Information processing system, information processing device, information processing method, and storage medium

Abstract

This information processing system comprises: an extraction unit that extracts iris information to be used for iris authentication from an iris image; and a quality score calculation unit that calculates a quality score indicating the quality of the iris image on the basis of the extracted iris information.

Description

Information processing system, information processing device, information processing method, and recording medium

[0001] The present disclosure relates to an information processing system, an information processing device, an information processing method, and a recording medium.

[0002] Iris authentication is generally used to authenticate a person using an iris image. In general iris authentication, for example, a matching is performed using information extracted from a registered iris image and an iris image for authentication. Depending on the result of the matching, for example, authentication is performed to determine whether the person shown in the iris image for authentication is the same as the person shown in the registered iris image.

[0003] For example, Patent Document 1 discloses a threshold calculation system including a first acquisition means, a second acquisition means, a storage means, a sample extraction means, a population estimation means, and a threshold calculation means. The first acquisition means acquires matching information used for biometric matching. The second acquisition means acquires attribute information indicating attributes of the biometric or success information. The storage means stores the matching information and attribute information for each biometric. The sample extraction means extracts multiple pieces of matching information as sample data from the storage means based on predetermined conditions related to the attribute information. The population estimation means estimates a population from the sample data. The threshold calculation means calculates a threshold related to the matching information based on the distribution of the estimated population.

[0004] Furthermore, Patent Document 1 describes that a storage means stores, as matching information, at least one of features extracted from an image including a living body and a matching score obtained by comparing the features between living bodies.

[0005] International Publication No. 2022 / 070349

[0006] The present disclosure aims to improve upon the techniques described in the prior art documents mentioned above.

[0007] The information processing system according to the present disclosure includes an extraction unit that extracts iris information to be used for iris authentication from an iris image, and a calculation unit that calculates a quality score indicating the quality of the iris image based on the extracted iris information.

[0008] The information processing device according to the present disclosure includes an extraction unit that extracts iris information to be used for iris authentication from an iris image, and a calculation unit that calculates a quality score indicating the quality of the iris image based on the extracted iris information.

[0009] In the information processing method disclosed herein, one or more computers extract iris information to be used for iris authentication from an iris image, and calculate a quality score indicating the quality of the iris image based on the extracted iris information.

[0010] The recording medium in the present disclosure is a recording medium having recorded thereon a program for causing one or more computers to extract iris information to be used for iris authentication from an iris image, and calculate a quality score indicating the quality of the iris image based on the extracted iris information.

[0011] 1 is a block diagram showing a configuration of a first information processing system according to the present disclosure. FIG. 1 is a block diagram showing a configuration of a first information processing device according to the present disclosure. FIG. 2 is a flowchart showing a processing operation of the first information processing device according to the present disclosure. FIG. 3 is a block diagram showing a detailed example of the configuration of the first information processing device according to the present disclosure. FIG. 4 is a flowchart showing a detailed example of the processing operation of the first information processing device according to the present disclosure. FIG. 5 is a diagram showing an example of the position of an iris image in a feature space according to the present disclosure. FIG. 6 is a block diagram showing a configuration of a first authentication processing unit according to the present disclosure. FIG. 7 is a flowchart showing the processing operation of the first authentication processing unit according to the present disclosure. FIG. 8 is a block diagram showing a physical configuration of the first information processing device according to the present disclosure. FIG. 9 is a block diagram showing a configuration of a first extraction unit according to the present disclosure. FIG. 10 is a flowchart showing the processing operation of the first extraction unit according to the present disclosure. FIG. 11 is a block diagram showing a configuration of a second information processing device according to the present disclosure. FIG. 12 is a flowchart showing the processing operation of a first learning unit according to the present disclosure. FIG. 13 is a flowchart showing a detailed example of the processing operation of the first learning unit according to the present disclosure. FIG. 14 is a block diagram showing a configuration of a third information processing device according to the present disclosure. FIG. 15 is a flowchart showing the processing operation of the third information processing device according to the present disclosure. FIG. 16 is a flowchart showing the configuration of a first calculation unit according to the present disclosure. FIG. 17 is a flowchart showing the processing operation of the first calculation unit according to the present disclosure. FIG. 18 is a block diagram showing a configuration of a fourth information processing device according to the present disclosure. Fig. 1 is a flowchart showing processing operations of a fourth information processing device according to the present disclosure. Fig. 2 is a block diagram showing a configuration of a fifth information processing device according to the present disclosure. Fig. 3 is a flowchart showing processing operations of the fifth information processing device according to the present disclosure. Fig. 4 is a block diagram showing a configuration of a second extraction unit according to the present disclosure. Fig. 5 is a flowchart showing processing operations of the second extraction unit according to the present disclosure.

[0012] Hereinafter, in this disclosure, the drawings relate to one or more embodiments. In addition, in all drawings, similar components are given similar reference numerals and descriptions thereof will be omitted as appropriate.

[0013] [Embodiment 1] (Overview)

[0014] According to Patent Literature 1, the smaller the threshold used in biometric authentication, the higher the false acceptance rate (FAR), while the larger the threshold used in biometric authentication, the higher the false rejection rate (FRR). Therefore, it is important to set an appropriate threshold when performing biometric authentication. Furthermore, since the calculated threshold varies depending on the distribution of sample data, if the sample data used to calculate the threshold is not appropriately extracted, the calculated threshold may be an inappropriate value.

[0015] Patent Document 1 describes the following technical effects of the threshold calculation system: Sample data for calculating a threshold is extracted based on population conditions related to attribute information. That is, appropriate sample data is extracted taking the attribute information into consideration. Using appropriate sample data, the population can be appropriately estimated. Appropriately estimating the population means that the distribution of the population including unknown others that are not stored can also be appropriately estimated. As a result, an appropriate threshold that assumes unknown others can be calculated from the estimated population distribution.

[0016] In general iris authentication, as described above, a registered iris image and an iris image for authentication are compared using information extracted from each, and authentication is performed based on the comparison result. If these iris images have low resolution or contain focus blur, motion blur, eyeglass reflection, noise, etc., authentication accuracy generally decreases. Therefore, in iris authentication, it is desirable to use an iris image that can be used for accurate authentication.

[0017] Here, the authentication accuracy may be expressed using, for example, at least one index related to an error rate in authentication. Examples of such indices include a false acceptance rate, a false rejection rate, an equal error rate (ERR), an are-under-the-curve (AUC) score, etc. The false acceptance rate may be, for example, a false acceptance rate (FAR) used in 1:1 authentication or a false positive identification rate (FPIR) used in 1:N authentication. The false rejection rate may be, for example, the FRR (False Reject Rate) used in 1:1 authentication, or the FNIR (False Negative Identification Rate) used in 1:N authentication.

[0018] However, even if Patent Document 1 can calculate an appropriate threshold value to be used for iris authentication, it does not disclose a technique for evaluating an iris image that can be authenticated with high accuracy.

[0019] One of the objectives of the present disclosure is to improve the accuracy of iris authentication.

[0020] As shown in FIG. 1 , the information processing system SYS includes an extraction unit 120 and a calculation unit 130 .

[0021] The extraction unit 120 extracts iris information to be used for iris authentication from the iris image.

[0022] The calculation unit 130 calculates a quality score indicating the quality of the iris image based on the extracted iris information.

[0023] According to this information processing system SYS, by performing iris authentication using a high-quality iris image, it is possible to reduce the error rate in iris authentication, thereby improving the accuracy of iris authentication.

[0024] As shown in FIG. 2, the information processing device 100 includes an extraction unit 120 and a calculation unit 130.

[0025] The extraction unit 120 extracts iris information to be used for iris authentication from the iris image.

[0026] The calculation unit 130 calculates a quality score indicating the quality of the iris image based on the extracted iris information.

[0027] According to the information processing device 100, the error rate in iris authentication can be reduced by performing iris authentication using a high-quality iris image, thereby improving the accuracy of iris authentication.

[0028] The information processing device 100 executes information processing as shown in FIG.

[0029] The extraction unit 120 extracts iris information to be used for iris authentication from the iris image (step S120).

[0030] The calculation unit 130 calculates a quality score indicating the quality of the iris image based on the extracted iris information (step S130).

[0031] According to this information processing, by performing iris authentication using a high-quality iris image, the error rate in iris authentication can be reduced, thereby improving the accuracy of iris authentication.

[0032] A detailed example of the information processing system SYS will be described below.

[0033] (Detailed Example) The information processing system SYS includes, for example, an information processing device 100 .

[0034] (Detailed Example of Configuration of Information Processing Device 100) As shown in FIG. 4, for example, the information processing device 100 includes an acquisition unit 110, an extraction unit 120, a calculation unit 130, and an authentication processing unit 140.

[0035] The acquisition unit 110 acquires an iris image.

[0036] The extraction unit 120 extracts iris information to be used for iris authentication from the iris image acquired by the acquisition unit 110 .

[0037] The calculation unit 130 calculates a quality score indicating the quality of the iris image based on the iris information extracted by the extraction unit 120 .

[0038] The authentication processing unit 140 performs authentication processing based on the iris information extracted by the extraction unit 120 and the quality score calculated by the calculation unit 130 .

