JP4412525B2 - Form recognition device and form recognition method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a form recognition apparatus and a form recognition method for discriminating a plurality of forms and reading data.
[0002]
[Prior art]
Conventionally, in office work such as a financial institution, a form such as national treasury money or tax is received, and an operator manually inputs data described in the form from the screen. In addition, the received form is designated as national treasury, and the image of the data is read and recognized from the corresponding position read by the scanner based on the format of the designated form, and is input.
[0003]
[Problems to be solved by the invention]
Conventionally, the operator has identified and specified the national treasury etc. for the accepted form, read the image from the relevant position according to the relevant format, recognized it as data, and entered the data. There is a problem that the operator needs to identify and specify each type of form.
[0004]
In order to solve these problems, the present invention extracts and collates the input form size, form ID, ruled line information, long capital letters, unique information, etc., determines the form type, and extracts character information. It is intended to input and tabulate, automatically and reliably determine the type of form from a plurality of types of forms, read the data, and input and tabulate.
[0005]
[Means for Solving the Problems]
Means for solving the problem will be described with reference to FIG.
[0006]
In FIG. 1, the processing device 1 executes various processes according to a program, and here is constituted by a recognition unit 3, a totaling unit 5, and the like.
[0007]
The recognition means 3 reads the size from the form, determines the corresponding form candidate based on the read size, extracts the form candidate ID, extracts the unique information of the form candidate, etc. Is determined.
[0008]
The tabulating means 5 tabulates data read from the form.
Next, the operation will be described.
[0009]
The form candidate is determined based on the form size read by the recognition unit 3 constituting the processing apparatus 1, the form candidate ID is extracted to determine the form, and the data is extracted based on the determined form information. To recognize it.
[0010]
In addition, a candidate for a form is determined based on the form size read by the recognition unit 3, a rule line of the form candidate is extracted to determine a form, and data is extracted and recognized based on the information on the determined form. I am doing so.
[0011]
Further, a candidate for a form is determined based on the form size read by the recognition means 3, a long upper case string is extracted for the determined form candidate, a form is determined, and data is obtained based on the information on the determined form. It is extracted and recognized.
[0012]
In these cases, the unique information is extracted from the form candidates and the form is determined.
[0013]
Further, data is extracted for each form on the basis of information on the type of form determined by the totaling means 5, and the amount is totaled for each date, each branch, and each storage location.
[0014]
Therefore, by extracting and collating the size of the input form, form ID, ruled line information, long capital letters, unique information, etc., the type of form is fixed, character information is extracted, entered, and totaled. The type of the form can be automatically and reliably determined from the form, and the data can be read and input / aggregated.
[0015]
DETAILED DESCRIPTION OF THE INVENTION
Next, embodiments and operations of the present invention will be described in detail sequentially with reference to FIGS.
[0016]
FIG. 1 shows a system configuration diagram of the present invention.
In FIG. 1, a processing device 1 executes various processes according to a program, and here includes an input means 2, a recognition means 3, a collation means 4, a totaling means 5, and the like.
[0017]
The input unit 2 captures an image (image) of the form read from the scanner 11.
[0018]
The recognizing means 3 measures the size of the form based on the captured form image, determines the corresponding form candidate based on the read size, extracts the form candidate ID, The form is determined by extracting unique information or the like (described later with reference to FIGS. 2 and 3).
[0019]
The collation means 4 collates with the corresponding ones in the DBs 6 to 9 (which will be described later with reference to FIGS. 2 and 3).
[0020]
The totaling unit 5 totals data read from the form (described later with reference to FIGS. 2 and 3).
[0021]
The form size DB 6 is a DB (database) in which the form sizes are registered (described later with reference to FIG. 4).
[0022]
The national treasury form DB 7 registers various information related to the national treasury form (described later with reference to FIG. 8).
[0023]
The tax public money OCR form DB 8 registers various information related to the tax public money form (which will be described later with reference to FIG. 8).
