JP3294995B2 - Form reader - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a form reading apparatus which forms a form based on an unfilled form image and reads the filled form.

[0002]

2. Description of the Related Art In order to read characters written on a form using a character recognition device, it is necessary to create form format data in which information necessary for reading characters, such as a reading area in the form and reading conditions, is described. is there. In order to create form format information in this type of conventional character recognition device, it is necessary to read an unfilled form with an image scanner, automatically extract the structure of the form from the read image, and make manual corrections was there.

FIG. 40 is a processing flow of a conventional form format creating apparatus disclosed in, for example, Japanese Patent Application Laid-Open No. 4-343190. In FIG. 40, 401 is an image input unit, 402 is an image storage unit, 403 is a frame / character recognition unit, 405 is a character data area recognition unit, 407 is a field processing procedure creation unit, and 409 is a character data recognition parameter creation unit. , 207 are format information, 20
8 is field information, 414 is a knowledge database, 41
5 is a character data recognition parameter, 404 is a correction process (1), 406 is a correction process (2), and 408 is a correction process (3). Reference numeral 602 denotes an example of field information.

Next, a processing flow of the above-described conventional form format creating apparatus will be described. First, the image input unit 401 includes an image scanner, a facsimile, and the like.
Read the blank form image. Next, image storage means 4
In 02, the read form image is stored in the buffer. Next, the frame / character recognizing unit 403 extracts a ruled line on the form image, recognizes a preprinted character (character printed in advance) on the form image, and then extracts the preprinted character recognized as the extracted ruled line. The information is stored in the format information 207. Next, in the correction means (1) 404,
The result of recognition of ruled lines and preprinted characters is displayed on the screen, and if there is any correction, the operator makes correction using a mouse or keyboard. Next, character data area recognition means 4
05 is the format information 207 and the knowledge database 4
14, the area enclosed by the black frame, the month,
A blank area near a keyword indicating entry of a character such as "date" is extracted as a field, and the field information is stored in the field information 208. Here, the field information is
As shown in the field information example 602, the starting point position and width / height of each field, the number of characters to be entered, the name of the field specified by a keyword such as “month / day”, the type of characters to be entered (numerical characters, alphabetic characters, etc.) Is attribute information that indicates next,
The correction means (2) 406 displays the field information on the screen, and if there is a correction, the operator makes the correction. Next, the field processing procedure creation means 407 creates a processing procedure with reference to the field information 208 and the knowledge database 414. Here, the processing procedure indicates the presence / absence of knowledge processing for each recognition result, and the contents of the arithmetic processing for the recognition result when the recognition result is a numerical value. Next, the correcting means (3) 408 displays the created field processing procedure on the screen, and if there is a correction, the operator makes the correction.
Finally, the character data recognition parameter creating unit 409 creates information necessary for recognizing the character data based on the field information 208, and generates the character data recognition parameter 415.
To be stored.

[0005] In the above-mentioned conventional form format creation apparatus, format information 207, field information 208 and character data recognition parameter 15 are created, and these data are put together to create form format data. However, the content of the information stored in the form format by the conventional form format creation apparatus is not limited to the information created by the form format creation apparatus described above, and the position and size information of the preprint character is added. Example in which the preprinted characters in the field are erased and only the entered characters can be recognized (Japanese Patent Laid-Open No. 7-13996)
Are also disclosed. Further, there is disclosed an example in which the presence or absence of a mark can be determined by adding feature data of a character pattern string to be marked (Japanese Patent Laid-Open No. Hei 6-266883). In general, when a plurality of types of forms are read by a character recognition device, IDs are stored in fixed positions on the forms to be read.
The number is printed, and the ID number is described in a form format.

[0006]

Since the conventional form format creation apparatus is configured as described above, keywords such as "month" and "day" indicating the entry of characters are not printed on the form. Extraction of a field for writing a mark without a blank area for writing a character in the vicinity has to be performed manually, and there is a problem that the user has to interactively specify the position of the field.

It is also possible to automatically determine a field type such as whether the extracted field is a field in which a character is to be entered, a field in which a figure is to be entered, or a field in which a mark is to be entered. However, there is a problem that it takes time and effort for the user to interactively specify the field type.

In addition, when reading a plurality of types of forms with a character recognition device, there is a problem that the user has to print an ID number on the form and write the ID number in a form format.

Further, when the line width of the preprint is large and the written characters come into contact with the preprint, the preprint cannot be removed or the written characters are damaged.

Further, it is possible to determine where the mark is written, but there is a problem that the shape of the written mark cannot be determined.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and reduces the information that must be specified by a user when creating a form format.
A form format can be created more automatically, and when a completed form is read using the created form format, characters in the preprint are correctly erased,
It is another object of the present invention to provide a character recognition device capable of determining the shape of a marked mark.

[0012]

A form reading apparatus according to a first aspect of the present invention comprises: a form image storing means for storing an entire form image when creating a form format; a ruled line extracting means for detecting a straight line portion from the stored form image; A preprint character string extracting means for extracting a preprint character string from a form image; a preprint character string recognizing means for recognizing characters in the extracted preprint character string image and converting the characters into a preprint character code string; Keyword matching means for matching the result of character string recognition with the keyboard dictionary, an area surrounded by ruled lines and an area including a preprinted character string serving as a keyword representing an option, and a preprinted character serving as a keyword representing a figure or image area A field candidate extracting means for extracting an area near a column as a candidate for an entry area; Among, or fields near the preprinted text character recognition result field type based on the
Column field, checkmark field, or
A field type determining means for automatically determining whether the field is a message field, a field attribute setting means for estimating a type of a character to be entered and setting it as a field attribute in the case of a field where a character string is entered, and a preprint character string Form title candidate extracting means for extracting form title candidates instead of form identification ID numbers by comparing the result of recognition with the title condition storage table, and a field for selecting an item from the options and writing a mark (check mark field Output format setting means for setting the output format so that not only the mark but also the preprinted character code string of the marked option can be output when the recognition result is output, and an image of the preprinted character string to be stored in the form format To extract and create preprinted character string shape information from And a string shape information creation means, the form data
The form title information obtained by the
Ruled line information obtained by line extraction means and preprinted character strings
The preprint character string information obtained by the extraction
Out of the field candidates obtained by the field candidate extraction means.
Ruled line information, embedded preprint character string information,
Field type obtained by the field type
Field attribute obtained by the field attribute setting means
Preprint obtained by the lint character string shape information creation means
Character string shape information and output obtained by output format setting means
Document format to output the format as a document format
Output means .

[0013] In the form reading apparatus according to a second aspect, the field type determining means determines whether or not the keyword type of the preprint character string included in the field represents an option. The field type is output as a check mark field.

According to a third aspect of the present invention, in the form reading apparatus, the field type determining means determines whether the keyword type of the preprint character string included in the field indicates an image area, and determines that the keyword indicates an image area. Is characterized in that the field type is output as an image field.

According to a fourth aspect of the present invention, there is provided a form reading apparatus, wherein the output format setting means writes a character string as a check mark field selection item in a check mark field for writing a mark in an option item. It is determined whether or not to include the field, and if it is determined to include, the output format of outputting a preprinted character string code representing an option and an image of a character string written in the character string field is determined. I do.

