JP3648270B2 - Form reading device and accounting form reading device - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to a form reading apparatus that reads forms for inputting data written in a document or form (hereinafter referred to as form), and more particularly, to a form reading apparatus using an optical character reader.
[0002]
[Prior art]
An optical character reader (OCR) is one type of input device that reads a document or a form and obtains input data. In principle, optical character readers optically scan forms to be read, obtain reflected light from characters or the like written on forms using a pencil or writing instrument, and generate electrical signals by photoelectric conversion. And A / D conversion is performed to obtain image data, and character recognition processing such as pattern recognition is performed on the image data. In such a character recognition device, technically, improvement of character reading accuracy and improvement of recognition rate of read characters have been attempted, but an important solution combined with improvement of reading accuracy and recognition rate. As a problem, there is a process when there is an entry mistake or a typo in a form, that is, a check and correction method of input data. As the most primitive method when detecting unrecognizable characters at the time of input, the read forms are discharged as hardware as unreadable, the operator inspects and corrects the discharged forms, A method of re-reading was originally developed.
[0003]
However, in the above-described method, it is necessary to determine how many characters are unrecognizable for the operator. For example, the operator's load is excessive and the data input efficiency is poor, and thereafter there are unrecognizable characters on the display device. A system has been developed that allows the operator to make corrections by indicating the position of the unreadable characters by showing the image pattern of the entry data each time the form is present and displaying the unreadable characters as a light spot. However, the above method has a disadvantage that it is unclear how the displayed image pattern was read (character recognition), and it is checked when only the reading result (recognition result) is displayed. Therefore, there is a disadvantage that it is necessary to match the form with each time.
[0004]
On the other hand, if there are unreadable characters in the form, each time the image is read, the image data and the reading result (the readable character is the reading result (recognition result), and the unreadable character is the unreadable character. There is a system (Japanese Examined Patent Publication No. 57-46106) that displays the information in correspondence with each other and displays an unreadable character on the basis of the display. .
[0005]
[Problems to be solved by the invention]
However, in the technique disclosed in the above Japanese Patent Publication No. 57-46106, when a large number of forms are read continuously, if a non-readable character is determined in a certain form, the form is displayed on the display device. Since the operator refers to the form and corrects the characters from the input device, the continuous reading of the form is often interrupted for a long time, and the input efficiency is lowered.
[0006]
In addition, accounting report data is numeric, and it may be easier for the entire device to handle characters (numerics) such as counterparty codes and subject codes as numeric values. In handling. For this reason, in the correction of accounting form data with an apparatus adopting such a method, the correction of only the characters that cannot be recognized as in the prior art causes inconveniences when performing a total check described later. .
[0007]
Furthermore, in the technique disclosed in Japanese Patent Publication No. 57-46106, when there are unreadable characters, the image data and the read result (the readable character indicates the read result (recognition result), and the unreadable character cannot be read. However, due to recent developments in reading accuracy and recognition technology, there are no characters that are virtually unreadable, and although the degree of discrimination accuracy is different, the read characters are different from correct or incorrect. As recognizable. Therefore, in the correction of input data in character recognition devices with excellent reading accuracy and recognition technology in recent years, it is hoped to develop a new method that replaces the conventional method for determining the suitability of entry data for forms due to the presence of unreadable characters. It was rare.
[0008]
The present invention has been made based on the above-mentioned problems and solutions of the prior art, and after inputting all target forms, the form of the read forms and the input image of the input data and the recognition result or An object of the present invention is to provide a form reading apparatus configured to display an estimation result and correct input data based on the estimation result displayed when the estimation is performed.
[0009]
[Means for Solving the Problems]
In order to achieve the above object, a form reading device according to a first aspect of the present invention includes a character reading unit that optically reads character data written on a form, converts it into an electrical signal, and outputs image data; A recognition unit that performs character recognition processing on image data from the character reading unit and outputs candidate characters as character data of the image data when there is a candidate character that is equal to or greater than a reference recognition threshold, and displays image data and / or character data A form reading apparatus having a display unit capable of storing data and a storage unit for storing data, wherein the image data obtained by the character reading unit is further applied to image data that is less than a reference recognition threshold value of the recognition unit. performs recognition processing to set a lower recognition threshold than recognition threshold, the first candidate character of a low recognition threshold or more candidate characters than the reference recognition threshold of the image data estimation sentences Storing a character estimating means obtained by the data, the storage unit in association with the character data output from the recognition unit with the image data when the image data is the image data of the above criteria recognition threshold, the image data reference When the image data is less than the recognition threshold, the image data and the estimated character data obtained by the character estimation means are stored in association with each other in the storage unit, and the image data and image data for each form from the storage unit The character data or the estimated character data stored in association with each other is taken out and displayed on the display unit in association with the image data, and when the estimated character data is displayed, the estimated character data is discriminated in a distinguishable manner. And a display means .
