JP2014238773A - Character recognition device, character recognition method, and character recognition program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically generate a definition body.SOLUTION: A character recognition device has an acquisition part, an extraction part, and a generation part. The acquisition part acquires an image of one or more items from an image of a paper surface. The extraction part extracts an image of the item with first size or larger from the image of one or more acquired items. The generation part generates a definition body for storing by associating a type of the paper surface from which the image of the item is acquired with the extracted image of the item.

Description

  The present invention relates to a technique for recognizing characters.

  In recent years, a character recognition device having an OCR (Optical Character Reader) function has been used to recognize characters written on a form (hereinafter also referred to as form character recognition). In the character recognition device, for example, when character recognition is performed on a form, a definition body that stores the position of the character written on the form, the type of character, and the like is used.

  In addition, when recognizing characters of multiple types of forms, the character recognition device stores the definition type used for character recognition of each form by further associating the type of form for character recognition with the specific figure of the form. To do. When the character recognition device recognizes characters in a form, it acquires a figure from the input form image and searches for a definition body that stores the same figure as the acquired figure from among a plurality of definition bodies. As a result, the character recognition device determines that the form for character recognition is the type of form indicated by the searched definition. Then, the character recognition apparatus executes a process for recognizing the character of the form by using the character position and the character type stored in the searched definition body.

  As a related technique, when the character written on the form is recognized by the character recognition unit, if the rejection rate is very high, the image data is stored in the image data storage unit and statistics of the recognition result are performed. If it is determined that the format information is necessary, the format information generation unit automatically generates the format information using the image data of the form. Also, a technique for re-registering font information is known for a reduction in recognition rate due to different fonts.

  As another related technique, the character recognition information storage means stores in advance character recognition information including position designation of a target range on a form for character recognition. Similarly, the syntax rule information storage means stores syntax rule information including a character recognition item represented by a character string in the target range and a rule for the character string in the target range. The character recognition means inputs the image information of the form optically read by the image scanner, extracts the image information of the target range based on the character recognition information read from the character recognition information storage means, and performs character recognition. The result of character recognition is sent to the parsing means. As the syntax analysis means, a technique is known in which a character string in a target range, which is a character recognition result, is analyzed based on syntax rule information and associated with a character recognition item.

  As another related technique, partial image data input by an image input device and stored in the image storage device is converted into ruled lines and character codes by image recognition processing and stored in the storage device. Format definition data is created from the ruled lines and character codes of the partial areas stored in the storage device, and stored in the format storage device. A technique for generating format definition data for the entire form by generating format data for each table from an input partial image and combining them is known.

  As another related technique, there is an image identification device including a master image input unit, a data image input unit, a mask area input unit, an image matching unit, a difference extraction unit, and a difference degree output unit. The master image input unit inputs a master image as the first image. The data image input unit inputs a data image as the second image. The mask area input unit inputs a set of mask areas designated for the master image. The image matching unit matches the master image and the data image. The difference extraction unit extracts a difference excluding the mask area between the matched master image and the data image. A technique for outputting the difference degree between the master image and the data image based on the extracted difference is known.

  As another related technique, the ruled line feature list extracted from the classified form is associated with the ruled line feature list of the reference form in the form format database, and the correction amount of the ruled line position is calculated by the correction amount detecting means, and the ruled line The ruled line position of the reference form is corrected by the feature correcting means. The ruled line feature collating means collates the ruled line feature list of the classified form with the ruled line feature list of the corrected standard form, and obtains the similarity. The classified form is classified as the same format as the reference form having the maximum similarity. A technique is known in which the correction amount is obtained from the correspondence of all combinations between the plurality of ruled line positions of the classified form and the plurality of ruled line positions of one reference form to be compared. (For example, Patent Documents 1 to 5).

JP-A-9-73500 Japanese Patent Laid-Open No. 2004-199529 JP-A-5-67189 JP 2013-61764 A JP 2003-109007 A

  In the above-described character recognition technology, for example, when the type of a form is determined using a specific figure, the user selects a figure specific to each form from the figures described in each form, and is used in each form. The selected figure is stored in advance in the definition body. Therefore, in the character recognition technology described above, when the types of forms for which characters are to be recognized increases, the task of generating a definition body may become complicated.

  As one aspect, the present invention provides a technique for automatically generating a definition body.

  One of the character recognition devices disclosed in this specification is a character recognition device having an acquisition unit, an extraction unit, and a generation unit. The acquisition unit acquires an image of one or more items from the image on the paper. An extraction part extracts the image of the item of 1st size or more from the acquired image of the 1 or more item. The generation unit generates a definition body that stores the type of the page from which the item image is acquired and the extracted item image in association with each other.

  According to one embodiment, the definition body can be automatically generated.

It is a functional block diagram which shows one Example of a character recognition apparatus. It is a flowchart which shows the process which discriminate | determines the kind of form. It is a flowchart which shows the process which discriminate | determines the kind of form. It is a flowchart which shows the process which produces | generates a definition body. It is a flowchart which shows the process which produces | generates a definition body. It is a flowchart which shows the process which produces | generates a definition body. It is a flowchart which shows the process which produces | generates a definition body. It is a flowchart which shows the process which produces | generates a definition body. It is a flowchart which shows the process which produces | generates a definition body. It is a figure which shows an example of a form. It is a figure which shows an example of a form. It is a figure showing an example of form discernment information. It is a figure which shows an example of character recognition information. It is a figure which shows an example of discrimination | determination data. It is a figure which shows an example of transaction data. It is a figure which shows an example of discrimination | determination data. It is a figure which shows an example of item data. It is a figure which shows an example of extraction data. It is a figure which shows an example of extraction data. It is a figure which shows an example of heading data. It is a figure explaining a recognition area. It is a figure explaining a recognition area. It is a block diagram which shows one Example of a computer apparatus.

The character recognition device of the embodiment will be described.
FIG. 1 is a functional block diagram showing an embodiment of a character recognition device.

A character recognition device 1 will be described with reference to FIG.
The character recognition device 1 includes a control unit 10, a storage unit 20, a reading unit 30, and a display unit 40. The character recognition device 1 is, for example, a computer device described later.

  The control unit 10 has functions of an acquisition unit 11, an extraction unit 12, a generation unit 13, a recognition unit 14, and a determination unit 15.

  The acquisition unit 11 acquires an image of one or more items from a paper image. The paper is, for example, a form, an answer sheet, a health checkup table, a questionnaire sheet, and the like, and is a sheet on which headings and data are described in association with each other. Items are, for example, figures and character strings written on paper. In the following description, a character string means a word including one or more characters.

  The extraction part 12 extracts the image of the item of 1st size or more from the acquired image of the 1 or more item. The first size is, for example, a threshold value used when extracting an image of an item that is a feature of the paper surface. The image of the item of the first size or larger is, for example, an image of an item that shows the characteristics of the page. In the following description, an item image indicating a feature of a paper surface is also referred to as a feature image.

  Further, the first size may include a vertical size and a horizontal size of the item image. At this time, the extraction unit 12 determines from the acquired image of one or more items that the vertical size is equal to or larger than the vertical size included in the first size, and the horizontal width is equal to or larger than the horizontal width included in the first size. Images of items may be extracted.

  When the acquired image of one or more items includes an image of one or more character strings, the extraction unit 12 includes a character string of a first character number or more including characters of a predetermined character size or more from the one or more character string images. Extract images. The predetermined character size is, for example, a character size threshold that is set to extract an image of a character string that is a feature of the page. The predetermined character size may include a vertical size and a horizontal size of the character. The predetermined character size may include a first character size and a second character size smaller than the first character size. The first number of characters is, for example, a threshold value used for extracting an image of a character string that is a feature of the page.