According to a fifth aspect of the present invention, there is provided a form reading apparatus, wherein the output format setting means writes a character string as a check mark field selection item in a check mark field for writing a mark in an option item. Determines whether to include the field, and if so, determines the output format that outputs the preprinted character string code representing the option and the recognition result of the character string entered in the character string field. And

According to a sixth aspect of the present invention, when the output format setting means determines that the check mark field is constituted by a continuous straight line and the entry mark can be entered at any position in the field, The output format in which the ratio of the length to the mark position with respect to the length of the mark is calculated and output is determined.

According to a seventh aspect of the present invention, the form title extracting means assigns an evaluation value to the form title condition stored in the form title condition storage table in advance, and evaluates the evaluation value for the preprint character string in the form. Is calculated, and the preprint character string having the largest total score of the evaluation value is output as the form title.

In the form reading device according to the present invention, the preprint character string shape information creating means outputs a directional component in a small area of the preprint character string as the preprint character string shape information stored in the form format. It is characterized by.

[0020]

[0021]

[0022]

BEST MODE FOR CARRYING OUT THE INVENTION

Embodiment 1 FIG. Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of a form reading apparatus according to Embodiment 1 of the present invention. In the figure, 1 is a form image storage means for storing an image of the entire form, 2
Is a ruled line extracting means for detecting a straight line portion from the entire form image; 3, a preprinted character string extracting means for extracting a preprinted character string portion from the whole form image; The print character string recognition means 5 is a character entry area in the form contents.
A keyword dictionary for storing a character string that can be a keyword representing an option / image area, 6 is a keyword matching unit that matches a preprinted character string after recognition with the keyword dictionary, and 7 is a field for extracting a field to be an entry area Candidate extracting means, 8 is a field type determining means for determining what is a field to be entered, 9 is a field attribute setting means for assigning a field attribute such as a type of a character to be entered in the field, 10 is an entry character string Output format setting means for setting the output format of the recognition result
Is a title condition storage table storing conditions for the preprint character string to be a form title, 12 is a means for comparing the recognized preprint character string with the title condition storage table to extract a form title candidate, 13 is Means for creating preprint character string shape information to be stored in the form format from the preprint character string image, and means 14 for outputting the form format.

Next, the operation when creating a form format is shown in FIG.
It is explained along. First, the form image storage unit 1 stores an unfilled form image input from an image scanner or a facsimile. FIG. 2 shows an example of the stored blank form image.

Next, the ruled line extracting means 2 analyzes the form image stored by the form image storing means 1 and extracts a ruled line portion. The ruled line is extracted, for example, as disclosed in
This can be realized by the conventional technique disclosed in Japanese Patent No. 3190.

FIG. 3 shows an example of ruled lines extracted from the blank form image of FIG. The ruled line extracting means 2 includes ruled lines 1201, 1202, 1203, 1204, and 120 shown in FIG.
5, 1206, the point 1207 is the start point of the ruled line 1201, the point 1208 is the end point of the ruled line 1201, the point 1209 is the start point of the ruled line 1202 and the intersection with the ruled line 1204, the point 1210 is the intersection of the ruled line 1202 and the ruled line 1205, The point 1211 is determined to be the end point of the ruled line 1202 and also the intersection of the ruled line 1206. Further, the ruled line extracting means 2 determines the ruled lines 1201, 1202, 1203, 120
After extracting 4, 1205 and 1206, position coordinates of a point 1207 as a start point of the ruled line 1201 and a point 1208 as an end point are calculated and output as ruled line information. Since the ruled line 1201 has no intersection with another ruled line, the position coordinates of the intersection are not set. Regarding the ruled line 1202, the point 120 is used as a starting point.
9. The position coordinates of the point 1211 as an end point and the points 1209, 1210, and 1211 as intersections with other ruled lines are calculated and output as ruled line information.

FIG. 4 shows an example of ruled line information extracted by the ruled line extracting means 2. In FIG. 4, 1225 is a ruled line number corresponding to the ruled line 1201 in FIG. 3, and 1226 is a ruled line number corresponding to the ruled line 1202 in FIG. In FIG. 4, reference numeral 1220 denotes the coordinates of point 1207 in FIG. 3, 1221 denotes the coordinates of point 1208 in FIG. 3, 1222 denotes the coordinates of point 1209 in FIG. 3, 1223 denotes the coordinates of point 1210 in FIG. 3, and 1224 denotes the point 1211 in FIG. It is an example of the coordinates of. Reference numerals 1220 and 1221 denote coordinate information of the start and end points of the ruled line 1201 in FIG.
Reference numerals 1223 and 1224 denote coordinate information of the start and end points and the intersection of the ruled line 1202 in FIG.

Next, the preprint character string extracting means 3 analyzes the form image stored by the form image storing means 1 and extracts and outputs the image of the preprint character string. This can be realized by a conventional technique disclosed in, for example, Japanese Patent Application Laid-Open No. 4-343190. Further, the preprint character string extracting means 3 calculates the start position coordinates, width, and height of the extracted preprint character string image and outputs it as preprint character string information. FIG. 5 shows an example of a preprinted character string extracted from the blank form image of FIG. In FIG. 5, reference numeral 1301 denotes a preprint character string “sample form”; 1302, a starting point of the preprint character string 1301; 1303, a preprint character string 1
The width 301 and 1304 are the height of the preprinted character string 1301.

FIG. 6 shows an example of a table storing preprint character string information. In FIG. 6, 1401, 1408, and 1
409, 1410, 1411 are the preprinted character strings 1301, 1305, 1306, 1307-1 shown in FIG.
This is the preprint character string number corresponding to 308. FIG.
Preprinted character string "sample form" 1301 shown in
Are the start position 1402 for the start point 1302 and the information 14 for the width 1303.
03, information 1404 for the height 1304.

Next, the preprint character string recognizing means 4 recognizes the preprint character string image output by the preprint character string extracting means 3 and converts it into a character code. The preprint character string image shown in FIG. Output as character string information. This can be achieved, for example, by conventional printed character reading techniques.

When the preprinted character string recognizing means 4 configured as described above receives the image of the preprinted character string "sample form" 1301 in FIG. 5, as a recognition result, a character code 1405 of "sample form" is obtained.
The font information 1406 “Gothic” and the character modification information 1407 “Bold” are detected and added to the preprinted character string information of the preprinted character string 1401 shown in FIG.

The keyword dictionary 5 is a character string that generally prompts entry on a form, for example, “year”, “month”, “day”, or a character string representing an option, for example, “yes”, “no”, “taisho”, “showa” "", "Heisei", and a table in which consecutive numbers and alphabets are stored in advance. FIG. 7 shows an example of the contents of the keyword dictionary.

The keyword collating means 6 searches the keyword dictionary for a character code of the recognition result of the preprinted character string shown in FIG. 6, determines whether or not it exists in the keyword dictionary, A flag in the preprint character string information. For example, the preprinted character strings 1305, 1306, 1 shown in FIG.
307, 1308 recognition results 1412, 1413, 14
14 and 1415 are 150 of the keyword dictionary shown in FIG.
1, 1502, 1503, and 1504, the keyword flags 1416, 1417, 1418, and 1419 of the preprint character string information shown in FIG. 6 are turned on.

