JP7597207B2 - Information processing device and information processing method - Google Patents

Description

The present invention relates to an information processing device that identifies target characters on a card.

Conventionally, methods for determining whether or not an identification document such as a driver's license or an Individual Number card is a forgery have been known that use human eyes or technology such as OCR (Optical Character Recognition) to check the contents of the document. Related technology includes the inventions disclosed in the following Patent Documents 1 and 2.

Patent Document 1 relates to an object verification device, an object verification program, and an object verification method for verifying the authenticity of an object. The object verification device is an object verification device for verifying the authenticity of an object, and is configured to include an image acquisition unit for acquiring an image of the object, a verification area identification unit for identifying a verification area in the object in which the authenticity should be verified, a described portion detection unit for detecting a plurality of described portions in the verification area in the image, and a verification unit for verifying the authenticity of the object by comparing the plurality of described portions in the image with each other.

Patent Document 2 relates to a recognition processing device and a program. The recognition processing device includes a means for recognizing features indicative of a classification of a document among documents of the same type from an image of the document, and a recognition means for applying a recognition process corresponding to the classification indicated by the recognized features to recognition target elements in the image, thereby recognizing the recognition target elements.

Japanese Patent Application Publication No. 2013-188935 Japanese Patent Application Publication No. 2020-042466

The inventions described in Patent Documents 1 and 2 have issues in terms of accurately identifying the characters printed on the face of the card.

One aspect of the present invention has been made in consideration of the above problems, and one example of its objective is to provide technology that can accurately identify characters printed on a card.

An information processing device according to one embodiment of the present invention includes an acquisition means for acquiring an image of a face of a card, an extraction means for extracting one or more target characters from the image, including images surrounding the one or more target characters, and an identification means for performing an identification process on the one or more target characters by inputting the image extracted by the extraction means to a trained model.

An information processing device according to one aspect of the present invention includes a training data acquisition means for acquiring training data including a plurality of pairs of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a card face, and a correct label for the one or more target characters, and a learning means for training a model using the training data, which receives as input an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a card face, and outputs a recognition result for the one or more target characters.

An information processing method according to one aspect of the present invention involves acquiring an image of a card face, extracting one or more target characters from the image, including the image surrounding the one or more target characters, and inputting the extracted image into a trained model, thereby performing a recognition process for the one or more target characters.

An information processing method according to one aspect of the present invention obtains training data including a plurality of pairs of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a certificate face, and a correct label for the one or more target characters; the training data is used to train a model that uses as input the image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a certificate face, and outputs a recognition result for the one or more target characters.

A method for manufacturing a trained model according to one aspect of the present invention involves obtaining training data including a plurality of pairs of images obtained by cutting out one or more target characters together with images surrounding the one or more target characters from an image of a certificate face, and correct labels for the one or more target characters; inputting the image obtained by cutting out one or more target characters together with images surrounding the one or more target characters from an image of a certificate face, and using the training data to train a model that outputs a recognition result for the one or more target characters.

A computer program according to one aspect of the present invention is a computer program for causing a computer to execute an information processing method, which involves acquiring an image of a face of a card, extracting one or more target characters from the image, including images surrounding the one or more target characters, and inputting the extracted images into a trained model, thereby performing a recognition process for the one or more target characters.

A computer program according to one aspect of the present invention is a computer program for causing a computer to execute an information processing method, the information processing method comprising: acquiring training data including a plurality of pairs of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a card face, and a correct label for the one or more target characters; inputting the image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a card face, and using the training data, training a model that outputs a recognition result for the one or more target characters.

According to one aspect of the present invention, the characters printed on the face of the card can be accurately identified.

1 is a block diagram showing a functional configuration of an information processing device according to a first exemplary embodiment of the present invention; 1 is a flow chart showing the flow of an information processing method according to an exemplary embodiment 1 of the present invention. FIG. 11 is a block diagram showing a functional configuration of an information processing device according to a second exemplary embodiment of the present invention. FIG. 11 is a flow chart showing the flow of an information processing method according to an exemplary embodiment 2 of the present invention. FIG. 11 is a block diagram showing a functional configuration of an information processing device according to an exemplary embodiment 3 of the present invention. FIG. 2 is a diagram showing an example of a face of a card. FIG. 11 is a flow chart showing the flow of an information processing method according to an exemplary embodiment 3 of the present invention. FIG. 11 is a block diagram showing a functional configuration of an information processing system according to an exemplary embodiment 4 of the present invention. 10 is a flowchart for explaining a processing procedure of a reading device according to an exemplary embodiment 4 of the present invention. 13 is a flowchart illustrating a processing procedure of a server according to an exemplary embodiment 4 of the present invention. FIG. 1 illustrates an example of the configuration of a computer.

