JPH07104942B2 - Fingerprint matching device - Google Patents

Description

The present invention relates to a fingerprint collation device adapted to be applied to, for example, a keyless entry system or an entrance management system for a special room such as a computer room.

[Prior Art] For security management in the above-mentioned entry system, entry management system, etc., it is considered to use different fingerprint collation systems for each individual.

As a fingerprint collation device used for such a purpose,
The fingerprint pattern of each person who is registered in advance is stored in the storage device as a registered fingerprint in advance, and the registered fingerprint pattern is compared with the input fingerprint pattern to determine whether they match or not. Therefore, various types of devices have been conventionally considered.

As such a fingerprint collation device, for example, Japanese Patent Laid-Open No. 56-246
As shown in Japanese Patent Laid-Open No. 75-75, it is considered to extract the end points or branch points (hereinafter referred to as minutiae) of ridges forming a fingerprint pattern as feature points of a fingerprint. And the positional relationship between the minutiae of the registered fingerprints,
The minutiae position, direction, and relative positional relationship of the input fingerprints are compared, or the number of ridges existing between the minutiae is used to compare the registered fingerprint and the input fingerprint,
The registration fingerprint and the input fingerprint are collated to determine whether they are the same fingerprint.

Further, as shown in Japanese Patent Laid-Open No. Sho 58-176781 or PRU87-31 of the Institute of Communication Engineers of Japan, a pattern matching method has been proposed which focuses on a partial pattern in the vicinity of a minutia to perform matching. There is.

However, in these fingerprint collation methods, since minutia is extracted and collation is performed based on this minutia, for example, the surface of the finger to be input is rough or the fingerprint is If the ridge forming the ridge has a low convex portion, the ridge is inevitably broken or crushed. Therefore, in such a state, it becomes difficult to stably extract the feature points, and the collation recognition rate is lowered.

[Problems to be Solved by the Invention] The present invention has been made in view of the above points, and for example, a fingerprint ridge is often cut or crushed without extracting minutiae as described above. Provide a fingerprint collation device that enables accurate collation even for low-quality input fingerprint images, further improves the recognition rate, reliably reduces the misrecognition rate, and ensures sufficient reliability. Is what you are trying to do.

[Means for Solving the Problem] That is, in the fingerprint collating apparatus according to the present invention, first, the first window area is set so as to correspond to the main part of the fingerprint image of the specific person to be recognized. Then this first
The registered fingerprint image corresponding to the window area of is stored in advance. Then, the input fingerprint image corresponding to the first window area of the input fingerprint image input after this registration is collated with the registered fingerprint image. At the time of this collation, the fingerprint input means first identifies the fingerprint image. A fingerprint image of a finger is input, and the fingerprint images corresponding to the first window region and the second window region set outside the fingerprint image are respectively cut out from the fingerprint image, and the registered fingerprint corresponding to the first window region is cut out. The number of pixels in the mismatched state between the image and the input fingerprint image corresponding to the first window region is measured. Then, when the number of non-matching pixels is smaller than the first set value, it is determined to be the collation matching, and when the number of non-matching pixels is larger than the second setting value that is larger than the first setting value, it is determined to be the collation non-matching. When the number of mismatched pixels is between the first setting value and the second setting value, the same collation is performed in the second window area set outside the first window area. ,
When the number of non-matching pixels is smaller than the third set value, it is determined to be a collation matching, and when the number of non-matching pixels is larger than the third setting value, it is determined to be a collation mismatch.