[0024]
The tax public money non-OCR form DB 9 is registered with various information related to the tax public money non-OCR form (to be described later with reference to FIG. 10).
[0025]
The display device 10 displays various information.
The scanner / OCR 11 reads various forms into images (images), recognizes information on positions where the forms are marked, and the like.
[0026]
The output device 12 is a variety of output devices, such as a printing device.
Next, the operation of the configuration of FIG. 1 will be described in detail according to the order of the flowcharts of FIGS.
[0027]
2 and 3 are flowcharts for explaining the operation of the present invention.
In FIG. 2, S1 inputs a form. This is because the input means 2 in FIG. 1 controls the scanner 11 to read and capture an image (image) of the form.
[0028]
In S2, noise removal and inclination correction are performed. This is because, after removing the noise from the image of the form read by the scanner 11 in S1, if the image of the form is inclined, the inclination is automatically corrected so that the outer frame of the form is formed in the horizontal and vertical directions. To.
[0029]
In S3, a form size is calculated. In S2, a horizontal width W and a vertical length (height) H are calculated from the image after noise removal and inclination correction in S2.
[0030]
In S4, the variable NC = 0 is initialized. This sets an initial value of 0 to a variable NC that counts the number of times a form type is determined based on its size. Here, as the types of forms, the following sizes are determined for three types of national treasury forms, tax public money OCR forms, and tax public money non-OCR forms.
[0031]
In S5, it is determined whether the form size of the national treasury money. This determines whether the size (W, H) of the form calculated in S3 is within the range of the national treasury size (see form size DB6 in FIG. 4). If YES, the process proceeds to S6. In the case of NO, since it has been found that it is not within the range of the national treasury form size, the process proceeds to S10.
[0032]
In step S6, the form ID is extracted / recognized / verified. This is within the national treasury bill size range in S5 YES, and since it has been found as a national treasury bill candidate, first, it is extracted from the position of the form ID in the form image and character recognition is performed, and the national treasury DB 7 in FIG. To the form ID registered in.
[0033]
In S7, it is determined whether it is OK. In the case of YES, since the form ID matches the ID of the national treasury form, the process proceeds to the processing after (2) in FIG. 3 to read and count the national treasury form data. On the other hand, in the case of NO in S7, it is determined that the form ID is not a national treasury form. Therefore, NC = NC + 1 is determined in S8, and NC <3 is determined in S9. If YES, the process proceeds to S10. The display is rejected at (1) in FIG.
[0034]
With the above S1 to S5 YES and S6 to S9, when the form size is within the range of the national treasury form size and the form ID matches the national treasury form form ID, it can be automatically determined as the national treasury form. Then, after (2) in FIG. 3, it is possible to read and aggregate data in the national treasury bill.
[0035]
In S10, it is determined whether or not the form size of the tax public money OCR. This is to determine whether the size (W, H) of the form calculated in S3 is within the range of the size of the public tax OCR form (see form size DB 6 in FIG. 4). If YES, the process proceeds to S11. In the case of NO, since it has been found that it is not within the range of the form size of the tax public money OCR, the process proceeds to S15.
[0036]
In step S11, the form ID is extracted / recognized / verified. This is within the form size range of the tax public money OCR in S10 YES, and since it has been determined as a tax public money OCR form candidate, first, it is extracted from the position of the form ID in the form image and character recognition is performed. It collates with the form ID registered in the tax public money OCR form DB8.
[0037]
In S12, it is determined whether it is OK. In the case of YES, since the form ID matches the ID of the tax public money OCR form, the process proceeds to the processing after (2) in FIG. 3, and the tax public money OCR form data is read and aggregated. On the other hand, in the case of NO in S12, since it is determined that the form ID is not a tax public money OCR form, NC = NC + 1 is determined in S13, and NC <3 is determined in S14. If YES, the process proceeds to S15, and if NO Is rejected and displayed in (1) of FIG.
