JPH0664624B2 - Optical character reading method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an optical character reading method, and more particularly to a character reading method that automatically detects a character entry format by an optical character reading device (hereinafter referred to as OCR) and outputs a reading result. It is a thing.

<Prior Art> In recent years, the data input efficiency of computers has been greatly improved due to the spread of OCR. However, in the past, a form that can be read by OCR had to be given an entry format, that is, a character entry position, the number of digits, etc., in detail in advance as an OCR reading program or a parameter.
For this reason, it is necessary to reset the parameters each time the entry format changes, and of course, the form must be changed each time, which makes the issue of dealing with diversity a problem.

In other words, with the spread of OCR, the range of people who use OCR has expanded, and the situation where it was initially used or managed by professionals has changed to a situation where non-specialists also handle it. Therefore, dealing with the variety of forms and reading programs as described above has become a big problem.

From this, it is desirable that the form can be prepared and the reading program can be created as easily as possible. As a concrete correspondence example, use PPC paper or create the reading program only by reading the actual form. Methods are emerging. However, since these improvement measures do not improve both the form and the reading program, OCR is still not always convenient for non-specialists.

In general, OCR is ideal for reading anything without specifying the format of the form, but technically this level is considered impossible for the time being.

To give a step back, it can be said that the OCR automatically reads the character groups that are filled in as freely as possible on a general-purpose form to alleviate the above problems and bring them closer to the ideal. To be as free as possible here means to set extremely simple constraints or rules, and then fill them in freely. However, if the rule is based on the habits of everyday life, it is considered that it will be generally accepted, and its practicality will increase. However, the rules here must be logically readable by OCR.

The degree of rules and the difficulty of automatic reading by OCR naturally have a correlation, but in the range where OCR can be logically deciphered, it can be treated simply as the degree of processing complexity. Here, it seems that the proof of the limit of rules that OCR can logically decipher is a problem, but the limits are set realistically within the range in which OCR users understand the rules and respond cooperatively. As it can be done, no rigorous proof is required, only the suitability of the rule is tested in the evaluation of the resulting utility value.

A general-purpose form that can be filled in as freely as possible can be one in which the filling frames are set vertically and horizontally, and can be designed and prepared in advance for the convenience of the device and user. The reading program for the entire general-purpose form is
A conventional type that can be incorporated in the OCR beforehand can be used.

<Object of Invention> An object of the present invention is to read all characters on a form including blanks and then temporarily store all character information when reading a general-purpose form filled in according to an entry rule, and to arrange the whole character arrangement. Details are obtained in advance by extracting and processing the entry layout information that appears on the form as intended by the character writer, and automatically obtaining the character data group and entry method information on the form according to the rules. It is to provide an optical character reading method that detects an entry format from a character array on a general-purpose format, removes unnecessary data, and outputs a necessary reading result intended by the writer without preparing a format program. .

<Structure of the Invention> According to the present invention, a writing rule that prescribes only the relative arrangement of the characters to be written is set in advance, and the characters and the like written in the general-purpose form according to the writing rule are also included, including blanks. Storage means for temporarily storing all character information obtained by reading in a character array that matches the format of the general-purpose form, and entry layout information indicating a character array on the general-purpose form from all character information read from the storage means. Layout information extracting means for extracting and outputting based on the entry rule, and a character data field belonging to the same group entered on the general-purpose form for a character array included in the entry layout information input from the layout information extracting means. Based on the entry rule, format information indicating the extracted data of and the entry method of the character data field is provided. Format information extraction means for extracting and outputting, and read result output means for reading out all character information stored in the storage means and outputting only required data based on the format information input from the format information extraction means. A characteristic optical character reader is obtained.

<Example> Next, the Example of this invention is described using drawing.
FIG. 1 is a block diagram of an embodiment of the present invention, in which 1 is an OC
R, 2 is a temporary storage device, 3 is a layout information extraction unit, 4
Is a format information extraction unit, and 5 is a read result output unit.

The reading result 11 of the general-purpose form read by the general OCR 1 is stored in the temporary storage device 2 in a character array that matches the format of the general-purpose form, including blanks. Temporary storage device 2
The output 21 of is output to the layout information extraction unit 3. The layout information extraction unit 3 extracts the input layout information from the character array on the form, that is, the contents of the temporary storage 2, and supplies it as the output 31 to the format information extraction unit 4. Next, the format information extraction unit 4 classifies the data groups belonging to the same group with respect to the character array,
The writing position is confirmed and right-handed and left-handed checks are performed, and the result is given to the read result output unit 5 as the format information 41. Finally, in the read result output unit 5, while reading the contents 21 of the temporary storage, unnecessary data is removed using the format information 41, and only the necessary data intended by the writer is output as the final output 51. , Read the data entered on the general-purpose form based on the entry rules.