[Example embodiment 1]
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first exemplary embodiment of the present invention will be described in detail with reference to the drawings. This exemplary embodiment is a basic form of the exemplary embodiments described below.

(Configuration of information processing device)
The configuration of an information processing device 10 according to this exemplary embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the functional configuration of the information processing device 10. The information processing device 10 includes an acquisition unit 11, a cutout unit 12, and an identification unit 13.

The acquisition unit 11 acquires an image of the face of the card. The acquisition unit 11 acquires, for example, an image of the face of the card captured by a camera, an image of the face of the card read by a scanner, etc. In this exemplary embodiment, specific examples of the face of the card include, but are not limited to, identification documents such as a driver's license, a My Number card, a passport, and a health insurance card. In this exemplary embodiment, the face of the card refers, in a broad sense, to an object having a predetermined format and including at least one character in the format. Examples of the material of the face on which the characters are printed or arranged include, but are not limited to, paper, resin, or a combination of these.

The size of the face of the card may be larger than an A4 size card, for example, and is not limited to the card size described above.

The cut-out unit 12 cuts out one or more target characters, including the image surrounding the one or more target characters, from the image of the face of the card acquired by the acquisition unit 11. As described below, the image surrounding the one or more target characters is the ruled lines surrounding the one or more target characters, at least one other character, etc.

Here, in this embodiment, a ruled line generally refers to a linear object that exists around the target character. More specifically, in this embodiment, a ruled line is a concept that includes any one or more of the lines on the top, bottom, left, or right of a rectangular frame such as the "Name" or "Type" field of a driver's license, and is not limited to rectangular frames such as the "Name" or "Type" field of a driver's license, but also includes cases where there is a line only above the target character, or a line only below the target character, etc.

The identification unit 13 performs identification processing on one or more target characters by inputting the image cut out by the cut-out unit 12 to the trained model. As described later, the trained model is, for example, a model generated by performing deep learning on a neural network (hereinafter, abbreviated as NN). Here, examples of the neural network include a CNN (Convolutional Neural Network) and an RNN (Recurrent Neural Network). Note that the model is not limited to these configurations, and may be other machine learning such as a SVM (Support Vector Machine), or may be a combination of these other machine learning and a neural network. Note that in this exemplary embodiment, the term "model" is often used simply, but the model may also be expressed as an inference model, an estimation model, an identification model, etc.

(Effects of the information processing device 10)
As described above, in the information processing device 10 according to this exemplary embodiment, the cut-out unit 12 cuts out one or more target characters from the image of the face of the card, including the image surrounding the one or more target characters, thereby enabling the characters to be identified more accurately than if the recognition process were performed on only the target characters.

(Flow of information processing method)
The flow of the information processing method according to this exemplary embodiment will be described with reference to Fig. 2. Fig. 2 is a flow diagram showing the flow of the information processing method. First, the acquisition unit 11 acquires an image of the face of a card (S1).

Next, the cutout unit 12 cuts out one or more target characters, including images surrounding the one or more target characters, from the image acquired by the acquisition unit 11 (S2). The recognition unit 13 then inputs the image cut out by the cutout unit 12 to the trained model, thereby executing a recognition process for the one or more target characters (S3).

(Effects of information processing methods)
As described above, in the information processing method according to this exemplary embodiment, one or more target characters are extracted from the image of the face of the card, including the image surrounding the one or more target characters, thereby enabling the characters to be identified more accurately than if identification processing were performed on the target characters alone.

Exemplary embodiment 2
(Configuration of information processing device)
3 is a block diagram showing a functional configuration of an information processing device 20 according to exemplary embodiment 2. The information processing device 20 includes a teacher data acquisition unit 21, a teacher data storage unit 22, and a learning unit 23. Note that in this exemplary embodiment, a case will be described in which the model to be learned includes a neural network.

The teacher data acquisition unit 21 acquires teacher data from the teacher data storage unit 22, the teacher data including a plurality of pairs of images obtained by cutting out one or more target characters together with images surrounding the one or more target characters from an image of the face of the card, and correct labels (correct data) for the one or more target characters.