[0038]
By the above S1 to S5NO and S10 to S14, when the form size is within the range of the tax public money OCR form size and the form ID matches the tax public money OCR form ID, it can be automatically determined as the tax public money OCR form. Become. Then, after (2) in FIG. 3, the data in the tax public money OCR form can be read and aggregated.
[0039]
In S15, ruled line information is extracted / recognized / checked. For this, ruled line information is extracted and recognized from the form image, and collated with, for example, the tax public non-OCR form DB 9 of FIG.
[0040]
In S16, it is determined whether it is OK. In the case of YES, the ruled line information coincides with the information in the tax public money non-OCR form DB 9, so the process proceeds to the processing after (2) in FIG. 3, and the tax public money non-OCR form data is read and aggregated. On the other hand, in the case of NO in S16, it is determined that the ruled line information is not a tax public money non-OCR form, so NC = NC + 1 is determined in S17 and NC <3 is determined in S18. At this time, the display is rejected at (1) in FIG.
[0041]
By the above S1 to S5NO, S10NO, and S15 to S18, when the ruled line information matches the ruled line information of the tax public money non-OCR form, it can be automatically determined as the tax public money non-OCR form. Then, after (2) in FIG. 3, it is possible to read and aggregate data in the tax public money non-OCR form.
[0042]
In FIG. 3, S21 selects a candidate form. This has been found to be a national treasury form when S2 in FIG. 2 is YES, a tax public money OCR form when S12 is YES, and a tax public money non-OCR form when S16 is YES. Selection is performed, and the processing after S22 is performed with reference to each form DB.
[0043]
In step S22, unique information is extracted / verified. This is based on the form image based on the national treasury form DB 8 shown in FIG. 8, the tax public money OCR form DB 9 shown in FIG. 9, or the tax public non-OCR form DB 10 shown in FIG. Extract, recognize and collate.
[0044]
In step S23, character extraction / recognition is performed. This is recognized by extracting other character information.
[0045]
In S24, the amount is summed for each form ID, for each date, for each branch, and for each storage destination.
[0046]
According to the automatically determined forms (national treasury form, tax public money OCR form, tax public money non-OCR form) according to S21 to S25 above, the respective DBs (national treasury form DB7, tax public money OCR form DB8, tax public money non-OCR) It is possible to extract and recognize unique information and various data from the form image read based on the form DB 9), and totalize them.
[0047]
FIG. 4 shows a form size DB example of the present invention. Here, the sizes (W, H) of the three types of forms described above are registered as shown below.
[0048]
Size (width / height) National treasury form Tax public money OCR form Tax public money non-OCR form W 19cm ± 2cm 13cm ± 1cm 8cm ± 1cm
H 10cm ± 1cm 11cm ± 1cm 18cm ± 2cm
As described above, the size (W, H) is registered for each type of form, and S5. In S10, it is possible to determine which form size range is within.
[0049]
FIG. 5 shows an example of a form (national treasury form) according to the present invention. The national treasury bill has a size as shown in the figure, and the data shown in the figure is entered and marked at the position shown in the figure.
[0050]
FIG. 8 shows an example of the national treasury form DB of the present invention. This is information in which data for reading data from the national treasury form in FIG. 5 is registered. The form ID, unique information, and character recognition information are registered as shown in the lower row, and the form ID (see FIG. 2 (S6), unique information (S22 in FIG. 3), and character information (S24 in FIG. 3). For example, a form image is extracted from the position (upper left and lower right coordinates) of the form ID of the national treasury form, and the form ID of the attribute (for example, 5 digits) is recognized. Then, in correspondence with the recognized form ID, the characters and symbols (national treasury money and its surroundings are rectangles and ellipses) specified by features from the positions (upper left and lower right coordinates) registered in the unique information column. The character corresponding to the attribute (12 digits, handwritten) is recognized from the position (upper left, lower right coordinates) registered in the character information recognition column. Similarly, it is recognized according to the contents registered in the other unique information and character recognition information fields.
