JPH1173472A - Format information registering method and ocr system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To lighten the burden of the registration of format information on an operator by providing an item search part which generates the formation information representing the coordinate information on a recognition area position corresponding to each item name by referring to a key item data base and layout information. SOLUTION: The key item DB 5 and layout information 4 are used to detect coordinate information on an area into which an item name 'address' is entered by referring to the layout information 4. It is judged that two areas positioned below the area are areas wherein information corresponding to the address is entered. Consequently, when the item name is 'address' as to a document having a sheet ID of 110, the coordinates of two stages of the areas below found with the layout information 4 are recorded as position information in the format information 6. This process can be performed by adding a specific judging process to the existent function that the OCR system has. Then those processes can be performed automatically, so the burden on the operator is lightened.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a format information registration method and an OCR system for acquiring and registering format information of a form to be read in advance when reading characters and marks recorded on the form. About.

[0002]

2. Description of the Related Art An OCR system is used for inputting information entered in forms used for various kinds of business transactions and office management into a computer and utilizing the information. In this system, an image of a form is read, a character written in a specified location is cut out and recognized, and converted into a character code to obtain data for information processing. In such a system, the position where information such as characters and figures on the form to be read is to be written needs to be registered in advance as format information. When two or more forms are mixed and necessary information is read and recognized, it is necessary to register format information for each form. An operator manually inputs such format information into the computer using a dedicated editor or the like.

[0003]

However, the above-mentioned prior art has the following problems to be solved.
If there are many types of forms to be read, the format information registration work by the operator becomes complicated. Further, even when the form has a complicated configuration, a heavy load is imposed on the registration work of the operator. In addition, even if a form is to be filled in with the same items, if there are several kinds of forms in which the entry positions are slightly different, it is necessary to perform predetermined registration work on all these forms, and It was a heavy burden.

[0004]

The present invention employs the following structure to solve the above problems. <Structure 1> An image input unit for reading an image of a form in which arbitrary information is entered, a recognition unit for recognizing the information and converting it into a character code, and an item name to be entered in the form and information about the item are entered. The key item database that displays the positional relationship between the recognition area and the item name for character recognition and the number of recognition area positions, and the ruled lines included in the image of the form read by the image input unit. By referring to the key item database and the layout information, the layout analysis unit generates coordinate information of the enclosed region and layout information displaying the recognition result of the character written in the region. An OCR system, further comprising an item search unit for generating format information displaying coordinate information of a corresponding recognition area position.

<Structure 2> In order to read an image of a form on which arbitrary information is entered, recognize the information, and convert it into a character code, format information for displaying the information entry position of the form is generated in advance. In advance, a key item database that displays the item names to be entered in the form, the positional relationship between the recognition area and the item names in which information on the items is written and character recognition is performed, and the number of recognition area positions is stored in advance. Prepare, read the image of the form, generate the coordinate information of the area surrounded by the ruled line included in the image of the form, and the layout information displaying the recognition result of the character written in the area, By referring to the key item database and the layout information, for each item name,
A format information registration method characterized by generating format information displaying coordinate information of a corresponding recognition area position.

<Structure 3> In order to read an image of a form on which arbitrary information is entered, recognize the information and convert it to a character code, format information for displaying the information entry position of the form is generated in advance. And the positional relationship between the item name of the hierarchical structure to be entered in the form, the recognition area in which information on the item at the lowest level in the hierarchical structure is written and character recognition, and the item name, Prepare a key item database that displays the number of area positions and
The image of the form is read, and coordinate information of an area surrounded by a ruled line included in the image of the form and layout information displaying a recognition result of a character written in the area are generated, and the key item is generated. A format information registration method characterized by generating format information displaying coordinate information of a corresponding recognition area position for each item name of the hierarchical structure by referring to a database and layout information.

<Structure 4> In order to read an image of a form on which arbitrary information is entered, recognize the information and convert it to a character code, the format information for displaying the information entry position of the form in advance is generated. The item name to be entered in the form, the positional relationship between the recognition area and the item name in which information about the item is written and character recognition is performed, the number of recognition area positions, and the information is pre-filled in the recognition area. A key item database displaying certain image information is prepared in advance, the image of the form is read, the coordinate information of the area surrounded by the ruled line included in the image of the form, and the character written in the area The layout information displaying the recognition result of the item is generated, the key item database and the layout information are referred to, and the area determined as the recognition area corresponding to the item name is generated. Advance by collating the image and the image information are entered, for each item name, format information registration method and generates format information displaying the coordinate information of the corresponding recognition region located in.

<Structure 5> In order to read an image of a form in which arbitrary information is entered, recognize the information and convert it to a character code, format information for displaying the information entry position of the form in advance is required. A key item database is prepared in advance, in which an item name to be entered in the form and information about the item are entered and a certain image information which is previously entered in a recognition area for character recognition is prepared. And generates the coordinate information of the area surrounded by the ruled line included in the image of the form and the layout information displaying the recognition result of the character written in the area, and displays the information in the layout information. The corresponding item name is determined by comparing the image information pre-filled in the designated area with the image information, and the corresponding item name is determined for each item name. Format information registration method and generates format information displaying the coordinate information of the recognition area position.