[0010]
Further, the form reading device according to the second invention is an optical character reading unit that optically reads numerical data written on an accounting form and converts it into an electrical signal and outputs image data, and an optical character reading unit. A recognition unit that performs character recognition processing on image data from and outputs a candidate character as numerical data of the image data when there is a character candidate greater than a reference recognition value, and a display capable of displaying image data and / or character data And an accounting form reading device having a storage unit for storing data, and among the image data obtained by the optical character reading unit, for the recognition target numerical data less than the reference recognition threshold value of the recognition unit, performs recognition processing to set a lower recognition threshold than the reference recognition threshold, the first candidate character of the reference below recognition thresholds recognition threshold or more candidate characters of the image data estimation Numerical value estimation means obtained as value data, and when the image data is image data greater than or equal to a reference recognition threshold value, the image data and numerical data output from the recognition unit are stored in association with each other, and the image data is the reference recognition threshold value If the image data is less than, the storage means for storing the image data and the estimated numerical data obtained by the numerical estimation means in association with each other, and when the reading of all accounting forms to be read is completed The image data corresponding to the form and the numerical data or estimated numerical data stored in association with the image data are extracted from the storage unit for each unit of the accounting form, and the format is determined for each type of accounting form. based on the display on the display unit in association with the image data, to be identified to the estimated character data when displaying the estimated character data Characterized by comprising a display means for another display.
[0011]
[Action]
With the above configuration, the form reading apparatus according to the first aspect of the invention sets a recognition threshold lower than the reference recognition threshold for image data obtained by the character reading unit by the character estimation unit and below the reference recognition threshold by the recognition unit. Since the estimated character data is obtained by performing recognition, unreadable characters do not occur.
[0012]
The accounting form reading device of the second invention obtains form type information for determining the type of form read from the image data obtained by the character reading unit, and the numerical estimation means includes Recognize by setting a recognition threshold lower than the reference recognition threshold for the recognition target numerical data below the reference recognition threshold by the recognition unit, and estimate the numerical value by estimating the numerical value based on the logical estimation criterion based on the form type information. As a result, an unreadable numerical value does not occur. Further, after all the accounting forms are read, the data is displayed in the form on the input correction screen.
[0013]
【Example】
FIG. 1 is a block diagram showing the configuration of an embodiment of a form reading device 10 according to the present invention. 1 is an optical character reading device (hereinafter referred to as OCR), which corresponds to a reading unit, and 2 is a character recognition unit. Reference numeral 3 denotes character estimation means, reference numeral 4 denotes storage means, reference numeral 5 denotes display means, and reference numeral 6 denotes a storage unit including a storage device for storing data, and corresponds to a magnetic disk, an optical disk, a floppy disk, and the like. Reference numeral 7 denotes a display unit including a display device such as a display, 8 denotes an input device such as a keyboard, and 9 denotes data correction means.
[0014]
FIG. 3 is an example in which the form reading apparatus 10 of FIG. 1 is applied to a financial processing system, and shows a start screen of the financial processing system (denoted as “OCR financial system” in the figure). Here, the menu “1” in FIG. 3 shows a batch input process of accounting forms by the form reading device 10, and the menu “2” shows a data correction process after the batch input process is completed.
[0015]
FIG. 2 is a flowchart showing the operation of the form reading apparatus 10 in the financial processing system of FIG. 3, and the operation of the form reading apparatus 10 will be described below with reference to FIGS.
[0016]
In step S 0, the accounting form to be read is set in the OCR 1 stacker, and the menu “1” is selected. In step S1, an accounting form is read by OCR1 in units of one sheet, and a form type code 31 printed in advance on a form as shown in FIG. 4 (withdrawal slip in the example of FIG. 4) and a handwritten company code (partner) Destination code) 32, date 33, account item code 34, auxiliary column 35, slip number 35, department code column 37, subject item codes 38 and 39, amount 40, 41 and total amount 42 etc. Or numerical values) are optically scanned and converted into electrical signals by photoelectric conversion, and an image data string is obtained from these electrical signals.
