JPH11213163A - Individual identifying device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize an individual identifying device capable of reducing personal refusal rate. SOLUTION: When the individual to be identified is judged not to be the said person by a recognizing means 1, image data of an object to be identified is divided into plural areas and a matching degree with dictionary data by every area is calculated by an area dividing means 3. The image data is rearranged in order of low matching degree by every calculated area by the area dividing means 3. The area with the lowest matching degree ordered by the area dividing means 3 is excluded for the image data of the object to be discriminated, the matching degree between the remaining areas and the dictionary data is calculated and the area are judged whether they satisfy a threshold value or not by a partial comparison means 5. When the threshold value is not satisfied by them, the areas with the lowest matching degree and the second lowest matching degree are excluded and the similar processing is performed. When the threshold value is not satisfied by the areas here, the areas with low matching degree are successively excluded and when the threshold value is satisfied by no area at all, the individual is judged not to be the said person.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an individual identification device for identifying an individual by comparing input image data with dictionary data stored in advance.

[0002]

2. Description of the Related Art There is known a technique for identifying an individual using biological information (a face, an iris, a fingerprint, etc.) of a human or an animal. As such a technique, for example, there is one disclosed in US Pat. No. 5,291,560 as a technique for performing individual identification using an iris.

In the technique disclosed in the above document, a circle circumscribing the pupil and a circle circumscribing the iris are detected, a coordinate system is set on the iris based on the two circles, and the coordinate system is set based on the set coordinate system. Then, the grayscale image of the iris area is binarized by filter processing (a code obtained by binarization is called an iris code). The iris code obtained in this way is compared with an iris code created in advance by a similar method to obtain a Hamming distance, and the identity is determined by threshold processing.

[0004]

However, according to the above-mentioned prior art, when the eyes are thin or the shape of the pupil is not a circle or the like, the rejection (rejection) is performed even if the identification target is the person himself / herself.
There was a risk of doing it.

FIG. 2 is an explanatory view showing such a conventional problem. In the figure, the example shown in [A] is a case where the iris region or part of the pupil is hidden by eyelids or eyelashes when the eyes are thin. In such cases, the pupil, iris,
The boundary of the sclera may become unclear and the result of circle position detection may be unstable. When the position detection result of the circle becomes unstable, the iris code fluctuates.
As a result, in the worst case, personal identification may fail due to fluctuations in the iris code.

[0006] Also, when the shape of the pupil is not a circle, the position detection result of the circle becomes unstable. For example, when the pupil is not a circle as in [B] in the figure, the position detection result of the circle changes depending on the state of the pupil boundary. For example, when the circle is applied to the lower boundary of the pupil or when the circle is applied to the upper boundary, the state changes depending on the state of the boundary. Also in this case, the iris code fluctuates, and in the worst case, the personal identification may fail. A similar problem occurs when the shape of the iris is not a circle.

Further, when wearing glasses, for example, if the lens is dirty as shown in [C] in the figure, noise is added to the image of the iris area. Therefore, if the dirty portion is directly coded, it is different from the case where the iris code is not dirty, and in the worst case, there is a possibility that the person identification may fail.

[0008] From such a point, it has been desired to realize an individual identification device that can reduce the false rejection rate.

[0009]

The present invention employs the following structure to solve the above-mentioned problems. <Structure of Claim 1> An input image of an individual to be identified,
Recognition means for comparing the dictionary data of a true individual provided in advance and determining whether the individual to be identified is a true individual,
If the recognition unit determines that the identification target is not a true individual, the identification target image is divided into a plurality of preset regions, and for each of the divided regions, the degree of coincidence with the true individual dictionary data is determined. The area dividing means to be obtained and the image to be identified are sequentially excluded from the areas having the lowest coincidence among the degrees of coincidence obtained by the area dividing means, and the degree of coincidence between the remaining area and the dictionary data of the true individual is determined. If the degree of coincidence satisfies a predetermined threshold value for determining a true individual at any point in time, a partial comparing means for determining that the individual to be identified is a true individual is provided. An individual identification device characterized in that:

<Explanation of Claim 1> The input image and the dictionary data are, for example, data relating to an iris image of a human eye.
The individual identification is identification based on comparison of iris codes, but may be applied to individual identification using other identification methods such as a face and a fingerprint. In addition, the present invention is not limited to humans, and may be individual identification of animals such as identification by the eyes of a horse.