[Operation] In the fingerprint collation device configured as described above, since a minutia is not used for collating a fingerprint, collation of a low-quality input fingerprint image is also effectively performed. . Then, if the number of unmatched pixels is smaller than the first set value by the pattern comparison in the first window area, it is directly determined as the matching match,
When the number of mismatched pixels is larger than the second set value,
As it is, it is determined that the collation does not match. Therefore, the collation reliability in this state can be effectively improved. When the number of mismatched pixels is between the first setting value and the second setting value in the collation in the first window area, and it is difficult to make a determination, the collation in the second window in the vicinity is performed. Since the matching match and the matching mismatch are determined by the number of mismatched pixels, the recognition rate is improved and the false recognition rate is surely reduced.
In particular, at the time of fingerprint input at the time of collation, the fingerprint image patterns corresponding to the first and second windows are simultaneously extracted from the fingerprint image input to the input means, and the first of them is accompanied by the collation action. Of the extracted fingerprint image patterns respectively corresponding to the first and second windows are counted, the number of mismatched pixels is counted, and the verification corresponding to each of the first and second windows is continuously performed by one fingerprint input operation. As a result, the collation with the registered fingerprint image is performed under the same condition with the same input image also in the re-collation with the second window. Therefore, the fingerprint collation accuracy can be effectively improved, and the "put your finger on the collation device" action that makes people feel the most annoying can be done only once, and the keyless entry system can be effectively used. It will be realized.

[Embodiment of the Invention] An embodiment of the present invention will be described below with reference to the drawings.
FIG. 1 shows the configuration of a fingerprint collation device. An input fingerprint to be collated is taken in from the image sensor 11 as an image of a fingerprint pattern.

The image signal of the fingerprint pattern captured by the image sensor 11 is supplied to the image input circuit 12, and the image output signal corresponding to the fingerprint image from the input circuit 12 is an image through the image bus 13. It is supplied to the memory 14 and stored and held. Then, the fingerprint data of the fingerprint image pattern stored in the image memory 14 is supplied to the image processing circuit 15 through the image bus 13 so that the image processing circuit 15 can appropriately execute the collation process and the like.

The image input circuit 12, the image memory 14, the image processing circuit 15, and the like are connected to the microcomputer 17 via the control bus 16.
It is designed to be connected to the image input circuit 12
The image processing circuit 15 is controlled by the microcomputer 17, and the result processed by the image processing circuit 15 is output.

A floppy disk 19 that constitutes an external storage device is further connected to the control bus 16 via a disk driver 18. The floppy disk 19 stores input fingerprint data, processing results, or a control program. In particular, registered fingerprint data for verification is stored.

FIG. 2 shows a detailed configuration of the fingerprint collation device configured as described above, particularly the image processing circuit 15 and the image memory 14 section. The image memory 14 stores the registered fingerprint image in the first section. The image storage unit 21, the second image storage unit 22 that stores the fingerprint image data input at the time of matching, and the third image storage unit 23 that further stores the processing result.

Here, in the first image storage unit 21, fingerprint image data of the first window area a, which is a main area corresponding to the central area of the fingertip specified first, is registered and stored. This corresponds to the registered image area in the conventional fingerprint collation device.

Then, in the first image storage unit 21, the first image
The fingerprint image data corresponding to the second window areas b and c located on both sides of the window area a are also registered and stored at the same time.

Then, at the time of matching, the fingerprint image data of the area corresponding to the first window area a is cut out from the input fingerprint image stored in the second image storage unit 22 according to a command from the cutting circuit 24 and output. , And the registered fingerprint data of the first window area a from the first image storage unit 21 are supplied to the matching circuit 25. Then, the registered fingerprint image corresponding to the first window area and the input fingerprint image cut out corresponding to the same area are matched. Then, the result of this matching is sent to the third image storage unit 23, and at the same time, the pixel number measuring circuit 26 counts and measures the number of unmatched pixels between the registered fingerprint image and the input fingerprint image. There is.

That is, in the fingerprint matching device, in order to enable matching in a low-quality fingerprint image with many cuts and crushes, the minutiae is not extracted as in the conventional case, and registration is performed in the matching circuit 25. The fingerprint image and the input verification fingerprint image are matched.

When such fingerprint pattern matching is performed, the larger the area of the image to be matched, the lower the probability of erroneous recognition, which is the rate of erroneous matching determination. However, when inputting a fingerprint image, it is difficult to accurately match the position of the input fingerprint image with the registered fingerprint image, and there is distortion in the input fingerprint image because the fingertip to be matched is soft. become. Therefore, if the matching area is simply enlarged, the rate at which the registered fingerprint image and the input fingerprint image are the same can be collated, that is, the recognition rate decreases.