<Structure 6> In order to read an image of a form in which arbitrary information is entered, recognize the information and convert it to a character code, the format information for displaying the information entry position of the form is generated in advance. An item name to be entered in the form, and a key item database in which predetermined character information is displayed in advance in a recognition area in which information relating to the item is entered and character recognition is performed.
The image of the form is read, and coordinate information of an area surrounded by a ruled line included in the image of the form and layout information displaying a recognition result of a character written in the area are generated, and the layout information is generated. The format information that displays the coordinate information of the corresponding recognition area position by determining the corresponding item name by comparing the character information previously written in the area displayed in the area with the character information, and for each item name A format information registration method characterized by generating a format information.

<Structure 7> In the method according to Structure 6,
A method for registering format information, characterized in that a recognition dictionary for collating characters previously written in an area is stored in a key item database.

<Structure 8> In order to read an image of a form on which arbitrary information is entered, recognize the information and convert it to a character code, the format information for displaying the information entry position of the form in advance is generated. The item name to be entered in the form, information on the item is entered, the positional relationship between the recognition area and the item name for character recognition, the number of recognition area positions, and information on the item are entered. A key item database displaying certain information previously written in a recognition area for character recognition is prepared in advance, the image of the form is read, and the image is surrounded by a ruled line included in the image of the form. The coordinate information of the area and the layout information displaying the recognition result of the character written in the area are generated, and the key item database and the layout information are referred to. By this, for each item name, the coordinate information of the corresponding recognition area position is displayed, and when the information pre-filled in the recognition area matches certain information in the key item database, the A format information registration method characterized by generating format information displaying weighting information indicating the likelihood of a determination.

<Structure 9> In order to read an image of a form in which arbitrary information is entered, recognize the information and convert it to a character code, the format information for displaying the information entry position of the form in advance is generated. An item name to be entered in the form and a key item database in which information on the item is entered and identification information previously entered in a recognition area for character recognition are prepared in advance, and an image of the form is prepared. To generate the coordinate information of the area surrounded by the ruled line included in the image of the form, and the layout information displaying the characters written in the area and the recognition result of the identification information, and generating the key item By referring to the database and the layout information, the column in which the identification information is entered is associated with the item name based on the identification information. Format information registration method characterized by generating display the format information coordinate information of the recognition area position.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below using specific examples. <Specific Example> FIG. 1 is an explanatory diagram of a format information registration method according to the present invention. In the present invention, an image of the form 1 as shown in the figure is read, and the format information 6 used in the OCR system is automatically generated. In this figure, the procedure is schematically illustrated. The form 1 has, for example, a configuration in which a sheet ID 2 is entered in the upper left corner. The sheet ID 2 is a unique symbol for each form representing the type of the form 1.

This form is provided with a column for entering, for example, an address or a name. These entry fields are surrounded by ruled lines 3 as shown in the figure. A general OCR system has a function of recognizing an area surrounded by ruled lines written therein when an image of the form 1 is read, and obtaining coordinate information of the area. This is to perform a process of extracting characters written in the area. This function is also used in the format information registration process.

That is, the layout information 4 as shown in the figure is obtained by this function. The layout information 4 obtains the coordinates of the upper left corner and the lower right corner of the three frames surrounded by the ruled line 3 in the case of, for example, a column for entering the address of the form 1 shown in the figure. Here, three sets of coordinate information are obtained. Further, when the three regions surrounded by the ruled lines are detected in this way, recognition processing of characters written inside each region is subsequently performed. This is also a function provided in a general OCR device.

In the state where the information to be actually taken in by the OCR is not entered in the form 1, the necessary characters are entered here only in the item name portion, and the other columns are blank. Therefore, layout information 4 as shown in the figure is obtained. Of course, as long as this format information generation processing is not hindered, appropriate characters may be entered in other columns.

Here, in order to automatically generate the format information of the form 1, a key item DB (hereinafter, all databases are referred to as DB) 5 is prepared in advance. The key item DB 5 stores an item name to be entered in the form 1, an area in which information corresponding to the item name is entered, that is, the position of the recognition area and the number thereof, as shown in FIG. This is an information group displaying the contents. For example, in the case of the form 1 shown in the figure, if there is an item called address, there are two columns in which information is to be entered side by side below the item name. The key item DB 5 displays information such as the position and the number of columns in which such item names and information are to be entered.

Note that the position in this case is different from the position information indicated by coordinates. This is information indicating the positional relationship between the item name indicating which direction the recognition area is located in the field of the item name and the recognition area. In the following description, this is simply referred to as a position. Such information
It can be easily created for each report. It is easy to input to a computer. The present invention uses this information to automatically generate format information that displays the exact image cutout position coordinates of the recognition area for each item.

