JPH0414185A - Identification method by fingerprint - Google Patents

Abstract

PURPOSE:To prevent an erroneous decision caused by noise by converting the degree of the coincidence of an image pattern into an evaluation value showing the degree of a person himself or the degree of another person himself respectively concerning plural decision areas set in an input image and totally judging the evaluation values concerning all the decision areas. CONSTITUTION:An input image for fingerprints is prepared by a TV camera 15. By a pre-processing circuit 16, the noise is eliminated and the image is binarized to store in an image memory 17. A controller 18 controls a pre- processing and a storage operation based on a synchronizing signal separated from the camera 15. A collation decision circuit 19 sets the plural decision areas in respect to the stored image of the memory 17, performing matching by using a corresponding template read out from a pattern memory 20 to calculate the degree of coincidence of the image pattern respectively. In such a case, the degree of coincidence for each decision area is converted into the evaluation value showing the degree of a person himself or the degree of another person himself and the identification is performed by totally deciding the evaluation values about all the decision areas to display the result. Therefore, the accuracy of the decision is improved.

Description

[Detailed Description of the Invention] <Industrial Application Field> The present invention provides a fingerprint-based personal authentication method in which a fingerprint image is input and the input image is compared with a specific person's fingerprint to authenticate whether or not the person is the person. Regarding.

<Conventional technology> In recent years, methods of authenticating individuals using fingerprints have been studied to control entry and exit from sensitive areas and to control the opening and closing operations of building and car doors. Identity authentication devices have been put into practical use in some areas.

As shown in FIG. 5, the personal authentication device uses the points where the ridge 1 of the fingerprint is interrupted or branched, that is, the end point 2 or the branch point 3 of the ridge 1, as a feature point, and uses multiple feature points as feature points. The position and direction are extracted, and the registered data and input data are compared with respect to the positional relationship to determine whether or not the person is the real person.

<Problems to be Solved by the Invention> Personal authentication devices using such matching algorithms have excellent fingerprint matching ability, but require preprocessing of images that are faithful to the original images, and are difficult not only during registration but also during verification. However, image processing becomes complicated, such as requiring feature point extraction processing. Therefore, the hardware and software become large-scale, and the manufacturing cost of the device becomes high.

Therefore, in recent years, attempts have been made to introduce template matching into fingerprint matching to authenticate individuals.

6 to 8 show the principle of this type of person authentication method.

FIG. 6 shows a reference image 4 obtained from a fingerprint of a specific person (principal), in which images of rectangular small areas containing a plurality of feature points 5 to 7 are cut out as templates 8 to 10, and each template 8 .about.10 is registered in advance in the pattern memory along with its position. The reference image 4 is a binary image, and for example, the ridge 1 is represented by black pixels, and the other parts are represented by white pixels.

FIG. 7 shows an input image 11 consisting of a binary image obtained from a fingerprint to be authenticated, and rectangular determination areas 12 to 14 are set at each position of the templates 8 to 10. Each determination area 12 to 14 is an input image 11 for the reference image 4.
Each template 8 corresponds to absorb the misalignment of
~10 is set to a sufficiently wide area.

Template matching is performed by scanning the input image 11 in each determination area 12 to 14 using the corresponding templates 8 to 10. FIG. 8 shows an example of template matching using template 8. be.

In the figure, the template 8 is scanned in the direction of the arrow,
The number of matching black pixels and white pixels between the input image 11 and the template 8 is measured at each scanning position, and the maximum value of this measured value is determined as the matching degree of pattern matching. Similar template matching is performed for all templates to determine the degree of matching, and it is determined whether each degree of matching exceeds a predetermined threshold.

As a result, if there are a predetermined number (for example, half) or more of determination areas that exceed the threshold, it is determined that the fingerprint in the input image 11 is that of the person himself/herself.

However, according to this method, when noise is distributed over a considerable range of the input image 11, a high degree of matching is obtained for the judgment area with little noise, but a low degree of agreement is obtained for the judgment area where there is a lot of noise. divide the threshold,
Moreover, the determination area that is below the threshold value may reach more than half of the area. In such a case, even if the input image 11 is a fingerprint of the person, the fingerprint of the input image 11 is erroneously determined not to be that of the person, and the discrimination accuracy decreases.