Therefore, in the fingerprint collation device shown in this embodiment, first, the first window area a, which is a part of the fingerprint image, is formed.
The fingerprint image corresponding to is registered, and the registered fingerprint image pattern in this area is matched with the fingerprint image input at the time of matching by pattern matching. Then, a pattern that best matches the pattern of the registered fingerprint image is searched in the input fingerprint image pattern, and the searched fingerprint pattern and the registered fingerprint image pattern are matched to determine the number of pixels determined to be inconsistent. measure. Then, the number of mismatched pixels is compared with a reference value set, and it is determined whether the input fingerprint matches the registered fingerprint.

The reference value serves as a determination threshold value, and the recognition rate and the erroneous recognition rate change depending on the setting state of the determination threshold value.
FIG. 3 shows the states of the recognition rate and the false recognition rate.
When the determination threshold is N1 and the same fingerprint is determined when the number of mismatched pixels is N1 or less, the false recognition rate indicated by the curve B can be completely eliminated. However, in such a state, the reliability is obtained when it is determined that the collation matches, but on the other hand, the recognition rate indicated by the curve A is low and the practicality is poor.

On the other hand, if the judgment threshold value N3 is sufficiently larger than the threshold value N1, the recognition rate increases, but at the same time, the erroneous recognition rate increases, resulting in poor reliability. That is, the improvement of the recognition rate and the reduction of the false recognition rate have the contradictory properties, and therefore, this problem cannot be solved only by operating the determination threshold value.

Therefore, in the device shown in this embodiment, it is possible to improve the recognition rate and decrease the false recognition rate at the same time.
Perform two-step verification. Therefore, the fingerprint images corresponding to the second window regions b and c located on both sides of the first window region are also registered with respect to the registration image similar to the conventional image corresponding to the first window region a. To do so.

Then, at the time of matching, matching is performed by pattern matching for the first window area a, and the number of unmatched pixels
When Na is equal to or less than the determination threshold N1 in FIG. 3, it is determined as a matching match as it is, and the number of unmatched pixels Na is the determination threshold.
If it is N3 or more, it is determined that the input fingerprint is a fingerprint of another person as a collation mismatch.

If the number of unmatched pixels Na is between the determination thresholds N1 and N3, it is determined that the determination cannot be performed in the first-stage matching, and the second determination is performed.
The step-by-step verification is performed, and the verification is performed with the registered fingerprint images of the second window regions b and c. That is, similarly to the above, the registered fingerprint images of the second window regions b and c are read from the first image storage unit 21, and the second fingerprint regions are read.
The fingerprint image is cut out by the cutout circuit 24 corresponding to the second window regions b and c and output, and the matching circuit 25 performs pattern matching. Then, the pixel number measuring circuit 26 measures the numbers Nb and Nc of mismatched pixels.

If the non-matching pixel numbers Nb and Nc are measured in this way, this is compared with the determination threshold N4 for the second-stage matching, and Nb
If both Nc and Nc are smaller than the judgment threshold value N4, it is determined that the fingerprints match each other and the fingerprints are judged to be the same.

Here, the registered fingerprint images of the second window regions Nb and Nc are fingerprint images at positions separated by a predetermined distance relative to the first window region Na that is generally adopted in the past, Matching using the fingerprint images of the second window regions Nb and Nc is equivalent to widening the area of the matching fingerprint image. Therefore, in the case of the same fingerprint, even if the input fingerprint image is distorted to some extent, when the input fingerprint image and the registered fingerprint image are best matched in the first-stage matching. The registered fingerprint image in the second window regions b and c and the input fingerprint image are matched with each other in a state where the positional shift amount is substantially the same as the positional shift amount of the input fingerprint image.