[0039] (Detailed Example of Processing Operation of Information Processing Device 100) The information processing device 100 executes information processing as shown in FIG. 5, for example.

[0040] The acquisition unit 110 acquires an iris image (step S110).

[0041] The extraction unit 120 extracts iris information to be used for iris authentication from the iris image acquired in step S110 (step S120).

[0042] The calculation unit 130 calculates a quality score indicating the quality of the iris image based on the iris information extracted in step S120 (step S130).

[0043] The authentication processing unit 140 performs authentication processing based on the iris information extracted in step S120 and the quality score calculated in step S130 (step S140).

[0044] (Regarding Acquisition Unit 110) Acquisition unit 110 acquires at least one iris image from a camera, another information processing device, etc. (not shown) via a wired, wireless, or a combination thereof network. Acquisition unit 110 may acquire at least one iris image from a storage unit (not shown) that stores one or more iris images in advance.

[0045] The acquisition unit 110 may acquire a plurality of iris images.

[0046] (Regarding Iris Images) An iris image is, for example, an image including a person's iris. An iris image is obtained by capturing an image of an area including a person's iris.

[0047] Note that the iris image may include an iris, but may also include other parts than the iris. For example, the iris image may be an iris region image showing only the iris region, a monocular image including one predetermined eye of the left or right eye, a binocular image including both eyes, a facial image including a face, an image including the upper body, an image including the whole body, etc. Furthermore, the iris image is not limited to a human, and may be an image including the iris of an animal such as a dog or a snake.

[0048] (Regarding the Extraction Unit 120) As described above, the extraction unit 120 extracts iris information from an iris image. The extraction unit 120 may include, for example, one or more engines for extracting iris information from an iris image, and may use these engines to extract iris information from the iris image. A machine learning model using a trained neural network may be applied as a technology for constructing the engine, or a general technology other than a machine learning model may also be applied. Examples of this machine learning model include, but are not limited to, a keypoint extraction model and a feature vector extraction model, which will be described later.

[0049] Iris information is information used for authentication using an iris image (i.e., iris authentication). The iris information is, for example, information indicating characteristics of an iris region included in an iris image. The iris region is a region of the iris image that indicates the iris.

[0050] In particular, for example, the iris information includes at least one of an iris feature vector and an iris keypoint, but does not necessarily include the iris image itself used to extract them.

[0051] It should be noted that the iris information does not exclude the iris image used to extract it from being associated with the iris information, for example, in processing, storing, etc., of the iris information.

[0052] The iris feature vector is a vector that indicates the features of the iris included in the iris image. For example, the iris feature vector is a vector that indicates the features of the iris pattern. Alternatively, for example, the iris feature vector may be a vector that includes information that identifies an individual.

[0053] The iris feature vector is, for example, a feature vector extracted from an iris region included in an iris image. An example of a configuration for extracting the iris feature vector will be described later in another embodiment.

[0054] Here, a vector may be composed of one or more values, and the same applies hereinafter. Each value constituting a vector is typically a numerical value. When a vector is composed of one value, the vector represents a scalar quantity.

[0055] The iris keypoints are information about characteristic points (predetermined locations) on the iris included in the iris image, for example, vectors that indicate the geometric characteristics of the points (predetermined locations) on the iris.

[0056] In detail, for example, the iris keypoint includes a parameter indicating at least one of the shape, position, size, etc. of a predetermined portion on the iris. The predetermined portion may be, for example, at least one of the iris, pupil, upper and lower eyelids, etc. The iris keypoint may include, for example, a parameter for specifying at least one shape, etc. of the outline or edge of the iris, the outline or edge of the pupil, the upper and lower eyelids, etc. The parameter for specifying the shape, etc. may be, for example, a parameter for specifying a circle, arc, ellipse, elliptical arc, curve, etc. representing the outline or edge of the iris, the outline or edge of the pupil, or the position, shape, size, etc. of the upper and lower eyelids.

[0057] Furthermore, for example, the iris key point may include coordinates indicating the position of a predetermined point related to at least one of the iris, pupil, upper and lower eyelids, etc. The point related to the iris may be, for example, the iris center, or one or more points may be determined on the outline or edge of the iris. The point related to the pupil may be, for example, the pupil center, or one or more points may be determined on the outline or edge of the pupil. The predetermined point related to the upper and lower eyelids may be, for example, one or more points determined on at least one of the upper and lower eyelids.

[0058] The iris keypoints are, for example, information about keypoints extracted from an iris image. An example of a configuration for extracting iris keypoints will be described later in another embodiment.

[0059] It should be noted that the iris information is not limited to the example given here.

[0060] (Regarding the calculation unit 130) As described above, the calculation unit 130 calculates the quality score based on the iris information. For example, the calculation unit 130 calculates the quality score using one or both of the iris feature vector and the iris keypoints included in the iris information.

[0061] That is, for example, the iris information may include an iris feature vector indicating features of the iris included in the iris image. In this case, the calculation unit 130 may calculate the quality score using the iris feature vector. That is, the quality score may be calculated using the iris feature vector.

[0062] Furthermore, for example, the iris information may include iris keypoints related to characteristic points of the iris included in the iris image. In this case, the calculation unit 130 may calculate the quality score using the iris keypoints. That is, the quality score may be calculated using the iris keypoints.

[0063] The iris information may include an iris feature vector indicating features of the iris included in the iris image and iris keypoints related to characteristic points of the iris included in the iris image. In this case, the calculation unit 130 may calculate the quality score using the iris feature vector and the iris keypoints. That is, the quality score may be calculated using the iris feature vector and the iris keypoints.

[0064] (Quality Score) The quality score is a vector that indicates the quality of an iris image. For example, the larger the value constituting the quality score, the higher the quality of the iris image.

[0065] The quality score is not limited to this, and may be a value in which, for example, the smaller the value, the higher the quality of the iris image. Furthermore, each value constituting the quality score is typically a numerical value, but may be, for example, one or a combination of numerical values, letters, symbols, etc., as long as it is defined so that the magnitude or high / low can be compared according to a predetermined standard.

[0066] The quality of the iris image may be defined according to, for example, the accuracy of iris authentication. That is, the quality score may be a score according to the accuracy of iris authentication. In detail, for example, the quality score of an iris image may be composed of one or more values ​​such that the higher the quality indicated by the quality score (for example, the larger the values ​​included in the quality score), the higher the accuracy of iris authentication using the iris image.

[0067] The index representing the accuracy of iris authentication may be, for example, at least one index relating to the error rate in iris authentication.

[0068] Examples of indices relating to the error rate include FAR, FRR, ERR, AUC score, etc. However, the indices relating to the error rate are not limited to these.

[0069] Generally, the error rate indicator values ​​are obtained by verifying the authentication results using the iris image. The lower the error rate indicated by these indicator values, the higher the accuracy of the iris authentication using the iris image. Therefore, the quality score of an iris image may be composed of one or more values ​​such that the higher the quality indicated by the indicator values ​​(e.g., the larger the values ​​included in the quality score), the lower the error rate indicated by the indicator values ​​obtained from the authentication results using the iris image.

[0070] In this way, the quality score may be any score that corresponds to the accuracy of iris authentication (i.e., for example, the error rate). Below, quality score examples 1 to 4 are described. For the quality score, any one of the examples described below may be used, or an appropriate combination of multiple examples described below may be used.

[0071] That is, the quality score of the iris image may include at least one of (Example 1) to (Example 4). (Example 1) An index value that indicates the accuracy of iris authentication (Example 2) A value based on the degree of deviation between the position of the iris image in the feature space that represents the iris feature vector and a predetermined high-quality position (Example 3) A value based on the norm of the iris feature vector (Example 4) A value based on the degree of deviation between the position of the iris image in the feature space that represents the iris feature vector and a predetermined low-quality position

[0072] The quality scores are not limited to Examples 1 to 4, but may include at least one of (Example 1) to (Example 4).

[0073] (Regarding Quality Score Example 1) In more detail, for example, the quality score of an iris image may be an index value that represents the accuracy of iris authentication when a matching process (described in detail below) is performed using the iris image in question and a reference iris image. The reference iris image is an iris image that serves as a reference. The reference iris image is, for example, an iris image included in personal information described below, or an iris image used to extract an iris feature vector included in registered personal information.

[0074] (Regarding Example 2 of Quality Score) For example, the quality score of an iris image may be a value based on the degree of deviation between the position of the iris image in the feature space and a predetermined high-quality position.

[0075] The feature space is a space representing the iris feature vector, and the same applies hereinafter. The position of the iris image is the position in the feature space of the iris feature vector extracted from the iris image, and the same applies hereinafter.

[0076] The predetermined high quality position in the feature space is the position of the predetermined high quality image in the feature space, i.e., the position in the feature space of the iris feature vector extracted from the high quality image.

[0077] The quality score of an iris image in this example is a value based on the degree of deviation between the position of the iris feature vector (extracted from the iris image) in feature space and a predetermined high quality position.

[0078] The degree of deviation is a value indicating the degree of separation in the feature space, and the same applies hereinafter. The degree of deviation may be at least one of a norm such as the L1 norm or the L2 norm, or a cosine similarity.