The teacher data storage unit 22 stores multiple pieces of teacher data that have been created in advance. Here, the teacher data refers to, as an example, a set of one or more target characters and their surrounding images cut out from an image of a genuine card face, and a correct label (correct answer data) for the one or more target characters.

The model is trained to learn the correspondence between one or more target characters and surrounding images cut out from an image of a genuine card face, and the correct label. For example, if there are 1000 types of target characters to be identified, 1000 output nodes are prepared, and the model is trained to output "1" (a value close to "1") to the output node corresponding to the target character, and to output "0" (a value close to "0") to the other output nodes. By training the model in this way, when one or more target characters to be identified and surrounding images are input to the model, the highest value is output to the target output node.

The model also outputs accuracy information indicating the accuracy of the classification process. As an example, the model outputs accuracy information corresponding to the value of each output node. As an example, the model outputs the value of the target output node as the accuracy information.

The learning unit 23 trains the model using the training data acquired by the training data acquisition unit 21. As described above, when using a NN, one or more target characters and surrounding images cut out from the image of the face of the card are input to the input layer. The model is then trained using the so-called back propagation (BP) method, which back propagates the difference (error) between the value output to the output node and the value of the corresponding correct label, and changes the weighting so as to reduce the error.

As described above, in the information processing device 20 according to this exemplary embodiment, the learning unit 23 trains a model using a pair of an image obtained by cutting out one or more target characters together with images surrounding the one or more target characters from an image of the face of a card, and a correct label (correct data) for the one or more target characters, making it possible to create a model that can accurately identify characters.

(Flow of information processing method)
The flow of the information processing method according to this exemplary embodiment will be described with reference to Fig. 4. Fig. 4 is a flow diagram showing the flow of the information processing method. The information processing method shown in Fig. 4 can be regarded as a model learning method using the information processing device 20, or as a model manufacturing method using the information processing device 20.

First, the teacher data acquisition unit 21 acquires teacher data including multiple pairs of images obtained by cutting out one or more target characters together with images surrounding the one or more target characters from an image of the face of the card, and correct labels for the one or more target characters (S11).

Next, the learning unit 23 trains the model using the teacher data acquired by the teacher data acquisition unit 21 (S12). Then, it is determined whether the learning of the model has been completed (S13). For example, when the model has been trained using all the teacher data stored in the teacher data storage unit 22, it is determined that the learning of the model has been completed.

If the model learning has not been completed (S13, No), the process returns to step S11 and repeats the subsequent steps. If the model learning has been completed (S13, Yes), the learning unit 23 outputs the learned model (S14).

As described above, in the information processing method according to this exemplary embodiment, a model is trained using a pair of an image obtained by cutting out one or more target characters from an image of a certificate face, together with an image surrounding the one or more target characters, and a correct label for the one or more target characters, making it possible to create a model that can accurately identify characters.

Exemplary embodiment 3
A third exemplary embodiment of the present invention will be described in detail with reference to the drawings. Note that components having the same functions as those described in the first exemplary embodiment are denoted by the same reference numerals, and the description thereof will not be repeated.

(Configuration example of information processing device)
5 is a block diagram showing a functional configuration of an information processing apparatus 10a according to a third exemplary embodiment of the present invention. The information processing apparatus 10a includes an acquisition unit 11, a cutout unit 12, an identification unit 13a, and a discrimination unit 14.

Figure 6 is a diagram showing an example of a card face. A driver's license will be described as an example of a card face, but it may also be an identification document such as a My Number card, a passport, or a health insurance card, and is not limited to these. In this exemplary embodiment, the card face refers in a broad sense to an object that has a predetermined format and includes at least one character in that format. Examples of materials for the surface on which the characters are printed or arranged include paper, resin, or a combination of these, but are not limited to these.

As shown in Figure 6, a driver's license has a name column, a date of birth column, an address column, a date of issue column, a type column, etc., but the positions of these columns are predetermined. Therefore, the cropping unit 12 can identify the columns for each item from their positional relationships within the overall image acquired by the acquisition unit 11.

For example, when extracting a name, the extraction unit 12 identifies the name field based on the coordinates in the entire image of the driver's license and extracts images of all character strings written in the name field. The method of extracting each character from the extracted character string image is not limited to this embodiment, but as an example, a conventionally known character extraction algorithm can be used.

