JP4160206B2 - Database registration method using character recognition device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a database registration method using a character recognition device for registering correct data used for evaluation of a character recognition algorithm in character recognition for recognizing characters on a postal letter in a database using the character recognition device.
[0002]
[Prior art]
In recent years, with the widespread use of computers, the Internet, etc., systems have been developed that use various information in a database. As a utilization form of such a database, it has become common to construct a database for evaluation and analysis in various developments. For example, in the development of an algorithm in a character recognition device, the development efficiency is improved by evaluating and analyzing the algorithm by comparing with correct answer data that is a standard for performance evaluation.
[0003]
As a target of the character recognition device, for example, when considering the case of reading character information described in a postal letter, when registering a large amount of correct data of a postal letter in a database, one piece of paper is conventionally used. The operator read the image data of one postal letter and visually displayed what was displayed on the display device, and entered the character code of a predetermined reading area as correct data from the keyboard.
[0004]
Further, along with the advancement of character recognition, in addition to the evaluation of the recognized character code at the final stage, it is important to evaluate the algorithm for detecting and cutting out the described character string in character units (hereinafter abbreviated as “censoring”). It is coming.
[0005]
In consideration of this, as an object of the character recognition device, for example, when considering the case of reading character information described in a postal letter, an algorithm for a censoring process that is a process preceding the character recognition process Coordinate data for the correct answer of the written character frame (area, character frame) for evaluation is required.
[0006]
Conventionally, when registering the correct coordinate data in the database, the image data of each postal letter is read and displayed on the display device by the operator, while the operator looks at it with a pointing device such as a mouse. The coordinate data of the character frame is input as correct coordinate data for each area or for each character.
[0007]
[Problems to be solved by the invention]
However, a large amount of data input work manually increases the fatigue level of the operator, resulting in a decrease in work efficiency and an increase in data input errors, and collecting the expected amount of correct data. There was a problem that it became difficult.
[0008]
Therefore, the present invention can perform data input more accurately and without taking time in the operation of registering correct data used for evaluation of the character recognition algorithm in the database. It is an object of the present invention to provide a database registration method using a character recognition device that enables efficient registration to a database with reduced degree.
[0009]
Further, the present invention can perform data input more accurately and without taking time in the work of registering correct coordinate data of written characters used for evaluation of the character recognition algorithm in the database. It is an object of the present invention to provide a database registration method using a character recognition device that enables efficient database registration with reduced fatigue of data input.
[0010]
[Means for Solving the Problems]
The database registration method using the character recognition device of the present invention is: Register in the database in the database registration method using the character recognition device that registers the coordinate data of the character area described on the paper used for the performance evaluation of the character recognition algorithm in automatic character recognition as the correct coordinate data in the database A data collection step for reading and collecting the data on the standardized paper by the data collection unit, and a description for performing the detection cutout of the written character area for the collected data with respect to the data collected by the data collection step Character recognition processing is performed using a character recognition device configured to have a character recognition unit that performs character recognition on the character region detection and extraction unit and the written character region detected and cut out by the written character region inspection and extraction unit , Process history information such as written character region extraction candidates in the character recognition process The character collection processing step and the data collection of the written character area extraction candidates obtained from the processing history information in the character recognition process output by the character recognition processing step, with the dialogue processing unit collecting the written character area extraction candidates. A correct coordinate data generating step for generating the correct coordinate data by the correct coordinate data teaching unit based on a processing operation performed interactively by the operator while being displayed together with the data collected in the step, and generated by the correct coordinate data generating step A tagging process step of adding tag information for each reading unit to be recognized by the tagging processor with respect to the correct coordinate data, and registering the correct coordinate data to which the tag information is added in the database; It has.
[0025]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0026]
First, the first embodiment will be described.
[0027]
FIG. 1 illustrates an example of a general database registration system. The operator visually observes the image of the postal letter displayed on the display device, and inputs the character code for each area to be recognized on the postal letter from the keyboard. Each tagging process is performed by the tagging processing unit 2 and then registered in the database 3.
[0028]
FIG. 2 illustrates the environment of a character recognition simulator that is usually developed to improve the performance of the character recognition algorithm. The data collection unit 4 collects image information of character data written on a data sheet to be recognized by the character recognition simulator 5. For example, in the case of a postal letter, a large amount of postal letter data is required for the performance evaluation of the character recognition algorithm, and the image data is collected and stored by dedicated hardware.