  The extraction unit 12 stores in the storage unit a definition body that stores the same ruled line layout as the layout of the ruled lines on the page, and the same item image as the item image extracted from the page using the first size. The image of the item of the second size or more smaller than the first size is extracted. The second size is a threshold used when, for example, an image of an item that is a feature of the paper surface is extracted.

  The extracting unit 12 stores in the storage unit a definition body that stores the same ruled line layout as the layout of the ruled lines on the page, and the same character string image as the character string image extracted from the page using the first number of characters. When stored, an image of a character string having a second character number or more smaller than the first character number is extracted. The second number of characters is, for example, a threshold used for extracting an image of a character string that is a feature of the page.

  The generation unit 13 generates a definition body that stores the type of the page from which the item image is acquired and the extracted item image in association with each other.

  The generation unit 13 generates a definition body that associates and stores the type of the paper surface from which the item image is acquired, the extracted item image, and an image area indicating the area in which the extracted item image is described.

  The generation unit 13 extracts the ruled line layout of the page from which the item image is acquired and the type of the page from which the item image was acquired when the definition unit storing the same ruled line layout is not stored in the storage unit. A definition body for storing the image of the item in association with the generated image is generated.

  When the data of the type corresponding to the item type is in the vicinity of the character string recognized as indicating the item type, the generation unit 13 recognizes the item type and an area in which the type of data corresponding to the item type is described. Generate a definition body that stores an area in association with it. The item type is, for example, the type of heading on the page. The type of the headline on the page means, for example, a bank name, a branch name, a deposit item, an account number, an amount, a payee, and a client when the page is a form. The data type refers to, for example, types such as kanji, kana, alphabetic characters, and numbers. In the following description, a character string recognized as indicating an item type is also referred to as a headline wording. The data corresponding to the item type is also referred to as item data.

  When the data is surrounded by ruled lines, the generation unit 13 sets the area surrounded by the ruled lines as a recognition area.

  When the data is not surrounded by ruled lines, the generation unit 13 surrounds the data and sets an area that does not include other words as a recognition area.

  When the acquired image of one or more items includes an image of a character string, the recognition unit 14 searches the item type associated with the same heading wording as the wording of the character string from the heading information, and the item type searched by the character string Recognize that

  The determination unit 15 searches for a definition body that stores an image of the same item as the acquired item image, and determines that the type of paper on which the item is acquired is the type of paper stored in the searched definition body. .

  The determination unit 15 searches for a definition body that stores the image of the same item as the acquired item image and the image area of the same item as the acquired item image area in association with each other. It is determined that the type of paper stored in the searched definition body.

  The storage unit 20 also stores definition body information 21, transaction information 22, discrimination information 23, item information 24, extraction information 25, heading information 26, and setting information 27.

  The definition body information 21 stores, for example, a definition body that stores information used for determining the type of paper for each type of paper for character recognition and for character recognition on the paper.

  The transaction information 22 stores, for example, transaction data that stores information regarding data corresponding to each item type read from the page for each type of the page on which character recognition is performed.

  In the discrimination information 23, for example, discrimination data storing information related to the result of discriminating the type of paper is stored.

  The item information 24 stores, for example, item data storing information about items acquired by the acquisition unit 11 from the paper for each type of paper on which character recognition is performed.

  In the extraction information 25, for example, extraction data storing information indicating items extracted by the extraction unit 12 is stored for each type of paper on which character recognition is performed.

  In the heading information 26, for example, heading data storing information on the heading wording used is stored for each item type on the page.

  The setting information 27 stores setting information such as a first size, a second size, a first character size, a second character size, a first character number, and a second character number.

  The reading unit 30 acquires a paper image. The reading unit 30 may be a scanner, for example. The reading unit 30 has a function of optically reading a paper image.

The display unit 40 displays information input from the control unit 10.
The character recognition device 1 will be described in more detail.

  In the following description, a form is used as an example of a paper surface on which the character recognition device 1 performs character recognition. However, the character recognition device 1 is not limited to a form, and can be applied to character recognition on various types of paper on which headings and data are described in association with each other.

2 and 3 are flowcharts showing processing for determining the type of form.
With reference to FIG. 2 and FIG. 3, processing for determining the type of form will be described. The process for recognizing item data will be described with reference to FIGS.

  In the following description, it is assumed that the definition unit information 21, the heading information 26, and the setting information 27 are stored in the storage unit 20 in advance. The form is assumed to be installed in the reading unit 30 by the user, for example. The reading unit 30 will be described as having read an image of a form 100 described later. Recognition of a figure, a character string, and a ruled line by the character recognition device 1 may be performed using, for example, an OCR function. Furthermore, for example, when the identifier or name is assigned to various data, the character recognition device 1 may assign an arbitrary identifier or name using an algorithm using a random number or an algorithm using a predetermined calculation. .

This will be described with reference to FIG.
The reading unit 30 reads the image of the form 100 shown in FIG. 10 (S101). Then, the reading unit 30 outputs the image of the form 100 to the acquisition unit 11.

The form 100 read by the reading unit 30 will be described with reference to FIG.
FIG. 10 is a diagram illustrating an example of a form.

  As shown in FIG. 10, the form 100 includes a bank name, a branch name, a deposit type, an account number, a transfer amount, a payee, and a client as headline words indicating item types. The bank name, branch name, deposit type, account number, transfer amount, payee, and client indicate the item name of the bank name, branch name, item, account number, amount, payee, and client, respectively. It is a headline wording. The form 100 includes figures SH1 and SH3 and character strings CH1 and CH2 as items having a size equal to or larger than the first size. Furthermore, the form 100 includes a graphic SH5 and a character string CH3 as items having a size smaller than the first size and equal to or larger than the second size.

This will be described with reference to FIG.
When the image of the form 100 is input from the reading unit 30, the acquisition unit 11 acquires the arrangement of ruled lines included in the image of the form 100 (S102). Then, the acquisition unit 11 outputs the acquired ruled line arrangement to the determination unit 15.

With reference to FIG. 11, processing for acquiring the arrangement of ruled lines by the acquisition unit 11 will be described.
FIG. 11 is a diagram illustrating an example of a form. FIG. 11 is an enlarged view of a part of the form 100 shown in FIG. In the following description, a process for acquiring the arrangement of the ruled line L1 will be described. The acquisition unit 11 similarly acquires the arrangement of ruled lines for other ruled lines (for example, ruled lines L2 to L7 shown in FIG. 11). However, the method by which the acquisition unit 11 acquires the arrangement of ruled lines is not limited to the method described below.

  The acquisition unit 11 extracts the ruled line L1 from the image of the form 100 using, for example, the OCR function. And the acquisition part 11 acquires ruled line coordinate (A1, B1)-(A2, B1) as arrangement | positioning of the ruled line L1. Ruled line coordinates (A1, B1)-(A2, B1) are information indicating that a straight line connecting coordinates (A1, B1) and coordinates (A2, B1) set on the form 100 is a ruled line L1. . The coordinates (A1, B1) are, for example, the coordinates of the starting point of the ruled line L1. The coordinates (A2, B1) are, for example, the coordinates of the end point of the ruled line L1.

This will be described with reference to FIG.
When the arrangement of the ruled lines of the form 100 is input from the acquisition unit 11, the determination unit 15 refers to the definition body stored in the definition body information 21 and defines the definition body (the same arrangement of the ruled lines as the form 100 ( Hereinafter, it is determined whether or not there is a definition body with matching ruled lines (S103). At this time, for example, the determination unit 15 receives an input of ruled line coordinates of each ruled line described in the form 100 from the acquiring unit 11, and receives the input ruled line coordinates and the definition body information stored in the definition object information 21. Judgment of coincidence with ruled line coordinates. Accordingly, the determination unit 15 may determine whether there is a definition body in which the form 100 and the ruled line match.