If the input fingerprint image is a fingerprint image of another person different from the registered fingerprint image, it means that the registered fingerprint image and the input fingerprint image in the first window area a are slightly coincident with each other. In other words, even if the number of unmatched pixels in the first-stage matching is between the determination thresholds N1 and N3, the second
It becomes possible to judge the non-coincidence by the step verification. That is, in the first-stage matching, the input fingerprint image of the first window area a has been searched and therefore has some degree of coincidence, but the first fingerprint area a predetermined distance away from the first window area a. In the second window regions b and c, since the fingerprints are originally different, the registered fingerprint image and the input fingerprint image do not match. Even if there is a match, it is one of the window regions b or c, or is a state in which the amount of positional shift is large due to a large movement of the input image, and it is significantly different from the amount of positional shift in the first-stage matching. Come to do.

Thus, when the number of mismatched pixels, which is the matching result in the first window area a, is between the determination thresholds N1 and N3, the two-step matching is performed again using the second window areas b and c. By performing the above, it becomes possible to simultaneously achieve the object of improving the recognition rate and the reduction of the erroneous recognition rate which have the contradictory properties.

FIG. 4 shows the flow of the fingerprint collating operation as described above. First, in step 11, the first step is performed along with the start of the collating operation.
The fingerprint image of the window area a of is roughly collated. That is, in this step 101, the registered fingerprint image corresponding to the first window region a is moved up and down, left and right at a pitch of, for example, half of the fingerprint ridge interval in order to speed up the matching process, and the first window is moved. The input fingerprint image cut out corresponding to the area and the registered fingerprint image are matched best. Then, the amount of positional deviation corresponding to the amount of movement of the registered fingerprint image when the registered fingerprint image and the input fingerprint image are in the most matched state is obtained.

If the rough collation for obtaining the positional deviation amount is performed in step 101, the process proceeds to step 102 for performing the next fine collation. In this step 102, the registered fingerprint image is set at a position corresponding to the positional deviation amount, that is, the registered fingerprint image and the input fingerprint image are made to match each other most, and the positional deviation amount is corresponded to. The registered fingerprint image is moved pixel by pixel around the position, and a location where the number of mismatched pixels is larger than the current number of mismatched pixels, regardless of whether the fingerprint image is moved in the vertical or horizontal direction, is obtained. That is, the location of the registered fingerprint image where the number of mismatched pixels is the smallest is required, and the position shift amount of this location is determined, and the next step 10
At 3, the number of unmatched pixels Na at this location is calculated.

In step 104, the first determination process of comparing the number of unmatched pixels Na obtained in step 103 with the first determination value N1 shown in FIG. 3 is performed.

If "N1 <Na <N3" is determined in step 104, the process proceeds to step 105. If "Na≤N1" is determined, the output of the "matching match" determination is obtained. , If “Na ≧ N3” is determined, an output of “collation mismatch” is obtained.

That is, in the first determination process in step 104, when it is not possible to determine whether the collation matches or the collation does not occur, the process proceeds to step 105. In step 105,
A registered fingerprint image corresponding to the second window area b, and an input image corresponding to the window area b, centered on the location corresponding to the final positional deviation amount obtained in the fine matching process of step 102. Step 10
At 6, the number Nb of mismatched pixels obtained by this matching is obtained. Then, in the next step 107, the registered fingerprint image in the second window area c and the input fingerprint image are precisely matched in the same manner as in step 105, and in step 108, the number Nc of mismatched pixels in this window area c is measured. To do.

In this way, if the non-matching pixel numbers Nb and Nc are obtained by fine matching in the second window regions b and c, the second determination processing is performed in step 109. That is, the determination threshold N4 set corresponding to step 109
And the above-mentioned non-matching pixel numbers Nb and Nc are respectively compared. If a judgment result of "Nb>N4" or "Nc>N4" is obtained, a judgment output of "matching mismatch" is obtained. To do. Further, if the determinations of “Nb ≦ N4” and “Nc ≦ N4” are obtained, the determination output of “matching match” is obtained.

In the above-described embodiment, in the second-stage matching process, two second window regions b and c are set on both sides of the first window region a, which is the main part of the matching, and these two windows are set. The number of mismatched pixels is measured in each of the regions b and c. However, in reality, only one of the areas b or c can provide a sufficiently reliable matching result. Therefore, in the second-stage matching, the right or left side of the first window area a can be obtained. Only, the window area for the second-stage collation processing may be set.

