JP3620938B2 - Personal identification device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a personal identification device for identifying an individual using human ecological characteristics, particularly an iris of an eyeball.
[0002]
[Prior art]
Existence of qualifications when conducting transaction operations using automatic transaction devices at financial institutions, accessing confidential data with computers, etc., and entering facilities that only specific persons can enter Personal identification is performed to check
[0003]
Various techniques for personal identification are currently conceived or implemented. One technique is to identify an individual using an iris that is a part of a human eyeball.
This iris is completed in early childhood, has little change throughout life, and is not easily affected by changes in physique such as fattening and thinning, so it has high identification accuracy, so it can be applied in various fields in the future. Expected.
[0004]
FIG. 12 is a functional block diagram showing a conventional general personal identification device using an iris.
In the figure, reference numeral 1 denotes a camera (video camera) for inputting video, 2 denotes a process for cutting out an iris from an image captured by the camera 1, that is, a preprocessing unit for performing preprocessing, and 3 denotes an iris cut out by the preprocessing unit 2. 4 is a feature extraction unit that extracts the features and codes the pattern, and 4 compares the iris pattern sent from the feature extraction unit 3 with the iris pattern registered in the collation dictionary unit 5 to The collation unit 5 obtains the similarity between them, and here, the collation dictionary unit 5 stores iris patterns for a plurality of people in advance.
[0005]
Reference numeral 6 denotes a determination unit that determines whether both iris patterns are of the same person based on the similarity calculated by the collation unit 4.
In this configuration, the preprocessing unit 2, the feature extraction unit 3, the collation unit 4, and the determination unit 6 are generally configured by a microprocessor or dedicated arithmetic hardware, and the collation dictionary unit 5 is a semiconductor. Generally, it is composed of a memory or a magnetic disk.
[0006]
Next, the effect | action of the structure mentioned above is demonstrated.
FIG. 13 is a flowchart showing a processing procedure in an iris pattern registration mode, and FIG. 14 is a flowchart showing a processing procedure in an iris pattern identification mode.
First, the registration mode of FIG. 13 will be described according to the steps indicated by S.
[0007]
An image of the upper body of the person to be registered is taken by the camera 1. That is, a video is input (S1).
Next, the photographed image is sent to the preprocessing unit 2, and the preprocessing unit 2 cuts out the head from the identified person's image (indexing of the position of the head) (S2). The eye cut-out (eye position determination) from the head image (S3) and the iris cut-out (iris position determination) process from the eye image are sequentially performed by the preprocessing unit 2 (S4).
[0008]
The cut out iris image is sent to the feature extraction unit 3, and the feature extraction unit 3 extracts features necessary for matching (S5).
Specifically, feature extraction is performed such as dividing the iris into necessary analysis bands or extracting a change in iris density by a Gabor filter or the like.
The extracted feature of the iris is normalized by the feature extraction unit 3 (S6) and coded as binary data (S7), and finally an iris code of about 256 bytes is generated as an iris pattern.
[0009]
Necessary data such as a registrant name, ID number, and date are added to the iris pattern generated in this way, and the iris pattern is registered in the collation dictionary unit 5 (S8).
Next, the identification mode of FIG. 14 will be described according to the steps indicated by S.
In this mode, it is assumed that iris patterns of a plurality of people are registered in the collation dictionary unit 5 as described above.
[0010]
First, the pre-processing unit 2 performs the video input (S1), head segmentation (S2), eye segmentation (S3), and iris segmentation (S4) processes of the identified person in the same manner as in the registration. In the same manner as in the case of the registration, the iris extraction (S5), normalization (S6), and encoding (S7) processes are performed in the feature extraction unit 3 in order. Generated.
[0011]
The generated iris pattern is sent to the collation unit 4, and the collation unit 4 collates (matches) the iris pattern sent from the feature extraction unit 3 with the iris pattern registered in the collation dictionary unit 5, and performs similar processing. The degree is obtained (S8).
In the collation calculation in this case, a distance value such as a Hamming distance is often used as a measure of similarity.
[0012]
The degree of similarity (distance value) of the iris pattern selected by the collation in the collation unit 4 is transferred to the determination unit 6, and the degree of similarity sent by the determination unit 6 is obtained in advance by experiments or statistical methods. It is compared with one or a plurality of threshold values, and it is determined whether or not the iris pattern obtained from the image of the identified person is the same person as the collated iris pattern in the collation dictionary unit 5 (S9).
