JP2680084B2 - Personal identification device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Outline] A personal identification device using a fingerprint is characterized in that characteristic points are read out from image information obtained by reading a fingerprint for the purpose of eliminating the troublesomeness of the pushing operation and eliminating the need for alignment. In the personal identification device that extracts and registers the matching data based on the extracted feature points, a reading unit that reads the image information at different time points within the time required for one pressing operation,
A storage unit that stores the read plurality of pieces of image information for each time point, and a feature point of each of the stored image information is extracted, and a weight value according to the appearance frequency of the same feature point And an extracting means for giving the feature points.

[Industrial applications]

The present invention relates to a personal identification device, and more particularly to a personal identification device using a fingerprint.

In recent years, with the introduction of information systems into society,
The problem is how to maintain the integrity (security) of the system. Until now, ID cards and personal identification numbers have generally been used as means for verifying identity when using information systems. However, ID cards can be lost or stolen,
In addition, it has been pointed out that problems such as personal identification numbers can be easily inferred from information around the individual.

Therefore, as a means of confirming the person's identity, attention is paid to fingerprints that have the two major characteristics of "universality" and "lifetime invariance", and development of individual collation systems using fingerprints is being promoted at various places.

[Conventional technology]

The personal identification system registers in advance information of characteristic points (points at which ridges terminate, so-called end points and branch points, so-called branch points) information of fingerprints extracted from a fingerprint image of a specific individual in advance, and this registration information is registered. And the feature point information of the fingerprint image, which has been appropriately input, are collated by, for example, template matching, and a match or non-match is determined. By the way, since the registration information is ID information of a specific individual, accuracy is required. However, it is unavoidable that false feature points may be generated or correct feature points may not be extracted depending on the state of the finger at the time of fingerprint registration, the imprinting condition, and the occurrence of noise. It wasn't enough.

Therefore, there has been proposed an individual identification device intended to improve the accuracy of registration information by adopting feature points having a high appearance frequency as registration information.

This is because when the fingerprint is registered, the image input operation with the same finger is repeated multiple times to count and weight the same feature points of each input image. This is the registration information. That is, since the feature point information having a high appearance frequency (high weighting value) is adopted as the registration information, the accuracy of the registration information is improved and the matching accuracy of the specific individual is improved.

[Problems to be solved by the invention]

However, in such a conventional personal identification device, since it is necessary to repeat the image input operation with the same finger a plurality of times, the pressing operation is troublesome, and the pressed finger is temporarily lifted and then pressed again. At that time, there is a problem that the feature points need to be aligned because the original position cannot be imprinted.

Therefore, an object of the present invention is to make it possible to obtain a plurality of input images by one pressing operation, eliminate the troublesomeness of the pressing operation, and eliminate the need for alignment.

[Means for solving the problem]

In order to achieve the above object, a personal identification device according to the present invention is a personal identification device that extracts feature points from image information obtained by reading a fingerprint and registers matching data based on the extracted feature points. A reading unit that reads the image information at different time points within the time required for the single embossing operation, a storage unit that stores the read image information of a plurality of sheets at each time point, and each of the stored image information. Extraction means for extracting the characteristic points of the image information and assigning a weight value according to the appearance frequency of the same characteristic point to each characteristic point is provided.

[Action]

According to the present invention, a plurality of pieces of image information read within the time required for one pressing operation are stored for each image, and feature points in descending order of appearance frequency are selected and registered from the stored image information. Information is created.

Generally, the finger pressing force at the time of fingerprint acquisition tends to be weak in the initial stage and then gradually increase, and the fingerprint image becomes unclear due to insufficient initial pressing force or excessive pressing force at the final stage. A proper pressing force can be obtained from the start to the end of the pushing operation, but it is practically impossible to specify the timing. This is because each person's imprinting movements cannot be patterned. In the present invention, the image information of the fingerprint is read at different times from the start to the end of the imprinting operation, and a weight value according to the appearance frequency of the same feature point of each image information is given to each feature point. This is nothing but reading the image information of various pressing forces, mutually evaluating each image information, eliminating the influence of excess and deficiency of the pressing force, and performing reliable feature point extraction. Therefore, since reliable feature point extraction is performed by performing only one embossing operation, it is not necessary to repeat the embossing operation many times as in the conventional example, and the annoyance is eliminated.

〔Example〕

 Hereinafter, the present invention will be described with reference to the drawings.

1 to 3 are diagrams showing an embodiment of a personal identification device according to the present invention.