[0079] In this case, the quality score of the iris image may be determined such that, for example, the smaller the deviation between the position of the iris image in the feature space and the position of a predetermined high-quality image, the higher the quality. In detail, for example, the quality score of the iris image may be, but is not limited to, the reciprocal of this deviation.

[0080] (Example 1 of Predetermined High-Quality Position) The predetermined high-quality position may be, for example, a position of a high-quality iris image selected by visual inspection.

[0081] (Example 2 of Predetermined High-Quality Position) The predetermined high-quality position may be, for example, a position of an iris image selected from a group of personal image images in a plurality of classes when the plurality of iris images are classified into a plurality of classes using iris feature vectors extracted from the plurality of iris images. The group of personal image images is an image group consisting of one or more iris images labeled with a personal class.

[0082] In this case, the iris image selected from the group of person images may be, for example, the iris image corresponding to the position that has the smallest deviation from the average position of the distribution in the feature space for each of the multiple iris images belonging to the group of person images.

[0083] (Example 3 of Predetermined High-Quality Position) The predetermined high-quality position may be, for example, a position of an iris image selected from a group of person images corresponding to a position where the deviation from the representative positions of other classes in the feature space is equal to or greater than a predetermined threshold.

[0084] The representative position of the other class may be determined by an appropriate method based on the positions in the feature space of iris images belonging to the other class (i.e., iris images other than the group of person images). The representative position of the other class may be, for example, the center of gravity (average) of the positions in the feature space of iris images belonging to the other class.

[0085] The representative position of the other class may be, for example, a position in the feature space of an iris image belonging to the other class that has the greatest degree of deviation from the positions of the iris images belonging to the class other than the other class. The iris images belonging to the class other than the other class may include, for example, a group of images of the person in question, but is not limited to this.

[0086] The representative position of another class may be, for example, the position that has the smallest degree of deviation from the iris images belonging to the other class.

[0087] The method for selecting an iris image from the group of personal images may be the same as the method described above. That is, the iris image selected from the group of personal images may be, for example, the iris image corresponding to the position that has the smallest deviation from the average position of the distribution in the feature space for each of the iris images belonging to the group of personal images.

[0088] (Quality Score Example 3) The quality score may be a value based on the norm (for example, L1 norm, L2 norm) of the iris feature vector included in the iris information.

[0089] Generally, in an iris image that can perform iris authentication with high accuracy (i.e., for example, a low error rate), the norm of the iris feature vector extracted from it often becomes large (see FIG. 6 ). Therefore, a value indicating higher quality may be used as a quality score corresponding to the accuracy of iris authentication (i.e., for example, the error rate), with the larger the norm of the iris feature vector included in the iris information. The value used as such a quality score (i.e., a value based on the norm) may be the norm itself, or may be a value obtained by performing an appropriate calculation on the norm.

[0090] (Quality Score Example 4) For example, the quality score of an iris image may be a value based on the degree of deviation between the position of the iris image in the feature space and the position of a predetermined low-quality image.

[0091] The predetermined low-quality location in feature space is the location of the predetermined low-quality image in feature space. The predetermined low-quality image location in feature space is the location in feature space of an iris feature vector extracted from the low-quality image.

[0092] That is, the quality score of the iris image in this example is a value based on the degree of deviation between the position of the iris feature vector (extracted from the iris image) in the feature space and a predetermined low-quality position.

[0093] In this case, the quality score of the iris image may be determined such that, for example, the greater the degree of deviation between the position of the iris image in the feature space and the position of a predetermined low-quality image, the higher the quality. In detail, for example, the quality score of the iris image is this degree of deviation, but is not limited to this.

[0094] The predetermined low-quality image may be one or more. When there are multiple predetermined low-quality images, the quality score of the iris image may be a value based on the degree of deviation between the position of the iris image in the feature space and the position of at least one of the multiple low-quality images.

[0095] (Example 1 of Low-Quality Position) The low-quality position may be, for example, a position of a low-quality iris image selected by visual inspection.

[0096] In general, the accuracy of iris authentication is often affected by various factors that degrade the iris image, such as those described above. Degradation factors include, for example, the presence or absence or level of blur, focus blur, motion blur, eyeglass reflection, lighting reflection, noise, etc., resolution, contrast, the size of the iris region included in the iris image, occlusion, iris clarity, and the use of hard or colored contact lenses. The size of the iris region depends on whether the entire iris is included in the iris authentication, and is often smaller when the eyes are not fully opened due to blinking, for example. Examples of occlusion include a hat, a hood, hair, eyelashes, etc., covering at least a portion of the iris region included in the iris image.

[0097] Therefore, an iris image that has been degraded due to a certain degradation factor may be visually selected, and the position of the selected iris image may be determined as the low-quality position.

[0098] (Example 2 of Low Quality Location) A low quality location may be determined based on a group of feature vectors corresponding to the low quality location.

[0099] Generally, the positions of images degraded by various degradation factors are located closer to the origin than the positions of high-quality images, as shown in Fig. 6. Furthermore, in the feature space, the positions of degraded iris images often form groups that are close together for each degradation factor.

[0100] In FIG. 6, circles, triangles, and crosses indicate the positions of the iris images of person A, person B, and person C, respectively, in the feature space.

[0101] Therefore, the low-quality position may be, for example, a position representative of a group of iris images degraded by each degradation factor, i.e., a representative position of the group corresponding to the low-quality position. More specifically, for example, one or more low-quality positions may be determined based on the positions of each of the iris images belonging to a group corresponding to one or more degradation factors. As described above, the position of an iris image is the position of a feature vector extracted from the iris image in the feature space. Therefore, such a low-quality position is an example of a low-quality position determined based on a group of feature vectors corresponding to the low-quality position.

[0102] Each of one or more feature vectors constituting the group of feature vectors corresponding to low-quality positions may be, for example, an iris vector that satisfies all of the following conditions 1 to 3. Note that each of one or more feature vectors constituting the group of feature vectors corresponding to low-quality positions may be an iris image that satisfies at least one of the following conditions 1 to 3.

[0103] Condition 1: Among the plurality of feature vectors extracted from the plurality of iris images, the norm is smaller than a predetermined first threshold value.

[0104] Condition 2: The distance in the feature space from the position indicated by another feature vector not belonging to the group of feature vectors is greater than a predetermined second threshold.

[0105] Condition 3: The feature vector indicates a position whose distance in the feature space to the positions indicated by other feature vectors belonging to the feature vector group is smaller than a predetermined third threshold value.

[0106] Here, the third threshold may be smaller than the second threshold.

[0107] For example, the low-quality position may be the center of gravity of each position of the iris images belonging to a group corresponding to a certain degradation factor. In this case, for example, the low-quality position may be the position of a vector formed by the average value of each component of the feature vector group corresponding to the low-quality position.

[0108] For example, the low-quality position may be the position of the iris image that is closest to the origin of the feature space among the positions of the iris image that belong to a group corresponding to a certain degradation factor. In this case, the low-quality position may be the position indicated by the feature vector with the smallest norm among the feature vectors that correspond to the low-quality position.

[0109] The calculation unit 130 may, for example, calculate and store low-quality positions in advance using the above-described method.

[0110] Alternatively, the degraded iris image may be created by image processing. The calculation unit 130 may perform image processing to degrade the iris image to create the degraded iris image. For example, the image processing to degrade the iris image may be one or a combination of Gaussian blur, white noise, resizing (reducing), rotation, contrast change, cutout, inpainting, and the like.

[0111] By using the quality scores described above, for example, it is possible to select a high-quality iris image and perform iris authentication. Therefore, it is possible to perform iris authentication with high accuracy (i.e., with a low error rate, for example). Note that, although examples of the quality scores will be described in other embodiments, the quality scores are not limited to the examples described above and in other embodiments.

[0112] As described above, the accuracy of iris authentication is generally affected by various degradation factors of iris images. However, it is difficult to identify the extent to which each degradation factor affects the accuracy of iris authentication. Furthermore, the extent to which each degradation factor affects the accuracy of iris authentication often depends on the characteristics of one or more specific engines (e.g., machine learning models described below) for extracting iris information used in iris authentication.

[0113] In other words, in iris authentication, it is desirable that the quality (quality score) of an iris image be determined by the relationship between one or more specific engines for extracting iris information and the iris image from which the iris information was extracted, rather than by the degree of each degradation factor in the iris image.

[0114] Therefore, the iris information used to calculate the quality score does not include the iris image itself, but is the iris information used for iris authentication. This means that the iris information used to calculate the quality score is extracted using the same engine or engines as those used to extract the iris information used for iris authentication. In more detail, for example, authentication may be performed using the iris information extracted by the extraction unit 120. Furthermore, for example, if another device (not shown) extracts iris information and performs authentication processing, the extraction unit 120 may include the same engine or engines as those used to extract iris information used for iris authentication in the other device.

[0115] It should be noted that the explanation here does not intend to exclude the fact that the quality of an iris image includes values ​​indicating the degree of each degradation factor in the iris image.

[0116] (Configuration Example 1 of Calculation Unit 130 for Calculating Quality Score) The calculation unit 130 may input iris information to a quality estimation model to calculate a quality score. The quality estimation model is a machine learning model for estimating the quality of an iris image. In other words, when iris information is input, the quality estimation model outputs a quality score of the iris image from which the iris information is extracted.