The dictionary data of a true individual is the dictionary data of a person in the case of a human. Judge it as a true individual,
Means that the user is identified. It is assumed that the number and shape of the areas to be divided by the area dividing means can be variously selected depending on the identification target, the use conditions, and the like. The coincidence is represented, for example, by the Hamming distance between the identification target and the dictionary data of the true individual, but other means may be used.

The partial comparing means excludes, for example, the area having the largest Hamming distance and obtains the Hamming distance between the remaining area and the dictionary data. As a result, if the value does not fall below the preset threshold, the area having the next largest hamming distance is also excluded, and the hamming distance between the remaining area and the dictionary data is obtained. Then, it is checked whether or not the Hamming distance has fallen below the threshold value. If the Hamming distance has not fallen below the threshold value, the above operation is repeated. Such an operation is repeatedly performed, and if the value does not fall below the threshold even once, it is determined that the user is not the principal.

According to such an operation, if there is a certain portion of the image data to be identified that matches the dictionary data, it is possible to determine that the user is a person, and therefore the rejection rate of the person can be reduced.

<Structure of Claim 2> In the individual identification apparatus according to claim 1, the smaller the number of regions in which the image of the individual to be identified is compared with the dictionary data, the more it is determined that the individual is a true individual. Threshold storage means for holding a threshold of a value requiring a high degree of coincidence, and the number of regions to be compared at that time which the threshold storage means has when determining whether an individual to be identified is a true individual Using the threshold corresponding to
An individual identification device comprising: a determination unit that determines whether the threshold satisfies the degree of coincidence; and, if so, a partial comparison unit that determines that the individual is a true individual.

<Explanation of Claim 2> In the invention of claim 2, in addition to the invention of claim 1, the threshold value is changed according to the number of regions to be compared. That is, the threshold value is set to a value that requires a higher degree of coincidence to determine a true individual as the number of regions to be compared is smaller. The value that requires a high degree of coincidence to determine that the individual is a true individual means that, for example, when the degree of coincidence is represented by a Hamming distance, the Hamming distance is a small value.
As a result, when the partial comparison unit performs comparison, the smaller the number of regions used for comparison, the more strict the personal identification is made. Therefore, even when the determination is performed with a small number of regions, it is possible to prevent erroneous identification in which a certain individual is determined to be a different individual, and it is possible to further improve the identification performance.

The relationship between the number of regions to be compared and the threshold is as follows:
The threshold may be set arbitrarily, for example, when the number of areas becomes smaller than a certain number, the threshold may be sharply sharpened.

According to a third aspect of the present invention, in the individual identification apparatus according to the first or second aspect, the image to be identified and the dictionary data are data relating to an iris region in an eye. Device.

<Explanation of Claim 3> In the invention of claim 3, the individual identification is performed by iris recognition processing. As a result, it is possible to reduce the possibility that the iris is obscured by the eyelids or eyelashes even if the input image is the person's own, or that the lens is dirty when wearing glasses, and that the person is rejected. .

[0019]

DESCRIPTION OF THE PREFERRED EMBODIMENTS << Outline of the Present Invention >> The present invention functions as post-processing of a conventional iris recognition technique. Therefore,
By adding the present invention to the conventional iris recognition technology, a high-performance iris recognition device with a reduced personal rejection rate can be realized.

According to the present invention, when the recognition result of the conventional iris recognizing means is "reject (reject)", that is, when it is determined that the user is not the person, the iris area is divided into a plurality of predetermined areas, and each of the divided iris areas is divided. A region having a low degree of matching with the dictionary data, that is, a region having a large Hamming distance is excluded from the regions, and the overall Hamming distance is calculated again to reduce the false rejection rate.

