JP4470263B2 - Fingerprint verification device and fingerprint verification method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a fingerprint collation apparatus and a fingerprint collation method for identifying an individual by collating a fingerprint.
[0002]
[Prior art]
In recent years, with the introduction of computers into a wide range of social systems, management of the system and ensuring safety have become problems. For example, it is generally performed that a user directly inputs a personal identification number or identifies each individual using a magnetic card or the like on which the personal identification number is recorded. Also, in everyday life, for example, a person is identified by using a key for managing a house or a car. However, the security code may be forgotten, and the magnetic card or key may be lost or forged, which is not sufficient for safety.
[0003]
Therefore, instead of these methods, a personal verification technique using biometric information has been developed. Among them, personal verification by fingerprint is an effective technique that can identify the person himself / herself reliably from the characteristic of the fingerprint that is unique to each individual and universal throughout the lifetime.
[0004]
Hereinafter, an example of a conventional fingerprint collation apparatus will be described.
FIG. 8 is a diagram showing the flow of processing of a conventional fingerprint collation apparatus, where (a) shows a fingerprint registration and (b) shows a fingerprint collation.
[0005]
When registering a fingerprint, as shown in FIG. 8A, a fingertip is first placed on the imaging surface and brought into close contact, and the fingerprint is formed as a single image by an optical camera using a CCD (Charge Coupled Device) or a semiconductor sensor. , And the fingerprint irregularities are converted into analog signals (S801). A specific range of the analog image signal is converted into a digital image signal, and image data binarized to 256 dots in the x direction and 128 dots in the y direction is created (S802).
[0006]
From this binarized data, three squares of 64 dots in the x direction and 64 dots in the y direction are extracted, and within each square, a straight line of 64 dots in the x direction and a straight line of 64 dots in the y direction, for example, 8 Eight dots are taken at a predetermined interval such as an equal interval of dots (S803). A total of 48 line data extracted in this way are extracted from each of the three squares, each line data that is 64 dot strings is recorded as a hexadecimal number string, and stored as a registered template (S804).
[0007]
In the case of fingerprint verification, as shown in FIG. 8B, the fingerprint is first read as an analog image (S811) and converted into binary data (S812), as in the case of registration, and then the collator's The straight line data of the person is read from the registered template using the identification number as a key (S813).
[0008]
Next, for all the straight line data, a search is made from the entire binarized data screen for a place where the arrangement of 64 dots matches (S814). At this time, considering that the binarized data may be slightly different even in the fingerprint image of the same person due to the difference in the imaging state of the fingerprint, for example, the number of matching dots out of 48 dots exceeded 48 At this time, it is regarded as the same straight line data and is extracted as a matching straight line. At this time, the top coordinates (x, y) of the extracted matching straight line are also read simultaneously.
[0009]
Since the straight line data is extracted only in the x and y directions, if the fingerprint image read at the time of verification is rotated at a different angle from the image at the time of registration, it matches. A straight line cannot be extracted. Therefore, if the coincidence line cannot be extracted in S814, the read image is rotated in both directions by a certain angle, and the coincidence line is extracted each time, and all the line data or a certain number of line data or more are extracted. When the coincidence curve can be extracted, the process proceeds to the next process. However, in FIG. 8B, the process of rotating such an image and extracting a matching straight line is omitted.
[0010]
When a coincidence straight line is extracted for all 48 straight line data, or for a certain number of straight line data, the relative position of the leading coordinates of each extracted coincidence curve and each registered straight line data Are compared with each other (S815). That is, when the registered straight line data is extracted at an 8-bit interval, it is compared whether or not the interval between adjacent matching curves is an 8-bit interval. Such comparison is performed for each of the 48 matching curves, and the number that matches the relative position of the straight line data is calculated as a score value (S816).
[0011]
Alternatively, the number of coincidence of relative positions is calculated for each straight line data of 8 squares each in the x-direction and y-direction at the time of registration. 2 points, 3 points for 7 points, 4 points for 8 points, 0 points for 4 points or less, and the number of coincidences is divided into appropriate stages and assigned points for each stage. Such a score assignment may be performed for each of the x and y directions of each square, and the score may be summed to calculate a scale with a total of 24 points as a limit, and this may be used as the score value.