[0013]
In the collation process in this identification mode, it is possible to freely select which (who) code registered in the collation dictionary 5 for collation or how many (how many) patterns are used. Although it is determined according to the system specifications, the following two types are generally adopted.
One of them is when the pattern to be verified is known among the iris patterns registered in the verification dictionary unit 5 with respect to the iris pattern obtained from the image of the identified person, that is, one-to-one verification. Yes, another one is that Iris pattern registered in the collation dictionary unit 5 is not known and the most probable iris pattern is selected from the registered iris patterns In other words, it is a one-to-n comparison.
[0014]
In the case of one-to-one matching, the matching unit 4 calculates the similarity using one iris pattern and outputs the calculation result to the determination unit 6.
In this case, the similarity is compared with one or a plurality of threshold values, and whether or not the iris pattern obtained from the image of the identified person is the same person as the collated iris pattern is determined. Identify.
[0015]
In the case of one-to-n collation, the collation unit 4 collates a plurality of iris patterns in the collation dictionary unit 5 to perform a collation operation, and the first or upper m-th (m is an arbitrary value) The similarity is sent to the determination unit 6.
In this case, the determination unit 6 determines the first or top m similarity by one or a plurality of threshold values, and the iris pattern obtained from the image of the identified person is the collated iris pattern. It is determined whether the iris pattern obtained from the image of the identified person is the same person by checking whether it is the most similar or not sufficiently similar. .
[0016]
[Problems to be solved by the invention]
However, in the above-described personal identification device using iris, the recognition device uses the instability of the image due to the person to be identified blinking at the time of photographing, the hair touching his eyes, or the like, or the image acquisition based on the surrounding lighting conditions. When accurate iris data cannot be obtained due to instability or the like, there is a problem that reliable identification becomes difficult and the identification rate decreases.
[0017]
In order to solve such problems, it is not preferable on the human interface to force an unreasonable posture on the person to be identified or to place severe restrictions on the surrounding lighting conditions in order to obtain stabilization of the image. Furthermore, if a camera with a large number of pixels is used in order to simply increase the recognition rate, an increase in processing amount, an increase in the size of the apparatus, an increase in price, an increase in power consumption, etc. are unavoidable, which is not realistic.
[0018]
Therefore, the object of the present invention is to enable reliable identification without forcing the identified person into an unreasonable posture, without severely limiting the surrounding lighting conditions, and without using a camera with a large number of pixels. Thus, it is to realize a personal identification device capable of obtaining a high identification rate.
[0019]
[Means for Solving the Problems]
Therefore, the present invention Taken by camera Identified person of From the image Iris recognition unit and facial recognition unit Extract features and create iris pattern and facial pattern As well as acquiring the iris pattern and facial pattern Check the pre-registered iris pattern and facial pattern for each similarity level, By similarity joint By combining the similarity of the iris pattern and the similarity of the facial pattern, and the iris pattern and facial pattern obtained from the image of the person to be identified based on the result of the combination, the iris pattern used for collation And whether it belongs to the same person as the facial pattern In the general judgment section It is characterized by determining.
[0020]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a block diagram showing a configuration of a first embodiment of a personal identification device according to the present invention.
This identification device includes a camera 1 that takes a picture of a person, a camera output distribution unit 11, an iris recognition unit 12, a facial recognition unit 13, a similarity combination unit 14, and a comprehensive determination unit 15.
[0021]
Here, the camera output distribution unit 11 receives the output from the camera 1 and sends the output to the iris recognition unit 12 and the facial recognition unit 13, respectively.
The iris recognition unit 12 includes a preprocessing unit 2, a feature extraction unit 3, a collation unit 4, and a collation dictionary unit 5, and the facial recognition unit 13 includes a preprocessing unit 7, a feature extraction unit 8, a collation unit 9, and It is comprised by the collation dictionary part 10. FIG.
