by the collation device, and if it is different from the planned color, The case where predetermined feature data is generated by inverting black and white of an input image has been described.

  However, the registered data storage unit stores the feature data extracted based on either the ridges or the valleys in the fingerprint, and the feature data is input from the input fingerprint image data accordingly. By extracting, desired feature data can be acquired from the input fingerprint image data without using a predetermined color as a reference.

  That is, a sweat gland in the input fingerprint image data can be detected, and the color of the pixel constituting the ridge in the fingerprint image data can be determined from the luminance of the pixel constituting the sweat gland. This makes it possible to specify the position of the ridges or valleys in the input fingerprint image data. Therefore, the feature data can be extracted from either the ridge or the valley line only by instructing the feature extraction unit whether the feature data is extracted from the ridge or the valley line.

  This makes it possible to extract desired feature data regardless of the light / dark relationship of the input fingerprint image data, as long as it is determined whether the feature data is extracted by ridges or valleys. It becomes. Further, it is not necessary to reverse the brightness of the input fingerprint image data regardless of the brightness of the input fingerprint image data.

It is a functional block diagram which shows the fingerprint collation apparatus provided with the fingerprint input image process part concerning one Embodiment of this invention. It is a figure which shows the fingerprint image data after performing a thinning process. It is a flowchart which shows an example of the whole process sequence of a fingerprint collation apparatus. It is a functional block diagram of the fingerprint input image determination processing unit according to the first embodiment of the present invention. It is a figure which shows the example of the determination image binarized by the binarization process part, and has shown two examples with which the light-dark relationship of the ridge and the valley line was reversed. 4 is a flowchart illustrating an example of a processing procedure of a fingerprint image determination processing unit 20. It is a flowchart which shows the reverse rotation detection process procedure concerning 2nd Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Fingerprint collation apparatus 11 Operation display part 12 Fingerprint input part 13 Data processing part
DESCRIPTION OF SYMBOLS 14 Registration data memory | storage part 15 Control part 16 Data memory | storage part 17 Feature extraction part 18 Collation part 19 Output part 20 Fingerprint input image determination process part 21 Input image determination part 22 Lightness inversion process part 23 Determination image generation part 24 Correction process part 25 2 Value processing unit 26 Reverse detection unit 27 Isolated region detection unit 28 Count unit 29 Determination unit 30 Inversion control unit 31 Lightness calculation unit 32 Image reading unit 40, 41, 42 End point 43 Branch point 50, 55 Ridge line 51, 56 Valley line 52, 57 Sweat glands 53, 58 Islands appearing in the valley line 60 Processing block example of determination image generation 61 Reverse rotation detection processing block example 62 Other examples of reverse rotation detection processing block

Claims (5)

Fingerprint input means for inputting fingerprint image data representing fingerprint ridge and valley fingerprints;
Determination image generation means for generating inverted image data obtained by binarizing the luminance of each pixel for all or part of the input fingerprint image data;
An isolated region extracting unit for extracting an isolated region in the binary determination image data generated by the determination image generating unit;
Of the pixels constituting the isolated region, the pixels having any one of the two luminance values are counted, and based on the counted number of pixels, whether or not the pixels counted in the isolated region are pixels constituting a sweat gland Sweat gland judging means for judging
When it is determined that the counted pixel is a pixel constituting a sweat gland, the brightness contrast indicating the ridge and valley line in the determination image data is expressed by the one brightness, and the ridge Is a light / dark relationship determination means for determining that the lightness relationship represented by the other luminance,
A fingerprint input image determination processing apparatus comprising: Furthermore, a light / dark relationship storage means for storing the predetermined light / dark relationship;
Reverse detection means for determining whether the light / dark relationship determined by the light / dark relationship determining means is the stored light / dark relationship or the reverse light / dark relationship;
An inversion processing unit that corrects the luminance of each pixel of the fingerprint image data or the determination image data so as to match the stored light / dark relationship when it is determined that the reverse light / dark relationship is present;
The fingerprint input image determination processing device according to claim 1.   The isolated region extracting unit scans each pixel of the determination image data with a matrix filter having a predetermined size, and when there are a predetermined number or more of pixels having different luminances in the peripheral region of the determination target pixel, The fingerprint input image determination processing device according to claim 1, wherein the pixel is determined to be a pixel constituting the isolated region.   The sweat gland determination means counts one of the binary luminance pixels among the pixels constituting the isolated region, and counts the one luminance pixel among the pixels in the determination image data, The pixel counted in the isolated region is determined to be a pixel constituting a sweat gland when a ratio between the number of pixels counted in the determination image and the number of pixels counted in the isolated region is equal to or greater than a predetermined threshold. The fingerprint input image determination processing device according to any one of claims 1 to 3. Fingerprint reading means for reading a fingerprint from a placed finger and generating fingerprint image data representing a fingerprint of a fingerprint ridge and valley;
Determination image generation means for generating determination image data obtained by binarizing the luminance of each pixel for all or part of the input fingerprint image data;
An isolated region extracting unit for extracting an isolated region in the binary determination image data generated by the determination image generating unit;
Of the pixels constituting the isolated region, the pixels having any one of the two luminance values are counted, and based on the counted number of pixels, whether or not the pixels counted in the isolated region are pixels constituting a sweat gland Sweat gland judging means for judging
When it is determined that the counted pixel is a pixel constituting a sweat gland, the brightness-darkness relationship indicating the ridge and the valley in the determination image data is represented by the one luminance, and the ridge Is a light / dark relationship determination means for determining that the light / dark relationship is represented by the other luminance,
A light-dark relationship storage means for storing the predetermined light-dark relationship;
Reverse detection means for determining whether the light / dark relationship determined by the light / dark relationship determining means is the stored light / dark relationship or the reverse light / dark relationship;
An inversion processing unit that corrects the luminance of each pixel of the fingerprint image data or the determination image data so as to match the stored light-dark relationship when it is determined that the reverse light-dark relationship is established;
A feature extracting means for identifying a pattern of ridges and valleys in the fingerprint image data based on the predetermined light-dark relationship and extracting fingerprint feature information;
Registered fingerprint storage means for storing pre-registered fingerprint feature information;
Fingerprint collation by calculating the similarity between the feature information extracted from the input fingerprint image data and the feature information stored in the registered fingerprint storage means, and comparing the similarity with a predetermined threshold value Matching means;
A fingerprint collation apparatus comprising:

Images