Further, when the keyword collating means 6 refers to the keyword dictionary 5, the "keyword indicating the character entry area" 1505 and the "keyword indicating the image area" 1
506, Detect and output which item of “keyword indicating option” 1507 is applicable. For example, the recognition result “month” 1412 of the preprinted character string “month” 1305 shown in FIG. 6 indicates that the keyword “month” 1501 in FIG.
, The keyword type of the preprint character string information is “character string” 1420 as shown in FIG.
And

The field candidate extracting means 7 extracts a candidate for an area (field) in which a writer enters characters and other information. The field candidate extracting means can be realized by a conventional technique such as a method disclosed in Japanese Patent Application Laid-Open No. 4-343190. That is, first, an area surrounded by the ruled line extracted by the ruled line extracting means 2 is set as a field candidate, and the ruled lines 1202, 1203, 1204, and 1205 shown in FIG.
After extracting the portion surrounded by the squares as a field 1702, the ruled lines 1202, 1203, 1204, 12
05 and output as field information. FIG. 9 shows an example of the field information. In FIG. 9, 1720 and 1721
1729 is the field 1701 1702 in FIG.
A field number corresponding to 1703. In FIG. 9, reference numeral 1722 denotes information indicating the outer peripheral ruled line of the field 1702 in FIG. 8 by the ruled line number.

Next, referring to the preprint character string information shown in FIG. 6, an area including the preprint character strings “month” 1305 and “day” 1306 with the keyword flag turned on is extracted. At this time, the preprint character string "month"
1305, the y-coordinate of the start position 1421 and the y-coordinate of 1422, and the heights 1424 and 1425, which are the pre-print character string information of "day" 1306, are the same, so that the pre-print characters 1305 and 1306 are in the same field. Is determined to be included. The x-coordinate of the field containing the preprinted character strings 1305 and 1306 is, for example, the preprinted character string 13 on the left side of the preprinted character string 1305.
Based on a predetermined condition that characters are to be entered with a width 1423 of 05, the x coordinate of the start position 1421 of the preprint character string “month” 1305 minus the width 1423 is set. The width of the field is the end point of the preprinted character string 1306 (the x coordinate of the start position 1422 of the preprinted character string “day” 1306 + the width 1427 of the preprinted character string 1306) −
The above-mentioned field x coordinate is used. As a result, FIG.
The area indicated by reference numeral 701 is set as a field candidate, and is output as field information together with the included preprint character strings "month" 1305 and "day" 1306. In FIG. 9, 1723
Indicates a preprint character string included in the field 1701 by a preprint character string number.

The field type judging means 8 estimates the contents to be written in the fields extracted by the field candidate extracting means 7, determines the field type, and sets the field type in the field information shown in FIG. For example,
A field 1701 illustrated in FIG. 8 includes a preprint character string serving as a keyword, “month” 1305 and “day” 1306.
Contains. On the other hand, the preprint character strings 1305, 1
Referring to the preprint character string information of FIG. 6 corresponding to 306, the keyword types of the recognition results “month” 1412 and “day” 1413 are “character string” as shown in 1420 and 1426.
Therefore, it is determined that the field 1701 in FIG. 8 is a character string field, and information 1724 of “character string” is output as the field type of the field of the field number 1720 in FIG. 9 corresponding to the field 1701.

The field 1702 in FIG. 8 is a field surrounded by ruled lines, and since there is no preprint character string serving as a keyword for determining the field type inside and adjacent to the field, the field type is defaulted to the field type. Is determined as a character string field, and information 1725 “character string” is output as the field type of the field of the field number 1721 in FIG. 9 corresponding to the field 1702.

If the field type is a character string field, the field attribute setting means 9 sets the character type to be entered in the field information. For example, regarding the field 1701 including the preprinted characters “month” 1305 and “day” 1306 shown in FIG.
7 is retrieved from the keyword dictionary of FIG. 7, it can be determined that the character type to be entered for the keyword “month” 1501 is “numeric” as shown by 1508 in FIG. It is determined that a numeral is to be entered in the preprinted character string 1306 in FIG. 8 by the same processing, and “numerical” 1726 is output as the field attribute of the field with the field number 1720 in FIG. The character type of the character field 1702 having no keyword shown in FIG. 8 is output as “all character types” 1727 as a field attribute of the corresponding field number 1721 in FIG.

The output format setting means 10 outputs a recognition result of the entered check mark and a keyword representing the checked item to a check mark field for entering a check mark in one of the options. Is set to the field information. For example, FIG.
In the field 1703 shown in FIG.
When a check mark is recognized at 04, the character string codes of “Yes” and “No” are recognized as the recognition results of 1704 and 1705 so that a character string code of “Yes” can be added to the recognition result and output. Is set to the output format 1728 of the field information shown in FIG.

The title condition storage table 11 is a table storing conditions under which a preprinted character string can be a title of a form. In this table, as shown in FIG. 10, “located at the top of the form image” 1801, “located at the top center of the form” 1802, “character size is the largest among the character sizes of the entire form” 1803, “The character thickness is the thickest among the character thicknesses of the entire form.” 1804
"The font is different from the character font of the entire form" 180
5, a character string of a condition for a form title such as “highlighted by underline” 1806 and “enclosed by ruled line” 1807 and a logical expression representing the condition are stored.

The form title candidate extracting means 12 refers to the information stored in the form title condition storage table 11 and extracts a character string that is a form title candidate from the preprinted character strings after character recognition. Of the preprinted character strings after character recognition shown in FIG.
Reference numeral 1405 denotes 1801, 1802, 18033.1 in FIG.
Since this applies to 804, 1805, and 1806, the preprint character string “sample form” 1405 in FIG. 6 is selected as the form title, and the start position 1402 and the width 14 are determined as form title information by referring to the preprint character string information.
03, height 1404, and recognition result character code 1405 are output.

Preprint character string shape information creating means 13
Creates the preprint character string shape information in order to store the preprint character string itself in a form format. The shape information of the preprinted character string is used to remove the image of the preprinted character string from the form image when reading the filled out form. In the first embodiment, this preprint character string information is image data in a rectangle circumscribing the preprint character string. For example, in FIG. 11, the preprint character strings 1704 and 1705 included in the field 1703 are extracted as image data within a circumscribed rectangle, and the image data is used as the shape information of the preprint character string in the field 1703 corresponding to the field 1703. 9 field number 1729
Preprinted character string shape information 1
730 and 1731 are set.

The form format output means 14 includes the form title information obtained by the form title candidate extraction means 12, the ruled line information obtained by the ruled line extraction means 2, and the preprinted character string obtained by the preprint character string extraction means 3. The character string information, the information on the outer peripheral ruled line for the field candidate obtained by the field candidate extracting means 7 and the included preprint character string information, and the field type obtained by the field type determining means 8 and the field attribute obtained by the field attribute setting means 9 The field attributes, the preprinted character string shape information obtained by the preprinted character string shape information creating means 13, and the output format obtained by the output format setting means 10 are output to a disk or memory as a form format.

FIG. 12 shows an example of a form format table to be output. In FIG. 12, reference numeral 1911 denotes a form title 19
10, upper left position, 1912 is the same width, 1913 is the same height,
Reference numeral 1914 denotes the same character string code. [Rule 1] 1915
The information on the ruled line corresponding to the start point 1916, the end point 1917,
[Rule 2] The information of the rule corresponding to 1918 is the starting point 191.
9, End point 1921, Intersection points 1919, 1920, 1921
It is. The information of the field corresponding to [Field 1] 1922 is a field type 1923, a field attribute 1924, preprint character string information 1925, a position 1926, a width 1927, a height 1928, and a shape information 1929. Further, a display means on a display may be provided to receive the correction of the form format by the operator.

According to the first embodiment, from the blank form image, the setting of the field type, the output format of the recognition result of the check mark field, and the setting of the form title corresponding to the ID number for form identification are automated. And a form reading format can be efficiently created.

