JP2014016762A - Form recognition apparatus and form recognition method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a form recognition apparatus capable of recognizing a character string recorded in a non-standardized form with high recognition accuracy while reducing processing time.SOLUTION: The form recognition apparatus obtains an image in a form, picks up and recognizes a piece of keyword information for identifying the form from the image and collates the keyword information with pieces of keyword information of records in a table data to determine which non-standardized form agrees with the form. When the keyword information is determined as not agree, the form recognition apparatus performs a predetermined recognition processing on the entire image and outputs an item name and an item value recorded in the form. When the keyword information is determined as agrees with any non-standardized form, the form recognition apparatus obtains a piece of area information from a record corresponding to the non-standardized form determined as agree and performs a predetermined recognition processing on the region determined by the area information to output the item name and the item value recorded in the form as a recognition result.

Description

  The present invention relates to a technique for recognizing a character string described in a form.

  The forms are classified into enacted forms that satisfy the form standard for character recognition and non-enacted forms that do not satisfy them. In the case of an enacted form, the character string can be recognized with high accuracy by “form definition” in which the description position of the character string to be read on the form image and its attribute are registered in the apparatus in advance. On the other hand, non-established forms do not have strict standards like established forms, so when setting a description position (also called a reading area) before operation, printing is performed between the form used and the form used in operation. There are often differences due to errors or cutting errors. If there is a difference between the forms, there is a difference between the preset reading area and the actual position on the form, which makes it difficult to recognize characters with high accuracy.

  Therefore, a technique that enables recognition of a non-established form without using a form definition has been proposed by the present applicant (see Patent Document 1). This technique detects a character string from the entire form image, calculates the item name likelihood and the item value likelihood for all character strings, and arranges the relationship with respect to a character string pair composed of the detected character string combinations. By calculating the validity of the item name-item value relationship, item names and item values are extracted, making it possible to recognize non-established forms with high accuracy.

JP 2011-248609 A

  In the technique of Patent Document 1, high-accuracy recognition is possible as described above. However, there is a demand for further shortening the processing time and further improving the recognition accuracy. Such a request is not limited to the technique of Patent Document 1, and is a problem common to techniques involving analysis of the contents of character strings.

  The present invention has been made to solve the above-described problems, and has as its object to reduce the processing time and improve the recognition accuracy for the recognition of character strings described in non-established forms.

SUMMARY An advantage of some aspects of the invention is to solve at least a part of the problems described above, and the invention can be implemented as the following configurations. That is, the present invention
A form recognition device for recognizing a character string described in a form,
A form image acquisition unit for acquiring an image of the form;
A keyword information recognition unit for recognizing and recognizing keyword information for identifying the form from the acquired image;
It is tabular data for classifying non-established forms, and for each record, includes keyword information for identifying the non-established form, and an area including a character string to be recognized in the non-established form A storage unit for storing table data including area information to be shown;
A form classification determination unit that determines which non-established form the form corresponds to by comparing the keyword information obtained by the keyword information recognition unit with the keyword information of each record in the table data;
When it is determined by the form classification determination unit that it does not correspond to any non-established form recorded in the table data, a predetermined recognition process is performed on the entire image acquired by the form image acquisition unit A first recognition processing unit that outputs the item name and item value described in the form as a recognition result,
When it is determined by the form classification determination unit that any non-established form recorded in the table data is acquired from the record corresponding to the non-established form determined to be applicable, the area information, A second recognition processing unit that outputs the item name and the item value described in the form as a recognition result by performing the predetermined recognition process on the area determined by the area information.

  According to the form recognizing apparatus having the above-described configuration, the non-established form recorded in the table data only needs to be subjected to a predetermined recognition process on the area determined by the area information included in the table data. For this reason, the processing time for the recognition of the character string described in the non-established form can be shortened, and the recognition accuracy can be improved.

It is explanatory drawing which shows the structure of the form recognition system of one Example of this invention. It is explanatory drawing which illustrates a non-established form. It is explanatory drawing which shows the form classification table 30 typically. It is explanatory drawing of the recognition item table. 4 is a flowchart showing a non-enacted form recognition routine included in the image processing software 20. It is explanatory drawing which shows an example of the upper area | region UA. It is explanatory drawing which shows the outline | summary of the form definition less recognition process in step S150. It is explanatory drawing which shows the outline | summary of the form definition less recognition process in step S160. It is explanatory drawing which shows the other example of the outline | summary of the form definition less recognition process in step S160.