[0051]
With respect to the image of the national treasury form of FIG. 5 described above, it is possible to automatically recognize the form ID, unique information, and character information sequentially in accordance with the national treasury form DB 7 of FIG.
[0052]
FIG. 6 shows an example of a form (tax public money OCR form) of the present invention. The tax public money OCR form has a size as shown in the figure, and the data shown in the figure is entered and marked at the position shown in the figure.
[0053]
FIG. 9 shows an example of the tax public money OCR form DB of the present invention. This is information in which data for reading data from the tax public money OCR form in FIG. 6 is registered. The form ID, unique information, and character recognition information are registered as shown below, and the form ID ( It is used when recognizing with S11 in FIG. 2), unique information (S22 in FIG. 3), and character information (S24 in FIG. 3), and is the same as in the case of the national treasury form DB. Omitted.
[0054]
With respect to the image of the tax public money OCR form shown in FIG. 6, the form ID, the unique information, and the character information can be automatically and sequentially recognized based on the tax public money OCR form DB 8 shown in FIG.
[0055]
FIG. 7 shows an example of a form of the present invention (tax public non-OCR form). The tax public money non-OCR form has a size as shown in the figure, and the data shown in the figure is entered at the position shown in the figure.
[0056]
FIG. 10 shows an example of the tax public money non-OCR form DB of the present invention. This is the information for reading data from the non-OCR form of tax public money shown in FIG. 7. The form type, ruled line information, unique information, and character recognition information are registered as shown in the lower row. The ruled line information (S15 in FIG. 2), unique information (S22 in FIG. 3), and character information (S24 in FIG. 3) are used for recognition. In the case of a tax public money non-OCR form, as described above, the type of tax public money non-OCR form is determined based on the ruled line information. The ruled line information includes the number, position (upper left) For example, in the upper column of the figure (xx city municipal tax), the number of ruled lines is 20 and the positions (upper left and lower right coordinates) are from (kxs111, kys111) ( kxe111, kye111) and a total of 20 ruled line start points and end points are registered. Based on the ruled line information, it is determined which type of form. Then, the unique information and character information are automatically recognized based on the determined unique information and character recognition information, respectively.
[0057]
With respect to the image of the tax public money non-OCR form shown in FIG. 7, the form type, unique information, and character information can be automatically recognized sequentially in accordance with the ruled line information of the tax public money non-OCR form DB 9 shown in FIG.
[0058]
(Appendix 1)
In a form recognition device that distinguishes multiple forms and reads data,
Means to read the size of the form,
A means for determining a form candidate based on the read form size;
With respect to the determined form candidate, means for extracting the form candidate ID and determining the form;
A form recognition apparatus comprising means for extracting and recognizing data based on the determined form information.
[0059]
(Appendix 2)
In a form recognition device that distinguishes multiple forms and reads data,
Means to read the size of the form,
A means for determining a form candidate based on the read form size;
With respect to the determined form candidate, means for extracting a ruled line of the form candidate and determining the form;
A form recognition apparatus comprising means for extracting and recognizing data based on the determined form information.
[0060]
(Appendix 3)
In a form recognition device that distinguishes multiple forms and reads data,
Means to read the size of the form,
A means for determining a form candidate based on the read form size;
With respect to the determined form candidate, means for extracting a long capital string of the form candidate and determining the form;
A form recognition apparatus comprising means for extracting and recognizing data based on the determined form information.
[0061]
(Appendix 4)
The form recognition apparatus according to any one of Supplementary Note 1 to Supplementary Note 3, further comprising means for extracting unique information from the candidate forms and determining the form.
[0062]
(Appendix 5)
In a form recognition method that distinguishes multiple forms and reads data,
Reading the size of the form,
Determining a form candidate based on the read form size;
For the determined form candidate, extracting the form candidate ID and determining the form;
A form recognition method comprising: extracting and recognizing data based on the determined form information.