[0017]
In step S2, the character recognition unit 2 performs character recognition processing based on the obtained image data string. Since the character recognition process is well known and the description other than the part related to the present invention is omitted, the recognition unit 2 performs character recognition process for each character on the normally input image data sequence and performs predetermined processing on the image data for one character. When there is a candidate character exceeding the recognition threshold value, the candidate character is recognized as character data of the image data, and the process proceeds to step S4. On the other hand, when there is no candidate character equal to or greater than the predetermined recognition threshold for the image data for one character, the process proceeds to step S3 and the image data is transferred to the character estimation means 3.
[0018]
In step S3, the character estimation means 3 estimates characters from the received image data as estimated character data, as will be described later (see FIG. 7 and FIG. 8). In step S4, the storage means 4 stores character data or estimated character data in a memory (not shown) of the form reading device 10 in association with the image data.
[0019]
The above processing is repeated for each character string for each input item of the form, and when the image data recognition process or estimation process for one form is completed, the image data for one form and the character associated with the image data are stored by the storage means. Data or estimated character data is written in the storage unit 6, and the next form reading operation by the OCR 1 is started.
[0020]
In the present embodiment, flags corresponding to the number of items are provided at the head or tail of the image data for one form stored in the storage unit. As a result of the recognition process and the character estimation process, when there is estimated character data corresponding to the image data character string of a certain input item, the input item of the form data is a character string composed of character data and estimated character data. In this case, a specific value (for example, “1”) is assigned to the above-described flag corresponding to the entire character string of the item as an estimated character string.
[0021]
In this embodiment, flags corresponding to the number of items are provided at the head or tail of image data for one form stored in the storage unit. However, the present invention is not limited to this method, and an estimated character string can be determined. A flag may be provided.
[0022]
In step S5, when all the forms to be input have been read, a menu screen as shown in FIG. Here, when the menu “2” is selected, the process proceeds to step S6. In step S6, the image data and the character string data string or the estimated character string stored in the storage unit 6 are extracted by the display means 5, and the corresponding items are displayed on the display unit 7 in a form format (input correction screen) as shown in FIG. Display at the item position. FIG. 5 shows the data (image data) and the character string data string or the estimated character string shown in FIG. 4 displayed at each item position on the input form (withdrawal slip) format display screen of FIG.
[0023]
In the example of FIG. 5, among the numbers displayed in the display items, the upper column is an image data string displayed in the form of a handwritten character itself, and the lower column is a (recognized) character data string or an estimated character data string (actually A data string in which at least one character of the estimated character data string is estimated character data), and the estimated character data is displayed in reverse for easy identification from other data (in FIG. 5, hatched lines indicate inverted display). Is done. For example, where the company code 32 should actually be “1234”, in the handwritten character, the third digit “3” from the beginning is misleading as “8”, and the estimation means “8” is estimated, so the estimated character of the company code 32 The data string is “1284”, and the company code column is highlighted as an estimated character data string.
[0024]
Similarly, in the date 33, “5 5 5” is “80803”, in the auxiliary code column 35 ““ is ””, in the department code 36 ““ is ””, and in the partner course 38 is “854”. “884” is estimated as “884” and “855” is estimated as “855” in the opponent subject 39 and is highlighted. In the embodiment, the total column is divided into three columns, upper, middle, and lower. The upper column is the image data, the lower column is the position where the character data string or the estimated character data string is displayed, and the middle column is the total check. As in the case of the estimated character data string, the sum of the numerical values in the amount data field (40, 41) consisting of the character data string or the estimated character data string is taken to facilitate identification with other data. Is highlighted.
[0025]
Thus, it is desirable that the display means 5 has an identification display addition means (not shown) for displaying an estimated data character string so that the operator can visually identify it easily. Further, before displaying the data for one form on the display unit 7, the display means 5 checks the form type code 31 of the record for the one form in order to display the form format corresponding to the form type. A form format based on the form format corresponding to the form type code is displayed on the display unit 7. Here, the form format corresponding to the form type code is stored in the storage unit 6 in advance, and the corresponding format is taken out and displayed by the display means 5.