Embodiments of the present invention will be described below in the following order based on examples.
A. Example:
A1. Form recognition system configuration:
A2. Non-enacted forms and table structure:
A3. Non-enacted form recognition routine:
A4. Example effect:
B. Variations:

A. Example:
A1. Form recognition system configuration:
FIG. 1 is an explanatory diagram showing the configuration of a form recognition system according to an embodiment of the present invention. The form recognition system includes a computer 100 for recognizing a form, a display 200 for displaying a recognition result, a scanner 300 for reading a form image, and a keyboard 400 for inputting various types of information. . The computer 100 can be connected to a host computer (not shown) via a line 500.

  The computer 100 includes a control unit 110, an output interface 120, an input interface 140, and a network interface 150.

  The control unit 110 includes a memory 111 such as a ROM or a RAM, and a CPU 112 that is a central processing unit. The memory 111 stores image processing software 20 and a form classification table 30.

  The image processing software 20 is a program for processing an image of a form read by the scanner 300. The image processing software 20 cuts out an image from the image to the size of the form, extracts the frame line information described in the form, Recognize the character string described. The image processing software 20 can handle both enacted forms and non-established forms as forms for recognizing character strings. In particular, the form image acquisition unit 21, the keyword information recognition unit 22, and the form classification determination unit 23 And the 1st recognition process part 24 and the 2nd recognition process part 25 are implement | achieved as a function, The character string described in the non-established form is made recognizable. The image processing software 20 includes a recognition item table 26. The recognition item table 26 is tabular data, and is used to refer to the item name, item name condition, and item value condition when the image processing software 20 recognizes the character string of the form.

  The form classification table 30 is tabular data, and is used to change the recognition process for each form classification when the image processing software 20 recognizes the character string of the form.

  In the control unit 110, when the CPU 112 executes the image processing software 20, control necessary for processing of the form including functions corresponding to the respective units 21 to 25 is performed. Each function is configured as software in the present embodiment, but may be configured as hardware using an integrated circuit or the like.

  The image processing software 20 is stored in a hard disk device (HDD) (not shown), and is loaded into a predetermined area of the memory 111 when a predetermined activation command is received. The form classification table 30 can be similarly loaded from the HDD into the memory 111. The image processing software 20 can be recorded on a recording medium (computer-readable recording medium) such as a CD-ROM or DVD-ROM, or can be downloaded to a computer via the Internet. Similarly, the form classification table 30 can also be recorded on a recording medium or downloaded to a computer via the Internet.

  The display 200 is connected to the computer 100 via the output interface 120. The display 200 receives information about the screen to be displayed from the computer 100 and displays the screen to provide the operator with the result of form recognition.

  The scanner 300 and the keyboard 400 are connected to the computer 100 via the input interface 140 and transmit information to the computer 100. The scanner 300 is a device for acquiring a form image, and corresponds to a non-contact scanner, a flat bed scanner, or the like. The keyboard 400 is a device that transmits an operator's request to the computer 100, and may be a device having the same function, such as a mouse. The computer 100 is connected to a host computer via a line 500 connected to the network interface 150, and exchanges various types of information.

A2. Non-enacted forms and table structure:
FIG. 2 is an explanatory diagram illustrating a non-enacted form. A non-enacted form is a form other than an enacted form that is a form of a form established by the bank (own bank), and can take a wide variety of formats. As shown in FIGS. 2A to 2C, examples of non-established forms include a public tax form such as a municipal tax, an automobile tax, or an electric bill. As can be seen from these, non-established forms have a wide variety of sizes, layouts, item names, item values, and the like. “Item name” is the name of the item described in the form, and “Item value” is the value for the item. For example, in FIG. 2B, “transfer account”, “transfer amount”, and “transfer deadline” correspond to the item names, “123-456”, “10,000 yen”, “2011 12 “Month 31” corresponds to the item value.

  FIG. 3 is an explanatory diagram schematically showing the form classification table 30. The data structure of the form classification table 30 is shown in FIG. The form classification table 30 stores fields 611 to 615 of “classification name”, “form type”, “form size information”, “keyword information”, and “area information” as one record. One record corresponds to one classification of the form. The “category name” field 611 records a classification name for classifying the form, the “form type” field 612 records the form type, and the “form size information” field 613 records the form size. Is recorded. In the “keyword information” field 614, keyword information for specifying the form described in the form is recorded. In the “area information” field 615, the name of the item described in the form and the position information indicating the position (including size) of the area in which the item is described are recorded.