[0022]
The components 2 to 4 in the iris recognition unit 12 are the same as the components shown in FIG. 13, and the components 7 to 10 in the facial recognition unit 13 are 2 to 2. 5 Corresponds to
On the other hand, the similarity combining unit 14 receives the similarities from the collation unit 4 of the iris recognition unit 12 and the collation unit 10 of the face recognition unit 13, calculates the overall similarity, and sends it to the comprehensive determination unit 6. The general determination unit 15 is a person whose person to be identified is registered in the collation dictionary unit 5 of the iris recognition unit 12 and the collation dictionary unit 10 of the face recognition unit 13 based on the total similarity sent from the similarity combination unit 14. Is output and the determination result is output.
[0023]
Next, the operation of the apparatus having such a configuration will be described.
This apparatus operates in two modes, a registration mode and an identification mode. FIG. 2 is a flowchart of the registration mode, and FIG. 3 is a flowchart of the identification mode.
First, the registration mode of FIG. 2 will be described according to the steps indicated by S.
[0024]
An image of the upper body of a person to be registered (registrant) is taken by the camera 1. That is, a video is input (S1).
The output of the camera 1 is sent to the camera output distribution unit 1, and images taken by the camera 1 by the camera output distribution unit 1 are distributed to the preprocessing units 2 and 7 of the iris recognition unit 12 and the facial recognition unit 13, respectively. (S2).
[0025]
In the iris recognizing unit 12, each of head segmentation (S3), eye segmentation (S4), iris segmentation (S5), iris feature extraction (S6), normalization (S7), and encoding (S8) The process is performed by the preprocessing unit 2 and the feature extraction unit 3 as in the conventional case, and the iris pattern is registered in the collation dictionary unit 5 (S9).
On the other hand, in the facial recognition unit 13, the preprocessing unit 7 detects the outline of the face from the registrant image sent from the camera output distribution unit 11 (S 10), and the face such as eyes, nose and mouth is detected from the outline. Detection of a region showing a unique feature (S11), and further, a relative relationship between these is detected (S12).
[0026]
The feature extraction unit 8 receives the output of the preprocessing unit 7, normalizes the features of the input face image (S13), encodes them, and generates a facial pattern (S14).
Necessary data such as a registrant name, ID number, and date are added to the facial pattern generated in this way and registered in the collation dictionary unit 10 (S15).
Since the iris pattern and the facial pattern are registered in the collation dictionaries 5 and 10 as described above, both patterns need to be linked to each other. Can be used.
[0027]
Moreover, the process of S3-S9 in the iris recognition part 12 mentioned above and the process of S10-S15 in the face recognition part 13 are performed in parallel.
Next, the identification mode of FIG. 3 will be described according to the steps indicated by S.
In this mode, it is assumed that iris patterns and facial patterns of a plurality of persons are registered in the collation dictionaries 5 and 10 as described above.
[0028]
First, from the video input of the identified person by the camera 1 to the iris coding in the iris recognition unit 12 (S1 to S8) and the facial coding in the facial recognition unit 13 (S1, S2, S10 to S14). Each process is performed in the same manner as in the registration.
The iris recognition unit 12 sends the iris pattern generated by the feature extraction unit 3 to the matching unit 4, and the matching unit 4 registers the iris pattern sent from the feature extraction unit 3 in the matching dictionary unit 5. The matching iris pattern is collated (matching), the similarity is obtained (S 9), and the result is output to the similarity combining unit 14.
[0029]
In this case, the similarity is calculated in the same manner as before.
On the other hand, in the facial recognition unit 13, the facial pattern generated by the feature extraction unit 8 is sent to the collation unit 9, and in the collation unit 9, the facial pattern sent from the feature extraction unit 8 is registered in the collation dictionary unit 10. The similarity is calculated by collating and calculating the facial pattern that is present (S15).
[0030]
A distance scale or the like is used for the collation calculation, and the collation unit 9 outputs the similarity (distance value) of the calculation result to the similarity combination unit 14.
In this way, each of the collation units 4 and 9 of the iris recognition unit 12 and the facial recognition unit 13 performs collation of the iris pattern and the facial pattern. The iris pattern and the facial feature of the collation dictionary unit 5 used for this collation are used. The pattern is controlled so that the pattern of the same person is always used.
[0031]
Next, the similarity combining unit 14 combines the two similarities received from the matching units 4 and 9 of the iris recognition unit 12 and the facial recognition unit 13 by the following equation (1) to calculate the total similarity. (S16)
T = k * R _I + (1-k) * R _f ・ ・ ・ ・ ・ ・ (1)
Where T: total similarity, R _I : Iris similarity, R _f : Similarity based on face, k: constant (1> k), and this constant k determines an appropriate value by experiments or statistical methods.