Information stored in the definition body will be described with reference to FIGS.
FIG. 12 is a diagram illustrating an example of the form determination information. FIG. 13 is a diagram illustrating an example of character recognition information. In the definition body, form discrimination information 200 shown in FIG. 12 and character recognition information 201 shown in FIG. 13 are stored. In the following description, information stored in a definition body corresponding to the form 100 will be described as an example. However, the information stored in the definition body is not limited to the form discrimination information 200 and the character recognition information 201, and is used when the character recognition device 1 discriminates the type of the page and recognizes the character on the page. Information only needs to be stored. Further, the character recognition device 1 may store a definition body of the same format for other forms for character recognition.

  As shown in FIG. 12, the form discriminating information 200 stores the form type, ruled line information, and feature information in association with each other.

  In the form type, a form identifier indicating the type of the form 100 is stored. The form N is a form identifier indicating the type of the form 100.

  The ruled line information is stored in association with ruled line identifiers and ruled line coordinates. The ruled line identifier indicates information for identifying each ruled line described in the form 100. The ruled line identifier may be assigned to each record when the character recognition device 1 generates ruled line information.

The feature information stores a feature identifier, an image area, and a feature image in association with each other.
The feature identifier indicates information for identifying each feature image described in the form 100. The feature identifier may be given to each record when the character recognition device 1 generates feature information.

  The image area indicates an area where a feature image is described in the form 100. The image regions (C3, D3)-(C4, D4) corresponding to the feature identifier SH1 are, for example, as shown in FIG. 11, the upper left coordinates (C3, D3) and lower right coordinates (C3, D3) of the rectangle surrounding the figure SH1. C4, D4). The image areas (C3, D3)-(C4, D4) are used as information indicating a rectangle having a diagonal line connecting the upper left coordinates (C3, D3) and the lower right coordinates (C4, D4). . Note that the rectangle indicated by the image area may be, for example, the smallest rectangle among the rectangles surrounding the feature image.

  The feature image is image data of an item unique to the form 100. The feature image IM2 is, for example, image data of the graphic 1 obtained by cutting out the area indicated by the image areas (C3, D3)-(C4, D4).

  In the character recognition information 201, as shown in FIG. 13, the item type, the recognition area, the data type, and the maximum number of digits are stored in association with the form type.

  The recognition area is information indicating an area in which item data corresponding to an item type is described in the form 100. As shown in FIG. 11, the recognition area (G1, H1)-(G2, H2) corresponding to the bank name (item type) is a coordinate (upper left corner of the rectangle surrounding Minami Tama, which is item data corresponding to the bank name). G1, H1) and lower right coordinates (G2, H2) are shown. Thereby, the recognition areas (G1, H1)-(G2, H2) are used as information indicating a rectangle whose diagonal line is a line connecting the upper left coordinates (G1, H1) and the lower right coordinates (G2, H2). It is done. Note that the rectangle indicated by the recognition area may be, for example, the smallest rectangle among the rectangles surrounding the item data.

  The maximum number of digits is information indicating the maximum number of characters of item data. And the recognition part 14 does not recognize the data using more characters than the maximum number of characters linked | related with the item kind as item data, for example, when performing the character recognition of the form 100. FIG.

This will be described with reference to FIG.
When the definition body that stores the same ruled line arrangement as the form 100 is not stored in the definition body information 21 (No in S103), the determination unit 15 determines that the type of the form 100 has no definition body that matches the ruled line. It is determined that it is a form (S104). Then, the determination unit 15 executes a process of S206 described later.

This will be described with reference to FIG.
When the definition body storing the same ruled line arrangement as the form 100 (hereinafter, also referred to as a definition body with matching ruled lines) is stored in the definition body information 21 (Yes in S103), the determination unit 15 The image area of each feature image stored in the matching definition body is acquired (S201).

  The determination unit 15 acquires an image of each item described in the image area acquired in S201 from the form 100 (S202). Note that, even if the item image described on the page and the feature image stored in the definition body are the same image, the determination unit 15 determines the same item as the feature image from the page when the image areas are different from each other. Do not get the image. In addition, when the item image described on the paper and the feature image stored in the definition body are the same image and the image areas are the same, the determination unit 15 obtains an image of the same item as the feature image from the paper. To do.

  Then, the determination unit 15 determines whether the image of each item acquired in S202 matches each feature image stored in the definition body (hereinafter also referred to as the feature image matches). (S203). In other words, the determination unit 15 executes S201 to S203, so that there is a definition body in which the same feature image as the acquired item image and the image region of the same feature image as the acquired item image are associated and stored. It is determined whether or not.

  When the image of each item acquired in S202 does not match all the feature images stored in the definition body (No in S203), the determination unit 15 has a definition body in which the ruled line matches the type of the form 100. However, it is determined that the definition form matches the ruled line and the new form does not match the feature image (S204). Then, the determination unit 15 executes a process of S206 described later.

  When the image of each item acquired in S202 and each feature image stored in the definition body all match (Yes in S203), the determination unit 15 determines that the type of the form 100 is an existing form. A determination is made (S205).

  Note that when the determination unit 15 determines in S103 that a plurality of definition bodies that match the form 100 and the ruled line are stored in the definition object information 21, the processing of S201 to S205 is performed on the definition bodies that match the ruled lines. May be executed.

  Then, the determination unit 15 associates the form identifier (form N) of the form 100, the image data (SPn) of the form 100, and the determination result (new: ruled line mismatch) of the form 100 and stores them in the determination data. (S206). At this time, the determination unit 15 may generate an arbitrary form identifier using a random number and store it in association with the image of the form 100.

With reference to FIG. 14, the information stored in the discrimination data will be described.
FIG. 14 is a diagram illustrating an example of the discrimination data.

  As shown in FIG. 14, the discrimination data 300 stores a form identifier, a form image, a discrimination result, a transaction identifier, and a ruled line matching form in association with each other.

  The form image stores image data of the form. The form image data is image data of the form read by the reading unit 30.

  The discrimination result is information indicating the discrimination result of the form type. “Existing” means that a definition body that stores ruled lines, feature images, and image areas of feature images stored in the definition body information 21 is stored in ruled lines, item images, and item image areas described in the form. Indicates that New: ruled line mismatch indicates that the ruled line described in the form and the definition body that matches the ruled line are not stored in the definition body information 21. New: Feature mismatch means that a ruled line described in a form and a definition that matches the ruled line are stored in the definition body information 21, but an image area of the feature image stored in the definition that matches the ruled line This indicates that no item matching the feature image is described in the corresponding area on the form.

  The transaction identifier is an identifier used when searching for transaction data corresponding to the form from the transaction information 22. In addition, when the character recognition apparatus 1 produces | generates the discrimination | determination data 300, you may provide the transaction identifier which shows corresponding transaction data to each record.

  The ruled line matching form is a form identifier of a definition that matches the form indicated by the form identifier and the ruled line.

This will be described with reference to FIG.
The recognition unit 14 acquires transaction data from the image of the form 100 (S207).

With reference to FIG. 15, information stored in the transaction data will be described.
FIG. 15 is a diagram illustrating an example of transaction data.

  As shown in FIG. 15, the transaction data 400 stores a form type, a transaction identifier, an item type, and item data in association with each other. Note that the transaction data 400 shows a transaction result corresponding to the description of the form 100 as an example.

  The transaction data 400 stores a form type, a transaction identifier, an item type, and item data.