[0063]
(Appendix 6)
In a form recognition method that distinguishes multiple forms and reads data,
Reading the size of the form,
Determining a form candidate based on the read form size;
For the determined form candidate, extracting a ruled line of the form candidate and determining the form;
A form recognition method comprising: extracting and recognizing data based on the determined form information.
[0064]
(Appendix 7)
In a form recognition method that distinguishes multiple forms and reads data,
Reading the size of the form,
Determining a form candidate based on the read form size;
For the determined form candidate, extracting a long upper case string of the form candidate and determining the form;
A form recognition method comprising: extracting and recognizing data based on the determined form information.
[0065]
【The invention's effect】
As described above, according to the present invention, the type of the form, the form ID, the ruled line information, the long capital string, the unique information, etc. are extracted and collated to determine the type of the form, and the character information is extracted. Therefore, it is possible to automatically and reliably determine the type of form from a plurality of types of forms, read the data, and input / aggregate.
[Brief description of the drawings]
FIG. 1 is a system configuration diagram of the present invention.
FIG. 2 is a flowchart (part 1) illustrating the operation of the present invention.
FIG. 3 is a flowchart (part 2) illustrating the operation of the present invention.
FIG. 4 is an example of a form size DB according to the present invention.
FIG. 5 is a form example (national treasury form) according to the present invention.
FIG. 6 is a form example (tax public money OCR form) of the present invention.
FIG. 7 is an example of a form of the present invention (tax public money non-OCR form).
FIG. 8 is an example of a national treasury form DB of the present invention.
FIG. 9 is an example of a tax public money OCR form DB of the present invention.
FIG. 10 is a tax public money non-OCR form DB example of the present invention).
[Explanation of symbols]
1: processing device 2: input means 3: recognition means 4: collation means 5: aggregation means 6: form size DB
7: National treasury book DB
8: Tax public money OCR form DB
9: Tax public money non-OCR form DB
10: Display device 11: Scanner / OCR
12: Output device

Claims (3)

In a form recognition device that distinguishes multiple forms with different sizes and reads data,
A group size table for registering the vertical and horizontal sizes of all forms that are subject to data reading, grouped according to the vertical and horizontal sizes of the forms, and assigned to the groups; ,
Register which information data is to be extracted from one or more pieces of information including form ID, ruled line, long capital string, and unique information read from the form image for reading data from the form. Form table to be
Means for reading the vertical and horizontal sizes of the form;
Based on the vertical and horizontal size of the document read above, means for determining the type of a form matching with reference to the document size table,
Referring to the form table corresponding to the type of the determined the form, the form ID read from in document image that is registered with the form table, borders, long uppercase column, one or more information including unique information Based on the image image read from the form, means for recognizing one or more of the form ID, ruled line, long upper case string, unique information,
Based on the recognition result, means for repeatedly determining until it matches which form in the determined type,
A form recognition apparatus comprising means for extracting and recognizing data based on the form information of the determined form.   2. The form recognition apparatus according to claim 1, wherein the acceptable ranges are registered in correspondence with the vertical and horizontal sizes as the vertical and horizontal sizes of the form in the form size table. In the form recognition method that distinguishes multiple forms with different sizes and reads the data,
A group size table for registering the vertical and horizontal sizes of all forms that are subject to data reading, grouped according to the vertical and horizontal sizes of the forms, and assigned to the groups; ,
Register which information data is to be extracted from one or more pieces of information including form ID, ruled line, long capital string, and unique information read from the form image for reading data from the form. And a form table to be
Reading the vertical and horizontal size of the form,
Based on the vertical and horizontal size of the document read above, determining the type of a form matching with reference to the document size table,
Referring to the form table corresponding to the type of the determined the form, the form ID read from in document image that is registered with the form table, borders, long uppercase column, one or more information including unique information Originally, recognizing at least one of a form ID, a ruled line, a long capital string, and unique information from an image read from the form;
Based on the recognition result, repeatedly determining until it matches which form in the determined type,
A form recognition method comprising: extracting and recognizing data based on the form information of the determined form.

Images