[0012]
When the score value is calculated, it is determined that the read fingerprints belong to the same person when this value is equal to or greater than a predetermined numerical value according to a collation level set in advance (S817). At this time, if the limit is 24 points, the score value range is divided into appropriate stages such as 0 points to 4 points, level 1 to 5 points to 9 points, and level 2. Each stage is set as a collation level, and collation may be determined for each stage.
[0013]
[Problems to be solved by the invention]
By the way, in order to increase the accuracy of fingerprint collation, it is necessary to register as good a template as possible at the time of registration. In other words, when performing fingerprint collation as described above, the shape of the ridges and valleys of the fingerprint in the binarized image read at the time of registration is read more clearly and in a wider range. is important.
[0014]
In this way, if straight line data can be extracted and registered from a good template in which a real fingerprint pattern is read more accurately, the score value calculated at the time of fingerprint matching will be higher. Thus, the probability of performing an incorrect collation determination that it matches the fingerprint of another person is reduced.
[0015]
However, in the case of a device that reads fingerprints optically or with a semiconductor sensor or the like as described above, the thickness of the fingerprint ridges differs depending on the wetness of the fingertips or changes in humidity depending on the season. There are variations in the image quality of the fingerprint image to be read, such as a partially missing portion, and the registered straight line data becomes inaccurate. For this reason, for a registrant who has a rough fingerprint image read at the time of registration, there is a high possibility that fingerprint matching will not be correctly determined even if the fingerprint of the same person is read at the time of collation.
[0016]
Conventionally, when fingerprints of a plurality of people are registered and the quality of data registered by the registrant varies, a threshold value (matching level) for determining a fingerprint match is set as the registered data. It was common to set it uniquely for registrants with poor quality. However, in this case, since fingerprint verification is determined at a low verification level for both people with good and bad registration data quality, the probability of verification error will be high for all registrants.
[0017]
The present invention has been made in view of the above-described points. A fingerprint in which a fingerprint collation level is increased and collation judgment errors are reduced regardless of variations in image quality of each registrant of a fingerprint image read at the time of registration. A verification device is provided.
[0018]
Another object of the present invention is to provide a fingerprint collation method in which a fingerprint collation level is increased and collation judgment errors are reduced regardless of variations in image quality of each registrant of a fingerprint image read at the time of registration. It is.
[0019]
[Means for Solving the Problems]
In order to solve the above-described problems, the present invention captures a fingerprint image in a fingerprint collation apparatus that collates fingerprints and identifies individuals. Read by An imaging unit, a binarizing unit for generating binarized data obtained by binarizing the fingerprint image, and a dot width from the binarized data First Straight line data extracting means for extracting a plurality of straight line data having a specific number of dot rows at a predetermined interval; and the binarized data And already extracted and registered The dot arrangement matches for each of the straight line data The straight line data already extracted and registered Coincident line Explore as A matching line search means; Explored A relative position of each of the coincident lines; Already extracted and registered Of the straight line data So Relative position comparison means for comparing each of the relative positions, score value calculation means for calculating the total number of the matched relative positions as a score value, and immediately after registration of the straight line data Test collation realized by executing a series of processes of the binarization means, the coincidence straight line search means, and the relative position comparison means for a fingerprint image read by the imaging means as a processing target is performed a plurality of times. It was repeated and obtained with the trial matching of each time Average value of the score values Calculate Collation level setting means for outputting as a collation level, storage means for registering and storing the straight line data and the collation level for each registrant, and fingerprint collation By the imaging means Read As processing target Determination means for determining that the fingerprint is that of the same person when the score value calculated from the fingerprint image is equal to or higher than the collation level of each registrant. In the case of the trial collation, the coincidence straight line search means is equal to or less than the first specific number of the binarized data and the dot rows constituting the straight line data already extracted and registered. And the already extracted and registered straight line data in which the dot arrangement for the second specific number of dots or more matches is searched as the matching straight line, and when the matching straight line cannot be searched, the processing object The fingerprint image is rotated by a certain angle, and the search for the matching straight line is repeated each time, and a search for a number of the matching straight lines or more is performed. A fingerprint collation device is provided.
[0020]
In such a fingerprint verification device, registration is performed by setting a verification level for each fingerprint registrant at the time of fingerprint registration, so even if the image quality of the fingerprint image varies for each registrant depending on the reading state at the time of registration, fingerprint verification Can improve the accuracy.