[0117] The quality estimation model may be configured, for example, by a neural network. In this case, the input to the quality estimation model is iris information. The iris information may be represented using a vector, and can generally be represented with a smaller amount of data than the iris image from which it was extracted. The quality estimation model can be realized by a relatively small-scale neural network. Furthermore, since the iris information is used for iris authentication, it contains information related to the accuracy of the authentication. Therefore, using the iris information, a quality score can be calculated quickly with a relatively small processing load.

[0118] An example of a method for learning the quality estimation model will be described in another embodiment.

[0119] (Configuration Example 2 of the Calculation Unit 130 for Calculating a Quality Score) When the quality score is a deviation or the like, the calculation unit 130 may calculate, for example, the deviation in feature space between each position of iris information extracted from an iris image and iris information of a predetermined high-quality image. Because the deviation is a norm or the like as described above, the quality score can be calculated using a general vector calculation or the like. Also, when the quality score is a value based on the norm of an iris feature vector, the calculation unit 130 can calculate the quality score using a general vector calculation or the like. Such calculation processing often requires even less processing load than calculation processing using a neural network. Furthermore, by using the deviation as exemplified, information related to authentication accuracy can be obtained. Therefore, the quality score can be calculated quickly using iris information with a relatively small processing load.

[0120] (Configuration Example 3 of Calculation Unit 130 for Calculating Quality Score) The calculation unit 130 may calculate, as the quality score, a vector that includes the quality scores described in Configuration Examples 1 and 2 above as components at predetermined positions.

[0121] (Regarding the authentication processing unit 140) As described above, the authentication processing unit 140 performs authentication processing based on the iris information and the quality score. For example, the authentication processing unit 140 selects an iris image or iris information based on the quality score, and performs authentication processing based on the selected iris image or the selected iris information.

[0122] The authentication process is a process performed to authenticate a person. The authentication process may include, for example, at least one of a first authentication process for registering a person and a second authentication process for confirming that the person is the registered person. Note that these authentication processes are examples of processes that use quality scores, and processes that use quality scores are not limited to these.

[0123] As shown in FIG. 7, the authentication processing unit 140 includes a selection unit 141, a person storage unit 142, a registration unit 143, and a matching unit 144, for example.

[0124] The selection unit 141 selects the iris image or iris information based on the quality score.

[0125] The personal information storage unit 142 is a storage unit for storing personal information, which includes at least one of an iris image and an iris feature vector of the person.

[0126] The registration unit 143 performs registration processing based on the selection result of the selection unit 141 .

[0127] The collation unit 144 performs collation processing based on the selection result of the selection unit 141 .

[0128] The authentication processing unit 140 executes a first authentication process (step S140a) as shown in FIG. 8, for example.

[0129] The selection unit 141 selects an iris image or iris information based on the quality score (step S140a_1).

[0130] The registration unit 143 performs a registration process based on the selection result of the selection unit 141 (step S140a_2).

[0131] The authentication processing unit 140 executes a second authentication process (step S140b) as shown in FIG. 9, for example.

[0132] The selection unit 141 selects the iris image or iris information based on the quality score (step S140b_1).

[0133] The collation unit 144 performs a collation process based on the selection result of the selection unit 141 (step S140b_2).

[0134] (Regarding the Selector 141) As described above, the selector 141 selects an iris image or iris information based on the quality score. The number of iris images or iris information to be selected may be one or more.

[0135] (Example 1 of Selection Method) For example, when the quality score is a scalar quantity, the selection unit 141 may compare the quality score with a predetermined threshold and select an iris image or iris information based on the comparison result.

[0136] In detail, for example, the selection unit 141 may extract iris information for which a quality score indicating a higher quality than a threshold has been calculated, or the iris image from which the iris information was extracted. That is, for example, the larger the quality score, the higher the quality of the iris image. In this case, the selection unit 141 may extract iris information for which a quality score greater than the threshold has been calculated, or the iris image from which the iris information was extracted.

[0137] (Selection Method Example 2) For example, when the quality score is a vector composed of multiple values, the selection unit 141 may calculate a weighted sum of the quality scores. The weighted sum is the sum of values ​​obtained by multiplying each of the multiple values ​​constituting the quality score by a predetermined weight. The selection unit 141 may compare the weighted sum with a predetermined threshold and the quality score, and select an iris image or iris information based on the comparison result.

[0138] In detail, for example, the selection unit 141 may extract iris information for which a quality score indicating that the sum is higher in quality than a threshold is calculated, or the iris image from which the iris information is extracted. That is, for example, the larger the sum calculated from the quality score, the higher the quality of the iris image. In this case, the selection unit 141 may extract iris information from which a sum larger than the threshold is calculated, or the iris image from which the iris information is extracted.

[0139] (Selection Method Example 3) For example, when the quality score is a vector composed of multiple values, the selector 141 may compare each of the multiple values ​​constituting the quality score with a predetermined threshold. The threshold here may be, for example, a different value for each component constituting the quality score (i.e., the position of the value in the vector), or the same value for some or all of the components.

[0140] The selection unit 141 may, for example, compare the number of values ​​that indicate a higher quality than a threshold among the multiple values ​​that make up the quality score with a predetermined value, and select the iris image or iris information based on the comparison result. Here, the predetermined value is, for example, half the number of components that make up the quality score, but is not limited to this and may be determined as appropriate.

[0141] In more detail, for example, it is assumed that the sum calculated from the quality scores indicates a higher quality than a threshold value. In this case, the selection unit 141 may extract iris information for which a quality score is calculated in which the number of values ​​exceeding the threshold value is equal to or greater than a predetermined value, or the iris image from which the iris information is extracted.

[0142] As explained in the selection method examples 1 to 3, when the quality score is composed of multiple values, the method for selecting the iris image or iris information can be made more complex than when the quality score is composed of a single value, and therefore, more appropriate iris image or iris information can be selected.

[0143] It should be noted that the selection method, that is, the method for selecting iris images or iris information based on the quality scores, is not limited to the example given here.

[0144] (Regarding the Registration Unit 143) As described above, the registration unit 143 performs registration processing based on the selection result of the selection unit 141.

[0145] For example, the registration unit 143 generates personal information based on the iris image or iris information selected by the selection unit 141. The registration unit 143 may store the generated personal information in the personal information storage unit 142.

[0146] (Regarding the Collation Unit 144) As described above, the collation unit 144 performs collation processing based on the selection result of the selection unit 141.

[0147] For example, the matching unit 144 performs matching based on the iris image or iris information selected by the selection unit 141 and the person's information. Based on the result of the matching, the matching unit 144 may determine whether the iris image or iris information selected by the selection unit 141 is that of the registered person. Common techniques may be used for the methods of matching and determination here.

[0148] The first authentication process (step S140a) and the second authentication process (step S140b) described here are examples of authentication processes, and the authentication processes are not limited to these.

[0149] (Example of physical configuration of information processing device 100) The information processing device 100 physically includes a bus 1010, a processor 1020, a memory 1030, a storage device 1040, a network interface 1050, an input interface 1060, and an output interface 1070, as shown in FIG.

[0150] The bus 1010 is a data transmission path for transmitting and receiving data among the processor 1020, memory 1030, storage device 1040, network interface 1050, input interface 1060, and output interface 1070. However, the method of connecting the processor 1020 and the like to each other is not limited to bus connection.

[0151] The processor 1020 is implemented by a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or the like.

[0152] The memory 1030 is a main storage device realized by a RAM (Random Access Memory) or the like.

[0153] The storage device 1040 is an auxiliary storage device realized by a hard disk drive (HDD), a solid state drive (SSD), a memory card, a read only memory (ROM), or the like. The storage device 1040 stores program modules for realizing the functions of the information processing device 100. The processor 1020 reads each of these program modules into the memory 1030 and executes them to realize the function corresponding to the program module.

[0154] The network interface 1050 is an interface for connecting the information processing device 100 to a network. The network is a communication network for transmitting and receiving information to and from other devices (not shown), and may be wired, wireless, or a combination of these.

[0155] The input interface 1060 is an interface for the user to input information, and is composed of, for example, a touch panel, a keyboard, a mouse, and the like.

[0156] The output interface 1070 is an interface for presenting information to the user, and is configured, for example, by a liquid crystal panel, an organic EL (Electro-Luminescence) panel, or the like.

[0157] The information processing system SYS may be configured with one or more devices that have the functions of the information processing device 100. In this case, the devices may be connected to each other via a network configured by wired, wireless, or a combination of these so as to be able to send and receive information to and from each other. Furthermore, each device in this case may be physically configured in the same way as the information processing device 100, for example.

[0158] (Operations and Effects) As described above, according to this embodiment, the information processing system SYS includes the extraction unit 120 and the calculation unit 130. The extraction unit 120 extracts iris information to be used for iris authentication from an iris image. The calculation unit 130 calculates a quality score indicating the quality of the iris image based on the extracted iris information.

[0159] This allows a high-quality iris image to be selected using the quality score, and the high-quality iris image to be used for iris authentication, thereby reducing the error rate in iris authentication and improving the accuracy of iris authentication.

[0160] Furthermore, since the iris information is used for iris authentication, the processing load for calculating the quality score can be relatively small, making it possible to calculate the quality score quickly with low calculation cost.