  The transaction identifier is information for identifying transaction data of each form. In addition, you may provide a transaction identifier to each record, when the character recognition apparatus 1 produces | generates transaction data.

  However, the information stored in the transaction data 400 is not limited to the form type, the transaction identifier, the item type, and the item data, but stores information related to the transaction result recognized by the character recognition device 1 from the page. It ’s fine. The character recognition device 1 may store transaction data of the same format for other forms for character recognition.

  With reference to FIG. 13, FIG. 15, the process in which the recognition part 14 acquires the transaction data 400 in S207 is demonstrated.

  In the following description, an example will be described using information stored in the transaction data 400 corresponding to the form 100.

  The recognition unit 14 acquires each recognition area stored in the character recognition information 201 illustrated in FIG. Then, the recognition unit 14 searches the image of the form 100, acquires item data described at the position indicated by each recognition area, and stores each item data in the transaction data 400 in association with the item type.

  When there is item data that cannot be acquired from the image of the form 100 because the form 100 is dirty or the like, the recognition unit 14 displays the image of the form 100 on the display unit 40 and inputs a value to the transaction data 400 to the user. May be encouraged. Thereby, the user may store item data of item types that the recognition unit 14 could not acquire in the transaction data 400 while referring to the image of the form 100 displayed on the display unit 40.

  Further, the recognizing unit 14 displays the image of the form 100 on the display unit 40 and displays the transaction data to the user when the definition body that stores the character recognition information 201 corresponding to the form 100 is not stored in the definition body information 21. The input of a value to 400 may be prompted. Thus, the user may input item data corresponding to each item type in the transaction data 400 while referring to the image of the form 100 displayed on the display unit 40.

  When recognizing each item data, the recognizing unit 14 refers to the data type associated with the item type in the character recognition information 201, and determines a character recognition algorithm corresponding to the type of data to be subjected to character recognition. May be used to perform character recognition. Further, when recognizing each item data, the recognizing unit 14 may refer to the maximum number of digits associated with the item type in the character recognition information 201 and recognize only the data having the number of characters equal to or less than the maximum number of digits. . Thereby, the recognition part 14 can improve the precision of character recognition.

This will be described with reference to FIG.
The recognition unit 14 stores the transaction identifier (TRn) of the transaction data 400 in the discrimination data 301 in association with the form identifier (form N) of the form 100 (S208). As a result, the recognition unit 14 generates a record corresponding to the determination result of the form 100 as indicated by the determination data 301 in FIG. In FIG. 16, the forms identifier (form N), the form image (SPn), the discrimination result (new: ruled line mismatch), and the transaction identifier (TRn) corresponding to the discrimination result of the form 100 are obtained by the processing of S206 and S208. The discriminating data 301 including the record that stores) is shown.

  As described above, the character recognition device 1 executes the process of determining the type of the form 100 and the process of recognizing each item data described in the form 100.

4 to 9 are flowcharts showing processing for generating a definition body.
4 to 6 are flowcharts showing processing for generating the form discrimination information 200. FIG. 7 to 9 are flowcharts showing processing for generating the character recognition information 201.

With reference to FIGS. 4-6, the process which produces | generates the document discrimination | determination information 200 is demonstrated.
In the following description, the character recognition device 1 executes the process of determining the form type described with reference to FIGS. 2 and 3, and the determination data 301 including the determination result of each form that is the object of character recognition is generated. It shall be. Further, in the following description, a process for generating a definition body of the form 100 is shown as an example. Note that the character recognition device 1 may generate the form discrimination information by executing the processing described below for other forms.

  The acquisition unit 11 acquires a record from the discrimination data 301 (S301). At this time, the acquisition unit 11 acquires a record corresponding to the form N.

  The acquisition unit 11 acquires a form image included in the record acquired in S301 (S302).

  Then, the acquisition unit 11 acquires the image of the item included in the form image acquired in S302 and stores it in the item data (S303). For example, when the form from which the item image is acquired is the form 100, the acquisition unit 11 uses the OCR function to extract the image area of the item and the graphic and character strings described in the form 100 illustrated in FIG. You may get

FIG. 17 is a diagram illustrating an example of item data.
The item data 500 indicates item data corresponding to the form 100 as an example.

  In the item data 500, as shown in FIG. 17, a form type, an item identifier, an item name, an image area, and an item image are stored in association with each other.

  The item identifier is information for identifying each item described in the form 100. The item identifier may be assigned to each record when the character recognition device 1 generates item data.

  The item name is information indicating the name of each item described in the form 100. In the item name, for example, when the item is a character string, a character string recognized by the recognition unit 14 may be stored.

  The image area indicates an area in which an item is described in the form 100. The image regions (C3, D3)-(C4, D4) corresponding to the item identifier SH1 are, for example, as shown in FIG. 11, the upper left coordinates (C3, D3) and the lower right coordinates (C3, D3) of the rectangle surrounding the figure SH1. C4, D4). Thus, the image areas (C3, D3)-(C4, D4) are used as information indicating a rectangle whose diagonal line is a line connecting the upper left coordinates (C3, D3) and the lower right coordinates (C4, D4). It is done. Note that the rectangle indicated by the image area may be, for example, the smallest rectangle among the rectangles surrounding the item.

  The item image is image data of an item described in the form 100. The item image IM2 is, for example, the image data of the graphic 1 obtained by cutting out the area indicated by the image areas (C3, D3)-(C4, D4).

This will be described with reference to FIG.
The acquisition unit 11 determines whether or not the determination result included in the record acquired in S301 is new (S304).

  When the determination result included in the record acquired in S301 is not new (existing) (No in S304), the acquiring unit 11 executes the process of S601 described later.

  When the determination result included in the record acquired in S301 is new (Yes in S304), the acquisition unit 11 determines whether the determination result included in the record acquired in S301 is a ruled line mismatch (S305). ).

  When the determination result included in the record acquired in S301 does not match the ruled line (No in S305), the acquiring unit 11 refers to the definition body corresponding to the form with the matched ruled line and acquires the ruled line coordinates (S306). Then, the acquisition unit 11 outputs the acquired ruled line coordinates to the generation unit 13. The generation unit 13 executes a process of S308 described later.

  When the determination result included in the record acquired in S301 does not match the ruled line (Yes in S305), the acquiring unit 11 acquires the ruled line coordinates included in the form image (S307). Then, the acquisition unit 11 outputs the acquired ruled line coordinates to the generation unit 13. At this time, for example, when the form is the form 100, the obtaining unit 11 may obtain the ruled line coordinates described in the form 100 shown in FIG. 11 using the OCR function. Note that the form 100 (form identifier: form N) corresponds to the case of Yes in S305 because there is no ruled line matching form as shown in the discrimination data 301 of FIG.

  When the ruled line coordinates are input, the generating unit 13 generates a new definition body that stores the ruled line identifier and the acquired ruled line coordinates in association with the ruled line information (S308). For example, when generating the definition body of the form 100, the generation unit 13 may store the ruled line identifier of the form determination information 200 illustrated in FIG. 12 and information indicated by the ruled line coordinates in the definition body. The form identifier stored in the form type of the definition body may store the form identifier included in the record acquired in S301. In the case of the form 100, for example, the form N may be stored in the form type.

The generation unit 13 stores the generated new definition body in the definition body information 21 (S309).
This will be described with reference to FIG.

  The extraction unit 12 acquires the first size from the setting information 27. Then, the extraction unit 12 searches the image area stored in the item data 500, and the vertical size is equal to or larger than the vertical size included in the first size, and the horizontal size includes the horizontal size included in the first size. The image of the above item is extracted (S401). In addition, when the item stored in the item data 500 is a character string, the extraction unit 12 uses the first character size and the vertical size is equal to or larger than the vertical width included in the first character size, and the horizontal width You may extract the image of the item more than the size of the width included in the 1st character size.