[0021]
According to the present invention, a fingerprint image is captured in a fingerprint collation method for identifying an individual by collating a fingerprint. To read The binarized data obtained by binarizing the fingerprint image is created, and the dot width is 1 dot from the binarized data. First Extract a plurality of straight line data with a specific number of dot rows at a predetermined interval, The binarized data and already extracted and registered The dot arrangement matches for each of the straight line data The straight line data already extracted and registered Coincident line Explore as , Explored A relative position of each of the coincident lines; Already extracted and registered The relative position of each of the straight line data is compared, and the sum of the number of matching relative positions is used as a score value. For the fingerprint image read immediately after registration of the straight line data, the trial collation is repeated a plurality of times, and the average value of the score values obtained in each trial collation is calculated and output as a collation level And the straight line data and Register and store the verification level for each registrant, Fingerprint In the case of Calculated from the fingerprint image as the read processing target The score value is , Climb The fingerprint is determined to be that of the same person when the collation level is equal to or higher than each recording person In the case of the trial collation, the binarized data and the second specific number that are equal to or less than the first specific number and the second specific number of the dot rows that constitute the straight line data that has already been extracted and registered. The previously extracted and registered straight line data in which the dot arrangement for the above dots matches is searched as the matching straight line, and when the matching straight line cannot be searched, the fingerprint image to be processed is set at a certain angle. Rotate each time and repeat the search for the matching line each time. A fingerprint collation method is provided.
[0022]
In such a fingerprint collation method, trial collation is performed at the time of fingerprint registration, and the collation level is set for each fingerprint registrant. Therefore, even if the image quality of the fingerprint image varies for each registrant depending on the reading state at the time of registration. The accuracy of fingerprint verification can be improved.
[0023]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 shows a principle diagram of a fingerprint collation apparatus according to the present invention. The fingerprint collation device 1 reads and registers a fingerprint, and collates the read fingerprint with the registered fingerprint to determine whether the fingerprint is of the same person.
[0024]
The imaging unit 2 captures a fingerprint and reads it as an analog image signal, and the binarizing unit 3 binarizes this image data and converts it into a digital image signal.
At the time of registration, the fingerprint is imaged a plurality of times, and a predetermined number of dots are arranged with a width of 1 dot in the vertical direction and the horizontal direction of the image from the binarized data of the first fingerprint image by the straight line data extraction means 4. Are extracted at predetermined intervals. This straight line data is stored in the storage means 9.
[0025]
Thereafter, the fingerprint is picked up again, and this straight line data and trial matching are performed a plurality of times. Similar to the reading of the first fingerprint, the fingerprint image read by the imaging unit 2 is converted into binarized data by the binarizing unit 3. The coincidence straight line search means 5 reads the straight line data registered from the storage means 9 and searches for the coincidence straight line in which the arrangement of dots coincides from the entire binarized data image for all the straight line data.
[0026]
The relative position comparison means 6 compares the relative positions of the extracted matching straight lines with the relative positions of the registered straight line data to check whether they match. The score value calculation means 7 uses the total number of the matching relative positions as a score value, and the collation level setting means 8 averages the score values calculated by a plurality of trial collations, and the same person at the time of collation This average value is set to the collation level indicating the threshold value to be determined. This collation level is stored for each registrant by the storage means 9 together with the straight line data.
[0027]
On the other hand, at the time of fingerprint collation, a score value is calculated in the same manner as trial collation at the time of registration. That is, a fingerprint is imaged by the imaging unit 2 and converted into binarized data by the binarizing unit 3. The coincidence line search means 5 reads the straight line data registered from the storage means 9 and finds a coincidence line that matches the dot arrangement of each line data from the entire binarized data image. The score value calculation means 7 compares the relative positions of the extracted matching straight lines with the relative positions of the registered straight line data, and the score value calculation means 7 counts the number of relative positions that match. Are summed and set to the score value.
[0028]
When the calculated score value satisfies the collation level registered in the storage unit 9, the determination unit 10 determines that the read fingerprint is that of the same person as the registered fingerprint.
[0029]
In the fingerprint collation apparatus 1, the higher the collation level is, the higher the accuracy of fingerprint collation is. Conversely, the lower the collation level is, the higher the possibility that an erroneous determination is made. Sometimes trial collation is performed a plurality of times, and an average of score values calculated at this time is set for each registrant. For this reason, a person who tends to read a fingerprint image relatively well has a higher collation level. When reading a fingerprint by the image pickup means 2, due to differences in the wetness of the fingertip, how to apply force, etc. Even if the image quality of fingerprint images varies among registrants, the accuracy of fingerprint collation can be improved as a whole.