[0032]
The total similarity calculated by the similarity combining unit 14 is sent to the total determination unit 15, and the total determination unit 15 compares the total similarity with one or a plurality of threshold values and is obtained from the input video. It is determined whether the iris pattern and the facial pattern are of the same person as the collated iris pattern and facial pattern in the collation dictionary units 5 and 10 (S17), and the determination result is output.
[0033]
In the verification process in the identification mode of FIG. 3 described above, the following two types are employed.
One of them is that the iris pattern and the facial pattern registered in the collation dictionary units 5 and 10 are respectively identified from the iris pattern and the facial pattern obtained from the image of the identified person. In other words, one-to-one collation is performed, and the other one is that the patterns to be collated among the iris patterns and facial patterns registered in the collation dictionary units 5 and 10 are not known, and a plurality of registered plural patterns are registered. When the most probable iris pattern and facial pattern are selected from the scan patterns, that is, one-to-n matching.
[0034]
In the case of one-to-one matching, the matching units 4 and 9 calculate the similarity using one iris pattern and facial pattern, respectively, and output it to the similarity combining unit 14.
In the case of one-to-n collation, the collation units 4 and 9 perform a collation operation using a plurality of iris patterns and facial patterns in the collation dictionary units 5 and 10, respectively. m is an arbitrary value) and is sent to the similarity combining unit 14.
[0035]
In this case, it is possible that the first or upper m-th similarity received from the matching units 4 and 9 is not the same person, but in that case, the similarity combining unit 14 is not sufficiently accurate. As a result, it is possible to determine that the iris pattern and facial pattern obtained from the image of the identified person are not the same person as the iris pattern and facial pattern used for the collation.
[0036]
According to the first embodiment described above, facial recognition is performed in addition to iris recognition, and individual identification is performed by using the similarity of both recognitions. Without being forced, or severely restricting ambient lighting conditions, and without using a camera with a large number of pixels, reliable identification can be performed and a high identification rate can be obtained.
[0037]
FIG. 4 is a block diagram showing the configuration of the second embodiment.
In the second embodiment, the comprehensive determination unit 15 is connected to each of the matching units 4 and 9 of the iris recognition unit 12 and the facial recognition unit 13, and the similarity combining unit 14 is connected to the comprehensive determination unit 15. The other configurations are the same as those in the first embodiment.
The second embodiment also operates in two types of modes, a registration mode and an identification mode. Since the registration mode is the same as that of the first embodiment, only the identification mode will be described here as an action.
[0038]
FIG. 5 is a flowchart showing the identification mode of the second embodiment, which will be described according to the steps indicated by S.
In this identification mode, the processing from S1 to S15 is performed in the same manner as in the first embodiment, so that the description thereof will be omitted. Of the patterns and facial patterns, the pattern to be collated is not known, and the most probable pattern is selected from a plurality of patterns, that is, one-to-n collation is applied.
[0039]
The difference in operation between the second embodiment and the first embodiment is a difference in determination method between the comprehensive determination unit 15 and the similarity combining unit 14.
That is, when the comprehensive determination unit 15 receives the similarity to the top n ranks of the iris pattern and the facial pattern from the matching units 4 and 9 of the iris recognition unit 12 and the facial recognition unit 13, respectively, Each difference of the similarity is calculated (S16).
[0040]
Next, the difference is determined based on a threshold value a determined in advance using experiments or statistical methods (S17). If the difference is greater than or equal to the threshold value a, a comprehensive determination is made using only the similarity of the iris pattern. The difference is the threshold a Less than In this case, the process proceeds to the next step of calculating the total similarity.
For example, when the similarity up to the second place is input from the collation unit 4 of the iris recognition unit 12 to the comprehensive determination unit 15, the comprehensive determination unit 15 determines the similarity between the input first and second ranks. The difference is obtained, and this difference is a preset threshold value a. Same as If it is larger, it is determined that the iris pattern indicating the first similarity is sufficiently certain, and the process proceeds to comprehensive determination.