First, using such a key item DB 5 and the layout information 4, coordinate information of an area in which an item name such as an address is entered is detected with reference to the layout information 4. Then, two areas located below the area are determined to be areas in which information corresponding to the address is written. As a result, as shown in the figure, in the format information 6, if the item name is an address for the form 1 with the sheet ID of 110, the position information is the coordinates of the lower two-level area obtained from the layout information 4. Be recorded. Such processing is OC
This can be performed by adding a predetermined judgment process to the existing functions of the R system. Since these processes can be automatically executed, the burden on the operator is reduced. The format information obtained in this way is checked by the operator thereafter, and registered in a form in which final correction and the like are added.

Hereinafter, specific examples of the present invention will be sequentially described with reference to block diagrams, flowcharts, and the like. FIG.
1 is a block diagram of an OCR system according to the present invention.
In order to implement each embodiment of the present invention described below, for example, an OCR system having a configuration as shown in FIG. In the system shown in the figure, an image input unit 11, an image memory 12, a recognition unit 13, a control unit 14, an item search unit 15, a layout analysis unit 16, a layout information memory 17, a key item DB 18, a format information DB 1
9, a format identification unit 20, a recognition result DB 21, an input unit 22, and a display unit 23 are connected.

The image input section 11 is composed of an image scanner or the like, and is a section for performing a process of converting characters, figures, and the like entered on a form into binary data. Image memory 12
Is an apparatus for storing the image data thus obtained for subsequent processing. The layout analysis unit 16
As described above, this section has a function of extracting an area surrounded by ruled lines from the image data stored in the image memory 12 and storing coordinate information of the extracted area in the layout information memory 17.

The recognition unit 13 has a layout information memory 17
A part that recognizes a character string written in a corresponding area of the image data stored in the image memory 12 with reference to the coordinate information of each area stored in the format information DB 19 and converts the character string into a character code. The recognition result is stored in the layout information memory 17 or the recognition result DB 21.

The layout information memory 17 is a section for storing the coordinate information of the area extracted by the layout analysis section 16 and the character string recognized by the recognition section 13. As described above, the key item DB 18 is a part for storing information such as an item name, a recognition area position, and a number. The item search unit 15 compares the coordinate information of the area and the character string stored in the layout information memory 17 with the key item DB1.
8 has a function of searching for information stored in the format information DB 8 and storing the searched result in the format information DB 19. This is the part for executing the format information automatic generation described above.

The format information DB 19 has a structure as described above, and is a part for displaying a result searched by the item search unit 15 in the form of a sheet ID, an item name, and position information. The format identification unit 20 has a function of reading a sheet ID on the image data stored in the image memory 12, searching the format information DB 19 with the read ID, and identifying the format of the image data. The recognition result DB 21 stores the character data output by the recognition unit 13. This is then taken into a computer or the like and used for various information processing.

The control section 14 has a function of controlling the entire system. The input unit 22 has a function of receiving an input from an operator, and includes devices such as a keyboard and a mouse. The display unit 23 includes devices such as a display for displaying information to the operator.

FIG. 3 shows a main operation flowchart of the system. The system executes a format information registration process (step S1) and a reading process (step S2) as shown in FIG. The format information registration process is a process of processing a blank form to be read as described above and reading the format information before reading the actual form. That is, the format information of various forms printed in a predetermined format in advance is associated with the sheet ID by this processing, and the format information DB 19
Register with. As described above, this form may be one in which information has been entered.

The reading process is a process of reading a form on which actual information is written, recognizing necessary information, and obtaining a result of the recognition. In this case, the sheet ID of the form is recognized, the corresponding format information is extracted from the format information 19, and the character registered in the format information is recognized to obtain necessary information. The result of the character recognition is converted into a character code and stored in the recognition result DB 21.

FIG. 4 shows an operation flowchart of the format information registration process. In step S1-1 in the figure, image input processing A is performed. This is a process of reading the image of the form shown in FIG. The required item names and the like are printed on this form 1, but other information is still not entered. In step S1-2, a layout analysis process for obtaining the layout information 4 as shown in FIG. 1 is performed. Then, step S1-3
In, recognition processing A of a character written in each area is performed based on the recognized coordinate information. Here, the recognition of the item names etc. already printed on the form 1 is performed. And
In step S1-4, an item search process is performed. This is a process for automatically obtaining the format information 6 using the key item DB 5 and the layout information 4 shown in FIG.

Thereafter, in step S1-5, a search result correction process is executed. That is, the automatically obtained format information is displayed on the display unit 23 shown in FIG. Thereby, the completed format information is re-registered in the format information DB 19 shown in FIG. Step S1--FIG.
The processes from 2 to S1-4 will be described below as specific examples.

FIG. 5 shows an operation flowchart of the reading process. This operation is a process in which, after the format information is registered, an image of a form on which various types of information are actually written is read and character recognition is performed. The operation of this part is exactly the same as that of the conventional device.
In the following specific example, only the registration processing of the format information will be described.

First, in step S2-1 in FIG.
Image input processing B for reading image data in which various information is written is performed. In step S2-2, the format information is referred to, the format information corresponding to the sheet ID is extracted, and the type of the form is identified.
Then, in step S2-3, the coordinate information of each area specified by the format information is referred to, and a character string written or printed in the area is recognized and converted into a character code. The result is the recognition result DB described in FIG.
21.