[0030]
Next, FIG. 2 shows a configuration example of the fingerprint collation device of the present invention.
The fingerprint image is picked up by an optical camera using the CCD 11. When the CPU 13 receives a fingerprint registration command from the host computer 20 through the RS232C interface 14, the CCD 11 images the fingerprint, and the read analog image signal is stored in the SRAM 15 through an ASIC (Application Specific Integrated Circuit) 12. This data is converted into binarized data by the ASIC 12, and straight line data is extracted by the processing of the CPU 13. The created straight line data is stored in the flash memory 16 and registered.
[0031]
Thereafter, the fingerprint is imaged again by the CCD 11 for trial collation, converted into binarized data by the ASIC 12, and collated with the straight line data stored in the flash memory 16 by the collation algorithm by the processing of the CPU 13 to obtain a score value. Is calculated, the collation level is set, and stored in the flash memory 16.
[0032]
When a fingerprint collation command is sent from the host computer 20, the fingerprint is read by the CCD 11. Similar to the trial collation at the time of fingerprint registration, the CPU 13 performs collation with the straight line data stored in the flash memory 16 by the collation algorithm, and the collation determination is performed based on the collation level registered from the calculated score value. Is done. The collation result is output from the CPU 13 to the host computer 20 via the RS232C interface 14.
[0033]
In some cases, a semiconductor sensor is used instead of the CCD 11 for imaging a fingerprint.
Next, an example of the fingerprint verification process will be described in detail.
[0034]
FIG. 3 shows the flow of fingerprint registration processing. FIG. 4 shows a method for creating registration data in detail. 4A shows an image of binary data, and FIG. 4B shows a method of extracting straight line data. Further, FIG. 5 shows a method for collating straight line data in detail. FIG. 5A shows a matching data search method, and FIG. 5B shows a relative position comparison method of matching data. The fingerprint registration process will be described with reference to FIGS. 3, 4 and 5.
[0035]
When the fingerprint registration operation is started, a fingerprint is imaged (S301), and the read image is converted into binarized data 41 (S302).
In S302, as shown in FIG. 4A, the specific range of the read analog fingerprint image is converted into binary data 41 of 256 dots in the x direction and 128 dots in the y direction. From this binarized image data, three 64-dot square areas are extracted, and as shown in FIG. 4B, in each square 41a, 41b, 41c, 64-dot linear data 42 is obtained. Eight lines are extracted at specific intervals in the x and y directions (S303). In FIG. 4B, the straight line data 42 is extracted at intervals of 8 dots. Each of the total 48 line data 42 is represented as a hexadecimal number string by arranging 64 dots, and is registered as a registration template together with the leading coordinates of each line data 42 (S304).
[0036]
Next, a fingerprint is imaged again for trial verification (S305), and binarized data 51 is created (S306). Then, the registered straight line data 42 is read (S307), and as shown in FIG. 5A, for each of the 48 straight line data 42, a matching straight line 52 in which the dot arrangement matches from the entire image of the binarized data 51. Is found (S308). Here, in actuality, considering that there are some differences in the binarized data 51 even in the fingerprint image of the same person due to the difference in the imaging state of the fingerprint, for example, there is a matching dot among 64 dots. When the number exceeds 48, it is regarded as the same straight line data 42 and is extracted as a matching straight line 52. At this time, the leading coordinates (x, y) of the extracted matching straight line 52 are also read at the same time.
[0037]
Since the straight line data 42 is data extracted only in the x direction and the y direction, if the fingerprint image read at the time of collation is rotated and read at a different angle from the image at the time of registration, The coincident line 52 cannot be extracted. Therefore, if the coincidence line 52 cannot be extracted in S308, the read image is rotated in both directions by a certain angle, and the coincidence line 52 is extracted each time. When the coincidence straight line 52 is extracted from the straight line data 42, the process proceeds to the next process. However, in FIG. 3, the process of rotating such an image and extracting the coincident line 52 is omitted.
[0038]
After the coincidence straight line 52 is extracted, as shown in FIG. 5B, the relative positions of the coordinates of the head part of the eight coincidence straight lines 52 in the x-axis and y-axis directions are registered for each square 51a. The relative position of the straight line data 42 is compared (S309). Then, for all square ranges, the sum of the numbers of matching relative positions is set as the score value (S310).