When cutting out extracted characters, the cutting unit 12 cuts out the image including the surrounding image of one or more target characters. In the case of the name field shown in FIG. 6, the cutting unit 12 cuts out the image including the ruled lines around one or more target characters included in the name. Specifically, the cutting unit 12 cuts out at least one of the ruled lines arranged above the one or more target characters and the ruled lines arranged below the one or more target characters. Note that the image may be cut out including the ruled lines to the left and right of one or more target characters included in the name.

Note that the "△" mark in the name column in Figure 6 is merely an abstract representation of kanji, hiragana, katakana, alphabet, etc., that represent a person's name, and does not limit this exemplary embodiment.

In addition, when cutting out extracted characters, the cutting unit 12 may cut out at least one character other than the one or more target characters. For example, since the combination of these two characters is fixed for "Showa" and "Nissei" in the date of birth column, the cutting unit 12 cuts out the character "Wa" as well when cutting out the character "Showa", as shown in Figure 6. Similarly, when cutting out the character "Ni", the cutting unit 12 cuts out the character "Natsu" as well as the character "Seika", as shown in Figure 6.

The identification unit 13a includes a name model 131, an address model 132, and a type model 133. The name model 131 is a model created using an information processing device according to the second exemplary embodiment shown in FIG. 3, and is a model that has been trained using training data created based on characters entered in the name field.

The address model 132 is a model created using an information processing device according to the second exemplary embodiment shown in FIG. 3, and the model was trained using training data created based on the characters written in the address field.

Similarly, the type model 133 is a model created using an information processing device according to the second exemplary embodiment shown in Figure 3, and the model was trained using training data created based on the characters entered in the type column.

In Figure 5, the recognition unit 13a only describes three types of models: a name model 131, an address model 132, and a type model 133, but it may also include models that have been trained on images of other items.

The identification unit 13a identifies which item the image cut out by the cut-out unit 12 corresponds to, and selects a model corresponding to the identified item from among a plurality of models (name model 131, address model 132, type model 133). The identification unit 13a then uses the selected model to perform an identification process on one or more characters included in the identified item.

The discrimination unit 14 determines whether the face of the note is genuine by referring to the results of the discrimination process. The results of the discrimination process also include one or more discriminated characters, and the discrimination unit 14 determines whether the face of the note is genuine by determining the format of the one or more discriminated characters. The discrimination unit 14 determines that the face of the note is genuine if the format satisfies predetermined requirements. As an example, the discrimination unit 14 determines whether the face of the note is genuine by checking the format of the character strings for each item shown below.

(1) If there is a "[" in the name column, is there a "]" after it? (2) Are the dates in the date of birth column, date of issue column, etc. valid? (3) Is the issue number in the date of issue column a five-digit number? The result of the identification process also includes one or more identified characters, and the discrimination unit 14 determines the authenticity of the certificate by determining the consistency of one or more identified characters. The discrimination unit 14 determines that the certificate is authentic when, as the consistency, a certain description on the certificate is not contradictory to another description (satisfies a predetermined condition). As an example, the discrimination unit 14 may be configured to determine the authenticity of the certificate by checking the consistency of the character strings in each of the items shown below.

(1) If the maiden name is written, whether the issue date is after a specified date or not. (2) If the Gregorian calendar is written, whether the issue date is after a specified date or not. (3) Whether the expiration date is the month following the date of birth or not. The result of the identification process may include one or more identified characters, and the discrimination unit 14 may be configured to determine the authenticity of the ticket by comparing one or more identified characters with a predetermined character group. For example, the type column lists types such as "large", "small", "medium", and "moped", and these types are predetermined. Therefore, the discrimination unit 14 can compare the characters written in the type column after identification by the identification unit 13a with these character groups (character groups including characters such as "large", "small", "medium", and "moped") and determine the authenticity of the ticket based on whether they match.

The discrimination unit 14 also discriminates the authenticity of the face of the note by referring to the accuracy information. As described above, a model is created so that accuracy information is output when an image cut out by the cut-out unit 12 is input to the trained model, and the identification unit 13a can output the accuracy information. If the accuracy information output from the identification unit 13a is lower than a predetermined value, the discrimination unit 14 determines that the one or more identified characters are forged. This utilizes the fact that misrecognition occurs more frequently when the characters on the face of the note are created using different types of fonts. The discrimination unit 14 determines the authenticity of all one or more identified characters that have been subjected to the identification process by the discrimination unit 13a, and determines that the face of the note is forged if the accuracy information is determined to be lower than a predetermined value.