Embodiment 2 Hereinafter, Embodiment 2 of the present invention (the operation of the field type determining means as a characteristic component) will be described with reference to FIG. 1 and FIGS. FIG.
FIG. 14 is an example of a check mark field, FIG. 14 is an example of a preprint character string information table corresponding to the preprint character string in the check mark field shown in FIG.
5 is an example of field information of the check mark field shown in FIG. The operation other than the field type determining means is the same as that of the first embodiment, and the description is omitted. Field type determination means 8 in the second embodiment
The preprint character string number 22 in the preprint character string information table shown in FIG. 14 corresponding to the preprint character strings 1704 and 1705 in the area 1703 in FIG.
Reference is made to the preprinted character string information of 01.2202 (the preprinted character string information in FIG. 14 is information created by the preprinted character string extracting means 3, the preprinted character string recognizing means 4, and the keyword matching means 6). .

Next, the field type judging means 8 shown in FIG.
For example, the recognition results 2203 and 2204 corresponding to the preprinted character strings 1704 and 1705 shown in FIG. In the example of FIG. 13, the preprint character strings shown in 1509 and 1510 in FIG. 7 are described in the item 1507 of the keyword representing the option, so that the field type determination unit 8 checks the field 1703 in FIG. It is determined that the field is a mark field, and the field type of the field with the field number 2301 shown in FIG. 15 corresponding to the field 1703 in FIG. 13 is output as “mark” 2302.

According to the second embodiment, even if the extracted field is a check mark field, the field type can be determined.

Embodiment 3 Hereinafter, Embodiment 3 of the present invention (the operation of the field type determining means as a characteristic component) will be described with reference to FIG. 1 and FIGS. FIG.
17 shows an example of the image field, FIG. 17 shows an example of preprint character string information corresponding to the preprint character string in the image field shown in FIG. 16, and FIG. 18 shows an example of field information of the image field shown in FIG. The operation other than the field type determining means is the same as that of the first embodiment, and the description is omitted. The field type determining means 8 according to the third embodiment includes a preprint character string 3102 in a field 3101 extracted by the field candidate extracting means 7 as an area surrounded by ruled lines as shown in FIG.
Reference is made to the preprinted character string information of the preprinted character string number 3201 in the preprinted character string information table shown in FIG. 17 (the preprinted character string information in FIG. This is information created by the preprint character string recognition means 4 and the keyword collation means 6).

Next, after obtaining the recognition result 3202 corresponding to the preprinted character string 3102, the field type determining means 8 searches whether or not the recognition result character string 3202 exists in the keyword dictionary shown in FIG. . In the example of FIG. 16, the preprint character string “drawing” shown at 1511 in FIG. 7 is described in the keyword item 1506 representing the image area. It is determined that the field is an image field, and the field type of the field with the field number 3301 shown in FIG.

According to the third embodiment, the field type can be determined even if the extracted field is an image field.

Embodiment 4 FIG. Hereinafter, Embodiment 4 (output format setting means as a characteristic component) of the present invention will be described with reference to FIG. 1 and FIGS. 19 shows an example of a check mark field including a field for entering a character string as a selection item, FIG. 20 shows an example in which a character string is entered in the character string entry field shown in FIG. 19, and FIG. 21 shows a check mark shown in FIG. It is an example of field information of a field and a character string field. The operation other than the output format setting means is the same as that of the first embodiment, and the description is omitted. Field candidate extracting means 7 in the fourth embodiment
And the field type determining means 8 performs a preprint character string 4102, 4104, 4106, 41 as shown in FIG.
After detecting a check mark field 4101 for writing marks in four options 08, preprint character strings 4103, 4105, 4107, 410 representing each option
9 and the preprinted characters "(" 4111 and ")" 41
It is determined whether or not a character string field 4110 sandwiched between the two exists. If it is determined that a character string field exists, the field candidate extracting unit 7 and the field type determining unit 8 extract the field 4110 included in the check mark field 4101 by the method described in the operation of the first embodiment.

Next, the output format setting means 10 checks the field 41 included in the check mark field 4101.
Preposition character string 41 indicating the position of 10 and each option
03, 4105, 4107, and 4109 are collated, and a preprint character corresponding to a field 4110 included in the check mark field 4101 is determined. In the example of FIG. 19, the pre-printed character string 4109 is associated with the coordinate position in the vertical direction.

Next, as the output format in the field information of the field with the field number 4301 shown in FIG. 21 corresponding to the field 4101 shown in FIG. 19, "recognition result of preprint character strings 4103, 4105, 4107 and recognition of 4109" Of the result obtained by adding the character string image 4201 of the field 4110 to the result, the preprinted character strings 4102 and 4104 that recognize the mark
"Output corresponding to 4106 and 4108" is set. At the time of form recognition, when an entry as shown in FIG. 20 is made, a character code of “others” and an image 4112 entered in the field 4110 are output as a recognition result.

According to the fourth embodiment, when a recognition result of a check mark field having a character entry field as an option is output, a preprinted character string representing the option and an image of an entry character are added to the recognition result of the mark. Items other than options can be smoothly input.

Embodiment 5 An embodiment of the present invention (output format setting means as a characteristic component) will be described below with reference to FIG.
This will be described with reference to FIGS. 19, 22 and 23. 19 shows an example of a check mark field including a field for entering a character string as a selection item, FIG. 22 shows an example of a character string entered in a character string entry field in the check mark field, and FIG. 23 shows a check shown in FIG. It is an example of field information of a mark field and a character string field.
The operation other than the output format setting means is the same as that of the first embodiment, and the description is omitted. In the fifth embodiment, the field candidate extracting means 7 and the field type determining means 8
Is a preprinted character string 4102 as shown in FIG.
After detecting a check mark field 4101 for writing marks in the four options 4104, 4106, and 4108, a preprinted character string 4103,
It is determined whether there is a character string field 4110 sandwiched between preprinted characters “(” 4111 and “)” 4112 in addition to 105, 4107, and 4109.
If it is determined that a character string field exists, the field candidate extracting unit 7 and the field type determining unit 8 extract the field 4110 included in the check mark field 4101 by the method described in the operation of the first embodiment.

Next, the output format setting means 10 outputs a field 411 included in the check mark field 4101.
Pre-printed character string 410 representing the position of 0 and each option
3, 4105, 4107, and 4109,
A field 4110 included in the check mark field 4101 determines a corresponding preprint character.
In the example of FIG. 19, the preprinted character string 4109 having the same coordinate position in the vertical direction is associated.

Next, the field 4101 shown in FIG.
As the output format in the field information of the field with the field number 4301 in FIG. 23 corresponding to “preprint character strings 4102, 4104, and 4106, the preprint character strings 4103, 4103,
105, the recognition result of 4107 is output, and when the preprint character string 4108 is marked, the recognition result of the preprint character string 4109 and the recognition result of the character string entered in the field 4110 are output. 5202 is described. If there is an entry as shown in FIG. 22 at the time of form recognition, a character code of “Others” and a character code string “when cold” 5102 of the recognition result of the input character string image 5101 are output as the recognition result.

According to the fifth embodiment, at the time of outputting the recognition result of the check mark field in which the character entry field is included in the option, the preprint character string representing the option and the recognition result of the input character are added to the recognition result of the mark. , Items other than options can be smoothly input.