  A specific example of the data stored in the form classification table 30 is shown in FIG. The first record R1 is for the non-established form shown in FIG. 2B, and the second record R1 is for the non-established form shown in FIG. In the first record R1, “Category 1” is recorded as the classification name, “Citizen tax” is recorded as the form type, “A5 size (x: 210, y: 148)” is recorded as the form size information, “Nagoya City” and “citizen tax” are recorded as keyword information, and “account number” which is an item name and position information “(x: 20-180, y: 40-100)” are recorded as area information. Has been. In the second record R2, “Class 2” is recorded as the classification name, “Automobile tax” is recorded as the form type, “A4 size (x: 210, y: 297)” is recorded as the form size information, “Aichi Prefecture” and “automobile tax” are recorded as keyword information, and “amount” as item name and position information “(x: 100-180, y: 80-280)” are recorded as area information. ing.

  In addition to the “class 1” and “class 2” records, records such as “class 3” are recorded in the form classification table 30, but the non-established form shown in FIG. The following description will be given assuming that is not recorded.

  FIG. 4 is an explanatory diagram of the recognition item table 26. The data structure of the recognition item table 26 is shown in FIG. The recognition item table 26 stores the fields 621 to 623 of “item name”, “item name condition”, and “item value condition” as one record. One record corresponds to one type of item. That is, in the recognized item table 26, “item name”, “item name condition”, and “item value condition” are recorded in association with each other in the type of item. An item name is recorded in the “item name” field 621, a condition that can be taken as another description of the same item name is recorded in the “item name condition” field 622, and a field 623 of “item value condition”. In, conditions that can be taken as separate descriptions of the same type of item values are recorded.

  A specific example of data stored in the recognition item table 26 is shown in FIG. In the first record R11, “amount” is recorded as an item name, and “amount”, “transfer amount”, “total amount”, and “tax amount” are recorded as item name conditions. That is, it is recorded as the item name condition that any one of “amount”, “transfer amount”, “total amount”, and “tax amount” is satisfied. As the item value condition, 9 digits, “¥” and “yen” are recorded. That is, it is recorded as the item value condition that any one of 9 digits or less, including “¥”, or including “yen” is satisfied.

  In the second record R12, “account number” is recorded as the item name, “account number”, “account”, and “transfer account” are recorded as the item name condition, and 9 digits of the item value condition, “¥ "," Yen "is recorded. In the third record R13, “expiration date” is recorded as an item name, and “expiration date”, “deadline date”, and “transfer due date” are recorded as item name conditions. As item value conditions, date (Japanese calendar), date (Western calendar), “year”, “month”, and “day” are recorded. In other words, the item name condition is to satisfy any one of the date of the Japanese calendar, the date of the Western calendar, the "year", the "month", and the "day". It is recorded.

A3. Non-enacted form recognition routine:
FIG. 5 is a flowchart showing a non-established form recognition routine included in the image processing software 20. This non-established form recognition routine is executed by the CPU 112, and is started when the form to be recognized is not an established form. When the form to be recognized is an enacted form, the process for recognizing the form is executed using the definition data registered by the form definition. When the form to be recognized is not the enacted form, Will be executed.

  When the process is started, the CPU 112 first acquires a form image read by the scanner 300 (step S110). Next, the CPU 112 calculates a form size from the acquired image (step S120). Here, the form size is a size defined by standards such as “A5 size” and “A4 size”. Subsequently, the CPU 112 selects an upper area from the image and cuts out and recognizes a character string from the upper area (step S130).

  FIG. 6 is an explanatory diagram illustrating an example of the upper area UA. In the form FR, for example, a form name such as “Osaka Prefecture Corporate Tax” is written. The form name is generally written at the top of the form. In step S130, the upper area UA of the form is obtained based on the form size obtained in step S120, and the character string is cut out from the upper area and recognized. The upper area UA is obtained by setting the length of 1 / n (n is an integer of 2 or more) of the vertical length of the form size as the height and the horizontal length of the form size as the width. Although the upper area UA is used here, it is not necessarily limited to the upper area UA. For example, when a lot of forms whose form names are described in the left area of the form are handled, You can also In short, it can be configured to select a predetermined area. In step S130, after selecting the upper area UA, a character string is extracted from the upper area UA. The character string is cut out and recognized using a known method for cutting out and recognizing a character string from an image area.