The present invention has been made in view of the above-mentioned problems, and aims to provide a fingerprint-based personal authentication method that does not require complex image processing and can prevent deterioration in discrimination accuracy due to the influence of noise. .

<Means for Solving the Problems> In the fingerprint-based personal authentication method of the present invention, a plurality of judgment areas are set at specific positions for a fingerprint input image, and a template corresponding to each judgment area is scanned. After performing matching and determining the degree of matching of image patterns between the input image and each template, the degree of matching for each judgment area is converted into an evaluation value indicating the degree of authenticity or degree of stranger, and Personal identification is performed by comprehensively judging evaluation values.

<Effect> When noise is distributed over a considerable range of the input image,
Judgment areas with a lot of noise will have a low degree of agreement and the evaluation value of the degree of authenticity will be low, but judgment areas with little noise will have a high degree of agreement and the evaluation value of the degree of identity will be high, so these are combined. If the judgment is made, it is correctly determined that the input image of the person's fingerprint is that of the person, and a decrease in discrimination accuracy is prevented.

<Embodiment> FIG. 1 shows the procedure of a fingerprint-based personal authentication method according to an embodiment of the present invention, and FIG. 2 is an example of a circuit configuration for carrying out the personal authentication method of the present invention.

In FIG. 2, a television camera 15 images the fingertip positioned on the fingerprint input section to generate a fingerprint input image.

The preprocessing circuit I6 receives the video signal of the input image and performs preprocessing such as noise removal and binarization. The binary input image obtained through this preprocessing (hereinafter simply referred to as “binary image”) is stored in the image memory 17. controller 18
controls the operation of the preprocessing circuit 16 and the reading and writing of image data into the image memory 17 based on the synchronization signal separated from the television camera 15.

The matching/determination circuit 19 sets a plurality of determination areas similar to the conventional example described above for the binary image stored in the image memory 17, and reads out the input image from the pattern memory 20 for the input image in each determination area. Template matching is performed using corresponding templates to determine the degree of matching of each image pattern. Up to this point, it is the same as the conventional example, but
In the case of this verification/determination circuit 19, the degree of matching for each determination area is converted into an evaluation value indicating the degree of identity or authenticity, and the evaluation values for all determination areas are comprehensively judged to perform identity authentication.
Output the result.

FIG. 3 shows a first conversion curve 25 for converting the above-mentioned -political change into an evaluation value indicating the degree of identity, and a second conversion curve 26 for converting the above-mentioned -political change into an evaluation value indicating the degree of otherness. shows. These conversion curves 25 and 26 are generated based on the experimentally determined distribution characteristics 22 and 23 shown in FIG.

On the other hand, the distribution characteristic 22 is obtained by setting a large number of determination areas as shown in FIG. 7 in the input image of the person's fingerprint, and performing template matching using the corresponding template for the input image in each determination area. The distribution of the degree of agreement obtained by performing is shown below. The other distribution characteristic 23 is
Matching degree obtained by setting a number of similar judgment areas for the input image of the person's fingerprint, and performing template matching on the input image in each judgment area using other templates that do not correspond to each of the input images. shows the distribution of Note that the horizontal axis in FIG. 4 indicates the degree of coincidence, and the degree of coincidence decreases in the direction of the arrow. Moreover, the vertical axis shows the frequency.

One distribution characteristic 22 has a matching degree of A-B (however, A>B)
The other distribution characteristic 23 is distributed in a range where the degree of coincidence is C-D (however, C0D), and the first conversion curve 25 has a degree of coincidence which is E close to C (however, E>C). The second conversion curve 26 starts from the position F (where F<B) where the degree of coincidence is close to B.

In other words, the evaluation value of the degree of identity and degree of otherness is -political change is E-
If it is within the range of F, all evaluation values are "0" and -
If the political upheaval is E or higher, the evaluation value of the degree of authenticity is 1°2.3 according to the first conversion curve 25. -0-0, and if the -political change is F or less, the evaluation value of the otherness degree is 1.2.3 according to the second conversion curve 26. ...and increases.