First, the configuration will be described. FIG. 1 is a block diagram showing the function of the personal identification device. In this figure, the personal identification device is configured to include, for example, a personal computer,
Broadly classified, the fingerprint sensor 1, the binarization circuit 2, the binarized image storage circuit 3, the thinning processing circuit 4, the feature point extraction circuit 5, the data file 6, the comparison circuit 7, the registered data selection circuit 8, and the dictionary circuit 9 And each function represented by the matching circuit 10.

The fingerprint sensor 1 captures a fingerprint image, and a known CCD, for example, is used as the image sensor. CCD consists of a large number of pixels arranged in a plane, and by scanning each pixel, 1
It is for generating a single image (frame). In the present embodiment, for example, one capable of capturing about 30 images per second is used. Therefore, the fingerprint sensor 1 functions as a reading unit that reads a plurality of pieces of image information within a predetermined time. The image information read by the fingerprint sensor 1 is binarized by the binarization circuit 2 and then binarized image storage circuit 3
Sent to The binarized image storage circuit 3 includes a plurality of frame memories FM _{1 to} FM _n , and each frame memory stores one piece of image information. That is, the first image information D ₁ read by the fingerprint sensor 1 is stored in FM ₁ and D ₂ is FM.
₂ , ... D _n is stored in FM _n . Note that FM _{1 to} FM _n may be one frame memory. Binary image storage circuit 3
The D _{1 to} D _n stored therein are sequentially read from D ₁ to D _n at the time of registration and sent to the thinning processing circuit 4. The thinning processing circuit 4 thins the ridges or valleys of D ₁ (or D ₂ ... D _n ) and compresses the information. The feature point extraction circuit 5 extracts feature points (end points and branch points) of the fingerprint based on the thinning information, sends information (feature point information) such as position coordinates of the extracted feature points to the comparison circuit 7, and It is also sent to the data file 6. The data file 6 is, for example, the second
The file format is as shown in the figure. In FIG. 2, in the coordinate column, the characteristic point information sent from the characteristic point extracting circuit 5, that is, the position coordinates (X ₁ , Y ₁ ) of each characteristic point,
(X ₂ , Y ₂ ), ... (X _m , Y _m ) are written, and weight values (W ₁ ), (W ₂ ), ...
… (W _m ) is written. The comparison circuit 7 is the feature point extraction circuit 5
The feature point information (coordinates) from the above is compared with each coordinate in the data file 6 to determine whether or not the feature points are the same. If they are the same, a weighting signal is output and the corresponding in the data file 6 is output. The value of the weight W ₁ (or W ₂ ... W _m ) of the feature point coordinates to be added is incremented by, for example, +1.

That is, for all the image information stored in FM _{1 to} FM _n in the binarized image storage circuit 3, the thinning process in the thinning process circuit 4, the feature point extracting process in the feature point extracting circuit 5, and the comparing circuit. By executing the comparison process in 7,
In the data file 6, all the characteristic point information is weighted and stored.

The registered data selection circuit 8 selects a plurality of higher ranks from the feature point information stored in the data file 6 in the order of large weight value, and outputs an image cutout instruction signal centered on the coordinates of the selected feature points. Output to the binarized image storage circuit 3. The dictionary circuit 9 stores the binarized information in the binarized image storage circuit 3 which is clipped by the image clipping instruction signal.
That is, the dictionary circuit 9 stores partial binarization information (registration information) centered on each of the plurality of feature points selected by the registration data selection circuit 8. Matching circuit 10
Is a fingerprint sensor 1 and a binarization circuit 2 at the time of matching.
The binarized information stored in the specific frame memory of the binarized image storage circuit 3 and the registered information stored in the dictionary circuit 9 are compared and collated by, for example, pattern matching.
Outputs a match / mismatch signal. The coincidence and non-coincidence signals are output to a security device (not shown), and the security device permits the entry into the room or the use of the terminal only when they coincide.

 Next, the operation will be described.

At the time of registration, first, the fingerprint sensor 1 of the finger of a specific individual
The fingerprint image is read by imprinting on the.
The shorter the time required for pressing at this time, the better. Now, for example, if it takes 1 second to print, the time of about 1 second does not seem to be that long. This is because the imprinting must be done carefully so as not to shift the fingerprint, and therefore the movement from the light contact of the finger surface to the contact of the entire fingerprint surface is performed slowly, and the time required for this is often at least 1 second. This is because it takes some degree. Therefore, for example, if the number of images (number of frames) of the fingerprint sensor 1 is 30 frames per second, 30 frames, that is, 30 pieces of image information are imaged during the pressing operation for 1 second. Alternatively, it may be stored for every plural frames.