Figure 7 is a flow diagram showing the flow of an information processing method according to exemplary embodiment 3 of the present invention. First, the acquisition unit 11 acquires an image of the face of a card (S21). Then, the cut-out unit 12 cuts out one or more target characters, including images surrounding the one or more target characters, from the image of the face of the card acquired by the acquisition unit 11 (S22).

Next, the recognition unit 13a identifies an item included on the face of the card based on its position in the image of the face of the card (S23) and selects a model corresponding to the identified item (S24). The recognition unit 13a then uses the selected model to perform a recognition process on one or more characters included in the identified item (S25).

Finally, the discrimination unit 14 determines the authenticity of the certificate by referring to the results of the discrimination process performed by the discrimination unit 13a (S26).

As described above, in the information processing device according to this exemplary embodiment, the cut-out unit 12 cuts out an image including the ruled lines surrounding one or more target characters. If the face of the note is forged, the distance between the target character and the ruled line may vary slightly, or the target character may be slightly tilted relative to the ruled line. Therefore, the identification unit 13a inputs the image cut out by the cut-out unit 12 to the trained model and executes an identification process for one or more target characters, thereby making it possible to identify a forged note face.

In addition, the cut-out unit 12 cuts out at least one character other than the one or more target characters. If the face of the note is forged, the distance between the target character and the other characters may be slightly different, or the target character may be slightly tilted relative to the other characters. Therefore, the recognition unit 13a inputs the image cut out by the cut-out unit 12 to the trained model and executes recognition processing on the one or more target characters, making it possible to recognize a forged note face.

In addition, the identification unit 13a identifies which item the image cut out by the cut-out unit 12 corresponds to, and selects a model corresponding to the identified item from among a number of models (name model 131, address model 132, type model 133). This makes it possible to have the model learn different types of fonts depending on the item, making it possible to more accurately identify counterfeit bills.

In addition, the discrimination unit 14 discriminates the authenticity of the face of the note based on the character string of each item after it has been identified by the discrimination unit 13a. Therefore, it becomes possible to discriminate the authenticity of the face of the note based on the content of the character string of each item after it has been identified.

In addition, the discrimination unit 14 determines the authenticity of the note by determining the format of one or more recognized characters. This makes it possible to easily discern counterfeit notes with formal defects.

In addition, the discrimination unit 14 determines the authenticity of the certificate by determining the consistency of one or more recognized characters. This makes it possible to easily determine whether a certificate is counterfeit or not, due to inconsistencies in the certificate.

In addition, the discrimination unit 14 discriminates the authenticity of the card by comparing one or more recognized characters with a predetermined character group. Therefore, in the case of an item in which a specific character group is written, such as the type field of a driver's license, if a character group other than the specific character group is written, it can be determined that the card is counterfeit.

Furthermore, if the accuracy information output from the recognition unit 13a is lower than a predetermined value, the discrimination unit 14 determines that the one or more recognized characters are forged. Therefore, it is possible to determine that the face of a note is forged when the characters on the note are written in different fonts, for example.

Exemplary embodiment 4
A fourth exemplary embodiment of the present invention will be described in detail with reference to the drawings. Note that components having the same functions as those described in the third exemplary embodiment are denoted by the same reference numerals, and the description thereof will be omitted as appropriate.

(Configuration of Information Processing System)
8 is a block diagram showing a functional configuration of an information processing system 1 according to an exemplary embodiment 4. The information processing system 1 includes a reading device 30 and a server 40, which are connected via a communication network 50. The reading device 30 includes an imaging unit 31 and a communication unit 32. The imaging unit 31 corresponds to a camera, a scanner, etc. provided in a smartphone, and captures an image of the face of a card.

The communication unit 32 communicates with a server 40 installed in a bank or the like via a communication network 50, which is a wide area communication network such as the Internet. The communication unit 32 transmits an image of the face of the card captured by the imaging unit 31 to the server 40 via the communication network 50.

The server 40 includes a communication unit 41 and an information processing device 10a. The communication unit 41 communicates with the reading device 30 via a communication network 50, which is a wide area communication network such as the Internet. The communication unit 41 receives an image of the face of the card from the reading device 30 via the communication network 50.

The acquisition unit 11 of the information processing device 10a acquires an image of the face of the card received by the communication unit 41. The subsequent processing is similar to that described in exemplary embodiment 3. Therefore, detailed description will not be repeated.

The communication unit 41 receives the results of the image recognition of the card face by the information processing device 10a and transmits them to the reading device 30 via the communication network 50.