In the actual processing, the Hamming distance is obtained for each divided area, the area having the largest Hamming distance is excluded from the processing target, and the Hamming distance of the remaining area is obtained.
A comparison is made with a preset threshold value for each comparison bit number. Here, if the value falls below the threshold value, it is determined that the user is the person himself, and the process is terminated.

If the value does not fall below the threshold value, both the region having the largest Hamming distance and the region having the next largest Hamming distance are excluded, and the Hamming distance of the entire remaining region is obtained. Similarly, personal identification is performed by comparing with a threshold value corresponding to the number of comparison bits. Here, when the person is not determined to be the person, the area having the third largest Hamming distance is also excluded from the comparison processing, and the same processing is repeated.

The above processing is repeated, and if the person is not determined as a person, it is determined that the person is not the person (reject).

By performing the above processing, even if the detection of the position of the circle is deviated, if the eyeglasses or the lens are dirty, or if the person identification is failed by the conventional iris recognition technique, the iris area can be detected. If a part that matches the recognition dictionary remains to some extent, the possibility that the user can be identified as a person increases.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. << Specific Example >><Configuration> FIG. 1 is a configuration diagram showing a specific example of the individual identification device of the present invention. The apparatus shown in the figure is constituted by a computer, and comprises a recognition unit 1, a recognition dictionary 2, an area division unit 3, an area information storage unit 4, a partial comparison unit 5, and a threshold storage unit 6. In this specific example, a case will be described in which iris recognition processing is performed as individual identification.

The recognizing means 1 is an iris recognizing means as disclosed in a prior art document. In other words, the function unit collates the iris image data to be identified with the personal dictionary data provided in the recognition dictionary 2 in advance, and determines whether the individual to be identified is the individual.

The recognition dictionary 2 is realized on an auxiliary storage device such as a semiconductor memory or a disk, and is a dictionary for storing feature data (iris code) to be recognized in advance.

The area dividing means 3 includes:
When it is determined that the identification target is not the person, the image to be identified is divided into a plurality of preset areas, and for each of the divided areas, the Hamming distance between the identification data and the person's dictionary data is obtained. This is a functional unit that performs rearrangement in the order of distance.

The area information storage means 4 is realized on an auxiliary storage device such as a semiconductor memory or a disk.
Is a functional unit for storing data such as the Hamming distance for each region obtained by the above.

The partial comparing unit 5 is a functional unit that operates only when the recognizing unit 1 determines that the user is not the user. That is, with respect to the iris area, the area having the highest ranking (the area having the largest Hamming distance) stored in the area information storage means 4 is sequentially excluded, and the Hamming distance between the remaining area and the dictionary data of the person is obtained. If the Hamming distance satisfies the threshold value of the threshold value storage means 6 at any time, the identification target is determined to be the individual.

The threshold value storage means 6 is realized on an auxiliary storage device such as a semiconductor memory or a disk, and holds a smaller threshold value as the number of areas for comparing the image of the individual to be identified with the dictionary data is smaller. It is a functional unit that performs

<Operation> The operation of each part of this embodiment will be described below in order. Since the recognizing means 1 has the same configuration as the conventional one, it performs the following operation. That is, recognition means 1
First, the position of a circle circumscribing the pupil and the iris in the input image is detected. Next, polar coordinates are set based on the two detected circles, filter processing (Gabor filter) is performed on the image data according to the set polar coordinates, and digital data of 0 and 1 is generated by threshold processing. The generated data is called an iris code.

Then, the generated iris code is compared with the dictionary iris code of the recognition dictionary 2 generated in a similar manner in advance, and the Hamming distance is obtained. If the obtained Hamming distance is smaller than a preset threshold value, it is determined that the user is the subject. Here, if the value exceeds the threshold, it is "not yourself"
Is determined.

Next, the partial comparing means 5 functions only when the recognizing means 1 determines that the user is not the user. Here, when the recognizing unit 1 determines that the user is the person,
No processing is performed, and the output result of the recognition means 1 is output as it is.