  Then, the extracting unit 12 determines whether or not the extracted item image is a character string image (S402).

  When the extracted image of the extracted item is not a character string image (a graphic) (No in S402), the extraction unit 12 executes the process of S404 described later.

  When the extracted item image is a character string image (Yes in S402), the extraction unit 12 determines whether the number of characters included in the character string image is equal to or greater than the first character number (S403).

  When the number of characters included in the image of the character string is not equal to or greater than the first character number (No in S403), the extracting unit 12 performs the process of S406 described later.

  When the number of characters included in the character string image is greater than or equal to the first number of characters (Yes in S403), the extraction unit 12 executes the process of S404. That is, when the image of the item extracted in S401 is a character string image, the extraction unit 12 extracts the character string image as a feature image candidate when the number of characters in the character string is equal to or greater than the first character number.

  Furthermore, the production | generation part 13 acquires the record containing the item extracted from item data (S404).

  And the production | generation part 13 stores the record acquired by S404 in extraction data (S405).

FIG. 18 is a diagram illustrating an example of extracted data.
In the extracted data 600, as shown in FIG. 18, an item identifier, an item name, an image area, and an item image are stored in association with each other. The extracted data 600 is data corresponding to the form 100. The extracted data 600 stores a result of extracting an image of an item of the first size or more from an item image described in the form 100. That is, the extracted data 600 stores feature image candidates for the form 100.

  Then, the extraction unit 12 determines whether or not the search has been completed for all image regions stored in the item data (S406). That is, the extraction unit 12 determines whether all items of the first size or larger have been extracted.

  When all the image areas stored in the item data are not searched (No in S406), the extraction unit 12 performs the process of S401.

  When all the image areas stored in the item data are searched (Yes in S406), the extraction unit 12 executes the process of S501. That is, the extraction unit 12 executes the processing of S501 when images of items that are candidates for all feature images are extracted from the form.

This will be described with reference to FIG.
The generation unit 13 determines whether the determination result included in the record acquired in S301 is a ruled line mismatch (S501).

  When the determination result included in the record acquired in S301 does not match the ruled line (Yes in S501), the generation unit 13 stores each record of the extracted data in a new definition body (S502). For example, the generation unit 13 adds the item identifier, the image area, and the item image of the extracted data 600 illustrated in FIG. 18 to the definition body of the form 100 generated in S308, respectively, the item identifier of the form determination information 200 illustrated in FIG. Store in the image area and feature image. That is, the generation unit 13 generates the form determination information 200 for the definition body of the form 100 by executing S308 and S502. And the process of S601 mentioned later is performed.

  When the determination result included in the record acquired in S301 is a ruled line match (No in S501), the generation unit 13 stores the image area stored in the definition body corresponding to the ruled line match form, and the image area of the extracted item Is matched (S503). At this time, the generation unit 13 sets a predetermined error in consideration of a reading error by the reading unit 30 and the image region stored in the definition body corresponding to the ruled line matching form if within the error range. Alternatively, it may be determined that the image area of the extracted item matches. The definition body corresponding to the ruled line matching form is, for example, a definition body corresponding to the form identifier stored in the ruled line matching form included in the record acquired in S301.

  When the image area stored in the definition body corresponding to the ruled line matching form does not match the image area of the extracted item (No in S503), the generation unit 13 executes the process of S502. At this time, in S502, the generation unit 13 associates the ruled line information stored in the definition body corresponding to the ruled line matching form with the feature information storing the extracted data generated in the processes of S401 to S406, and stores the new information. Generate a definition body.

  When the image area stored in the definition body corresponding to the ruled line matching form matches the image area of the extracted item (Yes in S503), the generation unit 13 stores the definition area corresponding to the ruled line matching form. It is determined whether or not the feature image being made matches the image of the extracted item (S504).

  When the feature image stored in the definition body corresponding to the ruled line matching form does not match the extracted item image (No in S504), the generation unit 13 executes the process of S502. At this time, in S502, the generation unit 13 associates the ruled line information stored in the definition body corresponding to the ruled line matching form with the feature information storing the extracted data generated in the processes of S401 to S406, and stores the new information. Generate a definition body.

  When the feature image stored in the definition body corresponding to the ruled line matching form matches the extracted item image (Yes in S504), the generation unit 13 determines the unused size threshold value and the unused size image. It is determined whether the threshold value for the number of characters is stored in the setting information 27 (S505). Here, the unused size threshold is, for example, a second size smaller than the first size stored in the setting information 27 when the first size is used in S401. The unused character count threshold is, for example, the second character count that is smaller than the first character count stored in the setting information 27 when the first character count is used in S403.

  When the unused size threshold value and the unused character count threshold value are not stored in the setting information 27 (No in S505), the generation unit 13 executes the process of S601 described later. At this time, the generation unit 13 may end the process of generating the definition body as an error. The generation unit 13 may display error information and a form image on the display unit 40 and prompt the user to generate the form determination information 200 by manual input. Then, after the user generates the form determination information 200, the generation unit 13 may execute the process of S601 described later.

  The generation unit 13 changes the size threshold and the character count threshold when the unused size threshold and the character count threshold are stored in the setting information 27 (Yes in S505). The process is executed (S506).

FIG. 19 is a diagram illustrating an example of extracted data.
The extracted data when the size threshold and the character count threshold are changed in S506 will be described with reference to FIG. 10, FIG. 18, and FIG. In the following description, it is assumed that the generation unit 13 changes the size threshold value from the first size to the second size, and changes the character number threshold value from the first character number to the second character number. The first number of characters is assumed to be 5 characters, for example. In addition, the second number of characters is assumed to be two characters, for example.

  The extracted data 600 shown in FIG. 18 is extracted data when the extracting unit 12 extracts an image of an item using the first size and the first number of characters in the form 100. The extracted data 600 stores the figures SH1 and SH3 that are extracted by the extracting unit 12 and have a size equal to or larger than the first size. Further, in the extracted data 600, a payment handling table (CH2) which is an image of a character string having a size equal to or larger than the first size and having a number of characters equal to or larger than the first number of characters (5 characters) extracted by the extracting unit 12. ) Is stored.

  The extracted data 600 illustrated in FIG. 19 is extracted data when the extracting unit 12 extracts an image of an item using the second size and the second number of characters in the form 100. The extracted data 600 stores a figure SH1, a figure SH3, and a figure SH5 that are extracted by the extraction unit 12 and have a size equal to or larger than the second size. Furthermore, the extracted data 600 includes Tokyo (CH1), which is an image of a character string having a size equal to or larger than the second size and having a number of characters equal to or larger than the second number (2 characters) extracted by the extraction unit 12. A payment handling table (CH2) is stored.

  In S505, when one or more of the unused size threshold and the character count threshold is stored in the setting information 27, the generation unit 13 determines in S506 which of the size threshold and the character count threshold. Or one or more may be changed. In addition, when the second character size that is smaller than the first character size is stored in the setting information 27, the generation unit 13 may change the first character size to the second character size in S506. .

  Thus, the character recognition device 1 ends the process of generating the form discrimination information 200. Subsequently, the character recognition device 1 executes processing for generating the character recognition information 201.

Processing for generating the character recognition information 201 will be described with reference to FIGS.
This will be described with reference to FIG. In the following description, generation of the character recognition information 201 of the form 100 will be described as an example. Note that the character recognition device 1 may generate character recognition information for other forms by executing the processing described below.

  The generation unit 13 acquires a character string (item name) from the item data 500 (S601). The generation unit 13 may acquire an unacquired character string from the item data 500 when executing the process of S601 following the process of S604 described later.