[0029]
The character recognition simulator 5 reads the image data collected by the data collection unit 4 and performs character recognition processing with hardware logic incorporating a character recognition algorithm. The recognition result is analyzed by the processing result analysis processing unit 6. . Here, in order to perform analysis processing, log information for each processing block in the character recognition process is stored in the form of intermediate information.
[0030]
FIG. 3 schematically illustrates the configuration of a database registration system that uses the character recognition apparatus according to the first embodiment. The basic configuration of this system is a combination of the components shown in FIGS. 1 and 2, and will be described in detail below.
[0031]
The data collection unit 11 is capable of collecting data on-site, and various collection condition files are stored in the hard disk device together with the collection of postal letter images. The character recognition simulator 12 performs the simulation of the character recognition algorithm using the postal letter image data stored in the hard disk device of the data collection unit 11 as an input, transfers the recognition result to the processing result analysis processing unit 13, and teaches the correct answer Transfer to the unit 14.
[0032]
The correct answer teaching unit 14 receives input data from the operator and the recognition result of the character recognition simulator 12 from the dialogue processing unit 15, generates correct answer data by performing correct answer teaching processing, and transfers the correct answer data to the tagging processing unit 16. Details of the correct answer teaching process will be described later. The tagging processing unit 16 adds tag information for each reading area to be character-recognized to the correct data from the correct answer teaching unit 14, and registers the correct data to which the tag information is added in the database 17.
[0033]
The processing result analysis processing unit 13 can also register necessary data among the processing results in the database 17.
[0034]
FIG. 4 shows an example of each area when character recognition on a postal letter is performed. In a character recognition algorithm for recognizing characters on a postal letter, the overall recognition rate is improved by making a special algorithm recognize each read area based on various attribute information.
[0035]
For example, “210-0915”, which is the contents of the postal code area R1, as the area of the numeric string above the postal letter, its coordinate data (X5, Y5) / (X6, Y6), and the contents of the address area R2 over a plurality of lines “Yukimachi, Kawasaki-shi / 50-2”, its coordinate data (X1, Y1) / (X2, Y2), “Taro Tokyo” as the contents of the address area R3, and its coordinate data (X3 , Y3) / (X4, Y4), etc., are extracted separately and recognized by a recognition algorithm in each area. The character recognition result can be obtained from the result file output by the character recognition simulator 12 such as reliability information of the recognition result for each region.
[0036]
FIG. 5 illustrates an example of tagging. A tag name is assigned to each area on the postal letter described in FIG. For example, the tag name of “210-0915” is “zip code character string”, the tag name of “Yukimachi Kawasaki City” is “address town name character string”, and the tag name of “50-2 address” is “address town block character” Add a tag name such as “Column”. Thereby, data registration in the database 17 can be automated.
[0037]
By the way, the reason why the address area is divided into “address town name character string” and “address town block character string” is due to the difference in the reliability of the character recognition result in the actual system. That is, the reliability of the character recognition result of the content of “address town block character string” tends to be lower than the reliability of the character recognition result of the content of “address town name character string”. For this reason, in automatic registration of data in the database 17, it becomes possible to operate more efficiently by dividing the data.
[0038]
FIG. 6 illustrates a general functional block diagram of the character recognition simulator 12. The image input unit 21 inputs a postal letter image as digital data to the system side. Usually, this is an apparatus independent of the character recognition simulator 12, and the data is stored in a hard disk device or the like and used as input data for the character recognition simulator 12 as necessary.
[0039]
The check processing unit 22 includes an area (line) check processing unit 23 and a character check processing unit 24. The character recognition processing unit 25 performs character recognition processing on the region extracted for each character region as a candidate by the cut-off processing in the cut-off processing unit 22. At this time, by performing matching processing with information stored as a dictionary in the dictionary unit 26, the character recognition result is finally output as a character code.
[0040]
As described above, the processing in each processing block is held as log data and used for analysis of character recognition results. For example, status information selected as an answer from candidate data is stored in the log data, and in this embodiment, this information is intended to be used effectively.
[0041]
FIG. 7 illustrates an operation for displaying a character recognition result on a display device and editing and confirming the content of the result. A postal letter image is displayed in the upper left image display area E1 of the operation screen. A list of tag names is displayed in the tag name display area E2 on the right side of the screen, and a list of tag name contents is displayed in the tag name content display area E3 on the right side.