  Returning to FIG. 5, after executing step S <b> 130, the CPU 112 collates the form size and character string obtained in steps S <b> 120 and S <b> 130 with the form classification table 30, so that the form image form is stored in the form classification table 30. It is determined whether or not the classification is recorded (step S140). The collation is performed by checking whether the form size and the character string obtained in steps S120 and S130 are sequentially matched with the form size information and the keyword information for each record recorded in the form classification table 30. . If both the form size and the character string obtained in steps S120 and S130 match the form size information and the keyword information of any record recorded in the form classification table 30, the form image form is classified into the form classification. It is determined that the classification is recorded in the table 30. On the other hand, if there is no record in which both the form size and the character string match, it is determined that the form image form does not correspond to the classification recorded in the form classification table 30.

  If “NO” in step S140, that is, if it is determined that the form does not correspond to the classification recorded in the form classification table 30, the CPU 112 executes a form definition-less recognition process for the entire form image. (Step S150). This form definition-less recognition process recognizes a non-established form without using a form definition, and uses the technique described in Japanese Patent Application Laid-Open No. 2011-248609 disclosed by the applicant of the present invention. Value.

  FIG. 7 is an explanatory diagram showing an outline of the form definition-less recognition process in step S150. In the figure, the form FR is the form of “Osaka Prefecture Corporate Tax” shown in FIG. 2A and is not the form of the classification recorded in the form classification table 30. Processing will be executed. Specifically, the form definition-less recognition process is performed in the following order.

(1). Character strings are searched for the entire form image P0 as a reading target (see FIG. 7A).
(2). By comparing each character string obtained in (1) with the item name condition stored in the recognition item table 26 (FIG. 4), an item name likelihood representing the item name likelihood for each character string is calculated. (Refer FIG.7 (b)).
(3). By comparing each character string obtained in (1) with the item value condition stored in the recognition item table 26 (FIG. 4), the item value likelihood representing the item value likelihood for each character string is calculated. (Refer FIG.7 (b)).
(4). For all the character string pairs formed of the two combinations of the character strings obtained in (1), an arrangement likelihood representing the validity as the item name-item value relation of the arrangement relation of the character string pair is calculated ( (Refer FIG.7 (b)).
(5). Based on the item name likelihood, the item value likelihood, and the placement likelihood obtained in (2) to (4), the character string pair is item name-item for all character string pairs in the form image. An evaluation value indicating whether or not there is a value relationship is calculated, and an item name and an item value are determined based on the evaluation value (see FIG. 7C).

As a result of the form definition-less recognition process in step S150, the following three items are obtained.
-The item name is “account number” and the item value is “987-6-54321”
-The item name is "Amount" and the item value is "50,000 yen"
-The item name is "Deadline" and the item value is "March 31, 2012"
Here, as the item name, the content recorded in the “item name” field 621 of the recognized item table 26 is adopted as the item name, not the content described in the form by character recognition.

  After executing step S150, the CPU 112 displays the recognition result obtained in step S150 on the display 200 (step S170), and then ends the non-enacted form recognition processing routine.

  On the other hand, if “YES” is determined in the step S140, that is, the form is determined to correspond to the classification recorded in the form classification table 30, the CPU 112 recognizes the form definition-less recognition process for a partial area of the form image. Is executed (step S160). In step S150, the form definition-less recognition process is performed on the entire form image P0. Instead, in step S160, the area recorded in the “area information” field 615 of the form classification table 30 is stored. As a partial area, a form definition-less recognition process is performed on the partial area. That is, the CPU 112 reads the position information recorded in the “area information” field 615 for the classification for which the form size and the character string are determined to match in step S140, and uses the area indicated by the position information as a partial area. The form definition-less recognition process is performed on the partial area.

  FIG. 8 is an explanatory diagram showing an outline of the form definition-less recognition process in step S160. In the figure, the form FR is a form of “Nagoya City Citizen Tax” shown in FIG. 2B, and is a form with a classification recorded in the form classification table 30. Therefore, the form definition is not recognized in step S160. Processing will be executed. More specifically, this form definition-less recognition process is performed in the following order (i) to (v).