[0161] According to this embodiment, the iris information includes an iris feature vector that indicates features of the iris included in the iris image, and the quality score is calculated using the iris feature vector.

[0162] This allows the quality score to be calculated using the iris feature vector used in general iris authentication, making it possible to calculate the quality score quickly with low calculation cost.

[0163] According to this embodiment, the iris information includes iris keypoints related to characteristic points of the iris contained in the iris image, and the quality score is calculated using the iris keypoints.

[0164] This allows the quality score to be calculated using iris keypoints that are commonly used in iris authentication, making it possible to calculate the quality score quickly and with low calculation costs.

[0165] According to this embodiment, the quality score of an iris image includes at least one of (1) to (4): (1) an index value representing the accuracy of iris authentication; (2) a value based on the degree of deviation between the position of the iris image in a feature space representing an iris feature vector and a predetermined high-quality position; (3) a value based on the norm of the iris feature vector; and (4) a value based on the degree of deviation between the position of the iris image in the feature space and a predetermined low-quality position.

[0166] This allows a high-quality iris image to be selected using a quality score according to the accuracy of iris authentication (i.e., for example, the error rate), and then the high-quality iris image is used to perform iris authentication, thereby reducing the error rate in iris authentication and improving the accuracy of iris authentication.

[0167] According to this embodiment, there are a plurality of low-quality positions, and the degree of discrepancy is calculated based on at least one of the plurality of low-quality positions in the feature space and the position indicated by the iris feature vector.

[0168] This allows a high-quality iris image to be selected using a quality score according to the accuracy of iris authentication (i.e., for example, the error rate), and then the high-quality iris image is used to perform iris authentication, thereby reducing the error rate in iris authentication and improving the accuracy of iris authentication.

[0169] According to this embodiment, low-quality positions are determined based on a group of feature vectors corresponding to the low-quality positions. Each of the one or more feature vectors constituting the group of feature vectors is a feature vector extracted from a plurality of iris images, and has a norm smaller than a predetermined first threshold. Furthermore, each of the one or more feature vectors constituting the group of feature vectors indicates a position whose distance in feature space to a position indicated by another feature vector not belonging to the group of feature vectors is greater than a predetermined second threshold. Furthermore, each of the one or more feature vectors constituting the group of feature vectors indicates a position whose distance in feature space to a position indicated by another feature vector belonging to the group of feature vectors is smaller than a predetermined third threshold. The third threshold is smaller than the second threshold.

[0170] This allows a high-quality iris image to be selected using a quality score based on the low-quality position quality, and the high-quality iris image to be used for iris authentication, thereby reducing the error rate in iris authentication and improving the accuracy of iris authentication.

[0171] According to this embodiment, the iris information includes an iris feature vector indicating features of the iris included in the iris image and iris keypoints relating to characteristic points of the iris included in the iris image. The quality score is calculated using the iris feature vector and the iris keypoints.

[0172] This allows a high-quality iris image to be selected using the quality score, and the high-quality iris image to be used for iris authentication, thereby reducing the error rate in iris authentication and improving the accuracy of iris authentication.

[0173] Second Embodiment In this embodiment, a detailed configuration example of the extraction unit 120 will be described.

[0174] As shown in FIG. 11, the extraction unit 120 includes a key point extraction unit 121 and a feature vector extraction unit 122.

[0175] The key point extraction unit 121 extracts, from the iris image acquired by the acquisition unit 110, iris key points relating to characteristic points of the iris contained in the iris image.

[0176] The feature vector extraction unit 122 extracts, from the iris image acquired by the acquisition unit 110, an iris feature vector that indicates the features of the iris included in the iris image.

[0177] The extraction unit 120 executes an extraction process (step S120) as shown in FIG. 12, for example.

[0178] The key point extraction unit 121 extracts iris key points relating to characteristic points of the iris contained in the iris image acquired in step S110 (step S121).

[0179] The feature vector extraction unit 122 extracts, from the iris image acquired in step S110, an iris feature vector that indicates the features of the iris included in the iris image (step S122).

[0180] (Keypoint Extraction Unit 121) As described above, the keypoint extraction unit 121 extracts iris keypoints from an iris image. For example, the keypoint extraction unit 121 may directly estimate iris keypoints from the iris image. For example, the keypoint extraction unit 121 may extract iris keypoints by inputting the iris image into a keypoint extraction model. The keypoint extraction model may be a general machine learning model for extracting iris keypoints from an iris image, or may be, for example, a deep learning model that performs iris keypoint regression. The keypoint extraction model may be one used in general iris authentication.

[0181] In general iris authentication, a two-stage estimation method is sometimes used to estimate iris keypoints from an iris image: segmenting the iris region and estimating iris keypoints from the segmentation results. As described above, by estimating iris keypoints directly from the iris image, it is possible to estimate iris keypoints more quickly than the two-stage estimation method.

[0182] The extraction unit 120 includes a key point extraction unit 121, and is therefore able to extract iris information including iris key points.

[0183] (Feature Vector Extraction Unit 122) As described above, the feature vector extraction unit 122 extracts an iris feature vector from an iris image. For example, the feature vector extraction unit 122 extracts the iris feature vector by inputting the iris image into a feature vector extraction model. The feature vector extraction model is a machine learning model for extracting the iris feature vector from the iris image. The feature vector extraction model may be one used in general iris authentication.

[0184] The extraction unit 120 includes the feature vector extraction unit 122, and is therefore able to extract iris information including an iris feature vector.

[0185] Here, the keypoint extraction model and the feature vector extraction model may both be those used in general iris authentication as described above, and therefore, the keypoint extraction model and the feature vector extraction model for iris authentication can be used to calculate the quality score.

[0186] To calculate the quality score, the extraction unit 120 may include only one of the key point extraction unit 121 and the feature vector extraction unit 122. In this case, the iris information for calculating the quality score may include only one of the iris key points and the iris feature vector, and the calculation unit 130 may calculate the quality score using either the iris key points or the iris feature vector included in the iris information.

[0187] As described above, according to this embodiment, the keypoint extraction unit 121 extracts iris keypoints by inputting an iris image into a keypoint extraction model.

[0188] This allows the extraction unit 120 to be constructed using a keypoint extraction model used in general iris authentication, thereby enabling the quality score to be calculated quickly with low calculation cost.

[0189] According to this embodiment, the feature vector extraction unit 122 extracts an iris feature vector by inputting an iris image into a feature vector extraction model.

[0190] This allows the extraction unit 120 to be constructed using a feature vector extraction model used in general iris authentication, thereby enabling the quality score to be calculated quickly with low calculation cost.

[0191] According to this embodiment, the extraction unit 120 includes a keypoint extraction unit 121 and a feature vector extraction unit 122. The keypoint extraction unit 121 extracts iris keypoints by inputting an iris image into a keypoint extraction model. The feature vector extraction unit 122 extracts iris feature vectors by inputting the iris image into the feature vector extraction model.

[0192] This allows the extraction unit 120 to be constructed using a keypoint extraction model and a feature vector extraction model that are used in general iris authentication, thereby enabling the quality score to be calculated quickly with low calculation cost.

[0193] Third Embodiment As described above, the calculation unit 130 may be configured to include a quality estimation model, which is a machine learning model for estimating the quality of an iris image. In this embodiment, an example of a learning method for the quality estimation model will be described.

[0194] As shown in FIG. 13, the information processing device 200 includes an acquisition unit 110, an extraction unit 120, a calculation unit 130, and an authentication processing unit 140 similar to those in the first embodiment, as well as a learning unit 210.

[0195] The learning unit 210 uses the correct answer data to learn a quality estimation model for evaluating the quality of the iris image.

[0196] The correct answer data includes training iris information extracted from the training iris image and a correct answer quality score for the training iris image.

[0197] The information processing device 200 executes, for example, a learning process as shown in Fig. 14. Note that the learning process may be included in the information processing executed by the information processing device 200.

[0198] The learning unit 210 uses the correct answer data to learn a quality estimation model for evaluating the quality of the iris image (step S210).

[0199] The correct answer data includes training iris information extracted from the training iris image and a correct answer quality score for the training iris image.

[0200] (Learning Unit 210) As shown in FIG. 15, the learning unit 210 may include, for example, a correct answer acquiring unit 210a, a predicted score calculating unit 210b, a loss calculating unit 210c, and a parameter updating unit 210d.

[0201] The correct answer acquisition unit 210a acquires correct answer data. As described above, the correct answer data may include training iris information extracted from the training iris image and the correct answer quality score of the training iris image.

[0202] The predicted score calculation unit 210b calculates a predicted quality score by inputting the training iris information into the quality estimation model being trained.

[0203] The loss calculation unit 210c calculates a loss based on the predicted quality score calculated by the predicted score calculation unit 210b and the correct quality score included in the correct data acquired by the correct answer acquisition unit 210a.

[0204] The parameter update unit 210d updates the parameters included in the quality estimation model being trained using the loss calculated by the loss calculation unit 210c.

[0205] The learning unit 210 may execute a learning process (step S210) as shown in FIG. 16, for example.

[0206] The correct answer acquiring unit 210a acquires correct answer data (step S210a). As described above, the correct answer data may include training iris information extracted from the training iris image and the correct answer quality score of the training iris image.