Embodiment 6 FIG. Hereinafter, a sixth embodiment (output format setting means as a characteristic component) of the present invention will be described with reference to FIG. 1 and FIGS. 24 shows an example of a check mark field composed of horizontal lines, FIG. 25 shows an example in which a mark is written in the check mark field shown in FIG. 24, and FIG. 26 shows an example of field information of the check mark field shown in FIG. . The operation other than the output format setting means is the same as that of the first embodiment, and the description is omitted. In the sixth embodiment, when the field candidate extracting unit 7 and the field type determining unit 8 detect a field where a plurality of short vertical lines intersect a continuous horizontal line as shown in FIG. It is determined that the check mark field 6101 is a check mark field 6101 which is formed of a straight line and in which a mark can be written at an arbitrary position on the straight line.

Next, the output format setting means 10 sets a field left end position 6202 shown in FIG. 25 to a check mark field 6101 which is formed of a continuous straight line and in which a mark can be written at an arbitrary position on the straight line. The right end position 6203 is detected, and the output format of the field is (mark center position−field left end position) /
An instruction 6302 of (field right end position-field left end position) is set.

At the time of form recognition, a circle mark 6201 in FIG.
If there is an entry as shown in the above, the field left end position 6
Distance 620 between 202 and field right end position 6203
5 is measured. Next, by erasing the straight line using a known technique, the mark 6201 is extracted to obtain the center coordinates, and the distance 6206 between the field left end position 6202 and the mark 6201 is measured. Then, the instruction 6 in FIG.
According to 302, the ratio of the mark position to the entire length is output as a recognition result.

According to the sixth embodiment, since a recognition result for a mark entered at an arbitrary position in a check mark field constituted by a continuous straight line can be output, a variety of check mark fields can be easily entered by a writer. can do.

Embodiment 7 FIG. Hereinafter, a seventh embodiment of the present invention (a form title candidate extracting unit as a characteristic component) will be described with reference to FIGS. 1, 6, 27, and 28.
FIG. 27 is an example of a form title condition storage table in which conditions that are form title candidates and evaluation values when the respective conditions are satisfied are previously stored. FIG. 28 is a diagram illustrating a preprinted character included in the form shown in FIG. It is an example showing an evaluation value as a form title for a column. The operation other than the form title candidate extracting means is the same as that of the first embodiment, and the description is omitted.

In the seventh embodiment, the form title candidate extraction means 12 stores, for example, evaluation values determined in advance for each item for all preprinted character strings after character recognition shown in FIG. By referring to a form title condition storage table as shown in FIG. 27, it is determined whether or not the preprint character string can be a form title. That is, it is checked whether or not the condition stored in the form title condition storage table is satisfied. For the items that satisfy the condition, the evaluation values described in the form title condition storage table are added, and the sum of the evaluation values is calculated. The largest preprint character string is determined as the form title.
FIG. 28 shows an example of the evaluation value of the preprinted character string in FIG. The preprint character string “sample form” 7201 is the condition 7101, 7103, 7105, 7106 in FIG.
7107 and 7108, the evaluation value “5”
7202, “5” 7203, “5” 7204, “5” 7
205, “3” 7206, and “5” 7207 are added to obtain a total value “28” 7208. Since “28” 7208 is the largest when compared with the total value of the other preprinted character strings, “sample form” 7201 is determined as the form title, and the processing is started with reference to the preprinted character string information shown in FIG. The position 1402, width 1403, height 1404, and recognition result character code 1405 are output as form title information.

According to the seventh embodiment, the form I
The form title as D can be detected.

Embodiment 8 FIG. Embodiment 8 of the present invention (preprinted character string shape information creating means as a characteristic component) will be described below with reference to FIGS. FIG.
Reference numeral 101 denotes a preprinted character string area image input to the preprinted character string shape information creating means, and 102 denotes an NxN
Small dot area 103 is preprinted character string shape information extracted from preprinted character string area image 101,
Reference numeral 4 denotes a small area having a vertical component, and 105 denotes a small area having a horizontal component. Operations other than the preprinted character string shape information creating means are the same as those in the first embodiment, and a description thereof will be omitted.

In the eighth embodiment, when the preprinted character string shape information creating means 13 receives a preprinted character string image 101 shown in FIG. 29, it divides the image into small areas 102 of N × N dot size. Next, when black pixels continue in the vertical direction in each divided small area, the small area is defined as a small area 104 having a vertical component. Similarly, when the black pixels are continuous in the horizontal direction, the small area 105 has a horizontal component. Here, each of the small areas has a “vertical component”, a “horizontal component”, and a “vertical + horizontal component”
With one of As described above, the result of obtaining the directional components for all the small areas on the preprinted character string area 101 is referred to as preprinted character string shape information 103.

According to the eighth embodiment, since the shape information of the preprinted character string is expressed by the directional component of each small area, the preprinted character string image is stored as it is as the shape information in the form format. Can also reduce the size of the form format.

Embodiment 9 Hereinafter, a ninth embodiment of the present invention will be described with reference to the drawings. FIG. 30 is a block diagram showing a process of reading a filled out form according to the ninth embodiment of the present invention. 30, 1 and 2 are the same as those in FIG. 30 in FIG. 30 is a form format group created in advance,
Reference numeral 17 denotes a form format selection unit that identifies the type of the input form according to the form title information described in the form format and selects one corresponding form format.

30 in FIG. 30 is a ruled line associating means for associating ruled lines extracted from the form image with ruled lines described in the form format, and 20 is a preprint character string on the form image and a pre-printed character string described in the form format. This is a preprint character string associating means for associating with a print character string.

30 is a field extracting means for extracting an entry field from a form image, 22 is a means for deleting a ruled line image on a form image by referring to ruled line information in a form format, and 23 is a preprinted character in a form format. This is a means for deleting the preprinted character string portion on the form image by referring to the column information.

Reference numeral 24 in FIG. 30 denotes an entry character string recognizing means for recognizing a character string image left on the character string entry field.
Reference numeral 5 denotes an entry mark recognition means for recognizing the mark left on the check mark field, and reference numeral 26 denotes a means for outputting the recognition result of the entry character string and the mark by referring to the output code information in the form format.

FIG. 31 shows a completed form image.
FIG. 32 illustrates the operation of the ruled line association means 19. In the figure, reference numeral 110 denotes a ruled line described in a form format, and 112 to 114 are part of the ruled lines described in the form format. Reference numeral 111 denotes a ruled line extracted from the filled-out form image.
Are part of the extracted ruled lines.

FIGS. 33 and 34 illustrate the operation of the preprinted character string associating means 20. Reference numeral 130 denotes a preprinted character string area on a completed form image.
1 is the schematic shape of the preprinted character string area, 133 is Nx
N dot small area, 140 is the shape information of the preprinted character string described in the form format, 132 is the schematic shape created from the shape information 140 of the preprinted character string. 134 to 139 and 177 are examples of small areas.

FIG. 35 illustrates the operation of the field extracting means 21, the ruled line erasing means 22, and the preprinted character string erasing means 23. Reference numerals 150, 151 and 154 denote field region images extracted by the field extracting means 21, 152 and 153 are preprint character string erasing means 2
As a result of removing the preprint by the step 3, the result 155 is the result of removing the ruled line by the ruled line erasing means 22, and the numeral 156 is a ruled line. FIG. 37 shows an example of the recognition result.