When the recognizing means 1 determines that the person is not the subject, the area dividing means 3 first divides the iris area into predetermined areas, and obtains the entire iris area obtained by the recognizing means 1. The comparison result is divided for each region.
That is, the Hamming distance is calculated again for each area.

FIG. 3 is a diagram showing an example of a method for dividing an iris region. FIG. 4 is a diagram illustrating an example of actual processing when dividing an iris region.

In the example of FIG. 4, when the output result of the recognizing means 1, that is, the entire iris area is compared with the recognition dictionary 2, the Hamming distance is 0.353, and the number of dissimilar (different) bits is compared. 722 (see FIG. 4A). When the threshold for the personal identification is set to 0.32, it is determined that the user is not the principal in the related art. Here, in this specific example, the output result of the recognizing means 1 is the same as the one shown in FIG.
It is divided into two regions, and a Hamming distance is obtained for each region. FIG. 4B shows the result.

In the example of FIG. 4B, the hamming distance is not a large value over the entire iris region, and the region having a large hamming distance is located on the right side of the iris region. Such a situation, for example, when the lens of the glasses is dirty,
This may occur, for example, when the position detection results of the pupil circle and the iris circle are shifted.

Next, the region dividing means 3 ranks (rearranges) the regions based on the Hamming distance obtained for each region.

FIG. 5 is an explanatory diagram when the regions are rearranged in the example of FIG. The table of FIG. 5A shows the state before the rearrangement, and the table of FIG. 5B shows the result of the rearrangement in the order of the region having the larger Hamming distance. The number of different bits in this table is the number of bits that did not match when comparing the bits. The total number of bits is the number of bits in the area. That is, the Hamming distance is obtained by (different number of bits) / (all number of bits).

By such rearrangement, an area having a low degree of coincidence with the recognition dictionary 2 (an area having a large Hamming distance) is ranked from an area having a high degree of coincidence with the recognition dictionary 2 (an area having a small Hamming distance). Can be

The information of each area obtained by the rearrangement (FIG. 5 (b)) is stored in the area information storage means 4. Then, the partial comparison unit 5 performs a partial comparison process on the iris region according to the information stored in the region information storage unit 4.

The partial comparing means 5 firstly outputs the area dividing means 3
Are excluded from the comparison process in order from the region having the largest Hamming distance stored in the region, and the Hamming distance of the region other than the excluded region is calculated.

FIG. 6 is a diagram showing an example of the partial comparison processing of the iris area. FIG. 6 shows the result of processing the table of FIG. In the example of FIG. 5, the regions having the larger Hamming distance are in the order of 7, 6, 3, 5, 2, 1,..., And the Hamming distance is recalculated by first excluding the region 7 from the comparison processing target. . As shown in FIG. 6, when the Hamming distance is recalculated excluding the region 7, the Hamming distance is 0.33
6 After the Hamming distance is obtained in this way, it is next determined whether or not the Hamming distance is smaller than a preset threshold. The threshold used for this determination is a threshold storage unit 6
Is stored in

FIG. 7 is a diagram showing the relationship between the number of bits compared in the threshold value storage means 6 and the threshold value. As shown in the figure, the threshold value for the personal identification is set to be smaller as the number of bits used for comparison is smaller. That is, it is set so that the smaller the number of bits compared, the more strict the determination.

The threshold value preset in the threshold storage means 6 is appropriately determined according to usage conditions (security level, number of users, environment, etc.). If the Hamming distance (0.336) when the region 7 which is the first process in the example of FIG. 6 is excluded and the total number of bits (1879) at that time are in the lower right region of the graph of FIG. It is determined that there is. On the other hand, if it is not in the lower right area (if it exceeds the threshold), then
The region having the second largest Hamming distance, that is, the region 6 is also excluded from the comparison process.

As shown in FIG. 6, when the regions 7 and 6 are excluded, the Hamming distance is 0.317 and the total number of bits is 1708. In this case as well, the threshold is compared with the threshold in FIG. 7, and if it is below the threshold, it is determined that the user is the person. .