[0207] The predicted score calculation unit 210b calculates a predicted quality score by inputting the training iris information into the quality estimation model being trained (step S210b).

[0208] The loss calculation unit 210c calculates a loss based on the predicted quality score calculated by the predicted score calculation unit 210b and the correct quality score included in the correct data acquired by the correct answer acquisition unit 210a (step S210c).

[0209] The parameter update unit 210d updates the parameters included in the quality estimation model being trained using the loss calculated by the loss calculation unit 210c (step S210d).

[0210] Here, the correct answer data may be prepared in advance, for example. The training iris information may include at least one of iris keypoints and iris feature vectors extracted from training iris images.

[0211] A common loss function may be used to calculate the loss, such as, but not limited to, a sum of squares error or a cross entropy error.

[0212] For updating the parameters, general optimization methods such as a gradient method, a grid search, or other search methods may be used. The gradient method is a technique in which the gradient of a parameter to be updated is calculated from a loss, and the parameter is updated using the gradient. Note that the parameter update method is not limited to these examples.

[0213] The learning process may be repeatedly executed until, for example, a predetermined termination condition is satisfied. The termination condition may include at least one of, for example, that the loss is equal to or less than a predetermined threshold, that a predetermined number of repetitions has been performed, etc. However, the termination condition is not limited to this.

[0214] The calculation unit 130 may use a quality estimation model that has undergone such learning.

[0215] As described above, according to this embodiment, the information processing device 200 further includes a learning unit 210 that uses supervised data to train a quality estimation model for evaluating the quality of an iris image. The supervised data includes training iris information extracted from the training iris image and a supervised quality score for the training iris image.

[0216] The learning unit 210 includes a correct answer acquisition unit 210a, a predicted score calculation unit 210b, a loss calculation unit 210c, and a parameter update unit 210d.

[0217] The correct answer acquisition unit 210a acquires correct answer data. The predicted score calculation unit 210b calculates a predicted quality score by inputting training iris information into a quality estimation model being trained. The loss calculation unit 210c calculates a loss based on the calculated predicted quality score and the correct quality score included in the correct answer data. The parameter update unit 210d uses the calculated loss to update the parameters included in the quality estimation model being trained.

[0218] This allows a quality estimation model to be constructed and applied to the calculation unit 130 to select a high-quality iris image. By performing iris authentication using such a high-quality iris image, the error rate in iris authentication can be reduced, thereby improving the accuracy of iris authentication.

[0219] [Embodiment 4] In embodiment 2, if the iris information includes both an iris feature vector and iris keypoints, the quality score may be calculated by treating these together, or the quality score may be calculated by treating the iris feature vector and iris keypoints separately.

[0220] In the latter case, the quality score may be calculated, for example, by combining the first score and the second score calculated from the iris feature vector and the iris keypoint, respectively.

[0221] In this embodiment, an example will be described in which, when iris information includes both an iris feature vector and iris keypoints, a quality score is calculated by integrating a first score calculated from the iris feature vector and a second score calculated from the iris keypoints. The first score is a score calculated from the iris feature vector that indicates the quality of the iris image. The second score is a score calculated from the iris keypoints that indicates the quality of the iris image.

[0222] As shown in FIG. 17, the information processing device 300 includes an acquisition unit 110, an extraction unit 120, and an authentication processing unit 140 similar to those in the first embodiment, a learning unit 210 similar to those in the second embodiment, and a calculation unit 330.

[0223] The calculation unit 330 calculates a quality score by integrating the first score and the second score. The first score is a score indicating the quality of the iris image calculated based on the iris feature vector extracted by the extraction unit 120. The second score is a score indicating the quality of the iris image calculated based on the iris keypoints extracted by the extraction unit 120.

[0224] The information processing device 300 executes information processing such as that shown in FIG.

[0225] Steps S110 and S120 are executed in the same manner as in the first embodiment.

[0226] The calculation unit 330 calculates a quality score by integrating the first score and the second score (step S330). The first score is a score indicating the quality of the iris image calculated based on the iris feature vector extracted in step S120. The second score is a score indicating the quality of the iris image calculated based on the iris keypoints extracted in step S120.

[0227] Step S140 is executed in the same manner as in the first embodiment.

[0228] As shown in FIG. 19 , the calculation unit 330 includes a first calculation unit 331 , a second calculation unit 332 , and a score integration unit 333 .

[0229] The first calculation unit 331 calculates a first score indicating the quality of the iris image based on the iris feature vector extracted by the extraction unit 120 (for example, the feature vector extraction unit 122).

[0230] The second calculation unit 332 calculates a second score indicating the quality of the iris image based on the iris keypoints extracted by the extraction unit 120 (for example, the keypoint extraction unit 121).

[0231] The score integrating unit 333 integrates the first score and the second score calculated by the first calculating unit 331 and the second calculating unit 332, respectively, to calculate a quality score.

[0232] The calculation unit 330 executes a calculation process (step S330) as shown in FIG. 20, for example.

[0233] The first calculation unit 331 calculates a first score indicating the quality of the iris image based on the iris feature vector extracted in step S120 (for example, step S122) (step S331).

[0234] The second calculation unit 332 calculates a second score indicating the quality of the iris image based on the iris keypoints extracted in step S120 (for example, step S121) (step S332).

[0235] The score integrating unit 333 integrates the first score and the second score calculated in steps S331 and S332, respectively, to calculate a quality score (step S333).

[0236] (First Calculation Unit 331) The first score may be, for example, a quality score calculated based on iris information including an iris feature vector (for example, including only an iris feature vector) in embodiment 1. Therefore, the first calculation unit 331 corresponds to, for example, the calculation unit 130 that calculates the quality score based on iris information including an iris feature vector (or including only an iris feature vector, or including other than iris key points).

[0237] The first calculation unit 331 may be configured to include a first quality estimation model for estimating the quality of an iris image based on an iris feature vector. The first quality estimation model is a machine learning model that, when an iris feature vector is input, outputs a quality score of the iris image from which the iris feature vector is extracted. In training the first quality estimation model, training iris information including the iris feature vector (e.g., including only the iris feature vector) may be used in the training method shown in FIG. 16 , for example.

[0238] (Second calculation unit 332) The second score may be, for example, a quality score calculated based on iris information including iris keypoints (e.g., including only iris keypoints) in embodiment 1. Therefore, the second calculation unit 332 corresponds to, for example, the calculation unit 130 that calculates the quality score based on iris information including iris keypoints (e.g., including only iris keypoints).

[0239] The second calculation unit 332 may be configured to include a second quality estimation model for estimating the quality of an iris image based on iris keypoints. The second quality estimation model is a machine learning model that, when iris keypoints are input, outputs a quality score of the iris image from which the iris keypoints are extracted. In training the second quality estimation model, training iris information including iris keypoints (e.g., including only iris keypoints) may be used in the training method shown in FIG. 16 , for example.

[0240] (Regarding the score integrating unit 333) The score integrating unit 333 may integrate the first score and the second score using any appropriate method, such as various statistical processes. For example, the score integrating unit 333 may calculate, as the quality score, the sum (weighted sum) of values ​​obtained by multiplying each of the first score and the second score by a predetermined weight. For example, the score integrating unit 333 may calculate, as the quality score, the average value of the first score and the second score. For example, the score integrating unit 333 may calculate, as the quality score, the larger of the first score and the second score.

[0241] (Actions and Effects) As described above, according to this embodiment, the quality score is calculated by combining a first score indicating the quality of the iris image calculated based on the iris feature vector and a second score indicating the quality of the iris image calculated based on the iris keypoints.

[0242] This allows a high-quality iris image to be selected using the quality score, and the high-quality iris image to be used for iris authentication, thereby reducing the error rate in iris authentication and improving the accuracy of iris authentication.

[0243] Fifth Embodiment An information processing device 400 includes an acquisition unit 110, an extraction unit 120, a calculation unit 130, and an authentication processing unit 140 similar to those of the first embodiment, as shown in FIG. 21, and a classification unit 410.

[0244] The classifier 410 uses the iris keypoints to classify the iris image.

[0245] The information processing device 400 executes information processing such as that shown in FIG.

[0246] Steps S110, S120, S130, and S140 are executed in the same manner as in the first embodiment.

[0247] The classification unit 410 classifies the iris image using the iris keypoints (step S410).

[0248] (Regarding the Classification Unit 410) The classification unit 410 may classify the iris image into a plurality of predetermined eye expression classes using, for example, iris key points. The plurality of eye expression classes are classes according to the state of the iris, such as the degree of eye opening, the angle of strabismus, the iris diameter, etc.

[0249] If the iris image is classified into a predetermined class as a result of such classification, the person to be authenticated can be prompted to take a higher quality iris image by displaying a message or the like using, for example, a display unit, a speaker, etc. (not shown). In detail, for example, if the iris image is classified into a class in which the eyes are closed and the iris is small, the person to be authenticated can be prompted to open their eyes by displaying a message or the like to the person to be authenticated.

[0250] As described above, according to this embodiment, the information processing device 400 further includes the classification unit 410 that classifies iris images using iris key points.

[0251] This allows the subject of authentication to be prompted to take a higher quality iris image when the iris image is classified into a predetermined class, thereby improving the accuracy of iris authentication.