Then, among the 30 pieces of image information, a plurality of pieces of image information excluding the low-reliability image information captured in the initial stage of the pushing operation are sequentially stored in FM _{1 to} FM _n for each image, and then stored. The plurality of pieces of image information are sequentially taken out for each piece of image information, and subjected to thinning processing, feature point extraction processing and comparison processing. By these processes, the feature point coordinates and their weight values are stored in the data file 6, and a data file as shown in FIG. 2 is completed, for example.

Then, the binarized information in the binarized image storage circuit 3 is cut out in accordance with a plurality of feature point coordinates in descending order of weight value in the data file, and the cut out partial binarized information is registered information. Is stored in the dictionary circuit 9 as 3 (a) to 3 (d) are diagrams showing a fingerprint image read during the embossing operation from (a) to (d) over time, and in FIG. 3, the ⊚ mark is common to all fingerprint images. The white circles indicate the feature points that appear in the fingerprint image from the middle to the final stage of the imprinting operation without appearing in the fingerprint image at the beginning of the imprinting operation. Also, the ∇ mark indicates a feature point that has appeared for a while but has been destroyed thereafter. That is, the appearance frequency is highest for the feature points indicated by ⊚, followed by the feature points indicated by ∘ and the feature points indicated by ∇. That is, the registration information of the present embodiment mainly employs the characteristic points indicated by the double circles.

At the time of matching, the image information for 30 frames read by the fingerprint sensor 1 is binarized and then stored in the frame memories FM _{1 to} FM _n of the binarized image storage circuit 3. Then, the binarized information stored in the specific frame memory in the binarized image storage circuit 3 is compared with the registration information stored in the dictionary circuit 9, and a match / mismatch signal is generated according to the comparison result. Is output. Here, the specific frame memory is a frame memory corresponding to a frame in which the highest quality image information is obtained among the image information of, for example, 30 frames read during the embossing operation. Incidentally, the above-mentioned high quality products are often obtained from the middle stage to the final stage of the pushing operation.

As described above, according to the present embodiment, for example, the fingerprint sensor 1 capable of outputting image information of 30 frames per second is provided, and the plurality of frame memories FM _{1 to} FM _n are provided, and the image information from the fingerprint sensor 1 is provided. Is binarized and then sequentially stored in FM _{1 to} FM _n for each image information.

Therefore, it is possible to store a plurality of pieces of image information in FM _{1 to} FM _n by performing only one pushing operation, and using this stored image information, extraction of feature points and weighting of each feature point are performed. Processing such as giving can be performed. That is, the registration information can be generated based on the feature points having a high appearance frequency, and the effect of improving the matching accuracy can be obtained. I) Since only one pushing operation is required, the pushing operation is troublesome. It is possible to obtain the unique effect that the finger is not aligned because it does not need to repeat the pushing operation.

〔The invention's effect〕

According to the present invention, it is possible to perform fingerprint registration based on a feature point having a high appearance frequency with one pressing operation.

Therefore, the complexity of the pushing operation at the time of registration is eliminated, and the alignment is not required, and as a result, the registration operation is simplified.

[Brief description of the drawings]

1 to 3 are diagrams showing an embodiment of a personal identification device according to the present invention, FIG. 1 is a functional block diagram thereof, FIG. 2 is a diagram showing an example of a data file thereof, and FIG. 10A to 10D are diagrams showing a fingerprint image read during the imprinting operation with time. Reference numeral 1 is a fingerprint sensor (reading means), 3 is a binarized image storage circuit (storage means), 6 is a data file (extraction means), and 7 is a comparison circuit (extraction means).

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Hiroyuki Ikeda 1015 Kamiodanaka, Nakahara-ku, Kawasaki City, Kanagawa Prefecture, Fujitsu Limited (72) Inventor, Yuki Yahagi, 1015, Kamiodanaka, Nakahara-ku, Kawasaki City, Kanagawa Prefecture, Fujitsu Limited

Claims (1)

(57) [Claims] 1. A personal identification device for extracting feature points from image information obtained by reading a fingerprint and registering collation data based on the extracted feature points is different within the time required for one embossing operation. A reading unit that reads the image information at each time point, a storage unit that stores the read plurality of pieces of image information at each time point for each image, and a feature point that extracts the characteristic points of each of the stored image information and is the same. A personal identification device, comprising: an extracting unit that gives a weight value to each feature point according to the appearance frequency of the feature point.