  Then, the generation unit 13 searches the heading data 700 and determines whether or not there is the same heading wording as the character string acquired in S601 (S602).

FIG. 20 is a diagram illustrating an example of heading data.
The heading data 700 stores a heading identifier, item type, heading wording, data type, and maximum number of digits.

The heading identifier is information for identifying each heading.
The generation unit 13 searches the heading data 700, and when there is the same heading wording as the character string acquired in S601 (Yes in S602), the generation unit 13 recognizes that the character string is a heading indicating the item type associated with the heading wording. (S603). Then, the generation unit 13 stores the character string in the setting information 27 as the headline wording indicating the item type. Thereby, in the setting information 27, each item type and the corresponding headline wording are stored in association with each other. And the production | generation part 13 performs the process of S604.

  The generation unit 13 searches the heading data 700, and when there is no heading wording that is the same as the character string acquired in S602 (No in S602), executes the processing of S604.

  The generation unit 13 determines whether all the character strings of the item data 500 have been acquired in S601 (S604).

  When the generation unit 13 has not acquired all the character strings of the item data 500 in S601 (No in S604), the generation unit 13 executes the process of S601.

  When the generating unit 13 acquires all the character strings of the item data 500 in S601 (Yes in S604), the generating unit 13 executes the process of S701. At this time, the generation unit 13 recognizes all the headline wordings corresponding to the item types described in the form 100. Note that when the generation unit 13 cannot recognize all the headline wordings corresponding to the item types described in the form 100, the generation unit 13 may end the process of generating the definition body as an error. At this time, the generation unit 13 may prompt the user to generate a definition body manually by displaying error information on the display unit 40.

This will be described with reference to FIG.
The generation unit 13 acquires a character string (item name) from the item data 500 (S701). The generation unit 13 may acquire an unacquired character string from the item data 500 when executing the processing of S701 following the processing of S703 described later.

  The generation unit 13 determines whether there is item data indicating the same character string as the character string acquired in S701 in the transaction data 400 (S702).

  When there is no item data indicating the same character string as the character string acquired in S701 in the transaction data 400 (No in S702), the generation unit 13 determines whether or not all the character strings of the item data 500 are acquired in S701. Determination is made (S703).

  When the generation unit 13 has not acquired all the character strings of the item data 500 in S701 (No in S703), the generation unit 13 performs the process of S701.

  When the generation unit 13 acquires all the character strings of the item data 500 in S701 (Yes in S703), the generation unit 13 ends the process of generating the definition body. At this time, if it is determined in S805, which will be described later, that the records corresponding to all the item types have not been generated, the generation unit 13 may end the process of generating the definition body as an error. Then, the generation unit 13 may cause the display unit 40 to display error information and a record of the character recognition information 201 that has not been generated, and prompt the user to generate a definition body by manual input.

  In S702, when the transaction data 400 includes item data indicating the same character string as the character string acquired in S701 (Yes in S702), the generation unit 13 is associated with item data indicating the same character string from the transaction data 400. The obtained item type is acquired (S704).

  The generation unit 13 refers to the setting information 27 and acquires a headline wording indicating the item type acquired in S704 (S705).

  Then, the generation unit 13 determines whether the character string acquired in S701 is in the vicinity of the headline wording indicating the item type acquired in S705 (S706). For example, the generation unit 13 acquires an image area corresponding to the headline wording indicating the item type acquired in S705 from the item data 500, and determines the positional relationship with the image area corresponding to the character string acquired in S701. Thereby, the generation unit 13 determines whether or not the character string acquired in S701 is near the heading indicating the item type acquired in S705. The vicinity of a headline differs depending on, for example, the type of paper and the type of headline. The vicinity of the heading means, for example, a predetermined area under the heading, the lower right, and the right side when the paper surface is the form 100.

  When the character string acquired in S701 is not in the vicinity of the heading indicating the item type acquired in S705 (No in S706), the generation unit 13 executes the process of S703.

  When the character string acquired in S701 is in the vicinity of the heading indicating the item type acquired in S705 (Yes in S706), the generation unit 13 recognizes that the character string is item data corresponding to the item type ( S707).

  Note that the generation unit 13 may acquire the data type corresponding to the headline wording acquired in S705 from the headline data 700 in S706. Then, the generation unit 13 may execute the process of S707 when the acquired data type corresponds to the character string type acquired in S701. Further, the generation unit 13 may execute the process of S703 when the acquired data type does not correspond to the character string type acquired in S701.

  The generation unit 13 acquires a recognition area surrounding the character string recognized as item data in S707 (S708). Then, the process of S801 is executed.

21 and 22 are diagrams illustrating the recognition area.
The setting of the recognition area will be described with reference to FIG. In the following description, processing for setting a recognition area for corresponding item data for a headline wording (transfer amount) corresponding to an item type (amount) will be described.

  As shown in FIG. 21, when the item data 800 is surrounded by ruled lines, the generation unit 13 coordinates the upper left corner (G9, H9) and the lower right coordinates (G10, H10) of the ruled line surrounding the item data 800. And get. Thereby, the generation unit 13 displays an image area (G9, H9)-(G10, H10) indicating a rectangle having a diagonal line connecting the upper left coordinates (G9, H9) and the lower right coordinates (G10, H10). To get. Then, the generation unit 13 sets the acquired image region (G9, H9) − (G10, H10) as a recognition region for recognizing the item data 800.

  When the item data 801 is not surrounded by ruled lines as shown in FIG. 22A, the generation unit 13 surrounds the item data 801 and does not include other words as shown in FIG. 22B. Creates a rectangle that encloses Then, the generation unit 13 acquires the upper left coordinates (G9, H9) of the generated rectangle and the lower right coordinates (G10, H10) of the rectangle. Thereby, the generation unit 13 displays an image area (G9, H9)-(G10, H10) indicating a rectangle having a diagonal line connecting the upper left coordinates (G9, H9) and the lower right coordinates (G10, H10). To get. Then, the generation unit 13 sets the acquired image region (G9, H9) − (G10, H10) as a recognition region for recognizing the item data 800. Note that the rectangle indicated by the recognition area may be, for example, the smallest rectangle among the rectangles surrounding the item data.

This will be described with reference to FIG.
The generation unit 13 determines whether or not the determination result included in the record acquired in S301 is new (S801).

  When the determination result included in the record acquired in S301 is new (Yes in S801), the generation unit 13 generates a record corresponding to the item type acquired in S704 in the character recognition information 201. Then, the generation unit 13 stores the recognition area acquired in S708 in the character recognition information 201 in association with the item type acquired in S704 (S802).

  The generation unit 13 searches the heading data 700 and acquires a record including the item type acquired in S704 (S803).

  The generation unit 13 stores the information included in the acquired record in the character recognition information 201 (S804). That is, the generation unit 13 acquires the data type and the maximum number of digits from the record including the item type acquired in S704, and stores the acquired information in the character recognition information 201 in association with the item type acquired in S704.

  Then, the generation unit 13 determines whether records corresponding to all item types have been generated (S805).

  When the generation unit 13 has not generated records corresponding to all the item types (No in S805), the generation unit 13 executes the process of S701.

  When the generation unit 13 generates records corresponding to all item types (Yes in S805), the generation unit 13 ends the process of generating the definition body.

  In S801, when the determination result included in the record acquired in S301 already exists (No in S801), the generation unit 13 determines from the definition corresponding to the ruled line matching form included in the record acquired in S301 in S704. A recognition area associated with the acquired item type (hereinafter also referred to as an existing recognition area) is acquired. Then, the generation unit 13 determines whether or not the recognition area acquired in S708 is different from the existing recognition area (S806).