Then, in step S2-4, the operator displays the recognition result on the display unit 23, and inspects and corrects the content. When the processing such as correction is completed, the recognition result DB 21 is updated thereafter, and the result is transferred to another information processing device or the like.

Hereinafter, various embodiments of the format information registration processing will be described in order with specific examples. <Specific Example 1> FIG. 6 is a diagram for explaining the contents of the format information DB generated in all the following specific examples including this specific example. As shown in FIG.
Is composed of a sheet ID, an item name, and position information.
The sheet ID is a character or a symbol for identifying the type of a form, and the item name is a name for specifying an item to be entered on the form. The position information is constituted by coordinate information of an area surrounded by each ruled line on the image of the form. This is as described with reference to FIG. The form 1 shown in FIG.
The sheet ID and other items described with reference to FIG. For each item, as shown in this figure, areas where the item names are printed are indicated as K1 and K2, and areas where information is written corresponding to the items are indicated as J1, J2 and J3.

FIG. 7 is a diagram for explaining the contents of the key item DB of the first embodiment. In the specific example 1, for example, an item DB as shown in this figure is used. As described above, the item DB 5 includes an item name, a recognition area position, and the number thereof. As shown on the right side of the figure, it is assumed that an item to be entered on a form is, for example, an address or a name.

In this case, there are areas J1 and J2 surrounded by two ruled lines below the item name K1 of the address. The address is entered here. Therefore, in the key item DB5 in the figure, the recognition area position of the item name of the address is below the column with the address, and the number is "2". For the item K2 of the name, the name is actually written in an area J3 surrounded by a ruled line on the right side. Therefore, in the key item DB5, the item name is represented by the name, the recognition area position is set to the right, and the number is expressed as "1".

The format information registration operation using the key item DB and the layout information 4 as described above will be described with reference to FIG. FIG. 8 is an operation flowchart of the first embodiment. First, in step S10, "1" is substituted for the counter 1, and the record number of the key item DB is substituted for RecNum. One record is a key item D
This is a set of information such as the recognition area position and the number corresponding to one item name of B. Step S20 is for all key item DBs.
It is a part for determining whether the processing has been completed for the record of.

In step S30, first, the key item DB
The first record of is read. That is, in the example shown in FIG. 7, the item name is the address, the recognition area position is down, and the number is “2”.
Is read. Then, step S4
At 0, the layout information memory is searched using the item name. When the corresponding area is obtained in step S50, the process proceeds to step S60, and the layout information memory is searched by the position information.

That is, when there is the layout information 4 as shown in FIG. 1, it is assumed that when the layout information memory is searched by the item name, the item name of which the recognition result is the address is obtained as the search result. In this case, in step S60, corresponding coordinate information is obtained. And key item DB
Is displayed on the lower side of the recognition area, the lower area is obtained from the coordinate information.

If it is determined in step S70 that there is a lower area, the flow advances to step S80 to obtain the coordinate information of the area by the number indicated by the number and obtain the format information 6. This format information is registered in the format information DB 19 shown in FIG. If the number is less than the required number, register as many as possible. If there is a shortage, it is added later by correction of the operator.

In step S90, the counter is incremented. As a result, the processes in steps S30 to S90 are repeated for the number of records in the key item DB. As described above, the format information is generated.

<Effects of Specific Example 1> A key item database displaying the names of items to be entered in the form, the positions of the recognition areas, and the number thereof is prepared in advance, and the area of the area surrounded by the ruled line included in the image of the form is prepared. The coordinate information and the recognition result of the character entered in the area are obtained, and the format information is automatically obtained by referring to the key item DB. Therefore, even if the entry is the same, the entry position is slightly changed. For such forms, format information can be registered very easily and automatically. Further, even if the result of the automatic generation is somewhat incomplete, it can be completed by simple correction work by the operator, so that the effect of reducing the burden on the operator is high.

<Embodiment 2> FIG. 9 is a view for explaining the contents of the key item DB of the embodiment 2. The item name, the recognition area position, and the number are also displayed in the key item DB of FIG. Here, as shown on the right side of FIG. 9, it is assumed that there is an area K3 in which an item name “name” is printed. At this time, the information relating to the name is entered separately under the areas K4 and K5 where the item names of last name and first name are printed. Areas in which information relating to names are written are areas J4 and J5.

In such a case, since the name and surname and the name and name each have a so-called hierarchical structure, a combination of these is displayed in the item name list. Therefore, information indicating that there is an area for putting a surname below the item name of the last name of the name is displayed in the key item DB. This relationship is called a hierarchical relationship. The same is applied to other areas. In the case of this example, the item names are displayed separated from each other by commas from the top of the hierarchy. Therefore, if there are several levels,
Some comma separated item names will be concatenated and displayed in the item name list.

FIG. 10 is a flowchart showing the operation of the second embodiment. Steps S10, S20, S3 shown in FIG.
The processing of 0, S60, S70, S80, and S90 is the same as the processing already described with reference to FIG. Therefore, only the unique part of the specific example 2 will be described in detail.