Next, the operation will be described. The operation will be described based on the block diagram of the process of reading the filled-out form shown in FIG. 30 and referring to a specific filled-out form example in FIG.
First, the form image storage unit 1 reads the completed form with an image scanner or a facsimile and acquires the form image shown in FIG. 31 in the same manner as shown in FIG.

Next, the form format selecting means 17 sequentially reads the form title information (position, width, height, character code) described in each form format from the form formats group 18 created in advance, and Recognize a character string on the form image existing near the position. At this time, when the character string recognition result matches the character code of the form title, the form format is assumed to be a form format corresponding to the input form image. In the subsequent processing steps, processing is performed using the form format. Here, if there is no report format in which the character string recognition result matches the character code of the report title, it is considered that the input report image is an unregistered report, and the entire process ends.

Next, given the form image shown in FIG. 31, the ruled line extracting means 2 extracts the ruled line group 11 as shown in FIG. 32 from the form image by the method shown in the first embodiment. Then, the start point / end point and intersection position coordinates of all the ruled lines are output.

Next, as shown in FIG. 32, the ruled line associating means 19 compares the ruled line 111 extracted by the ruled line extracting means 2 and the ruled line 110 described in the form format with the start point / end point of the ruled line and the intersection position coordinates. To the original. Thereby, for example, the ruled line 112 and the ruled line 115 are
And the ruled line 116, and the ruled line 114 and the ruled line 117 are associated with each other.
The correspondence between the ruled lines is described in, for example, Japanese Patent Application No. 06-07.
No.6795, “Method for processing a form including a table”, “Pattern recognition” (Noboru Funakubo, Kyoritsu Shuppan)
P. 75 can be realized by existing techniques such as the relaxation method.

Next, the preprint character string associating means 20
Associates a preprint character string described in a form format with a character string on a form image. In the association, first, the preprint character string position coordinates described in the form format are corrected based on the result of association between the ruled lines obtained by the ruled line associating means 19, so that preprinting on the form image is performed. Estimate the character string position. Then, an image near the estimated character string position is cut out.
Thus, for example, when associating the preprinted character string "month and day" described in the form format, the image 130 shown in FIG. 33 is cut out from the filled-out form image shown in FIG.

Next, the extracted image 130 is represented by N × N
The density of black pixels is obtained by dividing the dot into small areas 133 and inside each small area. Image 13 cut out by this
A general shape 131 of 0 is obtained. Here, the square inside the schematic shape 131 indicates the black pixel density in each small area, and the larger the square, the higher the black pixel density. Similarly, the black pixel density in each small area is obtained from the preprint character string shape information 140 (bitmap image of the preprint character string) stored in the form format, and the preprint character string described in the form format is obtained. Is obtained.

Thereafter, as shown in FIG. 34, each of the small regions constituting the schematic shape 131 of the preprinted character string created from the filled-out form image and the schematic shape 132 of the preprinted character string described in the form format are displayed. Each of the constituent small regions is optimally associated with each other in consideration of the black pixel density of the small region. Thereby, for example, the small area 134 and the small area 1
37, the small area 135 and the small area 138, and the small area 136 and the small area 139 are associated with each other. This association is performed for all fields including the preprint character string.

For example, the correspondence between such small regions is described in “Neural Network and Brain Theory” (MA Erbiff,
Science). MATC described in 382
It can be realized by the existing technology such as the H algorithm. As a result, even if there is a positional shift, distortion, or enlargement / reduction in the form image, the preprint character string position on the form image can be accurately specified.

Next, the field extracting means 21 first obtains the correspondence result between the ruled lines obtained by the ruled line associating means 19 and the result of association between the preprinted character strings obtained by the preprinted character string associating means 20. Then, the position (start point, width and height) of the entry field described in the form format is corrected, and the accurate entry field area position on the form image is specified. Thereby, for example, images 150, 151, 1 of the field area as shown in FIG.
Acquire 54.

Next, in the ruled line erasing means 22, based on the correspondence between the ruled lines obtained by the ruled line associating means 19, the image of the ruled line corresponding to the ruled line described in the form format is extracted by the field extracting means 21. Erase from the obtained field area image. As a result, an image 155 is obtained by removing the ruled line portion 156 from the image 154 of the field area including the ruled line.

Next, the preprint character string erasing means 23 deletes the preprint image from the field area image based on the correspondence between the character strings obtained by the preprint character string associating means 20. Here, the image area to be erased has a schematic shape 132 in association of small areas by the preprint character string association means 20 (FIG. 34).
The inside of the small area on the schematic shape 131 is associated with the upper small area. However, if any of the small areas located on the upper, lower, left, and right sides of the small area on the general shape 131 does not correspond to the small area on the general shape 132 as in the small area 177 on the general shape 131, the image inside the small area is not deleted.

As a result, even if a character entered as shown in image 150 in FIG. 35 contacts a preprinted character string having a large line width, the field area in which only the preprinted character string is erased without losing the entered character. An image 152 can be obtained. Similarly, as shown in an image 151 in FIG. 35, a field area image 153 in which only the preprint character string is erased can be obtained for the check mark field. The shape of the mark entered on the column can also be obtained.

Thereafter, the entered character string recognizing means 24 recognizes the entered character string with respect to the character entry field area images 152 and 155 from which the image of the ruled line and the preprinted character string has been deleted. Result image 15
The character code “114” is obtained from “2”, and “Yamada” is obtained from the image 155.

Next, the entry mark recognizing means 25 recognizes the entered mark with respect to the check field area image 153 from which the image of the ruled line and the preprinted character string has been deleted. As a result, the character code " Get "

Finally, the input character string / mark recognition result output means 26 outputs the character recognition result to the outside. Here, the mark recognition result obtained by the entry mark recognition means 25 is output according to the output format described in the form format 18. As a result, the character code obtained as the mark recognition result of the check field area image 153 "
Re "+" Yes ".

FIG. 36 shows a form title 180 and entry fields 181-1 extracted from the completed form shown in FIG.
89 is shown. The recognition results of these fields are shown in 190 to 199 in FIG. In FIG. 37, 184A corresponds to FIG.
6 is a field number corresponding to the field 184 in FIG.

As described above, according to the present embodiment, the correspondence between the ruled lines and the preprinted character strings are performed, and the field area position is corrected based on the result. Therefore, the field area can be correctly specified even when the form image is enlarged, reduced, distorted, or translated. In addition, the preprint character strings are associated with each other on a small area basis, and only the small area on the form image associated with the small area extracted from the preprint character string described in the form format is included in the small area. Delete the image of. However, in the small area on the form image, if any of the small areas located above, below, left and right does not correspond to the small area on the form image, the image inside the small area is not erased. This makes it possible to remove the preprinted part without losing the characters even if the characters to be written and the preprinted character string are in contact, and to know not only the presence or absence of the mark but also the shape of the mark when the mark is written. Can be.

Embodiment 10 FIG. Next, a tenth embodiment of the present invention will be described with reference to FIGS. FIG.
8 and FIG. 39 show the preprinted character string associating means 20 and the preprinted character string erasing means 2 in the tenth embodiment.
38 is described. Reference numeral 130 in FIG. 38 denotes a preprinted character string area 133 on the completed form image.
Is an N × N dot small area. Reference numeral 103 in FIG. 39 denotes shape information of a preprinted character string described in a form format. 161 is the preprint character string area 13
Shape information extracted from 0, 170 is a small area having a vertical component, 171 is a small area having a horizontal component, 164 to 16
9 and 172, 173 are examples of small areas.