[0026]
FIG. 6 is an explanatory diagram of the above-described form display format, and takes the form type code 31 (code = 122; withdrawal slip (see FIG. 4)) as an example. Actually, such a format is determined for each form and stored in the storage unit 5. In the figure, (1) to (20) are explanatory numbers corresponding to the description given below.
(1) Display format of company code 32, position (x, y) and range, number of digits (4 digits), ..., required flag = 1, ..., etc. (2) Display of date 33 Format (position (x, y) and range, number of digits (6 digits), ..., required flag = 1, ..., etc.) (15) Display format of the subject code, (15) to ( 19) is group data, the position (x, y) and range, the number of digits (4 digits), etc. (19) The display format of 42 in the amount column, (15) to (19) are group data Position (x, y) and range, number of digits (10 digits), mandatory flag = 1, ..., etc. (20) Display format of total amount 42, position (x, y), range, number of digits (11 digits), mandatory flag = 1,...
In step S <b> 7, the operator moves the cursor to the item position when there is an erroneous data item by looking at the data for one form displayed on the display unit 7, and inputs correct data from the input unit 8. For data input, in the case of a numerical item (for example, a money amount column) or a numerical item (for example, a subject code), all digits of the item are input. The input data is corrected by the correcting means 9, and when the operator confirms that all items for one form have been corrected (pressing the confirmation key), the storage unit 5 uses the corrected character data for one form. The corresponding record for one form is updated, and control is returned to step S6 to display the next form.
[0028]
In the above correction work, the estimated data character string for each item is displayed on the display means 5 so that it can be easily visually identified. Therefore, the operator compares the image data with the estimated character string and determines whether it is correct or incorrect. Can be easily corrected. Further, even if there is a character whose character data is not accurately estimated by the estimating means 3, the estimated data character string for each item is displayed on the display means 5 so that it can be easily visually identified. There is no particular problem because it can be simplified.
[0029]
FIG. 7 is a flow chart showing an embodiment of the character estimation means 3 of FIG. 1. The character estimation means 3 estimates characters from the image data received from the recognition unit 2 by the following operation to obtain estimated character data. In step S11, the image data for one character is checked for the presence or absence of a candidate character exceeding a predetermined recognition threshold in the character recognition process in the recognition unit 2, and if there is no candidate character, the process proceeds to step S12. In step S12, the recognition threshold value in the recognition unit 2 is stored, a threshold value equal to or lower than the recognition threshold value is set as the recognition threshold value, and the process proceeds to step S13. In step S13, recognition processing is performed based on the set recognition threshold. In step S14, the presence or absence of the candidate character is checked. If there is no candidate character, the process proceeds to step S12. The return process is terminated (the process proceeds to step S4 in FIG. 2).
[0030]
FIG. 8 is a flowchart showing another embodiment of the character estimation means 3 of FIG. 1. The character estimation means 3 estimates characters from the image data received from the recognition unit 2 by the following operation to obtain estimated character data. In step S21, the image data for one character is checked in the character recognition process in the recognition unit 2 for the presence of a candidate character that is equal to or greater than a predetermined recognition threshold. If there is no candidate character, the process proceeds to step S22. In step S22, the value (character) of the data is estimated according to a predetermined estimation criterion. As an example of the estimation criterion, an estimation criterion can be set for each item of the form type. For example, there is a certain rule (corresponding relationship) for the item code in accounting forms, and when the form type is a withdrawal slip, the credit item is “cash”, “deposit” or “check”, and the counterpart item (debit) (Course) is an expense item of “fare”, “packing cost”,..., And these item codes are usually set according to a certain code standard. For items having such a relationship, a correlation can be set as an estimation criterion in advance and used for the estimation process. In addition, in a document or the like, “Doshi ·.” Is “Did.”, The tail, prefix, and idiom can be estimated from the context, and can also be estimated by an associative method. Note that the estimation processes in FIGS. 7 and 8 are desirably performed in combination. 7 and 8, when no item is entered, 0 (zero) is estimated for numeric items, and space is estimated for character items.
[0031]
Although the present invention has been described above by way of the above embodiments, it goes without saying that the form reading apparatus of the present invention can be used not only for accounting forms but also for reading documents and other forms and creating input data. In the above description, the data such as the item code and the amount of money in the accounting form has been described as characters (numbers). However, it is preferable in terms of processing that the amount field is handled as a numerical item in response to a request for total check in the internal processing. Further, not only the amount column but also all the numerical entry items (for example, date, subject code, etc.) of the accounting form may be handled as numerical items.