9 is a flowchart for explaining the processing procedure of the reading device 30 according to the fourth exemplary embodiment of the present invention. First, the imaging unit 31 captures an image of the face of the card (S31). Then, the communication unit 32 transmits the image of the face of the card captured by the imaging unit 31 to the server 40 via the communication network 50 (S32).

Figure 10 is a flowchart for explaining the processing procedure of the server 40 relating to exemplary embodiment 4 of the present invention. First, the communication unit 41 receives an image of the face of the card from the reading device 30 via the communication network 50 (S41).

The information processing device 10a determines whether the image of the face of the card is genuine (S42). The communication unit 41 then receives the result of the determination of the image of the face of the card by the information processing device 10a and transmits it to the reading device 30 via the communication network 50 (S43).

As described above, the reading device 30 captures an image of the face of the note and transmits it to the server 40, and the server 40 determines whether the face of the note is genuine and transmits the determination result to the reading device 30. Therefore, there is no need for the reading device 30 to determine whether the face of the note is genuine.

In addition, since the cut-out unit 12 cuts out one or more target characters from the image of the face of the card, including the image surrounding the one or more target characters, the characters can be identified more accurately than if the recognition process were performed on only the target characters.

[Software implementation example]
Some or all of the functions of the information processing devices 10, 10a, and 20 may be realized by hardware such as an integrated circuit (IC chip), or may be realized by software.

In the latter case, the information processing devices 10, 10a, and 20 are realized, for example, by a computer that executes instructions of a program, which is software that realizes each function. An example of such a computer (hereinafter, referred to as computer 60) is shown in FIG. 11. The computer 60 has at least one processor 61 and at least one memory 62, which are connected via an internal bus 63. The memory 62 stores a program P for operating the computer 60 as the information processing devices 10, 10a, and 20. In the computer 60, the processor 61 reads and executes the program P from the memory 62, thereby realizing each function of the information processing devices 10, 10a, and 20.

The processor 61 may be, for example, a central processing unit (CPU), a graphic processing unit (GPU), a digital signal processor (DSP), a micro processing unit (MPU), a floating point number processing unit (FPU), a physics processing unit (PPU), a microcontroller, a general-purpose computing on graphics processing units (GPGPU), or a combination of these. The memory 62 may be, for example, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or a combination of these.

The computer 60 may further include a RAM (Random Access Memory) for expanding the program P during execution and for temporarily storing various data. The computer 60 may further include a communication interface for transmitting and receiving data to and from other devices. The computer 60 may further include an input/output interface for connecting input/output devices such as a keyboard, a mouse, a display, and a printer.

The program P can also be recorded on a non-transitory tangible recording medium 70 that can be read by the computer 60. Examples of such a recording medium 70 include a CD-ROM (Compact Disc-Read Only Memory), a DVD (Digital Versatile Disc), a tape, a disk, a card, a semiconductor memory, or a programmable logic circuit. The computer 60 can acquire the program P via such a recording medium 70. The program P can also be transmitted via a transmission medium. Examples of such a transmission medium include a communication network or broadcast waves. The computer 60 can also acquire the program P via such a transmission medium.

[Additional Note 1]
The present invention is not limited to the above-described exemplary embodiments, and various modifications are possible within the scope of the claims. For example, exemplary embodiments obtained by appropriately combining the technical means disclosed in the above-described exemplary embodiments are also included in the technical scope of the present invention.

[Additional Note 2]
Some or all of the above-described exemplary embodiments may be described as follows. However, the present invention is not limited to the following described aspects.

(Appendix 1)
An acquisition means for acquiring an image of a face of a card;
A cutting means for cutting out one or more object characters from the image together with an image surrounding the one or more object characters;
and a classification means for executing a classification process on the one or more target characters by inputting the image extracted by the extraction means to a trained model.

(Appendix 2)
The information processing device according to claim 1, wherein the surrounding image is a ruled line surrounding the one or more target characters.

(Appendix 3)
The information processing device according to claim 2, wherein the surrounding image is at least one of a ruled line arranged above the one or more target characters and a ruled line arranged below the one or more target characters.

(Appendix 4)
2. The information processing device according to claim 1, wherein the surrounding image is at least one character other than the one or more target characters.

(Appendix 5)
The model includes a plurality of models each trained corresponding to a plurality of items included on the face of the card,
The identification means is
identifying an item contained on the face of the card based on its location within an image of the face of the card;
selecting a model corresponding to the specified item from among the plurality of models;
5. The information processing device according to claim 1, further comprising: a step of: using the selected model, performing a classification process on one or more characters included in the identified item.