[0039]
At this time, for each direction of each square range, the range in which the score value is taken may be divided into appropriate stages, and the score value may be set at each stage. Here, for example, with respect to one direction of each square range, 0 points when the number of relative positions of the eight coincident lines 52 is 4 or less, 1 point when 5 are, 2 points when 6 are, 7 Scores are calculated in the same way in three squares in the x and y directions, with 3 points for books and 4 for 8 points, and a score value scale with a total of 24 points is created.
[0040]
Thereafter, the trial collation process from S305 to S310 is repeated a plurality of times, and the score value is calculated in each process (S311). After completing the predetermined number of trial collations, the score values are averaged (S312), and the average value is set as the collation level (S313).
[0041]
Further, when setting the collation level (S313), the range of the score value having the highest score of 24 may be divided into appropriate stages, and the collation level may be set for each stage. As an example, FIG. 6 is a graph showing the relationship between the calculated score value and the number of trial collations when a plurality of trial collations are performed.
[0042]
In FIG. 6, the number of each score value calculated for each time by performing trial collation for a plurality of registrants A and B is shown, and the average score value of registrant A is 9 points. It is assumed that the average score value of becomes 18 points. Here, the range of score values up to 24 points is divided into 5 levels: 0 points, 4 points, 5 points, 9 points, 10 points, 14 points, 15 points, 19 points, and 20 points, 24 points. Each stage is assigned with collation levels 1, 2, 3, 4, and 5. That is, in FIG. 6, a registrant A with a score value of 9 points is set to collation level 2, and a registrant B with a score value of 18 points is set to collation level 4.
[0043]
The collation level set for each registrant is registered together with the straight line data 42 (S314). In this way, a plurality of trial matching processes are performed at the time of registration to determine the tendency of the image quality of the registrant's fingerprint image to be read, and matching is performed for each registrant accordingly. The level is set.
[0044]
Next, FIG. 7 shows a flow of fingerprint collation processing.
The fingerprint of the collator is captured and the fingerprint image is read (S701), and binarized data is created (S702). Then, using the collator identification number as a key, straight line data is read from the registered template data (S703). Similar to the trial collation, each straight line data and dot data are extracted from the entire binarized data image. A matching straight line with a matching line is searched for (S704). At this time, as in the case of the trial collation at the time of fingerprint registration, when the coincidence straight line is not extracted, the image of the binarized data is rotated by a certain angle in both directions, and the coincidence data is searched each time. When coincident lines are extracted for straight line data or a certain number of straight line data, the process proceeds to the next process. However, in FIG. 7, the process of rotating such an image is omitted.
[0045]
Next, the relative position of the matching straight line is compared with the relative position of the registered straight line data from the leading coordinates of each of the extracted matching straight lines (S705), and a score value is calculated from the number of matching relative positions. (S706). The collation level registered here is read (S707), and when the calculated score value satisfies the collation level value, it is determined that the fingerprint belongs to the same person (S708).
[0046]
As in the case of trial collation at the time of registration, it is also possible to divide the range taken by the number of matches of the relative positions of the coincidence lines at the time of collation into a plurality of stages and assign score values to each stage. In addition, such a stepwise assignment method of score values and collation levels must be the same setting at the time of registration and collation.
[0047]
【The invention's effect】
As described above, in the fingerprint collation device according to the present invention, since the collation level is set and registered for each fingerprint registrant at the time of fingerprint registration, the fingerprint image quality varies depending on the registrant depending on the reading state at the time of registration. Even if it occurs, the accuracy of fingerprint verification can be improved.
[0048]
In the fingerprint collation method of the present invention, trial collation is performed at the time of fingerprint registration, and the collation level is set for each fingerprint registrant. Therefore, the image quality of the fingerprint image varies depending on the registrant depending on the reading state at the time of registration. However, the accuracy of fingerprint verification can be improved.
[Brief description of the drawings]
FIG. 1 is a diagram showing a principle diagram of a fingerprint collation device of the present invention.
FIG. 2 is a diagram illustrating a configuration example of a fingerprint collation device according to the present invention.
FIG. 3 is a flowchart showing a process flow of fingerprint registration.
4A and 4B are diagrams illustrating a registration data creation method, where FIG. 4A illustrates an image of binary data, and FIG. 4B illustrates a method of extracting straight line data.
5A and 5B are diagrams showing a method for collating straight line data, where FIG. 5A shows a method for searching for matching data, and FIG. 5B shows a method for comparing the relative positions of the matching data.