[0252] Sixth Embodiment Generally, low-quality iris images are prone to changes in their position in feature space due to slight image fluctuations. Therefore, a quality score can be calculated based on the distribution of positions in feature space indicated by multiple iris feature vectors extracted from multiple iris images that have been artificially shaken by image processing (e.g., the degree of variation expressed by variance, etc.).

[0253] As shown in FIG. 23, the information processing device 500 includes an acquisition unit 110 and an authentication processing unit 140 similar to those in the first embodiment, an extraction unit 520, and a calculation unit 530.

[0254] The extraction unit 520 extracts a plurality of iris feature vectors from each of a plurality of transformed images obtained by performing image processing on the iris image.

[0255] The calculation unit 530 calculates a quality score based on the distribution of positions of the extracted iris feature vectors in a feature space representing the iris feature vectors.

[0256] The information processing device 500 executes information processing such as that shown in FIG.

[0257] Step S110 is executed in the same manner as in the first embodiment.

[0258] The extraction unit 520 extracts a plurality of iris feature vectors from each of a plurality of converted images obtained by performing image processing on the iris image (step S520).

[0259] The calculation unit 530 calculates a quality score based on the distribution of positions of the extracted iris feature vectors in a feature space representing the extracted iris feature vectors (step S530).

[0260] Step S140 is executed in the same manner as in the first embodiment.

[0261] (Extraction Unit 520) As shown in FIG. 25, the extraction unit 520 includes an image conversion unit 521 and a feature vector extraction unit 522.

[0262] The image conversion unit 521 performs image processing on the iris image acquired by the acquisition unit 110 to generate a plurality of converted images.

[0263] The feature vector extraction unit 522 extracts a plurality of iris feature vectors from each of the plurality of transformed images generated by the image transformation unit 521 .

[0264] The extraction unit 520 executes an extraction process (step S520) as shown in FIG. 26, for example.

[0265] The image conversion unit 521 performs image processing on the iris image acquired in step S110 to generate a plurality of converted images (step S521).

[0266] The feature vector extraction unit 522 extracts a plurality of iris feature vectors from each of the plurality of transformed images generated in step S521 (step S522).

[0267] (Regarding the Image Transformation Unit 521) As described above, the image transformation unit 521 performs image processing on the iris image to generate multiple transformed images. Different image processing may be performed on the iris image for each of the multiple transformed images. This image processing may be predetermined processing, such as image processing that does not damage personal information shown in the iris image. In more detail, for example, the image processing may include at least one of luminance distribution transformation, addition of slight white noise, and image scaling.

[0268] (Regarding the feature vector extraction unit 522) As described above, the feature vector extraction unit 522 extracts a plurality of iris feature vectors from each of the plurality of transformed images. The method of extracting the iris feature vector from each of the transformed images may be the same as the method used by the feature vector extraction unit 122 to extract the iris feature vector from the iris image.

[0269] In addition, in order to simultaneously generate a plurality of transformed images and extract iris feature vectors for the plurality of transformed images, the information processing device 500 may include a plurality of image transforming units 521 and a plurality of feature vector extracting units 522 .

[0270] (Regarding the calculation unit 530) The calculation unit 530 calculates a quality score based on, for example, the distribution of the positions of the multiple iris feature vectors extracted by the feature vector extraction unit 522. The positions of the multiple iris feature vectors are positions in a feature space that represents the multiple iris feature vectors. The quality score is, for example, the degree of variation in the positions of the multiple iris feature vectors. In more detail, for example, the quality score is a value obtained by performing statistical processing such as variance of the positions of the multiple iris feature vectors.

[0271] As described above, according to this embodiment, the information processing device 500 includes the extraction unit 520 and the calculation unit 530. The extraction unit 520 extracts a plurality of iris feature vectors from each of a plurality of transformed images obtained by performing image processing on an iris image. The calculation unit 530 calculates a quality score based on the distribution of positions of the extracted iris feature vectors in a feature space representing the plurality of iris feature vectors.

[0272] This allows a high-quality iris image to be selected using a quality score according to the accuracy of iris authentication (i.e., for example, the error rate), and then the high-quality iris image is used to perform iris authentication, thereby reducing the error rate in iris authentication and improving the accuracy of iris authentication.

[0273] Although the present disclosure has been described above with reference to the embodiments, the present disclosure is not limited to the above-described embodiments. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure. Furthermore, each embodiment can be combined with other embodiments as appropriate.

[0274] In addition, although the flowcharts used in the above description show a sequence of steps (processes), the order of steps executed in each embodiment is not limited to the sequence shown in the flowcharts. In each embodiment, the order of steps shown in the diagrams can be changed as long as it does not cause any problems in terms of the content.