[0032]
【The invention's effect】
As described above, according to the form reading apparatus of the present invention, since the estimation means is provided, the input image of the input data and the recognition result or the estimation result are displayed in parallel in the form of the form, so that erroneous reading (estimation) ), The estimated character data string is displayed so as to be distinguishable from other data, so that it can be easily corrected. In the accounting form reading apparatus of the present invention, the numerical estimation means estimates the numerical value based on the logical estimation criterion based on the form type information for the recognition target numerical data below the reference recognition threshold value by the recognition unit among the recognition target numerical data. The estimated numerical data is obtained and the input image of the numerical data and the recognition result or estimation result are displayed in parallel in a slip format, so that even if erroneous reading (estimation) occurs, the estimated numerical data can be distinguished from other data Since it is displayed, it can be easily corrected. In addition, since all the accounting forms are read and the data is displayed on the input correction screen, the flow of processing is not interrupted and the input and input correction efficiency are improved.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an embodiment of a form reading apparatus according to the present invention.
FIG. 2 is a flowchart illustrating an operation example of the form reading apparatus in FIG. 1;
FIG. 3 is an example of a start screen of an example in which the form reading apparatus of the present invention is applied to a financial processing system.
FIG. 4 is an example of an input form.
FIG. 5 is an example of a correction screen.
FIG. 6 is an explanatory diagram of a form display format;
FIG. 7 is a flowchart showing an embodiment of character estimation means.
FIG. 8 is a flowchart showing an embodiment of character estimation means.
[Explanation of symbols]
1 Optical character reader (character reader)
2 Character recognition unit 3 Character estimation means (numerical value estimation means)
4 Storage means 5 Display means 6 Storage device (storage part)
7 Display device (display unit)
8 Input device (input unit)
10 Form reader

Claims (2)

Character reading unit that optically reads the character data written on the form, converts it into an electrical signal and outputs the image data, and performs character recognition processing on the image data from the character reading unit and exceeds the reference recognition threshold A document reader having a recognition unit that outputs the candidate character as character data of the image data, a display unit that can display image data and / or character data, and a storage unit that stores the data Because
Among the image data obtained by the character reading unit, image data that is less than the reference recognition threshold value of the recognition unit is further subjected to recognition processing by setting a recognition threshold value lower than the reference recognition threshold value. Character estimation means for obtaining a first candidate character of candidate characters having a low recognition threshold value or more as estimated character data of the image data ;
When the image data is image data equal to or higher than the reference recognition threshold, the image data and character data output from the recognition unit are stored in association with each other, and the image data is stored in the reference recognition threshold. Storage means for associating and storing the image data and the estimated character data obtained by the character estimation means in the storage unit when the image data is less than
The character data stored in association with the image data and the image data for each form and the estimated character data are extracted from the storage unit and displayed on the display unit in association with the image data. And a display means for displaying the estimated character data so that the estimated character data can be discriminated in a distinguishable manner. An optical character reading unit that optically reads numerical data written on an accounting form and converts it into an electrical signal and outputs image data, and a character recognition process for the image data from the optical character reading unit. A recognition unit that outputs the candidate character as numerical data of the image data when there is a candidate character greater than or equal to the recognition value;
An accounting form reader having a display unit capable of displaying image data and / or character data, and a storage unit for storing data,
Wherein among the image data obtained by an optical character reader, the relative recognized numerical data below threshold recognition threshold of the recognition unit, further performs the recognition process by setting a lower recognition threshold than the reference recognition threshold, the Numerical estimation means for obtaining a first candidate character of candidate characters that are equal to or higher than a recognition threshold lower than a reference recognition threshold as estimated numerical data of the image data ;
When the image data is image data equal to or higher than the reference recognition threshold, the image data and numerical data output from the recognition unit are stored in association with each other, and the image data is less than the reference recognition threshold. Storage means for associating the image data and the estimated numerical data obtained by the numerical value estimating means in association with each other and storing them in the storage unit;
When all the accounting forms to be read have been read, the image data corresponding to the forms and the numerical data stored in association with the image data are stored from the storage unit for each unit of the accounting form. Numeric data is extracted and displayed on the display unit in association with the image data on the basis of a format determined for each type of accounting form. When displaying the estimated character data, the estimated character data is discriminated so that it can be identified. Display means for displaying;
An accounting form reading apparatus comprising:

Images