FIG. 2 shows a general-purpose form and an example of character entry in one embodiment of the reading method according to the present invention. As shown in the figure,
First of all, the general-purpose form S is horizontally placed in A5 format and is in a general-purpose format with 20 characters per line and 12 lines in total. This 20 characters ×
The 12-line generic form S is programmed in the OCR1 to read in the generic OCR1 shown in FIG. Since such a general-purpose form can be sufficiently read by a commercially available OCR, detailed description is omitted.

In FIG. 2, the numbers 1 to 20 shown in the upper part of the form represent column numbers, and the numbers 1 to 20 shown in the left part show row numbers. Each square shown as a square represents a character entry box,
Each number, etc. in the entry frame represents a character entry example. The characters to be entered are not limited to numbers, and can be anything that can be read by OCR.

Now, the character entry in the figure is made according to the following entry rules in the embodiment.

Rule: Insert a blank line when the format changes line by line. (Example; line,) Rule 2: Blank lines in a line are used as field breaks. However, the left and right columns are cut even if there is no blank.

Rule 3: Left-justified and right-justified: A) If the left ends of each field are aligned, B) If the left ends of each field are not aligned, and if the right ends are aligned, then such rules The rules 1 to 3 are almost the same as the data entry method used in daily life, and are generally used even among human beings, and limiting the OCR orientation has little effect on daily business activities. In the explanation of the above rules 1 to 3, a field represents one set of data, and for example, "100" in columns 18 to 20 of the row in FIG. 2 is 1
It seems that there are two fields.

There are some supplementary rules regarding Rule 3. For example, this is a field of 5 digits for business purposes, but when filling in a certain form, if you happen to have only 3 digits of data in the right column, you should be aware of rule 3 and the left and right columns will not be confused. Need to do so. In other words, regarding the fields in columns 16 to 20 of the row in FIG. 2, this field has 5 digits, but the data is right-padded to "10".
Even if it is "0", "210", or "350", if you continue to enter 3 digits, it will be left-justified according to Rule 3A). At this time, if you enter "00100" for "100" as shown in the figure, it will be 5-digit data, and as a result, it will be right-justified. This is a supplementary rule.

Supplementary rule: Regarding right-padded data, when there is less data to fill in than the field length, "0" is added to the left by the required number of digits in accordance with Rule 3.

(However, only the top row is required.) Next, the relationship between the entered data and the rule and the definition of the data group will be described in detail with reference to FIG. FIG. 3 shows the entry example of FIG. 2 in more detail. In FIG. 3, line 1, line 2, and line 3 are all blank lines, and according to the rule 1, the data are three data groups D ₁ , D ₂ ,
Divided into D ₃ . That is, the line is group D ₁ , ~
The row is the group D ₂ and the rows are the group D ₃ .
The line is a blank space.

According to Rule 2, the group D ₁ is composed of three data fields D ₁₁ , D ₁₂ , and D ₁₃ , and according to Rule 3, D ₁₁ and D ₁₂ are 4-digit right-justified data and D ₁₃ are 3-digit left-justified data. . Here, Δ in D ₁₁ and D ₁₂ indicates a blank. For example, in D ₁₁ , "△△ 13" and "1658" are two data of the same quality with four digits, "△△ 13" is the same as "13", and △ (blank) is added for explanation. I am doing it. D ₁₂ ,
The same applies to D ₁₃ . Similarly, the data groups D ₂ and D ₃ are also configured as shown in the figure.

Therefore, the data shown in FIG. 3 includes two lines of 4-digit right-justified data consisting of 2 fields and 3-digit left-justified data consisting of 1 field, and 3-digit right-justified data consisting of 1 field and 5 digits. 3 lines of data consisting of 1 field of right padding data and 1 line of 13 digits left padding data and 3 lines of 3 digit left padding data are written. .

Next, the processing flow will be described with reference to FIG. The processing of FIG. 4 generally shows the processing after all characters are read by OCR and temporarily stored, and the layout information extraction processing is performed by the layout information extraction unit 3 of FIG. After the left and right ends of each block are detected line by line from the stored contents, blank lines are also detected, and then a data group is detected.

Next, the format information extraction process of FIG. 4 is performed by the format information extraction unit 4 of FIG. 1, and the left and right ends of each field are detected for each data group, and left-justified and right-justified checks are performed. Further, the reading result output processing of FIG. 4 is performed by the reading result output processing unit 5 of FIG. 1, and only the necessary data is extracted and output according to the format information from the contents stored in the temporary storage device 2. . Details of the series of processes will be described below.

FIG. 5 is an explanatory diagram showing an example of a layout table created in the process of extracting the layout information of FIG. 4, and in the form of FIG. 3, only the characters excluding blanks in the fields 1, 2 and 3 of each line are written. The results of detecting the number of columns at the left and right ends of the block are shown over the entire page. For example, in the information "5-6" of the block 1 in the row in FIG. 5, "5" indicates the leftmost column and "6" indicates the rightmost column. That is, the left end of the block "13" excluding blanks, that is, "1" in the field of "△△ 13" in D _{11 of} Fig. 3 is in the column 5, and this indicates the left end of this block, and the right end "3" is the column. 6 and this is the right edge of this block. The same applies to the following, but here, a block means a part of each field excluding blanks. Further, in the layout table, Δ indicates blanks in all columns, that is, blank lines.