  When the recognition area acquired in S708 is the same as the existing recognition area (No in S806), the generation unit 13 executes the process of S805.

  When the recognition area acquired in S708 is different from the existing recognition area (Yes in S806), the generation unit 13 changes the recognition area stored in the character recognition information 201 of the existing form to the recognition area acquired in S708. Update (S807). Then, the generation unit 13 executes the process of S805. As a result, when only the item data recognition area is changed in the existing form, the generation unit 13 automatically updates the definition area recognition area corresponding to the existing form.

FIG. 23 is a block diagram illustrating an embodiment of a computer device.
The configuration of the character recognition device 1 will be described with reference to FIG.

  23, a computer device 900 includes a control circuit 901, a storage device 902, a reading device 903, a recording medium 904, a communication interface 905 (communication I / F), and an input / output interface 906 (input / output I / F). ), A display device 907, and a network 908. Each component is connected by a bus 909.

  The control circuit 901 controls the entire computer device 900. The control circuit 901 is, for example, a CPU, a multi-core CPU, an FPGA (Field Programmable Gate Array), a PLD (Programmable Logic Device), or the like. For example, the control circuit 901 functions as the control unit 10 in FIG. Note that the CPU, FPGA, and PLD cache may store, for example, the setting information 27 shown in FIG.

  The storage device 902 stores various data. The storage device 902 includes, for example, a memory such as a ROM (Read Only Memory) and a RAM (Random Access Memory), an HD (Hard Disk), and the like. For example, the storage device 902 functions as the storage unit 20 in FIG. Then, the storage device 902 includes, for example, definition body information 21, transaction information 22, discrimination information 23, item information 24, extraction information 25, heading information 26, and setting information 27 shown in FIG. You may remember.

  The ROM stores a program such as a boot program. The RAM is used as a work area for the control circuit 901. The HD stores an OS, an application program, a program such as firmware, and various data.

  The storage device 902 stores, for example, a character recognition program that causes the control circuit 901 to function as the control unit 10.

  When the character recognition device 1 performs processing for determining the type of form or processing for generating a definition body, the character recognition program 1 reads the character recognition program stored in the storage device 902 into the RAM. Then, the character recognition device 1 executes a process for determining the type of the form and a process for generating a definition by executing the character recognition program read into the RAM by the control circuit 901.

  Note that the character recognition program may be stored in a storage device included in a server on the network 908 as long as the control circuit 901 is accessible via the communication interface 905.

  The reading device 903 is controlled by the control circuit 901 and reads / writes data on the removable recording medium 904. The reading device 903 includes, for example, an FDD (Floppy Disk Drive), a CDD (Compact Disc Drive), a DVDD (Digital Versatile Disk Drive), a BDD (Blu-ray (registered trademark) U USB Ver. Etc.

  The recording medium 904 stores various data. The recording medium 904 stores, for example, a character recognition program. Further, the recording medium 904 includes, for example, definition body information 21, transaction information 22, discrimination information 23, item information 24, extraction information 25, heading information 26, and setting information 27 shown in FIG. You may remember.

  The recording medium 904 is connected to the bus 909 via the reading device 903, and the control circuit 901 controls the reading device 903 to read / write data. The recording medium 904 is, for example, an FD (Floppy Disk), a CD (Compact Disc), a DVD (Digital Versatile Disk), a BD (Blu-ray (registered trademark) Disk), and a flash memory.

  The communication interface 905 connects the computer apparatus 900 and other apparatuses via a network 908 so that they can communicate with each other.

  The input / output interface 906 is connected to, for example, a keyboard, a mouse, a touch panel, a scanner, and the like. When signals indicating various types of information are input from the connected devices, the control circuit 901 receives the signals input via the bus 909. Output to. When a signal indicating various information output from the control circuit 901 is input via the bus 909, the input / output interface 906 outputs the signal to various connected devices. The input / output interface 906 may accept input of setting values for the first size, the second size, the first character size, the second character size, the first character number, and the second character number, for example. The scanner connected to the input / output interface 906 functions as the reading unit 30 shown in FIG. 1, for example.

  The display device 907 is connected to the input / output interface 906, for example, and displays various types of information based on signals input from the control unit 10. The display device 907 is, for example, a CRT (Cathode Ray Tube), an LCD (Liquid Crystal Display), a PDP (Plasma Display Panel), or an OELD (Organic Electroluminescence Display). The display device 907 functions as the display unit 40 in FIG.

  The network 908 is, for example, a LAN, wireless communication, the Internet, or the like, and connects the computer apparatus 900 and other apparatuses for communication.

  As described above, the character recognition device 1 according to the embodiment extracts an image having a predetermined size or more from an item image described on a sheet, and generates a definition body that stores the extracted item image as a feature image. Thereby, the character recognition device 1 of the embodiment can automatically generate the definition body.

  The character recognition device 1 according to the embodiment acquires an image of an item from a paper image read by the reading unit 30 and searches the feature image stored in the definition body for the same image as the acquired image of the item. Thereby, the character recognition device 1 of the embodiment can recognize that the type of the page read by the reading unit 30 is the type of the page stored in the definition body storing the searched feature image.

  The character recognition device 1 according to the embodiment extracts an image of an item having a vertical size and a horizontal size that are equal to or larger than a predetermined size from the item images acquired from the page, and features the extracted item image. Generate a definition body to be stored as an image. Therefore, the character recognition device 1 according to the embodiment does not extract a sentence, a line, or the like in which only one of the vertical width and the horizontal width is long. As a result, the character recognition device 1 according to the embodiment can accurately extract a feature image that is not on the other paper from the image of the item described on the paper.

  When the item image includes a character string image, the character recognition device 1 according to the embodiment generates a definition body that stores, as a feature image, an item image in which the characters included in the character string are a predetermined size or larger. Therefore, the character recognition device 1 according to the embodiment does not extract an image of an item whose character in the character string is smaller than a predetermined size as a feature image. As a result, the character recognition device 1 according to the embodiment can accurately extract a feature image that is not on the other paper from the image of the item described on the paper.

  When the item image includes a character string image, the character recognition device 1 according to the embodiment generates a definition body that stores, as a feature image, an item image in which the number of characters included in the character string is equal to or greater than a predetermined number of characters. Therefore, the character recognition device 1 of the embodiment does not extract an image of an item in which the number of characters in the character string is smaller than the predetermined number of characters as a feature image. As a result, the character recognition device 1 according to the embodiment can accurately extract a feature image that is not on the other paper from the image of the item described on the paper.

  The character recognition device 1 according to the embodiment generates a definition body that stores an image region indicating a region in which a feature image is described and a feature image in association with each other. When the character recognition device 1 according to the embodiment determines the type of the page, the character recognition apparatus 1 determines whether the image area corresponding to each of the image of the item described on the page and the feature image stored in the definition body matches. To do. As a result, when the image areas stored in the page and the definition body match, the character recognition device 1 of the embodiment stores the type of the page read by the reading unit 30 in the definition body in which the image area matches. It can be recognized that it is a type of paper.

  The character recognition device 1 according to the embodiment, when the storage unit 20 does not store the definition of the ruled line layout of the paper surface from which the item image is acquired and the same ruled line layout, the paper surface from which the item image is acquired. A definition body that stores the type of the item and the extracted item image in association with each other is generated. Thereby, since the character recognition apparatus 1 of embodiment does not perform the process which produces | generates the existing definition body redundantly, it can suppress complication of a process.