(I). A character string is searched for the partial area P1 of the form image as a reading target (see FIG. 8A). Here, the partial area P1 is indicated by the position information recorded in the “area information” field 615 (FIG. 3B) for the record (classification) for which the form size and the character string are determined to match in step S140. It is what That is, the partial area P1 is an area indicated by the position information (x: 20-180, y: 40-100) recorded in the “area information” field 615 of the first record R1 of the form classification table 30.
(Ii). By comparing each character string obtained in (i) with the item name condition stored in the recognition item table 26 (FIG. 4), the item name likelihood representing the item name likelihood for each character string is calculated. (See FIG. 8 (b)).
(Iii). By comparing each character string obtained in (i) with the item value condition stored in the recognition item table 26 (FIG. 4), an item value likelihood representing the item value likelihood for each character string is calculated. (See FIG. 8 (b)).
(Iv). For all the character string pairs composed of the two combinations of character strings obtained in (i), an arrangement likelihood representing the validity of the item name-item value relation of the arrangement relation of the character string pair is calculated ( (Refer FIG.8 (b)).
(V). Based on the item name likelihood, the item value likelihood, and the arrangement likelihood obtained in (ii) to (iv), the character string pair is an item name− for all the character string pairs in the partial region P1. An evaluation value indicating whether the item value is related is calculated, and an item name and an item value are determined based on the evaluation value (see FIG. 8C).

As a result of the form definition-less recognition process in step S160, the following items are obtained.
-Item name is "Account Number" and item value is "123-456"
Here, as the item name, the content recorded in the “item name” field 621 of the recognized item table 26 is adopted as the item name, not the content described in the form by character recognition. According to the form definition-less recognition process, a non-established form can be recognized with few errors.

  After executing step S160, the CPU 112 advances the process to step S170 and displays the recognition result obtained in step S160 on the display 200. After execution of step S170, the process returns to “return” to end the non-established form recognition routine.

FIG. 9 is an explanatory diagram showing another example of the outline of the form definition-less recognition process in step S160. In the figure, the form FR is the form of “Aichi prefecture automobile tax” shown in FIG. 2C. In this case as well, the form definition-less recognition in step S160 is performed in the same manner as the form of “Nagoya City municipal tax”. Processing will be executed. That is, the partial area indicated by the position information (x: 100-180, y: 82-280) recorded in the “area information” field 615 of the second record R2 of the form classification table 30 (FIG. 3B). The form definition-less recognition process is executed for P2. As a result, the following items are obtained.
・ The item name is "Amount" and the item value is "25,500 yen"

A4. Example effect:
As described above in detail, according to the form recognition system of the present embodiment, the non-established form recorded in the form classification table 30 is compared with the partial area determined by the area information included in the form classification table 30. All you need to do is a form definition-less recognition process. Conventionally, for non-established forms, item name likelihoods and item value likelihoods are calculated for all character strings obtained by reading the entire form image, and placement likelihoods for all character string pairs. Need to be calculated. For example, when the non-established form of “Nagoya City Municipal Tax” shown in FIG. 2B is recognized in the conventional example, the item name likelihood and the item value are respectively added to the character string “Nagoya” and the character string “Citizen Tax”. It is necessary to calculate the likelihood and further calculate the arrangement likelihood of the combination including both character strings. For example, when the non-established form of “Aichi Prefecture Automobile Tax” shown in FIG. 2C is recognized in the conventional example, it is included in the character string “Aichi Prefecture” and the character string “Automobile Tax”, Since the character string of “money”, the character string of “fee”, and the like exist outside the partial area, it is necessary to calculate more item name likelihoods, item value likelihoods, and placement likelihoods. On the other hand, as described with reference to FIGS. 8 and 9, only the partial areas P1 and P2 are the targets of the form definition-less recognition process. For this reason, according to the form recognition system of the present embodiment, it is possible to shorten the processing time for the recognition of the character string described in the non-established form, and to improve the recognition accuracy. Play.

  Further, in this embodiment, it is configured to determine whether or not the form image form corresponds to the classification recorded in the form classification table 30 based on both the form size information and the keyword information. . For this reason, it is possible to determine whether a form is registered in the form classification table 30 with high accuracy.

B. Variations:
The present invention is not limited to the above-described embodiments and modifications thereof, and can be carried out in various modes without departing from the gist thereof. For example, the following modifications are possible.

・ Modification 1:
In the above embodiment, the form classification table 30 is a fixed one prepared in advance. However, it is possible to sequentially add new classification records to the form classification table 30 by operating the keyboard 400 or the like.

Modification 2
In the above embodiment, the form definition-less recognition process is configured such that the item name likelihood, the item value likelihood, and the placement likelihood are calculated, and the item is recognized based on these likelihoods. But it is not limited to this. Any form definition-less recognition process can be used as long as the technique recognizes the contents by analyzing the contents of the character string without using the form definition.