[0041]
Also, if the difference is smaller than a, it can be said that what has risen as the first place is not sure enough, so the iris recognition unit 12 Lano Not judged only by similarity, General Degree of similarity To calculation Proceed with
When the difference between the top n ranks of the iris pattern is smaller than the threshold value a, the comprehensive determination unit 15 transfers the top n rank similarity of the iris pattern and the facial pattern to the similarity combining unit 14, In response, the similarity combining unit 14 combines both similarities using the equation (1) described in the first embodiment, calculates the overall similarity, and sends the calculation result to the overall determination unit 15. Transfer (S18).
[0042]
When the overall determination unit 15 receives the overall similarity from the similarity combining unit 14, the overall determination unit 15 determines the total similarity by comparing it with one or more threshold values (S19), and outputs the determination result.
That is, it is determined whether or not the iris pattern and facial pattern obtained from the image of the identified person belong to the same person as the collated iris pattern and facial pattern in the collation dictionary units 5 and 10. Output.
[0043]
In the second embodiment described above, it is possible to obtain a stable and high recognition rate as in the first embodiment, and in the second embodiment, it is determined that the iris pattern is sufficiently likely. In such a case, since the person to be identified is identified without using the similarity of the facial pattern, the processing speed can be increased.
FIG. 6 is a block diagram showing the configuration of the third embodiment.
[0044]
In this embodiment, the iris recognition unit 12 But The pre-processing unit 2 and the feature extraction unit 3 are configured, and the facial recognition unit 13 is also configured by the pre-processing unit 7 and the feature extraction unit 8, and both the feature extraction units 2 and 8 are connected to the feature coupling unit 14. Yes.
A collation unit 16 is provided between the feature combination unit 14 and the comprehensive determination unit 15, and a collation dictionary unit 17 is provided in the collation unit 16 and the feature combination unit 14.
[0045]
It should be noted that images taken by the camera 1 are pre-processed by the camera output distribution unit 11 in the respective pre-processing units 2 and 7 of the iris recognition unit 12 and the facial recognition unit 13. In Sending is the same as in the first and second embodiments.
The third embodiment also operates in two types of modes, a registration mode and an identification mode. FIG. 7 is a flowchart of the registration mode, and FIG. 8 is a flowchart of the identification mode.
[0046]
In both modes, there are features after the feature extraction of the iris recognition unit 12 and the facial recognition unit 13, respectively.
First, the registration mode will be explained And Each process from the video input of the person to be identified by the camera 1 to the iris coding in the iris recognition unit 12 (S1 to S8) and the facial coding in the facial recognition unit 13 (S1, S2, S9 to S13) Is performed in the same manner as in the first and second embodiments.
[0047]
The iris pattern and the facial pattern encoded by the feature extraction unit 3 of the iris recognition unit 12 and the feature extraction unit 8 of the facial recognition unit 13 are sent to the feature combination unit 14 and are combined by the feature combination unit 5 to form the feature pattern. It is made (S14).
As a coupling method in this case, a one-dimensional vector arrangement is adopted.
Necessary data such as a registrant name, ID number, date, and the like are added to the combined feature pattern and registered in the collation dictionary unit 17 (S15).
[0048]
Next, the identification mode will be described.
Also in this mode, it is assumed that a plurality of person feature patterns are registered in the collation dictionary unit 17.
Further, in this identification mode, the processing from the video input of S1 to the encoding of S13 is performed in the same manner as in the first and second embodiments and the registration mode of FIG. 7, and the characteristic combination of S14 is also shown. Since this is performed in the same manner as in the case of the registration mode in FIG.
[0049]
The feature pattern of the iris pattern and the facial pattern combined by the feature combining unit 14 is sent to the collation unit 16, and the collation unit 16 collates and calculates the feature pattern registered in the collation dictionary unit 17. A similarity is calculated (S15).
In this case, a distance measure such as a Hamming distance is also used for the collation calculation.
The similarity (distance value) calculated by the collation unit 16 is sent to the comprehensive determination unit 15, and the comprehensive determination unit 15 obtains the similarity obtained by the collation unit 16 in advance by experiments or statistical methods. Alternatively, it is compared with a plurality of threshold values, and it is determined whether or not the feature pattern obtained from the input video is the same person as the feature pattern collated in the collation dictionary unit 17 (S16). Is output.
[0050]
The following two types of forms are also adopted in the verification processing in the identification mode of FIG. 8 described above.