[0042]
Here, the contents of the tag name, that is, the result code obtained by character recognition is displayed. However, the data with high reliability of recognition is displayed using the recognition reliability information, and the normal display is performed. For data whose reliability is lower than the set threshold, blinking display or the like is performed. Thereby, the operator can perform correction editing while viewing the postal letter image display on the left side of the screen at this point. For example, when the display portion R11 of “address 50-2”, which has the tag name “address block character string”, blinks, the operator looks at the corresponding area of the postal letter image and recognizes the result. To do. If it is wrong, at this stage, the display section R11 of “address 50-2”, which is the content of “address block character string”, is corrected. In order to make confirmation easier, various processing methods such as displaying a corresponding area of a postal letter image in a rectangle can be considered.
[0043]
As can be seen from the above description, in the operation on this screen, the result data automatically processed by the character recognition simulator 12 is displayed as the first candidate, so that the operator can perform postal mail as conventionally performed. While looking at the letter, there is no need to input data from the beginning, and it is possible to confirm and edit operations that focus only on items that cannot be recognized and character codes are not displayed, or on data items with low recognition reliability. As a result, the data input efficiency is considered to be significantly improved.
[0044]
FIG. 8 explains the outline of the flow of the database registration process described above. First, an image of postal letter data is captured (S1), and the image is displayed on a display device (S2). On the other hand, the image is automatically recognized by the character recognition simulator 12 (S3), and the character recognition result is displayed on the operation screen shown in FIG. 7 (S4).
[0045]
Thereafter, the operator confirms or corrects the recognition result (S5), and then tags the result code recognized for each area on the postal letter (S6). The tag information is registered in the database 17 (S7).
[0046]
The postal letter image input to the character recognition simulator 12 is normally processed in units of several thousand sheets according to various simulation conditions and the like, and the number of processed sheets is checked in step S8.
[0047]
In the above description, the character recognition area with high recognition accuracy is easily noticed by the operator by blinking the character information area of the recognition result. At the same time, the character area subjected to the character recognition process is displayed. It can also be shown as a rectangle using coordinate data on the image. For example, if the recognition accuracy of the address block character code recognition result is low, the corresponding character code portion on the screen of the display device is displayed in reverse video or blinking, and a postal letter image is displayed. Similarly, a rectangle based on the coordinate data is displayed in a reversed area, blinking, or changing the color of the rectangle in a certain area so that the operator can easily notice it. Furthermore, it is also possible to enlarge the display of the postal letter image and use it for evaluating the censoring process between character lines or character strings.
[0048]
As described above, according to the first embodiment, the following effects can be expected. That is, in the character recognition development environment, a character recognition simulator is usually developed at the same time for the purpose of improving the verification performance of a recognition algorithm using a large amount of data. Using this character recognition simulator, the character code information of the recognition result is registered in the database. Of course, since the character code information recognized by the character recognition simulator is hardly recognized 100% correctly, the recognition result can be corrected so that an operator who inputs data in the environment of a conventional correct answer teaching process can correct it. The result is displayed on the display device as the first candidate data. This eliminates the need for the operator to input the correct character code from the beginning while looking at the image data, so that the data input efficiency is remarkably improved.
[0049]
Further, since the information output from the character recognition simulator holds the status data of the recognition process, it is possible to know the reliability of the recognition result. When the character code of the recognition result is displayed on the display device as the first candidate, the reliability information is displayed as blue, for example, as attribute information for weighting the reliability of the recognition result. If the recognition result cannot be held, it can be displayed in yellow, and the unrecognizable area can be displayed in red. Thus, there is an effect that the operator's fatigue level can be reduced.
[0050]
Next, a second embodiment will be described.
[0051]
FIG. 9 schematically illustrates the configuration of a database registration system that uses the character recognition apparatus according to the second embodiment.
[0052]
The data collection unit 31 is capable of collecting data on-site, and various collection condition files are stored in the hard disk device together with the collection of postal letter images. The character recognition simulator 32 simulates a character recognition algorithm using the postal letter image data stored in the hard disk device of the data collection unit 31 as input, and the character recognition code as the recognition result is input to another system (not shown). Information.