FIG. 6 is a graph showing a relationship between a calculated score value and the number of trial collations when a plurality of trial collations are performed.
FIG. 7 is a flowchart showing a flow of fingerprint matching processing.
FIGS. 8A and 8B are flowcharts showing a process flow of a conventional fingerprint collation apparatus, where FIG. 8A shows a fingerprint registration time, and FIG. 8B shows a fingerprint collation time.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Fingerprint collation apparatus, 2 ... Imaging means, 3 ... Binarization means, 4 ... Line data extraction means, 5 ... Matching line search means, 6 ... Relative position comparison means, 7 ... Score value Calculation means 8... Collation level setting means 9... Storage means 10... Determination means 11... CCD, 12... ASIC, 13. ... Flash memory, 20 ... Host computer

Claims (4)

In a fingerprint collation device that identifies individuals by collating fingerprints,
Imaging means for reading by taking a fingerprint image;
Binarization means for creating binarized data obtained by binarizing the fingerprint image;
Straight line data extracting means for extracting a plurality of straight line data that form a first specific number of dot rows with a dot width from the binarized data at a predetermined interval;
Matching line search means for searching the already-extracted and registered line data as matching lines for the binarized data and the already-extracted and registered line data in which the dot arrangement matches,
And searched respective relative positions of the matching linear already been extracted, the relative position comparing means for comparing the relative positions of their respective of the line data that is registered,
A score value calculating means having the total number of the matched relative positions as a score value;
Realized by executing a series of processes of the binarization means, the coincident line search means, and the relative position comparison means with the fingerprint image read by the imaging means immediately after registration of the straight line data as a processing target. The collation level setting means for repeating the trial collation performed a plurality of times , calculating an average value of the score values obtained by the trial collation of each time, and outputting as a collation level;
Storage means for registering and storing the straight line data and the collation level for each registrant;
When the score value calculated from the fingerprint image as a processing target read by the imaging unit at the time of fingerprint collation is equal to or higher than the collation level of each registrant, it is determined that the fingerprints belong to the same person A determination means;
I have a,
In the case of the trial matching, the matching line search means
The binarized data and the dot arrangement of dots that are not more than the first specific number and that are not less than the second specific number in the dot rows constituting the straight line data that has already been extracted and registered match. Search the straight line data already extracted and registered as the matching straight line,
When the matching straight line cannot be searched, the fingerprint image to be processed is rotated by a certain angle, and the search for the matching straight line is repeated each time to search for a certain number or more of the matching straight lines. Fingerprint verification device.   2. The fingerprint collation apparatus according to claim 1, wherein the score value calculation unit divides a value taken by the number of coincidences of the relative positions into a plurality of appropriate stages and sets the score value for each of the stages. .   2. The fingerprint collation apparatus according to claim 1, wherein the collation level setting means divides a value taken by the average value of the score values into a plurality of appropriate stages and sets the collation level for each of the stages. . In a fingerprint collation method for identifying an individual by collating a fingerprint,
Read by taking a fingerprint image,
Create binarized data by binarizing the fingerprint image,
Extracting a plurality of straight line data at a predetermined interval from the binarized data and forming a first specific number of dot rows with a dot width;
Search the already-extracted and registered straight line data in which the dot arrangement matches for each of the binarized data and the already extracted and registered straight line data as a matching straight line,
Comparing each relative position of the searched straight line with each relative position of the already extracted and registered straight line data;
The sum of the number of matching relative positions is the score value,
For the fingerprint image read immediately after registration of the straight line data, the trial collation is repeated a plurality of times, and the average value of the score values obtained in each trial collation is calculated and output as a collation level And
Register and store the straight line data and the matching level for each registrant,
During the fingerprint collation, said score value calculated from the fingerprint image to be processed is, the time of each registration Rokusha of the above matching level fingerprint is determined to be the same person,
In the case of the trial collation, the binarized data and the dot lines constituting the linear data already extracted and registered are not more than the first specified number and not less than the second specified number. The straight line data that has already been extracted and registered that matches the dot arrangement for the dots is searched as the matching straight line. When the matching straight line cannot be searched, the fingerprint image to be processed is rotated by a certain angle. In this case, the fingerprint matching method is characterized in that the search for the matching straight line is repeated each time to search for a certain number or more of the matching straight lines .

Images