One of them is a case where a feature pattern to be collated is known among the feature patterns registered in the collation dictionary unit 17, that is, one-to-one collation. The other one is that the feature pattern to be collated is Of course, the most probable feature pattern is selected from the registered feature patterns of a plurality of persons, that is, one-to-n matching.
[0051]
In the case of one-to-one collation, the collation unit 16 calculates the similarity using one feature pattern registered in the collation dictionary unit 17 and outputs it to the comprehensive determination unit 15.
In the determination by the comprehensive determination unit 15, the similarity is compared with one or a plurality of thresholds, and it is determined whether the feature pattern obtained from the input video is the same person as the collated feature pattern, that is, individual identification And output the result.
[0052]
In the case of one-to-n collation, the collation unit 16 performs collation calculation using a plurality of feature patterns registered in the collation dictionary unit 17, and the first rank or the upper m rank (m is an arbitrary value). Are output to the overall determination unit 15.
In this case, the determination by the comprehensive determination unit 15 compares the similarity degree of the first rank or the top m rank with one or a plurality of threshold values, and collates the feature pattern obtained from the image of the identified person. Depending on which feature pattern is most similar or not sufficiently similar, it is determined whether the feature pattern belongs to the same person, that is, individual identification is performed, and the result is output.
[0053]
According to the third embodiment described above, a stable and high recognition rate can be obtained in the same manner as in the first example. In addition, in this third embodiment, the collation unit and the collation dictionary unit are respectively Since only one is required, there is an advantage that the configuration is simplified.
FIG. 9 is a block diagram showing a configuration of the fourth embodiment.
In this embodiment, instead of the camera 1 and the camera output distribution unit 11 in the first embodiment, an iris camera 18 and a facial camera 19 are used, and the iris camera 18 is preprocessed by the iris recognition unit 12. Further, the facial camera 19 is connected to the preprocessing unit 7 of the facial recognition unit 13 and the other configuration is the same as that of the first embodiment.
[0054]
Therefore, in this embodiment, in the registration mode and the identification mode, the image of the person to be identified is photographed by the iris camera 18 and the facial camera 19, respectively, and sent separately to the preprocessing units 2 and 7. Is performed in the same manner as in the first embodiment.
In the fourth embodiment as described above, the same effect as that of the first embodiment can be obtained, and the camera for image input is divided into two for iris recognition and for facial recognition. A camera adapted for recognition can be used.
[0055]
For example, a high-resolution and narrow-angle camera can be used for iris recognition, and a low-resolution and wide-angle camera can be used for facial recognition. Therefore, compared to the case of using one expensive and large camera, The cost and size can be reduced without degrading the identification performance of the device.
FIG. 10 is a block diagram showing the configuration of the fifth embodiment, and FIG. 11 is a block diagram showing the configuration of the sixth embodiment.
[0056]
In both of these embodiments, instead of the camera 1 and the camera output distribution unit 11 in the second and third embodiments, an iris camera 18 and a facial camera 19 are used, and the iris camera 18 is replaced with an iris recognition unit 12. The facial processing camera 19 is connected to the preprocessing unit 7 of the facial recognition unit 13, and other configurations are the same as those of the second and third embodiments.
[0057]
Accordingly, in both the embodiments, the image of the person to be identified is captured by the iris camera 18 and the facial camera 19 in the registration mode and the identification mode, respectively, and sent separately to the preprocessing units 2 and 7, but thereafter Processing is performed in the same manner as in the second and third embodiments.
In the fifth and sixth embodiments as described above, the same effects as those of the second and third embodiments can be obtained, and the video input camera can be used for iris recognition as in the fourth embodiment. By dividing the camera into two for facial recognition, a camera suitable for each recognition can be used, so that the price and size can be reduced without degrading the identification performance of the apparatus.
[0058]
【The invention's effect】
As described above, according to the present invention, the iris pattern and facial pattern obtained by photographing the identified person and extracting features from the image are collated with the previously registered iris pattern and facial pattern. Each similarity is obtained, and the similarity of the iris pattern and the similarity of the facial pattern are combined, and the iris pattern and facial pattern obtained from the image of the person to be identified based on the combination result. It is determined whether or not the screen belongs to the same person as the iris pattern and facial pattern used for collation.