[0053]
The character recognition simulator 32 is composed of processing modules such as a letter image input unit 33, a written character area check extraction unit 34, and a character recognition unit 35. Processing history information in each processing module includes processing history information. It is output and stored as one log file in the information storage unit (log file) 36.
[0054]
The letter image database 37 is a database of postal letter images collected by the data collection unit 31. The storage medium is usually stored on a CD-ROM. The dialogue processing unit 38 operates and manages letter images from the letter image database 37 and displays them on a GUI (graphical user interface) display unit 39.
[0055]
The correct coordinate data teaching unit 40 generates correct coordinate data based on the processing history information in the processing history information storage unit 36 for the letter image displayed on the GUI display unit 39 by the operation management of the dialogue processing unit 38. Transfer to the tagging processor 41. Details of the correct answer teaching process will be described later. The tagging processing unit 41 adds tag information for each reading area to be recognized to the correct coordinate data from the correct coordinate data teaching unit 40, and adds the correct coordinate data to which the tag information is added to the letter attribute database 42. sign up.
[0056]
The correct coordinate data registered in the letter attribute database 42 is used as input data for the character recognition simulator 32, and it is possible to evaluate the censoring processing algorithm for the described character area in the previous stage of the character recognition processing.
[0057]
FIG. 10 is a diagram for explaining the operation by each area attribute icon in order to perform the character censoring process in accordance with the characteristics of each character description area. By using the log information of the processing history information in the character recognition simulator. The extraction candidates for each character description area are displayed on the letter image. That is, the postal code area candidate R21, the address area candidate R22, and the name area candidates R23 and R24 are respectively displayed in a rectangle in the letter image display area E4 of the operation screen.
[0058]
In FIG. 10, on the right side of the letter image display area E4 of the operation screen, as each extraction area, an icon A1 indicating a zip code area, an icon A2 indicating an address area / horizontal, an icon A3 indicating an address area / vertical, a name area An icon A4 indicating “” and other icons are prepared. The operator can determine the attribute of each extraction region by moving the icon onto the corresponding extraction region on the display image and using a position designation device such as a mouse. .
[0059]
That is, the zip code area icon A1 is moved to the display frame area R21, the address area / vertical icon A3 is moved to the display frame area R22, and the name area icon A4 is moved to the display frame area R23. Here, the name area candidate R24 is affected by the advertisement image of the area R25, and an incorrect area is extracted as the name area, and the operator judges it and moves the name area icon A4 to the area R23. It was decided. In addition, as a position designation apparatus, the method by the direct designation | designated with a touchscreen other than a mouse | mouth is also possible.
[0060]
FIG. 11 explains the contents of the attribute information related to the icon corresponding to each written character area. For example, by using information such as horizontal writing, numeric characters, and a description frame as the attribute information of the postal code area, it is possible to perform a cut-off process of the written characters according to the attribute information.
[0061]
In addition, by using information such as vertical writing, stamped characters (printed characters), and a plurality of lines as the attribute information of the address area, it is possible to perform a cut-off process of written characters according to the attribute information. Incidentally, the address area icon A2 for horizontal writing is also prepared on the right side of the operation screen. In this case, the address area icon A3 for vertical writing is used.
[0062]
By such processing, correct coordinate data of each written character can be obtained by the cut-off processing according to the attribute information in each region with high-performance and stable written character check-out processing. Of course, the cut-off processing algorithm for each character in each area is performed by the method used in the character recognition simulator.
[0063]
A direct mail letter is considered as an example of a similar letter. Since these direct mail letters have a uniform way of writing address information, their similarity can be checked as an image. In addition, as a characteristic of direct mail letters, many are uniquely designed outside the address area. It is quite possible to check the similarity of these images.
[0064]
FIG. 12 illustrates an example of a similarity pattern for such a direct mail letter. In FIG. 12, a region P1 is an example in which an address and address using a label often used in a direct mail letter or the like are described. In the label, all information such as the postal code, address, and address is described in the same label, and the similarity can be checked using the label frame size information.
[0065]
In FIG. 12, a region P2 shows an example of a frequently used advertisement logo mark outside the address description region of the direct mail letter. By capturing the advertisement logo mark area as an image and checking its similarity, it is possible to determine the same type of direct mail letter.