  The character recognition device 1 of the embodiment stores a first size and a second size as threshold values used when extracting a feature image. The character recognition device 1 according to the embodiment stores a definition of a ruled line that is the same as the layout of the ruled line on the page, and an image of the same item as the image of the item extracted from the page using the first size. Is stored in the storage unit 20, an image of an item of the second size or larger is extracted. Thereby, the character recognition device 1 according to the embodiment gradually reduces the size of the item image to be extracted as the feature image, so that the number of feature images to be extracted increases and the processing becomes complicated, and the like. It is possible to extract feature images that are not in the form.

  The character recognition device 1 according to the embodiment includes a definition body that stores the same ruled line arrangement as the arrangement of the ruled lines on the page, and the same character string image as the character string image extracted from the page using the first number of characters. Is stored in the storage unit 20, an image of a character string having a number of characters equal to or larger than a second character number smaller than the first character number is extracted. Thereby, since the character recognition device 1 of the embodiment reduces the number of characters of the character string extracted as the feature image step by step, the number of feature images to be extracted increases and the processing becomes complicated, and It is possible to extract feature images that do not exist in other forms.

  In the character recognition device 1 of the embodiment, when the type of data corresponding to the item type is in the vicinity of the character string recognized as indicating the item type, the item type and the type of data corresponding to the item type are described. A definition body is generated that stores an association with a recognition area indicating the area. Thereby, the character recognition device 1 of the embodiment can automate the creation of a definition body indicating a recognition area of data corresponding to the item type described on the paper.

  When the data corresponding to the item type is surrounded by ruled lines, the character recognition device 1 of the embodiment sets the area surrounded by the ruled lines as a recognition area. Thereby, the character recognition device 1 of the embodiment can automatically set the recognition area.

  When the data is not surrounded by ruled lines, the character recognition device 1 of the embodiment surrounds the data and sets a region that does not include other words as a recognition region. Thereby, the character recognition device 1 of the embodiment can automatically set the recognition area.

  The character recognition device 1 according to the embodiment searches for a definition body that stores the same feature image as the image of the item acquired from the page, and determines the type of the page from which the item is acquired as the type of the page stored in the searched definition body. It is determined that Thereby, the character recognition apparatus 1 of embodiment can discriminate | determine the kind of paper surface using a definition body.

  The character recognition device 1 according to the embodiment searches for a definition body in which the same feature image as the acquired item image and the image area of the same feature image as the acquired item are stored in association with each other, and acquires the item. Is determined to be the type of the paper stored in the searched definition body. Thereby, the character recognition apparatus 1 of embodiment can improve the precision which discriminate | determines the kind of paper surface using a definition body.

  In addition, this embodiment is not limited to embodiment described above, A various structure or embodiment can be taken in the range which does not deviate from the summary of this embodiment.

DESCRIPTION OF SYMBOLS 1 Character recognition apparatus 10 Control part 11 Acquisition part 12 Extraction part 13 Generation part 14 Recognition part 15 Discrimination part 20 Storage part 21 Definition body information 22 Transaction information 23 Discrimination information 24 Item information 25 Extraction information 26 Information 27 Setting information 30 Reading part 40 Display unit 100 Form 200 Form discrimination information 201 Character recognition information 300, 301 Discrimination data 400 Transaction data 500 Item data 600 Extraction data 700 Heading data 800, 801 Item data 900 Computer device 901 Control circuit 902 Storage device 903 Reading device 904 Recording medium 905 Communication interface 906 Input / output interface 907 Display device 908 Network 909 Bus

Claims (14)

An acquisition unit that acquires an image of one or more items from an image on a paper surface;
An extraction unit that extracts an image of an item of a first size or more from the acquired image of the one or more items;
A generating unit that generates a definition body that associates and stores the type of the page from which the image of the item is acquired and the image of the extracted item;
A character recognition device comprising: The first size is
Including the vertical and horizontal dimensions of the item image,
The extraction unit includes:
From the acquired image of one or more items, an image of an item whose vertical size is equal to or larger than the vertical size included in the first size, and whose horizontal width is equal to or larger than the horizontal width included in the first size. The character recognition device according to claim 1, wherein the character recognition device is extracted. The extraction unit includes:
When the acquired image of one or more items includes an image of one or more character strings, an image of a character string of a first character number or more including characters of a predetermined character size or more from the one or more character string images. The character recognition device according to claim 1, wherein the character recognition device is extracted. The generator is
Generating a definition body that associates and stores the type of the page from which the image of the item is acquired, the image of the extracted item, and an image area indicating an area in which the image of the extracted item is described. The character recognition apparatus as described in any one of Claims 1-3. The character recognition device further includes:
A storage unit for storing one or more definition bodies;
The definition body further includes:
Store the paper type and ruled line layout in association with each other.
The generator is
The layout of the ruled lines of the paper surface from which the image of the item is acquired and the definition body storing the same layout of the ruled lines are not stored in the storage unit, the type of the paper surface from which the image of the item was acquired, and the extracted The character recognition apparatus according to claim 1, wherein a definition body that stores an image of an item in association with the image is generated. The character recognition device further includes:
A storage unit for storing one or more definition bodies;
The definition body further includes:
Store the paper type and ruled line layout in association with each other.
The extraction unit includes:
A definition body storing an arrangement of the same ruled line as the arrangement of the ruled line on the page and an image of the item extracted from the page using the first size is stored in the storage unit. 6. The character recognition device according to claim 1, wherein an image of an item of a second size or larger that is smaller than the first size is extracted. The character recognition device further includes:
A storage unit for storing one or more definition bodies;
The definition body further includes:
Store the paper type and ruled line layout in association with each other.
The extraction unit includes:
The definition unit storing the same ruled line arrangement as the ruled line arrangement on the page and the same character string image extracted from the page using the first number of characters is stored in the storage unit. 4. The character recognition device according to claim 3, wherein an image of a character string equal to or greater than a second character number smaller than the first character number is extracted. The storage unit
Store heading information to store the heading item type and heading wording in association with each other,
The character recognition device further includes:
When the acquired image of one or more items includes an image of a character string, the item type associated with the same headline wording as the wording of the character string is searched from the heading information, and the item type searched by the character string is A recognition unit that recognizes when
The generator is
When the type of data corresponding to the item type is in the vicinity of the character string recognized as indicating the item type, the item type and a recognition area indicating an area in which the type of data corresponding to the item type is described; The character recognition device according to claim 1, wherein a definition body that stores the association is generated. The generator is
The character recognition device according to claim 8, wherein when the data is surrounded by ruled lines, an area surrounded by the ruled lines is used as the recognition area. The generator is
The character recognition device according to claim 8, wherein when the data is not surrounded by ruled lines, an area that surrounds the data and does not include other words is used as the recognition area. The character recognition device further includes:
A storage unit for storing one or more definition bodies;
A determination unit that searches for a definition body that stores an image of the same item as the acquired image of the item, and determines that the type of paper on which the item is acquired is the type of paper stored in the searched definition body When,
The character recognition device according to claim 1, further comprising: The character recognition device further includes:
A storage unit for storing one or more definition bodies;
The definition item stored in association with the image of the same item as the acquired image of the item and the image area of the same item as the image area of the acquired item is searched, and the type of the page on which the item is acquired is searched. A discriminating unit for discriminating that the type of paper stored in the defined definition body is;
The character recognition device according to claim 4, further comprising: A character recognition method executed by a computer,
The computer
Acquire an image of one or more items from a paper image,
Extracting an image of an item of a first size or larger from the acquired image of the one or more items;
A character recognition method, comprising: generating a definition body that associates and stores the type of the page from which the item image is acquired and the extracted item image. Acquire an image of one or more items from a paper image,
Extracting an image of an item of a first size or larger from the acquired image of the one or more items;
A character recognition program that causes a computer to execute a process for generating a definition body that stores a type of a page on which an image of the item is acquired and an image of the extracted item in association with each other.

Images