[0059]
Therefore, according to this, facial recognition is performed in addition to iris recognition, and individual identification is performed by using the similarity of both recognitions. Without any strict restrictions and without using a camera with a large number of pixels, it is possible to perform reliable identification and obtain an effect of obtaining a high identification rate.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a first embodiment of a personal identification device according to the present invention.
FIG. 2 is a flowchart showing the operation of the first embodiment.
FIG. 3 is a flowchart showing the operation of the first embodiment.
FIG. 4 is a block diagram showing a second embodiment of a personal identification device according to the present invention.
FIG. 5 is a flowchart showing the operation of the second exemplary embodiment.
FIG. 6 is a block diagram showing a third embodiment of the personal identification device according to the present invention.
FIG. 7 is a flowchart showing the operation of the third exemplary embodiment.
FIG. 8 is a flowchart showing the operation of the third exemplary embodiment.
FIG. 9 is a block diagram showing a fourth embodiment of a personal identification device according to the present invention.
FIG. 10 is a block diagram showing a fifth embodiment of a personal identification device according to the present invention.
FIG. 11 is a block diagram showing a sixth embodiment of a personal identification device according to the present invention.
FIG. 12 is a block diagram showing a conventional technique.
FIG. 13 is a flowchart showing the operation of the prior art.
FIG. 14 is a flowchart showing the operation of the prior art.
[Explanation of symbols]
1 Camera
2 Pre-processing section
3 Feature extraction unit
4 verification part
5 collation dictionary
7 Pre-processing section
8 Feature extraction unit
9 Verification part
10 collation dictionary
11 Camera output distributor
12 Iris recognition unit
13 Facial recognition part
14 Similarity coupling part
15 Comprehensive judgment part
16 Verification part
17 collation dictionary
18 Iris camera
19 Facial camera

Claims (10)

The iris recognition unit and the facial recognition unit extract features from the image of the identified person photographed by the camera to acquire the iris pattern and facial pattern, and the iris pattern and facial pattern are registered in advance. Check the similarity with each of the iris pattern and facial pattern,
The similarity combiner combines the similarity of the iris pattern and the similarity of the facial pattern, and the iris pattern and facial pattern obtained from the image of the identified person based on the combination result are used for collation. A personal identification device characterized in that the comprehensive judgment unit judges whether or not the same iris pattern and facial pattern are used. The personal identification device according to claim 1,
A camera to shoot people,
A pre-processing unit that extracts an iris from an image captured by the camera, a feature extraction unit that extracts an iris feature extracted by the pre-processing unit to obtain an iris pattern, and an iris pattern obtained by the feature extraction unit at the time of registration An iris recognition unit having a collation dictionary unit to be registered, and a collation unit that collates an iris pattern obtained by the feature extraction unit at the time of identification with an iris pattern registered in the collation dictionary unit to obtain a similarity;
Feature extraction to obtain a down - the camera processing unit before detecting the like thereof relative relationship between the indicated area of contours and features of the face from the photographed image, facial pattern by extracting a feature of the facial detected by the preprocessing unit A collation dictionary unit for registering the facial pattern obtained by the feature extraction unit at the time of registration, and a similarity degree obtained by collating the facial pattern obtained by the feature extraction unit at the time of identification with the facial feature registered in the collation dictionary unit A facial recognition unit having a matching unit for obtaining
A similarity combining unit that combines the similarity of the iris pattern and the similarity of the facial pattern sent from the respective matching units of the iris recognition unit and the facial recognition unit;
Whether the iris pattern and facial pattern obtained from the image of the person to be identified based on the result of the combination sent from the similarity combination unit is the same person as the iris pattern and facial pattern used for collation A personal identification device comprising a comprehensive judgment unit for judging whether or not. The iris recognition unit and the facial recognition unit extract features from the image of the identified person photographed by the camera to acquire the iris pattern and facial pattern, and the iris pattern and facial pattern are registered in advance. Check the similarity with each of the iris pattern and facial pattern,
The comprehensive determination unit determines whether the iris pattern obtained from the image of the identified person is the same person as the iris pattern used for the collation based on the similarity of the iris pattern.