The table shown in FIG. 5 is extracted by the left and right edge detection processing of each block shown in the flowchart of FIG. 7, and the read during each processing means that the contents of the temporary storage processing 2 are read out. , Column indicates each digit. A set is a line number for creating the table shown in FIG.
The block number is registered, and the memory is written in the table.

The process shown in FIG. 8 follows the process shown in FIG. 7, and the process shown in FIG. 8 is performed using the layout table shown in FIG. 5 created by the process shown in FIG. . In the process of FIG. 8, the information of each row of the layout table is read and checked, the registration of each data group and the target row number belonging to each group are extracted while using the blank row as the separation information, and as shown in FIG. The left T ₁ part of a simple format table. That is, blank line detection and data group detection are performed here.

FIG. 9 is a flow chart showing the details of the format information extraction processing shown in FIG. 4, which is performed subsequent to the processing shown in FIG. 8 and is on the right side T _{2 of the} format table as an example shown in FIG. Is to create. In Fig. 6
Regarding the T ₂ portion, fields 1, 2, and 3 are included in each data group in blocks 1 and 2 of the layout table in FIG.
It is the result extracted from the left end minimum value and the right end maximum value of 2 and 3, and consists of the left and right end values (left-right), left-justified, and right-justified discrimination results (left / right). For example, the field 1 of the table group D ₁ has left and right end values “3-6” and “right-justified” in the row. This is because the data "5-" of the block 1 for the row in the layout table of FIG.
6 "and" 3-6 "to the left end minimum value" 3 ", the right end maximum value" 6 ", and the right end having the same right, that is, the right end has the same value, the processing shown in FIG. Have been extracted by. In addition, "Entry error" in Fig. 9
Means that a body that violates the above-mentioned rules 1 to 3 is detected, and it is better for practical use that the operator can be notified by a display or the like.

After the processing of FIG. 9, the reading result output processing of FIG. 4 is performed. In this process, the contents of the temporary storage device 2 are extracted and output using the target line number of the format table of FIG. 6 and the format information of each field, especially the column numbers indicated by the left and right end values. That is, in the form shown in FIG. 3, for example, in the row of the data group D ₁ , columns 3 to 6 are used as field 1 information having 4 digits and columns 10 to 13 are used as field 2 information having 4 digits. , Columns 18 to 20 are fetched and output as information of field 3 consisting of 3 digits. Below data group D ₂ , D
_By doing the same for _3, it is possible to output only the necessary data intended by the writer per page.

In this case, for example, the data in the field 1 on the line is "ΔΔ13", but this is to automatically output a blank in order to align the digits with those on the second line, and the same applies to the others.

In the above description, the tables of information extracted in the process of various processes are expressed as an example according to the form example, but as can be seen from the flowcharts of each process, the process itself is a general flow. Obviously, it is possible to process other than those shown in the form example.

Further, as described above, the entry rule may not be the one described in the embodiment as long as it can be algorithmized as a process, but it is desirable to preserve the naturalness of data entry.

<Effects of the Invention> Since the character reading method of the present invention does not need to individually design the form according to the application, it is extremely useful when the OCR is simply used.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of a character reading system according to the present invention, FIG. 2 is a diagram showing an example of a general-purpose form and character entry, and FIG.
FIG. 4 is a schematic diagram showing the primary storage contents of the read result of what is shown in FIG. 2, FIG. 4 is a diagram showing the outline of the processing flow in the present invention, and FIG. 5 is a table showing an example of the extracted layout information. 6 and 6 are tables showing examples of detected format information, FIG. 7 is a flowchart of a process of extracting layout information, and FIGS. 8 and 9 are flowcharts of a format information detection process. Indicate. Reference numeral 1 is an OCR, 2 is a temporary storage device, 3 is a layout information extraction unit, 4 is a format information extraction unit, and 5 is a read result output unit.

Claims (1)

[Claims] 1. A writing rule that prescribes only a relative arrangement of characters to be written is set in advance, and all the characters obtained by reading characters including a blank written in a general-purpose form according to the writing rule are obtained. Storage means for temporarily storing information in a character array conforming to the format of the general-purpose form, and entry layout information indicating a character array on the general-purpose form based on the entry rule from all character information read from the storage means. And a layout information extraction unit for outputting the extracted layout data extracted from the character data field belonging to the same group entered on the general-purpose form and the character array included in the entry layout information input from the layout information extraction unit. Format information indicating how to fill in the data fields is extracted and output based on the above-mentioned writing rules. Format information extraction means, and read result output means for reading out all character information stored in the storage means and outputting only required data based on the format information input from the format information extraction means. Character reader.