The above processing is repeated, and if the value does not fall below the threshold value, it is determined that the user is not the principal. Note that a minimum value may be set for the number of regions to be compared, for example, up to three, and if the number of regions to be compared does not fall below the threshold even once, it may be determined that the user is not the user.

As described above, even if it is determined that the user is not the principal by comparing the entire iris region of the prior art, if the region having a high degree of coincidence with the recognition dictionary 2 partially exists in the iris region, the present invention is applied. The possibility of determination is increased, and the problems of the prior art are solved.

In the above specific example, the iris region is divided into 12 along the circumference and two in the radial direction, but the division number of the region is not limited to 12. ,
The following various forms are also possible for the dividing method.

● When the size of each area is equal or not equal ● Change the shape of the area such as dividing the iris area horizontally and vertically

In the above description, the case where the present invention is applied to human iris recognition has been described. However, the present invention is not limited to this case. The present invention can be applied.

The operation of the individual identification device as described above is as follows.
It can be realized by control by a program of a computer that plays the role of an individual identification device. Therefore, the individual identification device of the present invention is realized by means of recording the program on various recording media such as a floppy disk or a CD-ROM, and then installing the program on a general computer or downloading the program via a network. be able to.

<Effects> As described above, according to the above-described specific example, even if personal identification fails in the conventional iris recognition technology, an area having a low degree of matching with the recognition dictionary in the iris area is subjected to comparison processing. , And the degree of coincidence is calculated again, so that an improvement in the personal identification performance can be expected.

Further, since the present invention can be added as post-processing of the prior art, when the present technology is introduced into an apparatus using the conventional technique, a minimum change is required. Is possible. Therefore, since the problems of the individual identification device of the related art can be solved, individual identification that is easier to use can be realized.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a specific example of an individual identification device of the present invention.

FIG. 2 is an explanatory diagram of a problem of the related art.

FIG. 3 is a diagram illustrating an example of a method for dividing an iris region in a specific example of the individual identification device of the present invention.

FIG. 4 is a diagram illustrating an example of actual processing when an iris region is divided in a specific example of the individual identification device of the present invention.

FIG. 5 is an explanatory diagram in a case where the degree of coincidence of regions is ranked in a specific example of the individual identification device of the present invention.

FIG. 6 is an explanatory diagram of a partial comparison process of an iris region in a specific example of the individual identification device of the present invention.

FIG. 7 is an explanatory diagram of a threshold value of an area information storage unit in a specific example of the individual identification device of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Recognition means 2 Recognition dictionary 3 Area division means 4 Area information storage means 5 Partial comparison means 6 Threshold storage means

Claims (3)

[Claims] A recognition unit that checks an input image of an individual to be identified with dictionary data of a true individual provided in advance and determines whether the individual to be identified is a true individual; If the recognition unit determines that the identification target is not a true individual, the identification target image is divided into a plurality of preset regions, and for each of the divided regions, dictionary data of the true individual is included. A region dividing means for calculating the degree of coincidence of the images to be identified, and sequentially excluding the regions having the lowest degree of coincidence among the degrees of coincidence obtained by the area dividing means, and the remaining region and the true individual When the degree of coincidence with the dictionary data is obtained, and at any point in time, the degree of coincidence satisfies a preset threshold for determining a true individual, the individual to be identified is a true individual. And a partial comparing means for determining Individual identification apparatus according to symptoms. 2. The individual identification device according to claim 1, wherein the smaller the number of regions in which the image of the individual to be identified is compared with the dictionary data, the higher the degree of coincidence is required to determine that the individual is a true individual. Threshold storage means for holding a threshold of a value to be, when determining whether the individual to be identified is a true individual, the threshold storage means corresponding to the number of regions to be compared at that point in time An individual identification device, comprising: a threshold value; determining whether the threshold value satisfies the degree of coincidence; and if so, a partial comparing unit that determines that the individual is a true individual. 3. The individual identification device according to claim 1, wherein the image to be identified and the dictionary data are data relating to an iris region in an eye.