Now, in step 1 of FIG. 1 (indicated by rsTIJ in the figure), when an input image of a fingerprint is generated from the television camera 15, in the next step 2, the preprocessing circuit 16 inputs the video signal of the input image and removes noise. Preprocessing such as removal and binarization is performed to generate a binary image. Next, in step 3, the matching/determination circuit 19 sets a plurality of determination areas at specific positions for this binary image, scans the template corresponding to each determination area, performs template matching, and performs template matching between the input image and each template. Find the degree of match between the image pattern and the image pattern.

In the next step 4, each degree of matching is calculated from the first step. The second conversion curves 25 and 26 are used to convert into an evaluation value of the degree of authenticity or degree of stranger. Note that a broken line 24 in the figure indicates that each process of step 3.4 is sequentially executed a number of times corresponding to the set number of determination areas.

Once the evaluation values for all determination areas have been determined, the process proceeds to step 5, where these evaluation values are comprehensively evaluated to determine whether or not the fingerprint in the input image is that of the person himself/herself. In this example, the evaluation value of the degree of identity is set as a positive value, the evaluation value of degree of otherness is set as a negative value, the sum of all evaluation values is calculated, and if the calculated value is a positive value, the input It is determined that the fingerprint in the image belongs to the person himself/herself, and if it is a negative value, it is determined that the fingerprint in the input image is not that of the person himself/herself, and the judgment results are outputted (step 6).

Assuming that the fingerprint in the human image is that of the person himself/herself, and that noise is distributed over a considerable range of the input image, the degree of matching will be low in the noisy judgment areas, and the evaluation value of the degree of authenticity will be low. This is a small value, and depending on the degree of noise, evaluation values of different degrees will appear. On the other hand, in the determination area with less noise, a high degree of matching is obtained, and the evaluation value of the degree of authenticity is high. In this case, if the evaluation value of the degree of authenticity and degree of stranger for the judgment area with a lot of noise is around 0, but if there are a suitable number of judgment areas with little noise, the sum of these evaluation values will be a positive value. Therefore, a correct judgment is made that the fingerprint in the input image is that of the person himself/herself. In such cases, according to the conventional personal authentication method, the person is authenticated based on the number of judgment areas divided by the threshold value. If there are a large number of fingerprints, it will be determined that the input image is not the fingerprint of the person himself/herself, even if there are several judgment areas with a high degree of matching. According to the person authentication method G of the present invention, erroneous determination in such cases can be prevented and the accuracy of the judgment M1 can be improved.

<Effects of the Invention> As described above, the present invention converts the degree of matching of image patterns for a plurality of judgment areas set in an input image into an evaluation value indicating the degree of authenticity or degree of stranger, and calculates the evaluation value for all judgment areas. Since the person is authenticated based on a comprehensive judgment, it is possible to prevent erroneous judgments from occurring due to the influence of noise appearing in the input image, and improve discrimination accuracy.

Further, since complicated image processing is not required, large-scale hardware and software are not required, and the device can be manufactured at low cost, achieving the remarkable effect of achieving the purpose of the invention.

[BRIEF DESCRIPTION OF THE DRAWINGS] FIG. 1 is a flowchart showing the steps of a personal authentication method according to an embodiment of the present invention, FIG. 2 is a block diagram showing an example of a circuit configuration for carrying out the invention, and FIG. is an explanatory diagram showing a method for converting the degree of matching of image patterns into evaluation values of identity and identity, Figure 4 is an explanatory diagram showing a method for obtaining the conversion curve of Figure 3, and Figure 5 is a diagram showing the feature points in a fingerprint. 6 is an explanatory diagram showing a plurality of templates used for template matching, FIG. 7 is an explanatory diagram showing the determination area set in the input image, and FIG. 8 is an explanatory diagram showing the situation of template matching. FIG. 8-9... Template 11... Input image 12-14... Judgment area 25, 26... Conversion curve

Claims (1)

[Claims] Multiple judgment areas are set at specific positions for the fingerprint input image, and template matching is performed by scanning the corresponding template for each judgment area to evaluate the degree of image pattern matching between the input image and each template. After the determination, the degree of matching for each judgment area is converted into an evaluation value indicating the degree of identity or authenticity, and the evaluation values for all judgment areas are comprehensively judged to perform identity authentication. Identity verification method.