(Appendix 6)
The information processing device further comprises:
An information processing device according to any one of claims 1 to 5, further comprising a discrimination means for discriminating the authenticity of the face of the bill by referring to the result of the discrimination process.

(Appendix 7)
a result of the classification process including one or more classified characters;
The information processing device described in Appendix 6, wherein the discrimination means determines the authenticity of the face of the certificate by determining the format of the one or more recognized characters.

(Appendix 8)
a result of the classification process including one or more classified characters;
The information processing device described in Appendix 6 or 7, wherein the discrimination means determines the authenticity of the face of the certificate by determining the consistency of the one or more recognized characters.

(Appendix 9)
a result of the classification process including one or more classified characters;
The information processing device according to any one of appendices 6 to 8, wherein the discrimination means discriminates the authenticity of the face of the certificate by comparing the one or more identified characters with a predetermined group of characters.

(Appendix 10)
The result of the identification process includes accuracy information of the identification process,
The information processing device according to any one of Appendices 6 to 9, wherein the discrimination means discriminates the authenticity of the face of the certificate by referring to the accuracy information.

(Appendix 11)
a training data acquisition unit that acquires training data including a plurality of pairs of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a face of a card, and a correct label for the one or more target characters;
The information processing device according to any one of appendices 1 to 10, further comprising a learning unit that uses the teacher data to learn the model.

(Appendix 12)
The information processing device according to any one of appendices 1 to 11, wherein the face of the card is an identification document and is one of a driver's license, a My Number card, a passport, and a health insurance card.

(Appendix 13)
a training data acquisition unit that acquires training data including a plurality of pairs of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a face of a card, and a correct label for the one or more target characters;
An information processing device comprising: a learning unit that inputs an image obtained by cutting out one or more target characters together with images surrounding the one or more target characters from an image of a card face, and uses the training data to learn a model that outputs a recognition result for the one or more target characters.

(Appendix 14)
Take an image of the card,
Extracting one or more target characters from the image, including images surrounding the one or more target characters;
An information processing method comprising: inputting the extracted image into a trained model, and performing a classification process on the one or more target characters.

(Appendix 15)
Acquire training data including a plurality of pairs of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a face of a card, and a correct answer label for the one or more target characters;
An information processing method in which an image obtained by cutting out one or more target characters from an image of a card face together with images surrounding the one or more target characters is input, and a model that outputs recognition results for the one or more target characters is trained using the training data.

(Appendix 16)
Acquire training data including a plurality of pairs of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a face of a card, and a correct answer label for the one or more target characters;
A method for manufacturing a trained model, in which an image obtained by cutting out one or more target characters from an image of a card face together with images surrounding the one or more target characters is input, and a model that outputs a recognition result for the one or more target characters is trained using the training data.

(Appendix 17)
A computer program for causing a computer to execute an information processing method,
The information processing method includes acquiring an image of a face of a card;
Extracting one or more target characters from the image, including images surrounding the one or more target characters;
A computer program that executes a classification process for the one or more target characters by inputting the extracted image to a trained model.

(Appendix 18)
A computer program for causing a computer to execute an information processing method,
The information processing method includes acquiring training data including a plurality of pairs of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a face of a card, and a correct answer label related to the one or more target characters;
A computer program that inputs an image obtained by cutting out one or more target characters from an image of a card face together with images surrounding the one or more target characters, and uses the training data to train a model that outputs a recognition result for the one or more target characters.

(Appendix 19)
A computer-readable recording medium having a program recorded thereon that causes a computer to function as an information processing device,
The computer includes an acquisition means for acquiring an image of a face of a bill;
A cutting means for cutting out one or more target characters from the image, including an image surrounding the one or more target characters;
a classification means for executing a classification process for the one or more target characters by inputting the image extracted by the extraction means to a trained model;
A recording medium on which a program that functions as a

(Appendix 20)
A computer-readable recording medium having a program recorded thereon that causes a computer to function as an information processing device,
The computer includes a training data acquisition means for acquiring training data including a plurality of pairs of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a face of a card, and a correct label for the one or more target characters;
a learning means for learning a model using the training data, the model inputting an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a face of a card, and outputting a recognition result for the one or more target characters;
A recording medium on which a program that functions as a

(Appendix 21)
The method includes at least one processor, the processor comprising:
A process of extracting one or more target characters from the image, including images surrounding the one or more target characters;
An information processing device that performs a process of performing a recognition process on the one or more target characters by inputting an image extracted by the extraction means to a trained model.