・ Modification 3:
In the above-described embodiment, the computer 100, the display 200, the scanner 300, and the keyboard 400 are all separate devices. However, the computer 100, the display 200, the scanner 300, and the keyboard 400 may all be configured as dedicated machines. By doing so, it is possible to recognize the form with high accuracy by a single device.

  The present invention can be realized in various forms other than the forms described in the claims. For example, each process included in the form processing system including the form recognition apparatus and the form recognition method is realized as a function. The present invention can be realized in the form of a computer program, the form of the computer program, a print medium on which the computer program is recorded, or the like.

  It should be noted that elements other than those described in the independent claims among the constituent elements in each of the above-described embodiments and modifications are additional elements and can be omitted as appropriate.

DESCRIPTION OF SYMBOLS 20 ... Image processing software 21 ... Form image acquisition part 22 ... Keyword information recognition part 23 ... Form classification | category determination part 24 ... 1st recognition processing part 25 ... 2nd recognition processing part 26 ... Recognition item table 30 ... Form classification table 100 ... Computer 110 ... Control unit 111 ... Memory 112 ... CPU
120 ... Output interface 140 ... Input interface 150 ... Network interface 200 ... Display 300 ... Scanner 400 ... Keyboard 500 ... Line

Claims (5)

A form recognition device for recognizing a character string described in a form,
A form image acquisition unit for acquiring an image of the form;
A keyword information recognition unit for recognizing and recognizing keyword information for identifying the form from the acquired image;
It is tabular data for classifying non-established forms, and for each record, includes keyword information for identifying the non-established form, and an area including a character string to be recognized in the non-established form A storage unit for storing table data including area information to be shown;
A form classification determination unit that determines which non-established form the form corresponds to by comparing the keyword information obtained by the keyword information recognition unit with the keyword information of each record in the table data;
When it is determined by the form classification determination unit that it does not correspond to any non-established form recorded in the table data, a predetermined recognition process is performed on the entire image acquired by the form image acquisition unit A first recognition processing unit that outputs the item name and item value described in the form as a recognition result,
When it is determined by the form classification determination unit that any non-established form recorded in the table data is acquired from the record corresponding to the non-established form determined to be applicable, the area information, A form recognition apparatus comprising: a second recognition processing unit that outputs an item name and an item value described in the form as a recognition result by performing the predetermined recognition process on the area determined by area information. The form recognition device according to claim 1,
The table data is
Provided with form size information indicating the size of the non-enacted form,
The form recognition device
Further comprising a form size calculation unit for calculating the size of the form from the acquired image,
The form classification determination unit
By comparing the form size obtained by the form size calculation unit with the form size information of each record in the table data, together with the keyword information, which non-established form corresponds to the form A form recognition device for judging. The form recognition device according to claim 1 or 2,
The predetermined recognition process is:
Recognize by cutting out the character string from the image to be recognized,
Calculate the item name likelihood representing the item name likelihood for each of the character strings,
Calculate the item value likelihood representing the likelihood of the item value for each character string,
For the character string pair for the character string, calculate an arrangement likelihood indicating whether the arrangement relationship of the character string pair is valid as the item name-item value relationship,
A form recognition apparatus that determines an item name and an item value based on the item name likelihood, the item value likelihood, and the placement likelihood. A form recognition method for recognizing a character string described in a form,
It is tabular data for classifying non-established forms, and for each record, includes keyword information for identifying the non-established form, and an area including a character string to be recognized in the non-established form Prepare table data with area information to show,
Obtain an image of the form,
Extract and recognize keyword information that identifies the form from the acquired image,
By comparing the obtained keyword information with the keyword information of each record in the table data, it is determined which non-enacted form the form corresponds to,
When it is determined that it does not correspond to any of the non-established forms, by performing a predetermined recognition process on the entire acquired image, the item name and item value described in the form are recognized as a recognition result. Output,
When it is determined that one of the non-established forms is determined, the area information is acquired from a record corresponding to the non-established form determined to be applicable, and the predetermined area is determined for the area determined by the area information. A form recognition method for outputting an item name and an item value described in the form as a recognition result by performing a recognition process. The form recognition method according to claim 4,
Determining whether the acquired image is that of the enacted form,
A form recognition method for executing a process after the extraction of the keyword information when it is not of the established form.

Images