[0066]
FIG. 13 illustrates that the similarity of postal letter images is checked, and similar letters are automatically acquired as correct coordinate data in a lump and registered in the database.
[0067]
In FIG. 13, first, the similarity of letters is checked in step S11. Here, the similarity of the letter is checked by the method described in FIG. Note that the letter pattern data that is the target of the similar letter is registered in advance in the form of a template in the letter pattern template data (storage unit) 43.
[0068]
As a result of the similarity check in step S11, for the letter whose similarity exceeds the set value, each correct coordinate data is automatically acquired using various processing parameters attached to the template data having high similarity. (S12).
[0069]
On the other hand, as a result of the similarity check in step S11, for a letter whose similarity is equal to or less than the set value, an interactive processing operation for obtaining correct coordinate data is performed in the form described in FIG. This will be described in detail below.
[0070]
First, using the extracted region candidate information for each written character region existing in the processing history information of the character recognition processing process stored in the processing history information storage unit 36, the result is displayed as a letter image. A process of displaying the extracted rectangle on the operation screen is performed (S13). Thereafter, the GUI is operated on the icon corresponding to each corresponding written character area to determine its attribute (S14).
[0071]
Next, a character-by-character check-out process is performed using attribute information depending on the extraction area linked to the icon assigned to each written character area (S15). Thereby, it is possible to perform an optimum character inspection process suitable for the characteristics of each written character area. Next, in order to identify a plurality of attribute information in the letter attribute database 42, the correct coordinate data of each extracted character area extracted as correct coordinate data, for example, a zip code area, an address area, a name area, etc. Are added to the letter attribute database 42 by automatic processing (S16).
[0072]
FIG. 14 illustrates an image of a postal letter, and shows each character description area when character recognition on the postal letter is performed. In the character recognition algorithm for recognizing characters on a postal letter, the overall character recognition rate is improved by causing each reading area to recognize with a dedicated algorithm based on various attribute information.
[0073]
For example, “210-0915”, which is the contents of the zip code area R21, and its coordinate data (X5, Y5) / (X6, Y6) as the contents of the numerical string above the postal letter, and the contents of the address area R22 over a plurality of lines. A certain “Yukimachi, Kawasaki city / 50-2 address”, its coordinate data (X1, Y1) / (X2, Y2), “Taro Tokyo” as the contents of the address area R23, and its coordinate data (X3 Y3) / (X4, Y4), etc. are extracted separately, and characters are recognized by a recognition algorithm in each area. The character recognition result can be obtained from the result file output by the character recognition simulator 32, such as reliability information of the recognition result for each region.
[0074]
FIG. 15 illustrates an example of tagging the correct coordinate data of each written character area, in which a tag name is assigned to each area on the postal letter described in FIG. For example, the tag name of the correct coordinate data of “210-0915” is “zip code coordinates”, the tag name of the correct coordinate data of “50-2 Yanagimachi, Saiwai-ku, Kawasaki-shi” is “address coordinates”, “Taro Tokyo” The tag name of the correct coordinate data is given as “name coordinate”. Thus, data registration in the letter attribute database 42 can be automated.
[0075]
In the letter attribute database 42 shown in FIG. 9, a plurality of types of attribute information are registered with a structure, and the correct coordinate data is one attribute information. A tag name is defined to identify the correct coordinate data.
[0076]
As described above, according to the second embodiment, the following effects can be expected. In other words, in a character recognition development environment, a character recognition simulator is usually used at the same time for the purpose of improving the verification performance of a recognition algorithm based on a large amount of data, the preceding character area, and the written character censoring processing algorithm. We are developing. This character recognition simulator accumulates information on the process of character input, cut-off processing, and recognition processing as a log information file. Utilizing this character recognition simulator, the described character area and the described character extraction rectangle information, that is, the information of the coordinate data indicating the extraction rectangle is registered in the database. Of course, the rectangular character data for extracting the written character region and the written character is hardly extracted 100% correctly. Therefore, the extracted coordinate data is added to the letter image by adding the above-described interactive processing method. Since it is not necessary to generate correct coordinate data from the beginning while watching the above, the efficiency of generating correct coordinate data is greatly improved.
[0077]
For example, a direct mail letter is considered to be the same letter form. By checking the similarity of letter images against similar letter forms, and using similar template information for similar letters, correct coordinates of a large number of letters at once. Since the data is automatically acquired, there is an effect that it is possible to further improve the efficiency of generating correct coordinate data and extremely reduce the fatigue level of the operator.