[Effects of the Invention] As described above, in the fingerprint collation device according to the present invention,
Matching or non-matching is determined by the first-stage matching process that matches the main part of the fingerprint by pattern matching. The determination threshold set during the first-stage matching is the match or mismatch determination. Are set to different values that can be ensured. Therefore, the recognition rate at this stage can be easily set to a high value, and at the same time, the false recognition rate can be sufficiently reduced. Further, in the case where the determination cannot be made in the first-stage matching, the matching in the second window regions located on both sides of the main part of the fingerprint is the same as the positional deviation amount in the first-stage matching. This is performed in the state, that is, the second-stage collation is performed in a state in which the area used for the collation is enlarged, and the collation coincidence or collation non-coincidence is determined based on the collation result. Therefore, even in a low-quality state where the input fingerprint is broken or crushed, highly accurate fingerprint collation can be executed.

[Brief description of drawings]

FIG. 1 is a block diagram illustrating a fingerprint collation device according to an embodiment of the present invention, FIG. 2 is a diagram illustrating the structure of an image processing circuit unit and an image memory unit in the above embodiment, and FIG. FIG. 4 is a diagram showing the relationship between the threshold value, the recognition rate and the erroneous recognition rate, and FIG. 4 is a flow chart for explaining the flow of the recognition operation of the fingerprint recognition apparatus. 11 ... Image sensor (fingerprint image input), 12 ... Image input circuit, 14 ... Image memory, 15 ... Image processing circuit, 17 ...
... microcomputer, 19 ... external storage device (floppy disk), 21 to 23 ... first to third image storage units (registered fingerprint image storage, input fingerprint image storage, processed image storage), 24 ... cutout circuit , 25 ... Matching circuit, 26 ...
… Pixel number (mismatch) measurement circuit.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor, Michicho Nagura, 1-1, Showa-cho, Kariya city, Aichi prefecture, Nihon Denso Co., Ltd. (72) Inventor, Ri: Eguchi, 1-1, Showa-cho, Kariya city, Aichi prefecture, Nihon Denso Co., Ltd. Incorporated (56) References JP-A-60-15779 (JP, A) JP-A-61-98483 (JP, A) Actually developed-62-195874 (JP, U)

Claims (1)

[Claims] 1. A first fingerprint image of a first window region corresponding to a main part of a fingerprint to be collated, and the first fingerprint image.
Registered fingerprint image storage means for storing and setting fingerprint images of at least one second window region that are set to be located outside the window region, and a fingerprint input for reading a fingerprint image of a finger specified at the time of fingerprint matching. Means, an input fingerprint image storage means for cutting out a fingerprint image pattern corresponding to each of the first and second window areas from the input fingerprint image input to the fingerprint input means, and storing the fingerprint image patterns, and the registered fingerprint image storage. Of the registered fingerprint image and the input fingerprint image corresponding to the first window area from the fingerprint images stored in the means and the input fingerprint image storage means, respectively, and the number of mismatched pixels is measured. The number-of-pixels measuring means and the number of non-matching pixels measured by the first number-of-pixels measuring means are set to the first set value and the first set value. Compared with a second set value larger than the set value, the number of mismatched pixels is smaller than the first set value, or larger than the second set value, and further the first set value and the second set value. And the registered fingerprint in the state where the first determination means determines that the number of mismatched pixels is between the first and second set values. From the fingerprint images stored in the image storage means and the input fingerprint image storage means, respectively, a registered fingerprint image and an input fingerprint image corresponding to the second window area are extracted and collated, and the number of non-matching pixels is measured. The second pixel number measuring means and the second judging means for comparing the number of mismatched pixels measured by the second pixel number measuring means with the third set value, and the first judging means does not match. The number of pixels is less than the first set value , And the second determining means determines a match when the number of non-matching pixels is less than the third set value, and the first determining means determines the number of non-matching pixels is more than the second set value, and 2. A fingerprint collation device, wherein the determination means of 2 determines the mismatches in a state where the number of mismatched pixels is larger than the third set value.