[0275] Some or all of the above embodiments can be described as, but are not limited to, the following supplementary notes. 1. An information processing system comprising: extraction means for extracting iris information to be used for iris authentication from an iris image; and calculation means for calculating a quality score indicating the quality of the iris image based on the extracted iris information. 2. The information processing system described in 1., in which the iris information includes an iris feature vector indicating features of the iris included in the iris image, and the quality score is calculated using the iris feature vector. 3. The information processing system described in 1., in which the iris information includes iris keypoints related to characteristic points related to the iris included in the iris image, and the quality score is calculated using the iris keypoints. 4. 4. The information processing system according to any one of 1. to 3., further comprising: learning means for using supervised data to learn a quality estimation model for evaluating the quality of the iris image, wherein the supervised data includes training iris information extracted from a training iris image and a supervised quality score for the training iris image, and the learning means includes: supervised acquisition means for acquiring the supervised data; predicted score calculation means for calculating a predicted quality score by inputting the training iris information into the quality estimation model being trained; loss calculation means for calculating a loss based on the calculated predicted quality score and the supervised quality score included in the supervised data; and parameter update means for updating parameters included in the quality estimation model being trained using the calculated loss. 2. The information processing system described in 2, wherein the quality score of the iris image includes at least one of an index value representing the accuracy of the iris authentication, a value based on the degree of deviation between the position of the iris image in a feature space representing an iris feature vector and a predetermined high-quality position, a value based on the norm of the iris feature vector, and a value based on the degree of deviation between the position of the iris image in the feature space and a predetermined low-quality position. 6. The information processing system described in 5, wherein there are a plurality of low-quality positions, and the degree of deviation is calculated based on at least one of the plurality of low-quality positions in the feature space and a position indicated by the iris feature vector.7. The information processing system described in 5. or 6., wherein the low-quality positions are determined based on a group of feature vectors corresponding to the low-quality positions, and each of the one or more feature vectors constituting the group of feature vectors is a feature vector, among a plurality of feature vectors extracted from a plurality of iris images, whose norm is smaller than a predetermined first threshold, whose distance in the feature space to a position indicated by another feature vector not belonging to the group of feature vectors is greater than a predetermined second threshold, and whose distance in the feature space to a position indicated by another feature vector belonging to the group of feature vectors is smaller than a predetermined third threshold, and the third threshold is smaller than the second threshold. 8. The information processing system described in 3., further comprising classification means for classifying the iris image using the iris keypoints. 9. The iris information includes iris feature vectors indicating features of an iris included in the iris image and iris keypoints related to characteristic points related to the iris included in the iris image, and the quality score is calculated using the iris feature vectors and the iris keypoints. 10. The information processing system described in 9., wherein the quality score is calculated by integrating a first score indicating the quality of the iris image calculated based on the iris feature vector and a second score indicating the quality of the iris image calculated based on the iris keypoints. 11. The information processing system described in 1., wherein the extraction means extracts a plurality of iris feature vectors from each of a plurality of processed images obtained by performing image processing on the iris image, and the calculation means calculates the quality score based on a distribution of positions of the extracted plurality of iris feature vectors in a feature space representing the plurality of iris feature vectors. 12. The information processing device comprising: extraction means for extracting iris information to be used for iris authentication from an iris image; and calculation means for calculating a quality score indicating the quality of the iris image based on the extracted iris information. 13. The information processing device described in 12., wherein the iris information includes an iris feature vector indicating features of an iris included in the iris image, and the quality score is calculated using the iris feature vector.14. The information processing device according to any one of 12. to 14., wherein the iris information includes iris keypoints related to characteristic points related to the iris included in the iris image, and the quality score is calculated using the iris keypoints. 15. The information processing device according to any one of 12. to 14., further comprising: learning means for using correct answer data to train a quality estimation model for evaluating the quality of the iris image, wherein the correct answer data includes training iris information extracted from a training iris image and a correct answer quality score for the training iris image, and the learning means includes: correct answer acquisition means for acquiring the correct answer data, predicted score calculation means for calculating a predicted quality score by inputting the training iris information into the quality estimation model being trained, loss calculation means for calculating a loss based on the calculated predicted quality score and the correct answer quality score included in the correct answer data, and parameter update means for updating parameters included in the quality estimation model being trained using the calculated loss. The information processing device described in 13., wherein the quality score of the iris image includes at least one of an index value representing the accuracy of the iris authentication, a value based on the degree of deviation between the position of the iris image in a feature space representing an iris feature vector and a predetermined high-quality position, a value based on the norm of the iris feature vector, and a value based on the degree of deviation between the position of the iris image in the feature space and a predetermined low-quality position. 17. The information processing device described in 16., wherein there are a plurality of low-quality positions, and the degree of deviation is calculated based on at least one of the plurality of low-quality positions in the feature space and a position indicated by the iris feature vector.18. The information processing device according to 16. or 17., wherein the low-quality positions are determined based on a group of feature vectors corresponding to the low-quality positions, and each of the one or more feature vectors constituting the group of feature vectors is, among a plurality of feature vectors extracted from a plurality of iris images, a feature vector whose norm is smaller than a predetermined first threshold, whose distance in the feature space to a position indicated by another feature vector not belonging to the group of feature vectors is greater than a predetermined second threshold, and whose distance in the feature space to a position indicated by another feature vector belonging to the group of feature vectors is smaller than a predetermined third threshold, and the third threshold is smaller than the second threshold. 19. The information processing device according to 14., further comprising classification means for classifying the iris images using the iris keypoints. 20. The information processing device described in any one of 12. to 19., wherein the iris information includes an iris feature vector indicating features of the iris included in the iris image and iris keypoints related to characteristic points related to the iris included in the iris image, and the quality score is calculated using the iris feature vector and the iris keypoints. 21. The information processing device described in 20., wherein the quality score is calculated by integrating a first score indicating the quality of the iris image calculated based on the iris feature vector and a second score indicating the quality of the iris image calculated based on the iris keypoints. 22. The information processing device described in 12., wherein the extraction means extracts a plurality of iris feature vectors from each of a plurality of processed images obtained by performing image processing on the iris image, and the calculation means calculates the quality score based on the distribution of positions of the extracted plurality of iris feature vectors in a feature space representing the plurality of iris feature vectors. 23. 24. An information processing method in which one or more computers extract iris information to be used for iris authentication from an iris image, and calculate a quality score indicating the quality of the iris image based on the extracted iris information. 25. The information processing method described in 23., in which the iris information includes an iris feature vector indicating characteristics of the iris included in the iris image, and the quality score is calculated using the iris feature vector.25. The information processing method according to 23., wherein the iris information includes iris keypoints related to characteristic points related to the iris included in the iris image, and the quality score is calculated using the iris keypoints. 26. The information processing method according to any one of 23. to 25., further comprising: learning means for using supervised data to train a quality estimation model for evaluating the quality of the iris image, wherein the supervised data includes training iris information extracted from training iris images and a supervised quality score for the training iris images, and the learning means includes: supervised answer acquisition means for acquiring the supervised answer data; predicted score calculation means for calculating a predicted quality score by inputting the training iris information into the quality estimation model being trained; loss calculation means for calculating a loss based on the calculated predicted quality score and the supervised quality score included in the supervised answer data; and parameter update means for updating parameters included in the quality estimation model being trained using the calculated loss. 27. The information processing method according to 24., wherein the quality score of the iris image includes at least one of an index value representing the accuracy of the iris authentication, a value based on the degree of deviation between the position of the iris image in a feature space representing an iris feature vector and a predetermined high-quality position, a value based on the norm of the iris feature vector, and a value based on the degree of deviation between the position of the iris image in the feature space and a predetermined low-quality position. 28. The information processing method according to 27., wherein there are a plurality of low-quality positions, and the degree of deviation is calculated based on at least one of the plurality of low-quality positions in the feature space and a position indicated by the iris feature vector.29. The information processing method according to 27. or 28., wherein the low-quality positions are determined based on a group of feature vectors corresponding to the low-quality positions, and each of one or more feature vectors constituting the group of feature vectors is, among a plurality of feature vectors extracted from a plurality of iris images, a feature vector whose norm is smaller than a predetermined first threshold, whose distance in the feature space to a position indicated by another feature vector not belonging to the group of feature vectors is greater than a predetermined second threshold, and whose distance in the feature space to a position indicated by another feature vector belonging to the group of feature vectors is smaller than a predetermined third threshold, and the third threshold is smaller than the second threshold. 30. The information processing method according to 25., further comprising classification means for classifying the iris images using the iris keypoints. 31. The information processing method according to any one of 23. to 30., wherein the iris information includes an iris feature vector indicating features of the iris included in the iris image and iris keypoints relating to characteristic points related to the iris included in the iris image, and the quality score is calculated using the iris feature vector and the iris keypoints. 32. The information processing method according to 31., wherein the quality score is calculated by integrating a first score indicating the quality of the iris image calculated based on the iris feature vector and a second score indicating the quality of the iris image calculated based on the iris keypoints. 33. The information processing method according to 23., wherein the extraction means extracts multiple iris feature vectors from each of multiple processed images obtained by performing image processing on the iris image, and the calculation means calculates the quality score based on the distribution of positions of the extracted multiple iris feature vectors in a feature space representing the multiple iris feature vectors. 34. A program for causing one or more computers to execute the information processing method according to any one of 23. to 33. 35. A recording medium having recorded thereon a program for causing one or more computers to execute the information processing method according to any one of 23. to 33.

[0276] SYS Information processing system 100, 200, 300, 400, 500 Information processing device 110 Acquisition unit 120, 520 Extraction unit 121 Key point extraction unit 122, 522 Feature vector extraction unit 130, 330, 530 Calculation unit 140 Authentication processing unit 141 Selection unit 141 Sorting unit 142 Personal memory unit 143 Registration unit 144 Matching unit 210 Learning unit 210a Correct answer acquisition unit 210b Predicted score calculation unit 210c Loss calculation unit 210d Parameter update unit 331 First calculation unit 332 Second calculation unit 333 Score integration unit 410 Classification unit 521 Image conversion unit

Claims

1. An information processing system comprising: an extraction means for extracting iris information to be used for iris authentication from an iris image; and a calculation means for calculating a quality score indicating the quality of the iris image based on the extracted iris information.

2. The information processing system according to claim 1, wherein the iris information includes an iris feature vector indicating characteristics of the iris contained in the iris image, and the quality score is calculated using the iris feature vector.

3. The information processing system according to claim 1, wherein the iris information includes iris keypoints relating to characteristic points of the iris contained in the iris image, and the quality score is calculated using the iris keypoints.

4. An information processing system as claimed in any one of claims 1 to 3, further comprising a learning means for using correct answer data to train a quality estimation model for evaluating the quality of the iris image, wherein the correct answer data includes training iris information extracted from a training iris image and a correct answer quality score for the training iris image, and wherein the learning means comprises: correct answer acquisition means for acquiring the correct answer data; predicted score calculation means for calculating a predicted quality score by inputting the training iris information into the quality estimation model being trained; loss calculation means for calculating a loss based on the calculated predicted quality score and the correct answer quality score included in the correct answer data; and parameter update means for updating parameters included in the quality estimation model being trained using the calculated loss.

5. The information processing system of claim 2, wherein the quality score of the iris image includes at least one of an index value representing the accuracy of the iris authentication, a value based on the degree of deviation between the position of the iris image in a feature space representing the iris feature vector and a predetermined high-quality position, a value based on the norm of the iris feature vector, and a value based on the degree of deviation between the position of the iris image in the feature space and a predetermined low-quality position.

6. The information processing system according to claim 5, wherein the low-quality positions are multiple, and the degree of deviation is calculated based on at least one of the multiple low-quality positions in the feature space and the position indicated by the iris feature vector.

7. The information processing system of claim 5 or 6, wherein the low-quality position is determined based on a group of feature vectors corresponding to the low-quality position, and each of one or more feature vectors constituting the group of feature vectors is a feature vector, among a plurality of feature vectors extracted from a plurality of iris images, whose norm is smaller than a predetermined first threshold, whose distance in the feature space from a position indicated by another feature vector not belonging to the group of feature vectors is greater than a predetermined second threshold, and whose distance in the feature space from a position indicated by another feature vector belonging to the group of feature vectors is smaller than a predetermined third threshold, and wherein the third threshold is smaller than the second threshold.

8. The information processing system according to claim 3, further comprising classification means for classifying the iris image using the iris keypoints.

9. An information processing system according to any one of claims 1 to 8, wherein the iris information includes an iris feature vector indicating characteristics of the iris contained in the iris image and iris keypoints relating to characteristic points of the iris contained in the iris image, and the quality score is calculated using the iris feature vector and the iris keypoints.

10. The information processing system of claim 9, wherein the quality score is calculated by combining a first score indicating the quality of the iris image calculated based on the iris feature vector and a second score indicating the quality of the iris image calculated based on the iris keypoints.

11. The information processing system of claim 1, wherein the extraction means extracts a plurality of iris feature vectors from each of a plurality of processed images obtained by performing image processing on the iris image, and the calculation means calculates the quality score based on the distribution of positions of the extracted plurality of iris feature vectors in a feature space representing the plurality of iris feature vectors.

12. An information processing device comprising: an extraction means for extracting iris information to be used for iris authentication from an iris image; and a calculation means for calculating a quality score indicating the quality of the iris image based on the extracted iris information.

13. An information processing method in which one or more computers extract iris information to be used for iris authentication from an iris image, and calculate a quality score indicating the quality of the iris image based on the extracted iris information.

14. A recording medium having recorded thereon a program for causing one or more computers to extract iris information to be used for iris authentication from an iris image, and calculate a quality score indicating the quality of the iris image based on the extracted iris information.

Images