In the tenth embodiment, the preprinted character string shape information stored in the form
This is a method for reading a filled-out form when it is represented by a directional component for each small area as described in. Here, the operation of the tenth embodiment will be described with reference to FIG. 1 showing the operation of the form format creation processing, but the preprint character string associating means 20 and the preprint character string erasing means 23
Operations other than the above are the same as those in the ninth embodiment, and a description thereof will be omitted. First, the print character string associating means 20 associates the preprint character string described in the form format with the character string on the form image. In the association, first, the preprint character string position coordinates described in the form format are corrected based on the result of association between the ruled lines obtained by the ruled line associating means 19, so that preprinting on the form image is performed. Estimate the character string position. Then, an image near the estimated character string position is cut out. The preprint character string described in the form format
When associating "month and day", an image 130 shown in FIG. 38 is cut out from the completed form image (FIG. 31).

Next, the extracted image 130 is converted into an N × N
The dot is divided into small regions 133, and the directional component is obtained inside each small region. This direction component can be obtained by the method described in the eighth embodiment. Image 1 cut out by this
From 30, a small area 170 having a vertical component and a small area 171 having a horizontal component are obtained. Then, the small areas having these directions are defined as shape information 161.

Next, the pre-print character string shape information 10
3 is read from the form format. As described in the eighth embodiment, the shape information 103 is also composed of a directional component for each small area.

Thereafter, as shown in FIG. 39, each small area forming the preprint character string shape information 161 created from the filled-out form image and the preprint character string shape information 103 described in the form format are formed. Are optimally associated with each of the small regions in consideration of the directionality of the small region. In the existing technology described in the preprinted character string associating means 20 in the ninth embodiment, such small regions are associated with small regions having the same directional component having high similarity and having different directions. This can be realized by defining that small regions having components have low similarity. Thereby, for example, the small area 164 and the small area 167, the small area 16
5 and the small region 168, the small region 166 and the small region 169, and the small region 172 and the small region 173 are associated with each other. This association is performed for all fields including the preprint character string.

Next, after executing the field extracting means 21 and the ruled line erasing means 22 in the same manner as in the ninth embodiment, the preprinted character string erasing means 23 outputs the character strings obtained by the preprinted character string associating means 20. Based on the correspondence of
The image in the preprint area is deleted from the field area image. Here, the image area to be erased is associated with the small areas by the preprint character string associating means 20 (see FIG. 3).
In 9), the form information 1 of the form image associated with the small area on the form information 103 described in the form format
It is inside the small area on 61. However, like the small area 172 on the form information 161 of the form image,
When the small area 173 associated with has a direction component orthogonal to the direction component of the small area 172, the image inside the small area 173 is not erased.

As a result, even if the entered character and the preprinted character string having a large line width come into contact with each other as shown in image 150 in FIG. 35, the field area in which only the preprinted character string is erased without losing the entered character. An image 152 can be obtained. Similarly, as shown in an image 151 in FIG. 35, a field area image 153 in which only the preprint character string is erased can be obtained for the check mark field. The shape of the mark entered on the column can also be obtained.

As described above, according to the present embodiment, in preprinted character string associating means 20, small areas having the same directional component have a high similarity and small areas having different directional components have a high similarity. Are defined as having low similarity, and the small areas are associated with each other. Thus, even when the preprint shape information shown in the eighth embodiment is used, it is possible to associate preprint character strings with each other. Also,
The preprint character string erasing means 23 erases the image inside the small area on the form image associated with the small area in the form format. However, if the small area on the form image and the corresponding small area in the form format have directional components orthogonal to each other, the image is not erased. As a result, even when the preprinted shape information shown in Embodiment 8 is used, the preprinted portion can be removed without losing the written characters even if the written characters touch the preprinted character string, and the mark is written. At this time, not only the presence or absence of the mark but also the mark shape can be known.

[0102]

According to the form reader of the first aspect, the output format of the recognition result of the field type and the check mark field from the blank form image and the ID for form identification
By automating the setting of the form title corresponding to the number, the form reading format can be efficiently created.

According to the form reader of claim 2, even if the extracted field is a check mark field,
The field type can be determined.

According to the form reader of claim 3, even if the extracted field is an image field, the field type can be determined.

According to the form reading apparatus of the present invention, when the recognition result of the check mark field having the character entry field in the option is output, the preprint character string representing the option and the image of the entry character are added to the recognition result of the mark. This makes it possible to smoothly input items other than options.

According to the form reader of the present invention, when the recognition result of the check mark field having the character entry field in the option is output, the preprint character string representing the option and the recognition result of the input character are added to the recognition result of the mark. By doing so, it is possible to smoothly input items other than options.

According to the form reader of claim 6, since the recognition result for the mark entered at an arbitrary position in the check mark field constituted by a continuous straight line can be output, a variety of check marks can be easily entered by the entry person. Fields can be configured.

According to the form reader of claim 7, the form title as the form ID can be detected more accurately.

According to the form reading device of the present invention, since the shape information of the preprinted character string is represented by the directional component for each small area, the preprinted character string image is stored as it is as the shape information in the form format. It is possible to reduce the size of the form format as compared with the case of performing the above.

[0110]

[0111]

[Brief description of the drawings]

FIG. 1 is a block diagram showing an operation of a form reading device according to a first embodiment of the present invention.

FIG. 2 is a diagram showing an example of a blank form of the form reading device according to the first embodiment of the present invention;

FIG. 3 is a diagram showing extracted ruled lines of the form reading device according to the first embodiment of the present invention;

FIG. 4 is a diagram showing format information of ruled lines extracted by the form reading apparatus according to the first embodiment of the present invention;

FIG. 5 is a diagram showing an extracted preprinted character string of the form reading apparatus according to the first embodiment of the present invention.

FIG. 6 is a diagram showing information of a preprinted character string extracted by the form reading apparatus according to the first embodiment of the present invention.

FIG. 7 is a diagram showing a keyword dictionary of the form reading device according to the first embodiment of the present invention.

FIG. 8 is a diagram showing extracted field candidates of the form reading apparatus according to the first embodiment of the present invention.

FIG. 9 is a diagram showing an example of extracted field information of the form reading device according to the first embodiment of the present invention.

FIG. 10 is a diagram showing a form title condition storage table of the form reading device according to the first embodiment of the present invention.

FIG. 11 is a diagram showing an example of a preprinted character string for creating preprinted character string shape information of the form reading apparatus according to the first embodiment of the present invention.

FIG. 12 is a diagram showing a form format of the form reading device according to the first embodiment of the present invention.

FIG. 13 is a diagram showing an example of a check mark field of the form reading device according to the second embodiment of the present invention.

FIG. 14 is a diagram showing preprint character string information for the fields in FIG. 13 of the form reading apparatus according to the second embodiment of the present invention.

FIG. 15 is a diagram showing field information for the fields in FIG. 13 of the form reading apparatus according to the second embodiment of the present invention.

FIG. 16 is a diagram showing an example of an image field of the form reading device according to the third embodiment of the present invention.

FIG. 17 is a diagram showing preprinted character string information for the fields in FIG. 16 of the form reading apparatus according to the third embodiment of the present invention.

FIG. 18 is a diagram showing field information corresponding to the fields in FIG. 16 of the form reading apparatus according to the third embodiment of the present invention.

FIG. 19 is a diagram showing an example of a form (not filled) in which a character string field exists in a check field of the form reading apparatus according to the fourth embodiment of the present invention.