When the first record of the key item DB is read in step S30, the parameter J is set to "1" in step S31. When the item name list is formed by linking two or more item names, this parameter J is set in steps S32 to S3 in order from the first item name.
This is for performing control for executing the processing of No. 6.

For example, if the item name is a name and a last name, and J = 1, a character string having the item name as a name is read in step S32. Then, step S33
Then, the parameter J is monitored to determine whether or not the processing has been completed for all item names. If the name in the item name of the name and the last name is read, the process proceeds to step S3
Then, the process proceeds to step S4 to search for layout information having the item name as a recognition result.

If there is a corresponding area in step S35, the process proceeds to step S36, and J is incremented this time so that the next item name is to be processed.
Then, the process returns to step S32, and reads a character string of last name. Since the J-th character string exists in the judgment of step S33, the process proceeds to step S34, and the layout information memory is searched by the item name. If there is a corresponding area, the process returns from step S36 to step S32.

Here, J is incremented, and 3
If you try to read the th string,
There is no third string other than the last name. Therefore, step S3
The process proceeds from Step 3 to Step S60. Then, the layout information memory is searched based on the recognition area position of the key item DB. As a result, the position information of the area in which the information J4 below the last name item K4 in FIG. 9 is entered is obtained.

When the position information is obtained in this way, step S80
To register the format information in the database by the number specified by the number. That is, the item name is the last name of the name, and the position information is an area displaying the information J4 shown in FIG. When these processes are completed, the process proceeds to step S90, where I is incremented and the process for the next item name list is performed.

That is, the above steps S30 to S3
Through the processing up to 6, the item names of the hierarchical structure are checked, and the area where the last item name is entered is found. Then, it is determined that there is an area in which information corresponding to the item is entered at a location indicated by the information on the recognition area position with respect to the area. According to the result, the processing of steps S60 to S90 is executed to generate layout information. The format information obtained in this way is inspected by the operator as in the case of the first embodiment, and is registered again with necessary corrections.

<Effect of Specific Example 2> As described above, when the item names have a hierarchical structure, the item at the bottom of the hierarchy is associated with the position of the recognition area and the number of the areas, and the key item database , It is possible to cope with a case where a complicated item is displayed. Other effects are similar to those of the first embodiment.

<Third Embodiment> FIG. 11 is a diagram for explaining the contents of the key item database in the third embodiment. In this example, information called image information is provided in addition to the item name, the recognition area position, and the number of key items described in the above specific examples. On the right side of the figure, an example is shown in which, for example, information corresponding to an area surrounded by a ruled line on the right side is entered for an item called an address. Usually, the postal code is displayed on the address. In this example, the area J in which information is to be entered
The information corresponding to the postal code entry frame included in No. 1 is included in the key item database as image information. Therefore, in the image information of this key item DB,
The image data of the area J6 for writing the postal code is stored as it is.

FIG. 12 shows an operation flowchart of the third embodiment. The processing from step S10 to step S90 excluding step S71 in this flowchart is the same as the operation of the specific example 1 described using FIG. In this specific example, after the determination in step S70 in the figure, step S7
One judgment processing is added.

Here, the item K1 of the address is detected,
When the position of the recognition area on the right side is detected, it is determined whether or not the image information registered in the key item DB exists therein. Here, it is determined whether or not the postal code area J6 exists in the area J1. If there is, there is no doubt that the address is in the area where the address is to be entered, so the process proceeds to step S80, and registration of the corresponding number of pieces in the format information is performed. If there is no information that matches this image information, it is determined that the area is not an area where an address is to be entered, so that registration of format information is not performed.

In this example, for the address, a frame for entering a postal code is exemplified as image information. For example, a symbol for entering an amount of money, and whether or not the area is an area for entering a predetermined item. If there is an important image for distinguishing whether the image is present or not, the image information can be stored in the key item DB.

<Effect of Specific Example 3> As described above, when an image supporting the recognition area exists in the recognition area corresponding to the item name, the image information is included in the key item DB. In addition, the reliability of the format information registration can be improved, and the registration work burden on the operator can be reduced.

In the above case, the image information is used to confirm whether or not it is an area for writing an item corresponding to an address. However, if it is possible to determine that the area in which the image information is present is an area in which an address is to be written reliably, for example, even if the key item DB does not include information regarding the recognition area position and the number, the address item It is also possible to determine whether or not this image information is included in all the areas adjacent to the area in which the name is entered, and find the corresponding area.

<Embodiment 4> FIG. 13 is an explanatory view of the contents of the key item DB of the embodiment 4. In the specific example 3, image information is added to the key item DB in addition to the item name, the recognition area position, and the number. On the other hand, in this specific example 4,
Recognition area character information is added. That is, information about characters printed in advance included in the recognition area is stored as character codes.

For example, information corresponding to the name is written below the item K1 of the name. This area is defined as J1. At this time, characters such as "sama" are generally printed at the end of the name. That is, those characters printed before information entry are recognized and converted into character codes are shown in FIG.
In the layout information 4 shown in FIG. by this,
As in the case of the specific example 3, it can be confirmed that, for example, a name is entered in this area. In the lower right part of FIG. 13, an area J immediately below the area K2 in which the item of the bank name is printed.
2 has a letter M2 of "bank". Therefore, the character "bank" is displayed as the recognition area character information of the item "bank name" in the key item DB.