[0078]
【The invention's effect】
As described above in detail, according to the present invention, in the operation of registering correct data used for evaluation of the character recognition algorithm in the database, data can be input more accurately and without taking time. As a result, it is possible to provide a database registration method using a character recognition device that enables efficient database registration with reduced data input fatigue.
[0079]
Further, according to the present invention, in the operation of registering correct coordinate data of written characters used for evaluation of the character recognition algorithm in the database, data can be input more accurately and without taking time, As a result, it is possible to provide a database registration method using a character recognition device that enables efficient database registration with reduced data input fatigue.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating an example of a general database registration system.
FIG. 2 is a block diagram illustrating a configuration example of a character recognition simulator system.
FIG. 3 is a block diagram schematically showing a configuration of a database registration system using the character recognition device according to the first embodiment of the present invention.
FIG. 4 is a diagram for explaining an example of each area when character recognition on a postal letter is performed.
5 is a diagram for explaining an example in which recognition results for each area shown in FIG. 4 are classified by tags.
FIG. 6 is a diagram illustrating functional blocks of a character recognition simulator.
FIG. 7 is a diagram showing an example of an operation screen for explaining that an output result of a character recognition simulator is displayed on a display device with a tag name assigned to each region, and an operator performs an editing or confirmation operation.
FIG. 8 is a flowchart for explaining an outline of the flow of database registration processing;
FIG. 9 is a block diagram schematically showing a configuration of a database registration system using a character recognition device according to a second embodiment of the present invention.
FIG. 10 is a diagram showing an example of an operation screen for explaining correct coordinate data editing and confirmation operations in each written character area on a postal letter.
FIG. 11 is a diagram for explaining an example of attribute information related to an icon corresponding to each written character area on a postal letter.
FIG. 12 is a diagram for explaining a pattern example of similarity to a postal letter image;
FIG. 13 is a flowchart for explaining an overview of the flow of correct coordinate data generation and database registration processing;
FIG. 14 is a diagram for explaining an example of each area when character recognition on a postal letter is performed.
15 is a diagram for explaining an example in which the extracted coordinate result for each area shown in FIG. 14 is divided by tags.
[Explanation of symbols]
11 …… Data collection part
12. Character recognition simulator (character recognition device)
13 …… Processing result analysis processing section
14 …… Correct answer teaching part
15 ... Dialogue processing part
16 …… Tagging processing part
17 …… Database
21 …… Image input section
22 …… Cut-off processing section
23 …… Line check cut-off processing section
24 …… Character cut-off processing section
25 …… Character recognition processing section
26 …… Dictionary
31 …… Data Collection Department
32 …… Character recognition simulator (character recognition device)
33 ... Letter image input section
34 …… Described character area check extraction part
35 …… Character recognition part
36 …… Processing history information storage (log file)
37 ... Letter image database
38 …… Dialogue processing section
39 …… GUI display
40 …… Correct answer data teaching part
41 …… Tagging processing section
42 ... Letter attribute database

Claims (1)

In the database registration method using the character recognition device that registers the coordinate data of the character area described on the paper used for the performance evaluation of the character recognition algorithm in automatic character recognition in the database as the correct coordinate data,
A data collection step of reading and collecting data on the standardized paper to be registered in the database by a data collection unit;
A written character area detection / extraction unit for detecting and cutting out a written character area for the collected data with respect to the data collected by the data collection step, and a written character detected and cut out by the written character area detection / extraction unit Character recognition that performs character recognition processing using a character recognition device that includes a character recognition unit that performs character recognition on a region and outputs processing history information such as written character region extraction candidates in the character recognition process Processing steps;
For the written character area extraction candidates obtained from the processing history information in the character recognition process output in the character recognition processing step, the data collected by the data collection step for the written character area extraction candidates in the dialog processing unit A correct coordinate data generation step for generating the correct coordinate data by the correct coordinate data teaching unit based on a processing operation performed interactively by the operator while displaying with
Tag for adding tag information for each reading unit to be recognized by the tagging processing unit to the correct coordinate data generated in the correct coordinate data generating step, and registering the correct coordinate data with the tag information in the database Processing steps;
A database registration method using a character recognition device.

Images