The information processing device may further include a memory, and the memory may store a program for causing the processor to execute the acquiring process, the extracting process, and the executing process. The program may also be recorded on a computer-readable, non-transitory, tangible recording medium.

(Appendix 22)
The method includes: acquiring training data including a plurality of pairs of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a face of a card, and a correct answer label for the one or more target characters;
An information processing device that inputs an image obtained by cutting out one or more target characters from an image of a card face together with images surrounding the one or more target characters, and uses the training data to train a model that outputs recognition results for the one or more target characters.

The information processing device may further include a memory, and the memory may store a program for causing the processor to execute the acquisition process and the learning process. The program may also be recorded on a computer-readable, non-transitory, tangible recording medium.

REFERENCE SIGNS LIST 1 Information processing system 10, 10a, 20 Information processing device 11 Acquisition unit 12 Extraction unit 13, 13a Identification unit 14 Discrimination unit 21 Teacher data acquisition unit 22 Teacher data storage unit 23 Learning unit 30 Reading device 31 Imaging unit 32, 41 Communication unit 40 Server 50 Communication network 60 Computer 61 Processor 62 Memory 63 Internal bus 70 Recording medium 131 Name model 132 Address model 133 Type model P Program

Claims (11)

An acquisition means for acquiring an image of a face of a card;
A cutting means for cutting out one or more object characters from the image together with an image surrounding the one or more object characters;
A classification unit that executes a classification process for the one or more target characters by inputting the image extracted by the extraction unit to a trained model,
the peripheral image is a ruled line around the one or more target characters,
The information processing device, wherein the trained model is a model that has been machine-learned to learn an image including the one or more target characters and the ruled lines . The information processing device according to claim 1 , wherein the peripheral image is at least one of a ruled line arranged above the one or more object characters and a ruled line arranged below the one or more object characters. The information processing device according to claim 1, wherein the surrounding image is at least one character other than the one or more target characters. The model includes a plurality of models each trained corresponding to a plurality of items included on the face of the card,
The identification means is
identifying an item contained on the face of the card based on its location within an image of the face of the card;
selecting a model corresponding to the specified item from among the plurality of models;
The information processing apparatus according to claim 1 , further comprising: a step of: executing, using a selected model, a classification process for one or more characters included in the specified item. The information processing device further comprises:
The information processing device according to claim 1 , further comprising a discrimination means for discriminating the authenticity of the face of the card by referring to a result of the discrimination process. a result of the classification process including one or more classified characters;
The information processing apparatus according to claim 5 , wherein the determining means determines whether the face of the card is genuine by determining a format of the one or more identified characters. a result of the classification process including one or more classified characters;
The information processing device described in claim 5, wherein the discrimination means determines that the face of the card is genuine when the content of the one or more identified characters and the content of identified characters contained on the face of the card that are different from the one or more identified characters satisfy predetermined conditions. the trained model is trained to output accuracy information of the one or more target characters by inputting an image cut out by the cutout means,
The information processing device according to claim 5 , wherein the determining means determines that the face of the card is counterfeit when the certainty information is lower than a predetermined value. A training data acquisition means for acquiring training data including a plurality of pairs of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a face of a card, and a correct label for the one or more target characters;
a learning means for learning a model using the training data, the model receiving an input of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a face of a card, and outputting a recognition result relating to the one or more target characters ;
the peripheral image is a ruled line around the one or more target characters,
The model is a model that is machine-learned to learn an image including the one or more target characters and the ruled lines . Take an image of the card,
Extracting one or more target characters from the image, including images surrounding the one or more target characters;
Executing a classification process for the one or more target characters by inputting the extracted image to a trained model ;
the peripheral image is a ruled line around the one or more target characters,
An information processing method, wherein the trained model is a model trained by machine learning on an image including the one or more target characters and the ruled lines . Acquire training data including a plurality of pairs of an image obtained by cutting out one or more target characters together with an image surrounding the one or more target characters from an image of a face of a card, and a correct answer label for the one or more target characters;
an image obtained by cutting out one or more target characters from an image of a face of a card together with an image surrounding the one or more target characters is input, and a model that outputs a recognition result for the one or more target characters is trained using the training data ;
the peripheral image is a ruled line around the one or more target characters,
An information processing method, wherein the model is a model that has been machine-learned to learn an image including the one or more target characters and the ruled lines .

Images