FIG. 14 shows an operation flowchart of the fourth embodiment. The processes from step S10 to step S90 except for step S72 in the specific example 4 are exactly the same as those in the specific example 3. In step S72, while the comparison with the image information was performed in step S3, the matching with the recognition area character information is performed here. Then, when the two match, the format information is generated by the number corresponding to the number. If they do not match, no format information is generated.

<Effect of Specific Example 4> In the specific example 4, similar to the specific example 3, based on the information written in the recognition area, the reliability of whether or not the recognition area is the corresponding area is improved.

<Fifth Embodiment> FIG. 15 is an explanatory view of the contents of the key item DB of the fifth embodiment. In this specific example, a dictionary corresponding to the item name is prepared in the key item DB. That is, if the item name is a name, characters such as a last name, a first name, a reading, and the like are written in a column where the name is entered and columns around the name. These characters are all recognized when generating layout information. However, in some cases, the operation may proceed to the format information generating operation without being recognized or with a recognition error. In the specific example 5, such specific character information is stored in the key item DB as a recognition dictionary.

FIG. 16 is an operation flowchart of the fifth embodiment. The processing before and after the flowchart in this figure is the same as the operation so far, and is not shown. In the example of the specific example 4, for the part before step S70, the specific example 5 performs the same processing. Then, the processing of steps S70 to S77 is executed, and the operation proceeds to step S90 of the specific example 4 shown in FIG.

First, in step S70 of FIG. 16, it is determined whether or not there is an area at the recognition area position. This is performed by referring to the position information of the key item DB5.
Here, the process proceeds to step S73, and it is determined whether or not characters or the like are entered in the area. If characters or the like are entered, the process proceeds to step S74, and it is determined whether or not the recognition result is included in the layout information.

If it is included, the process proceeds to normal processing. If not included, the process proceeds to step S75, the recognition dictionary in this specific example is extracted, and recognition and collation of these characters are executed. Since the target is limited, the accuracy of recognition is improved, and the object that cannot be recognized by the normal character recognition processing can be recognized. In addition, a deformed character, a decorative character, and the like may not be recognized at the time of generating the layout information, but such recognition is possible here. When such recognition processing is executed and the layout information is prepared, processing exactly the same as that of the above specific example is executed.

In step S76, it is determined whether or not this is a recognition area. For example, in the column immediately below the name,
If it is determined that the item “sama” has been entered, the process proceeds to step S77, and registration in the format information DB is executed. In other cases, the generation of the format information is passed, and the processing proceeds to the next record.

<Effect of Specific Example 5> Since a recognition dictionary corresponding to an item name is prepared in advance, a character or the like that cannot be easily recognized by general processing unique to the item name is surely recognized. Collation and determination of the corresponding recognition area can be ensured.

<Embodiment 6> FIG. 17 is a diagram for explaining the contents of the format information DB. As explained above,
In the case where the position coordinates of the character recognition area corresponding to the item name are automatically obtained and registered as the format information, the reliability becomes a problem. If the automatically generated format information has reliability, the burden on the operator is further reduced.

Therefore, weighting for indicating reliability is performed on each position information of the generated format information. In this weighting, for example, it is highly likely that an area located at a position close to an area in which an item name is written is a recognition area related to the item name. In such a case, the point information is increased, and the point information is decreased as the distance increases. Also, if a character such as "sama" is entered in the column, there is a high probability that it is a recognition area in which a name should be entered. In this case, weighting is performed by adding point information.

In this way, the format information is added with information that the reliability is high, and is useful when the operator performs a correction process. Format information DB4 shown in FIG.
For this purpose, additional information is written in addition to the sheet ID, the item name, and the position information. The configuration of the form 1 is the same as that described above.

FIG. 18 is an explanatory view of a modified example of a form.
In the case of this figure, there are words such as "prefecture" and "city" in the recognition area J1 corresponding to the address of the item K1. Thereby, additional point information is added.

FIG. 19 is a flowchart showing the operation of the sixth embodiment. First, according to this flowchart, specific example 6
Will be described. In FIG. 19, the processing from step S10 to step S90 except for the processing shown in steps S51 to S54 is exactly the same as the operation described using the specific example 3 or the specific example 4 so far.

Here, in step S50, a corresponding area is checked based on the position information of the corresponding item name, and if it is determined that there is an area, the flow advances to step S51 to search for a surrounding area. And, by judging that it was a close area,
For example, the additional information "1" is added (step S5).
2). Next, character information included in the area is checked.
Here, in the case of a name, for example, if a character such as "sama" is included, "1" is newly added to the point information. Therefore, the point information is “2” if the item is adjacent to the area in which the item of name is entered, and the character “sama” exists in the area. The result of such processing is determined in step S54.
, The information is registered in the format information DB.

While looking at the format information DB shown in FIG. 17, the operator refers to, for example, the position information corresponding to the item name of the address and the additional information, and determines whether or not these information are automatically generated without error. to decide. It is also possible to focus on correction of other information, etc., as information having high additional point information has high reliability.

<Effect of Specific Example 6> As described above, FIG.
As shown in (1), if characters and the like are present in a column where information is to be entered, the reliability is further increased. The information to be added may be, for example, an area for entering a postal code as described above, or characters such as "sama" or "bank".

<Specific Example 7> In the above example, in order to determine the recognition area corresponding to the item name, the position of the item name, characters written in the area, and the like are used as a reference. In this specific example, in order to generate format information, a special form in which information serving as a guide is previously written in a recognition area is prepared.

FIG. 20 is an explanatory diagram of the contents of the key item DB of the seventh embodiment. As shown in this figure, it is assumed that character strings such as ADRESS1, ADRESS2, and NAME1 are written in each recognition area of a form in which the item names are an address, a name, a bank, etc., as shown in the figure. . This is this key item DB
To be displayed.

FIG. 21 is an explanatory view of a modified example of a form. As shown in this figure, in each of the items K1, K2, and K3 of the form 1, characters as displayed in the key item DB are written in their recognition areas. By preparing such a form for format information generation in advance, it is possible to accurately generate reliability format information.

FIG. 22 is a flowchart showing the operation of the seventh embodiment. Steps S10, S30, S50, S90
Is the same processing as the specific example described so far. Here, when reading the first record of the key item DB in step S30, a search of the layout information memory using the recognition area designation character string is performed in step S31. This character string is accumulated and stored as layout information 4 as shown in FIG. 1 when collecting layout information of the form as shown in FIG. 21 in advance.

Therefore, by searching this, a corresponding area is found. If there is a corresponding area, the process proceeds to step S54, and it may be registered in the format information DB. This completes the registration process. Thereafter, the process proceeds to a registration process for another item. According to such a method, extremely reliable and accurate format information can be easily generated.

As described above, in this specific example, the key item DB
Can be extremely simplified. In the above specific examples, even if a part of the information is missing from each key item DB, the search may be omitted and the format information may be generated. This is because if the format information is incomplete, the operator may correct and complete it. In the specific examples 3, 4, 5, etc., data such as image information, recognition area, character information, and recognition dictionary are prepared, but these need not necessarily be included in the key item DB. That is, other independent dictionary files may be prepared. Of course, the key item DB itself does not need to be an integrated database, and may be divided into several.

The image data of the form described above may be read from a scanner or the like, or may be transferred from a host device (not shown). Further, in addition to identifying a form by a sheet ID or the like, it is also possible to identify the type of sheet by matching ruled line information or the like.

<Effect of Specific Example 7> According to the specific example 7, the recognition area designation character string is written in the recognition area corresponding to each item name, and the format information is generated using the identification information. Thus, it is possible to quickly generate extremely reliable format information.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram of a format information registration method according to the present invention.

FIG. 2 is a block diagram of an OCR system according to the present invention.

FIG. 3 is a main operation flowchart of the system.

FIG. 4 is a flowchart of a format information registration process.

FIG. 5 is an operation flowchart of a reading process.

FIG. 6 is an explanatory diagram of the contents of a format information DB.

FIG. 7 is an explanatory diagram of the contents of a key item DB of a specific example 1.

FIG. 8 is an operation flowchart of the first embodiment.

FIG. 9 is an explanatory diagram of the contents of a key item DB of a specific example 2;

FIG. 10 is an operation flowchart of a specific example 2.

FIG. 11 is an explanatory diagram of the contents of a key item DB of a specific example 3.

FIG. 12 is an operation flowchart of the third embodiment.

FIG. 13 is an explanatory diagram of the contents of a key item DB of a specific example 4.

FIG. 14 is an operation flowchart of a specific example 4.

FIG. 15 is an explanatory diagram of the contents of a key item DB of a specific example 5.

FIG. 16 is an operation flowchart of a specific example 5.

FIG. 17 is an explanatory diagram of the contents of a format information DB.

FIG. 18 is a diagram illustrating a modification of a form.

FIG. 19 is an operation flowchart of a specific example 6.

FIG. 20 is an explanatory diagram of the contents of a key item DB of a specific example 7.

FIG. 21 is an explanatory diagram of a modification of a form.

FIG. 22 is an operation flowchart of Example 7.

[Explanation of symbols]

 1 Form 2 Sheet ID 3 Ruled Line 4 Layout Information 5 Key Item DB 6 Format Information

Claims (9)

[Claims] 1. An image input unit for reading an image of a form in which arbitrary information is entered, a recognition unit for recognizing the information and converting it into a character code, an item name to be entered in the form, and information on the item A key item database that displays the positional relationship between the recognition area and the item name to be filled in for character recognition, and the number of recognition area positions; and a ruled line included in a form image read by the image input unit. A layout analysis unit that generates coordinate information of an area surrounded by a circle and layout information displaying a recognition result of characters written in the area, and an item name by referring to the key item database and the layout information. An OCR system comprising: an item search unit for generating format information in which coordinate information of a corresponding recognition area position is displayed for each recognition area. 2. The method according to claim 1, further comprising: reading an image of a form in which arbitrary information is entered, and generating format information for displaying an information entry position of the form in advance in order to recognize and convert the information into a character code; A key item database is prepared in advance that displays the item names to be entered in the form, the positional relationship between the recognition areas and the item names in which information on the items is written and character recognition is performed, and the number of recognition area positions. Reading the image of the form, generating coordinate information of an area surrounded by ruled lines included in the image of the form, and layout information displaying a recognition result of characters written in the area, and generating the key By referring to the item database and the layout information, it is possible to generate format information in which the coordinate information of the corresponding recognition area position is displayed for each item name. How to register format information. 3. A method of reading an image of a form in which arbitrary information is entered, and generating format information for displaying an information entry position of the form in advance in order to recognize and convert the information into a character code. Item name of the hierarchical structure to be entered in the form, the positional relationship between the recognition area and the item name for information on the item at the lowest level in the hierarchical structure to perform character recognition, and the position of the recognition area A key item database is prepared in advance, which displays the number of items, and the image of the form is read, and the coordinate information of the area surrounded by the ruled line included in the image of the form and the recognition of characters written in the area are read. By generating layout information displaying the result, and referring to the key item database and the layout information, a corresponding recognition area position is obtained for each item name of the hierarchical structure. A format information registration method characterized by generating format information that displays coordinate information of the format information. 4. A method of reading an image of a form in which arbitrary information is entered, and generating format information for displaying an information entry position of the form in advance in order to recognize and convert the information into a character code. Item names to be entered in the form, the positional relationship between the recognition area and the item name in which information about the item is to be written and used for character recognition, the number of recognition area positions, and the constants pre-filled in the recognition area A key item database displaying the image information of the form is prepared in advance, the image of the form is read, and the coordinate information of the area surrounded by the ruled line included in the image of the form and the recognition of the character written in the area are recognized. The layout information displaying the result is generated, and the key item database and the layout information are referred to, and the area is determined in the area determined as the recognition area corresponding to the item name. A format information displaying coordinate information of a corresponding recognition area position for each item name by collating the entered image with the image information. 5. A method of reading an image of a form in which arbitrary information is entered, and generating format information for displaying an information entry position of the form in advance in order to recognize and convert the information into a character code. An item name to be entered in the form and a key item database displaying certain image information which has been previously entered in a recognition area in which information relating to the item is entered and character recognition is prepared in advance, the image of the form To generate the coordinate information of the area surrounded by the ruled line included in the image of the form and the layout information displaying the recognition result of the character written in the area, and displayed the layout information. By comparing the image information pre-filled in the area with the image information, a corresponding item name is determined, and a corresponding recognition area is determined for each item name. A format information registration method characterized by generating format information displaying coordinate information of an area position. 6. A method of reading an image of a form in which arbitrary information is entered, and generating format information for displaying an information entry position of the form in advance in order to recognize and convert the information into a character code. An item name to be entered in the form and a key item database in which information relating to the item is entered and certain character information previously entered in a recognition area for character recognition are prepared in advance, and the image of the form is prepared. To generate the coordinate information of the area surrounded by the ruled line included in the image of the form and the layout information displaying the recognition result of the character written in the area, and displayed the layout information. By comparing the character information previously written in the area with the character information, the corresponding item name is determined, and for each item name, the coordinates of the corresponding recognition area position A format information registration method characterized by generating format information displaying information. 7. The format information registration method according to claim 6, wherein the key item database stores a recognition dictionary for collating characters previously written in the area. . 8. A method of reading an image of a form in which arbitrary information is entered, and generating format information for displaying an information entry position of the form in advance in order to recognize and convert the information into a character code. The item names to be entered in the form, the positional relationship between the recognition area and the item name in which information on the item is written and character recognition is performed, the number of recognition area positions, and the information on the item is written and the characters are written. A key item database displaying predetermined information previously written in a recognition area for recognition is prepared in advance, an image of the form is read, and an area surrounded by a ruled line included in the image of the form is read. By generating coordinate information and layout information displaying a recognition result of characters written in the area, referring to the key item database and layout information Thus, for each item name, the coordinate information of the corresponding recognition area position is displayed, and when the information pre-filled in the recognition area matches certain information in the key item database, the determination is made. A format information registration method characterized by generating format information displaying weighting information indicating the likelihood of the format information. 9. A method for reading an image of a form in which arbitrary information is entered, and generating format information for displaying an information entry position of the form in advance in order to recognize and convert the information into a character code. A key item database is prepared in advance, which displays the item names to be entered in the form and the identification information in which information on the items is entered and character recognition is performed in advance in a recognition area for character recognition, and the image of the form is read. Generating coordinate information of a region surrounded by ruled lines included in the image of the form, and layout information displaying a character written in the region and a recognition result of the identification information. By referring to the layout information, the column in which the identification information is entered is associated with the item name based on the identification information. A format information registration method characterized by generating format information displaying coordinate information of a knowledge area position.