If they are not the same person, the similarity combining unit combines the similarity of the iris pattern and the similarity of the facial pattern, and the iris pattern and the face obtained from the image of the person to be identified based on the combination result A personal identification device characterized in that a comprehensive judgment unit judges whether or not the pattern is the same person as the iris pattern and facial pattern used for collation. The personal identification device according to claim 3,
A camera to shoot people,
A pre-processing unit that extracts an iris from an image captured by the camera, a feature extraction unit that extracts an iris feature extracted by the pre-processing unit to obtain an iris pattern, and an iris pattern obtained by the feature extraction unit at the time of registration An iris recognition unit having a collation dictionary unit to be registered, and a collation unit that collates an iris pattern obtained by the feature extraction unit at the time of identification with an iris pattern registered in the collation dictionary unit to obtain a similarity;
Feature extraction to obtain a down - the camera processing unit before detecting the like thereof relative relationship between the indicated area of contours and features of the face from the photographed image, facial pattern by extracting a feature of the facial detected by the preprocessing unit A collation dictionary unit for registering a facial pattern obtained by the feature extraction unit at the time of registration, and a facial pattern obtained by the feature extraction unit at the time of identification with a facial pattern registered in the collation dictionary unit A facial recognition unit having a matching unit for obtaining a degree;
A similarity combining unit that combines the similarity of the iris pattern and the similarity of the facial pattern sent from the matching unit of each of the iris recognition unit and the facial recognition unit,
It is determined whether or not the iris pattern obtained from the image of the identified person is the same person as the iris pattern used for matching, based on the similarity of the iris pattern sent from the matching unit of the iris recognition unit,
If they are not of the same person, the similarity of the iris pattern and the similarity of the facial pattern sent from the collation unit of the facial recognition unit are combined by the similarity combining unit, and based on the combination result And a comprehensive judgment unit for judging whether or not the iris pattern and facial pattern obtained from the image of the identified person belong to the same person as the iris pattern and facial pattern used for collation. Personal identification device. In the personal identification device according to claim 1 or 3,
Iris pattern similarity and facial pattern similarity
T = k * R _I + (1−k) * R _f
T: Overall similarity R _I : Iris similarity R _f : Facial similarity k: Constant (1> k)
The personal identification device characterized by combining by the formula. The feature combining unit combines the iris pattern and the facial pattern obtained by extracting the features of the iris recognition unit and the facial recognition unit from the image of the identified person photographed by the camera to create a characteristic pattern,
This feature pattern is collated with the feature pattern registered in advance by the collation unit to obtain the similarity,
A personal identification device characterized in that a comprehensive judgment unit judges whether or not a feature pattern obtained from an image of a person to be identified is the same person as the feature pattern used for collation based on the similarity . The personal identification device according to claim 6,
A camera to shoot people,
An iris recognition unit having a pre-processing unit that extracts an iris from an image captured by the camera, and a feature extraction unit that extracts an iris feature extracted by the pre-processing unit to obtain an iris pattern;
Processing unit prior to detection of the relative relationship among the region indicating the contours and features of the face from the image the camera is taken, and extracted with facial pattern features of facial detected by the preprocessing section - characterized to obtain a down A facial recognition unit having an extraction unit;
A feature combining unit that combines the iris pattern and the facial pattern sent from the feature extraction unit of each of the iris recognition unit and the facial recognition unit, and creates a characteristic pattern;
A collation dictionary part for registering the feature pattern created by this feature combination part at the time of registration;
A collation unit for collating the feature pattern sent from the feature coupling unit at the time of identification with the feature pattern registered in the collation dictionary unit to obtain a similarity;
A comprehensive determination unit for determining whether or not the feature pattern obtained from the image of the identified person is the same person as the feature pattern used for verification based on the similarity sent from the verification unit; A personal identification device characterized by that. The personal identification device according to claim 6,
A feature identification unit is a personal identification device that combines an iris pattern and a facial pattern with a vector. In the personal identification device according to claim 2, claim 4, or claim 7,
A personal identification apparatus comprising a camera for photographing a person and a camera output distribution unit for distributing the output of the camera to both an iris recognition unit and a facial recognition unit. In the personal identification device according to claim 2, claim 4, or claim 7,
Equipped with an iris camera and facial camera, the image taken with the iris camera is sent to the iris recognition unit,
Personal identification device characterized by sending the facial recognition unit an image captured by the face 貌用 camera.

Images