FIG. 20 is a diagram showing an example in which characters are entered in the fields shown in FIG. 19 of the form reader according to the fourth embodiment of the present invention.

FIG. 21 is a diagram showing field information of the fields shown in FIG. 19 of the form reading device according to the fourth embodiment of the present invention.

FIG. 22 is a diagram showing an example of characters entered in a character string field in a check mark field of the form reader according to the fifth embodiment of the present invention.

FIG. 23 is a diagram showing field information of the fields shown in FIG. 19 of the form reading apparatus according to the fifth embodiment of the present invention.

FIG. 24 is a diagram showing a check mark field of the form reading device according to the sixth embodiment of the present invention.

FIG. 25 is a diagram showing an example in which a mark is entered in the field shown in FIG. 24 of the form reading apparatus according to the sixth embodiment of the present invention.

FIG. 26 is a diagram showing an example of field information output from the form reading device according to the sixth embodiment of the present invention.

FIG. 27 is a diagram showing a form title condition storage table with evaluation values of the form reading apparatus according to the seventh embodiment of the present invention.

FIG. 28 is a diagram showing an example of a form title evaluation value output from the form reading apparatus according to the seventh embodiment of the present invention.

FIG. 29 is a diagram illustrating preprinted character string shape information of the form reading apparatus according to the eighth embodiment of the present invention.

FIG. 30 is a block diagram showing an operation of a form reading apparatus according to Embodiment 9 of the present invention.

FIG. 31 is a diagram showing an example of a filled-out form of the form reading apparatus according to Embodiment 9 of the present invention.

FIG. 32 is a diagram showing an example of association of ruled lines in a form reading apparatus according to Embodiment 9 of the present invention.

FIG. 33 is a diagram showing an example of shape information of a preprinted character string extracted from a filled-out form of the form reading apparatus according to Embodiment 9 of the present invention.

FIG. 34 is a diagram showing an example of association of preprinted character string shape information of the form reading apparatus according to Embodiment 9 of the present invention.

FIG. 35 is a diagram illustrating an example of an image in which a preprinted character string has been deleted by the form reading apparatus according to Embodiment 9 of the present invention;

FIG. 36 is a diagram showing all fields extracted from the filled-out form shown in FIG. 31 of the form reading apparatus according to Embodiment 9 of the present invention.

FIG. 37 is a diagram illustrating an example of a recognition result of the filled-in form illustrated in FIG. 31 by the form reading apparatus according to the ninth embodiment of the present invention;

FIG. 38 is a diagram showing an example of shape information of a preprinted character string extracted from a filled-out form of the form reading apparatus according to the tenth embodiment of the present invention.

FIG. 39 is a diagram showing an example of association of preprinted character string shape information of the form reading apparatus according to Embodiment 10 of the present invention.

FIG. 40 is a diagram showing a processing flow of a conventional form reading apparatus.

[Explanation of symbols]

1 form image storing means, 2 ruled line extracting means, 3 preprint character string extracting means, 4 preprint character string recognizing means, 5 keyword dictionary, 6 keyword collating means, 7 field candidate extracting means, 8 field type discriminating means, 9 fields Attribute setting means, 10 output format setting means, 11 title condition storage table, 12
Form title candidate extracting means, 13 preprint character string shape information creating means, 14 form format output means,
17 form format selecting means, 18 form format, 19 ruled line associating means, 20 preprint character string associating means, 21 field extracting means, 22
Ruled line erasing means, 23 preprinted character string erasing means, 2
4 Entry character string recognition means, 25 Entry mark recognition means,
26 Entry character string / mark recognition result output means.

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-4-343190 (JP, A) JP-A-5-28310 (JP, A) JP-A-7-13996 (JP, A) (58) Field (Int.Cl. ⁷ , DB name) G06K 9/20

Claims (8)

(57) [Claims] 1. A form image storage means for storing a form image to be read, a ruled line extraction means for extracting a ruled line constituting a form from the form image stored in the form image storage means, and a form line stored in the form image storage means A preprinted character string extracting means for extracting a character string (preprinted character string) printed in advance on the form from the formed form image, and recognizing the character string image extracted by the preprinted character string extracting means and converting the character string into a character code A preprinted character string recognizing means for comparing the preprinted character string coded by the preprinted character string recognizing means with a word registered in a keyword dictionary storing a word string to be pre-filled in the form. Keyword matching means, using ruled line information obtained by the ruled line extracting means and a matching result obtained by the keyword matching means. A field candidate extraction unit determines whether or not the reading area (field), the type of the field the field by analyzing the contents of the field candidate obtained by the candidate extracting means, a character string field
Or check mark field or image field
Field type determining means for automatically determining whether the field is a field, and determining an item to be entered in the field from the field type determined by the field type determining means and the matching result obtained by the keyword matching means. Field attribute setting means for setting reading conditions; and a result output format setting means for determining a format of a result to be output at the time of reading a form from the field type determined by the field type determination means and the matching result obtained by the keyword matching means. Comparing the recognition result obtained by the preprint character string recognition means with a title condition storage table describing characteristics of a form title added to the form, and selecting the form title candidate from the preprint character string. A form title candidate extracting means for extracting the preprint character string extracting means, A preprint character string shape information creating means for extracting shape information of the preprint character string to be recorded in form format information from the preprint character string image , and a form title obtained by the form title candidate extracting means
Information, the ruled line information obtained by the ruled line extracting means,
Preprint sentence obtained by preprint character string extraction means
The character string information and the file obtained by the field candidate extraction means.
Information on Outer Ruled Lines for Field Candidates and Inclusive Preprinting
Character string information and the field type
Obtained by the field attribute setting means.
Field attributes and the preprinted character string shape
Preprint character string shape information obtained by the information creation means
And an output format obtained by the output format setting means.
Form format output means to output as a format
And a form reader. 2. The method according to claim 1, wherein the field type determining means includes a
The keyword type of the preprint character string included in
Judge whether it represents the option.
If specified, check the field type
2. The form reading apparatus according to claim 1, wherein the form is output as a field . 3. A field type determining unit detects whether there is a word that prompts input of a figure or an image in a field, and when it is determined that there is a word that prompts input of a figure or an image, the upper, lower, left and right of the word is determined. 2. The form reading apparatus according to claim 1, wherein the predetermined neighboring area is an image field. 4. An output format setting means determines whether or not there is an area for entering a character string in the check mark field in addition to a preprinted character string representing an option,
2. The form according to claim 1, wherein when it is determined that there is a document, an output format is determined in which a preprint character string representing an option and an image of a character string written in a character area are added when the recognition result is output. Reader. 5. The output format setting means determines whether or not there is an area for entering a character string in the check mark field in addition to the preprinted character string representing an option. 2. The form reading apparatus according to claim 1, wherein an output format in which a preprint character string representing an option and a character code of a recognition result of a character string written in a character area are added when the recognition result is output is determined. 6. An output format setting means, wherein the check mark field is constituted by a continuous straight line, and
If it is determined that can be entered at any position in the field
2. The form reading apparatus according to claim 1, wherein an output format is set in which a ratio of a length from a field end to a mark entry position with respect to an entire length of the field is added. 7. The form title extracting means determines whether a preprint character string obtained from the form image matches a condition stored in a form title condition storage table. 2. A character string is output as a form title.
Form reader. 8. The preprinted character string shape information creating means outputs a position and a size of the preprinted character string and a directional component in a divided small area of an area where the preprinted character string exists. 2. The form reading device according to claim 1, wherein: