JP2003323586A - Document form registering method and document recognizing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a document form registering method extensively reducing labor of a user in form registration, and allowing even an inexperienced user to easily carry out form registration. <P>SOLUTION: In preregistering a form of a document and recognizing characters written on the document on the basis of the registered form, an image of a blank document with nothing written is captured by using an image inputting means, positions, thicknesses, line types or the like of ruled lines are extracted from the captured image, and information of the extracted ruled lines is registered as form information. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to a form for registering a form on a form in advance and recognizing characters written on the form based on the registered form, such as a form recognizing form on the form. The present invention relates to a form registration method and a form recognition method for recognizing characters written on a form.

[0002]

2. Description of the Related Art Generally, general form recognition is roughly divided into a form registration process for registering a form on a form and a character recognition process for recognizing characters written on the form based on the registered form. Divide. 16 shows an example of a basic form.
Shown in. The blank form shown in FIG. 16 (a) means a pre-printed form in which nothing has been written on the form. 16B is a completed form in which characters corresponding to the respective items are entered. The form recognition technique is a technique for recognizing the characters written in each entry field and storing the recognition result in a storage unit that stores the recognition result of each item. At that time, the form information of the form is required. As shown in FIG. 16C, the form information is information that pre-designates what kind of information is written at which position on the form. In addition to this, the form information also stores information such as which position the ruled line is drawn, and which position should be used as the reference for character recognition alignment. The reference for alignment is called a target mark. In addition, the ruled line information is used to efficiently cut out a character when the character is hung on the ruled line during character recognition.

Further, in the form information, attribute information indicating where in the database the recognition result of the character written in the entry field should be stored is also registered. For example, the result of recognizing the character written in the entry field having the attribute of "name" is stored in the portion of the database representing "name".

[0004]

However, all of the conventional form registration processing has been manually performed by humans. In the early version, positions such as entry fields in the form were manually measured in advance, and the positions were manually input to the form recognition device. In addition, in the method of using a GUI (graphical user interface), registration is performed by displaying the captured blank form image on the screen and carefully pointing the inside of the entry field with a mouse pointer or the like. However, in the form information, the recognition rate may change greatly with the accuracy of one dot, and the registration takes cost and time. Also, if the ruled lines have a special shape, the complexity cannot be expressed simply by registering them as ruled line information, and it may be difficult to separate the ruled lines from the entered characters during character recognition. There is.

Therefore, an object of the present invention is to provide a form form registration method in which the user's time and effort for registering a form can be greatly reduced, and even an inexperienced user can easily perform form registration. Another object of the present invention is to provide a form recognition method capable of improving character recognition accuracy.

[0006]

According to the form form registration method of the present invention, when a form on a form is registered in advance and the characters written on the form are recognized based on the registered form, nothing is written. The feature is that an image of a blank form that has not been captured is captured using the image input means, the position of the ruled line, the thickness, the line type, etc. are extracted from the captured image, and the information of the extracted ruled line is registered as form information. And

The form form registration method of the present invention is
In a form in which a form on a form is registered in advance and characters recognized on the form are recognized based on the registered form, an image of a blank form on which nothing is described is captured using the image input means, and this capture is performed. It is characterized in that an area surrounded by a solid line or a broken line is extracted from the image as an entry field, and the information in the extracted entry field is registered as form information.

Further, in the form recognition method of the present invention, a form on a form is registered in advance and the characters written on the form are recognized based on the registered form. When the form is registered, nothing is written. An image of an empty blank form is captured using the image input means, at least a part of the captured image is registered together with the form information, and at the time of character recognition, the image of the input recognition target form is registered first in the above. It is characterized in that ruled lines and characters printed in advance on the form are removed by using the image of the blank form, and then character recognition is performed on the form.

Further, in the form recognition method of the present invention, in the form recognition method for recognizing the characters described in the image on the form, the image on the form in which no character is described is registered in advance, and the character recognition is performed. At this time, the image on the form is removed from the input image of the form to be recognized by using the previously registered image, and then character recognition is performed on the form.

[0010]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. First, the first embodiment will be described. The first embodiment reduces the form registration work of ruled lines by automatically detecting the ruled lines of the blank form. A schematic configuration of the ruled line extraction device is shown in FIG. 1, which will be described below.

First, set a blank form on the reading unit,
The grayscale image of the blank form is captured by the scanner 11 as an image input unit and stored in the grayscale image storage unit 12. Next, the image of the blank form stored in the grayscale image storage unit 12 is binarized by the binarization unit 13, and the result is stored in the binarized image storage unit 14.

Next, with respect to the binarized image stored in the binarized image storage unit 14, the vertical projection unit 15 and the horizontal projection unit 16 project black pixels in each direction, and the result is obtained. The results are stored in the vertical projection result storage unit 17 and the horizontal projection result storage unit 18, respectively. Next, the ruled line extraction unit 19 stores in the binarized image storage unit 14 based on the information of the vertical projection and the horizontal projection stored in the vertical projection result storage unit 17 and the horizontal projection result storage unit 18. The black pixels of the binarized image are traced to find the start and end points of the ruled line. The ruled line extraction unit 19 also detects how the black pixels of the ruled line are connected,
It also identifies whether it is a solid line or a broken line.

The information on the ruled lines extracted by the ruled line extraction unit 19 may be registered as the form information as it is, but it is desirable to perform registration work on the GUI in order to avoid erroneous extraction and improve accuracy. On the GUI, for example, as shown in FIG. 2, the extracted ruled lines are displayed in different colors. Also, the color is displayed in different colors depending on the line type of the ruled line (solid line, broken line, etc.). The user selects a ruled line by clicking this color-changed portion with a mouse or the like, and registers the ruled line information in the form information.

Next, a second embodiment will be described. In the second embodiment, the position of the entry column and its attributes are identified by automatically detecting the entry column of the blank form, and the form registration work is reduced. 3 and will be described below.

First, set a blank form on the reading unit,
The grayscale image of the blank form is captured by the scanner 21 as image input means and stored in the grayscale image storage unit 22. Next, the image of the blank form stored in the grayscale image storage unit 22 is binarized by the binarization unit 23, and the result is stored in the binarized image storage unit 24. Binary image storage unit 2
An example of the image stored in No. 4 is shown in FIG.

Next, the ruled line information stored in the binarized image storage unit 24 may be interrupted due to blurring or the like, and may not be linearly connected like a broken line. Therefore, these interruptions are eliminated by the ruled line extension portion 25. That is, the ruled line extension unit 25 examines the black pixels in both the vertical and horizontal directions, and first obtains information on which direction the ruled line is facing. After that, the line segment is searched up to the break. If the line segment intersects another line segment and reaches the end, the search ends there. If a line segment is searched for in the middle, search continues in that direction for a while, and if there is an endpoint of another line segment or another line segment in the vertical direction, it is regarded as another line segment in that direction. Stretch the black pixels until they touch. The result thus obtained is stored in the binarized image storage unit 26. An example of the image stored in the binarized image storage unit 26 is shown in FIG.

Next, the inverted image creating unit 27 inverts the image stored in the binarized image storage unit 26. The inverted image is stored in the inverted image storage unit 28. An example of the image stored in the reverse image storage unit 28 is shown in FIG. Next, the labeling unit 29 uses the inverted image in the inverted image storage unit 28 to perform the labeling process. In this case, the black pixels of the binarized image may be regarded as the background and the labeling may be performed directly without the need to generate the inverted image. An example of the position of the label extracted by the labeling unit 29 is shown in FIG.

Next, the entry section extraction section 31 has label information (size, position, etc.) obtained by the labeling section 29.
Based on, the process of detecting the entry column is performed. That is, the entry column extraction unit 31 searches the labeling result based on the preset entry column size standard, and when an appropriate label is found, checks the inverted image stored in the inverted image storage unit 28. Find out if most of the label's position is filled with black pixels. If the conditions are met,
The position of the label is determined as the position of the entry field candidate.

The entry field candidates thus extracted may be directly registered as form information, but it is desirable to manually register them on the GUI in order to improve accuracy. On the GUI, in the same way as the display after detecting the ruled line, the circumscribed rectangle of the extracted entry field is displayed in different colors and thicknesses, or the entry field itself is displayed in different colors and lightness.
The user registers it by clicking it with the mouse.

Next, a third embodiment will be described. In the third embodiment, the attributes of the entry fields extracted in the entry field extraction process are detected and automatically registered in the form information.
Types can be considered.

(1) A method for recognizing the item name printed near the entry field (2) A method for recognizing the item name specified by the user in the entry field (3) A number predetermined by the user for each attribute A method of recognizing a number or a mark in the entry field and recognizing the number or the mark First, a method of recognizing an item name printed near the entry field will be described. In this method, the attribute of the entry field extracted in the second embodiment is acquired by recognizing the item name printed near the entry field and registered in the form information.

FIG. 5A shows an example of a specific entry field on the form, for example, a name entry field. In this embodiment,
When the name entry field is extracted, the characters printed in the vicinity are automatically recognized by the character recognition processing. Figure 5
In the example of (a), since "name" is written on the left side of the entry field, "name" is recognized. This is registered in the form information as an attribute of the entry field.

In this case, when the entry field is long horizontally, the characters written on the left side are preferentially recognized as attributes, and when the entry field is vertically long, the characters written on the upper side are given priority. Recognized as an attribute. When a plurality of characters are printed near the entry field, the recognition result of the print character closest to the entry field is registered as an attribute. If the printed characters have an attribute that is not registered in the database in advance, the user is requested to register again using the GUI.

Next, a method for the user to recognize the item name specified in the entry field will be described. This method is shown in FIG.
As shown in (b), when a character previously entered by the user is detected near the entry field or in the entry field, the character is recognized and the recognition result is registered as an attribute of the entry field. It is a thing. In this case, if the character described by the user has an attribute that is not registered in the database in advance, the user is requested to register it again using the GUI.

In addition, using a pen of a specified color,
By describing the attribute character, the accuracy of character extraction can be improved and the automatic attribute registration can be smoothly performed. In this case, a filter for extracting a specific color is applied to the color image captured by the scanner, and character extraction and character recognition are performed using the image information.

Next, a method will be described in which the user specifies a predetermined number or mark for each attribute in the entry field and recognizes the number or mark. This method is shown in FIG.
As shown in (c), the user writes a mark or a number predetermined according to the type of attribute in the vicinity of the entry field or in the entry field and recognizes it.
The attributes of the entry fields are automatically registered.

To the user, for example, a table of numbers and marks corresponding to the attributes as shown in FIG. 6 is handed in advance, and the user can apply to the inside or the vicinity of the entry field according to the table. Enter the numbers and letters you want. The system recognizes the numeral or mark, refers to the previously registered mark or character and attribute relation table, determines the attribute of the entry field, and registers it as form information. Also in this case, by writing the attribute character using a pen or the like of a designated color, it is possible to improve the accuracy of extraction of numbers and marks and to smoothly perform automatic attribute registration.

Next, a fourth embodiment will be described. The above-mentioned automation of form registration cannot be performed without a blank form. In addition, it is easily affected by noise and the like that occur when the blank form image is input. Therefore, in the fourth embodiment, stable form registration is performed using a plurality of blank forms or completed forms. A schematic configuration of the processing device is shown in FIG. 7, which will be described below. .

First, a plurality of blank forms or completed forms are set in the supply unit and are sequentially supplied to the reading unit one by one, so that the scanner 41 as an image input unit.
The grayscale images of the form are sequentially fetched by the, and the fetched images are respectively stored in the form image storage sections 42a and 42a.
b, 42c, 42d, ...

Next, the image comparison unit 43 is connected to the image storage unit 42.
Based on each image in a, 42b, 42c, 42d, ..., a comparison result image is created in which only preprinted ruled lines, characters, background patterns, etc. are created and stored in the comparison result image storage unit 44. . The system uses this comparison result image to extract entry fields and ruled lines and register the results as form information.

The processing of the image comparison unit 43 will be described in more detail. The density values of pixels at the same position in a plurality of input form images are added for all forms. As a result, the density value is high in the parts such as ruled lines, characters, and background patterns that have been printed in advance, and on the other hand, the characters written in the entry field are not written in the same place each time, so the density value is not so high. . Therefore, only the portion having a high density value is stored in the comparison result image storage unit 44 as a black pixel. This makes it possible to stably extract preprinted parts (ruled lines, characters, background patterns, etc.). By extracting the periphery of the pixel with the highest value among the addition results of the densities of each form as the alignment target mark,
The target mark specified by the user through experience can be registered more efficiently.

Next, a fifth embodiment will be described. In the fifth embodiment, when registering form information,
By inputting the images of a plurality of completed forms and recognizing the characters, the attributes of the entry fields written in the form are obtained, and the schematic configuration of the processing device is shown in FIG.
This will be described below.

First, a plurality of completed forms are set in the supply unit and sequentially supplied to the reading unit one by one, so that the grayscale images of the form are sequentially acquired by the scanner 51 as an image input means. Each of the incorporated images is stored in the form image storage units 52a, 52b, 52c, 52d, ...
To store.

Next, the character recognition unit 53, based on the written position information stored in the written position information storage unit 54, the character writing unit of each image in the form image storage units 52a, 52b, 52c, 52d ,. Character recognition is performed on the inside, and the character recognition result is stored in the character recognition result storage unit 55 for each form. Next, the attribute determination unit 56 determines the attribute of each entry field based on the written position information stored in the written position information storage unit 54 and the character recognition result stored in the character recognition result storage unit 55.

The attribute determining unit 56 determines the attribute based on the character recognition result of the same entry field of a plurality of forms. For example,
If the character recognition result of the entry field includes many words representing the address name such as "Tokyo" or "Kawasaki City", the attribute of the entry field is determined to be "address". If there are many numbers of 10 digits or more starting from "0", the attribute of the entry field is "telephone number". In addition, "man", "child", "husband", "Kimura", "ro" often used for names
When a character such as “” appears frequently, the attribute of the entry field is “name”.

In addition, if a character or character sequence that is likely to appear for each element of the item is registered in advance in the database and the registered character or character sequence is written in many cases, the attribute of the entry field is stored in the database. It is also possible to use the elements registered in. For example, the characters "male" and "female" are registered in the database of the item "gender column", and the approximate size of the entry column is also registered. Numbers, "pieces", "tsu", "ko", etc. are registered in the "quantity column" database. As a result, the attributes of each entry field of the form are automatically acquired even if the user does not specify one by one.

Next, a sixth embodiment will be described. In the sixth embodiment, an image of a blank form is registered as an image at the time of form registration, and the image is used at the time of form recognition. A schematic configuration of the processing device is shown in FIG. This will be explained below.

First, the form to be recognized is set in the reading unit,
The scanner 61 as an image input means captures a grayscale image of the form and stores it in the recognition target form image storage unit 62. Next, the image of the recognition target form stored in the recognition target form image storage unit 62 is binarized by the binarization unit 63, and the result is stored in the recognition target form binary image storage unit 64.

The registered blank form image storage section 65 stores an image (binarized) of the blank form taken together with the form information at the time of form registration. Therefore, the alignment processing unit 66 uses the two image storage units 64 and 6
Based on the target mark of each image stored in No. 5, the two images are aligned, and the positional deviation information is stored in the alignment result storage unit 67.

Here, the target mark is a characteristic mark that is always printed on the form to be recognized. The position of this mark is registered in the form information in advance, and the form to be recognized is binarized. The most likely target mark part is extracted from the image, and the vertical and horizontal shifts of the two images (registered blank form image and recognition target form image) are estimated.

Next, as shown in FIG. 10, the pre-print erasure section 68 uses the registration blank form image storage section 6 as a reference, based on the positional deviation information stored in the registration result storage section 67.
Using the registered blank form image stored in 5, the recognition target form image stored in the recognition target form binarized image storage unit 64 is used to create ruled lines and characters (pre-printed portion) that are printed in advance on the form. Only the character image to be recognized (pre-print erased image) is extracted. This is performed by calculating the pixel in the binarized image of the recognition target form, which corresponds to the same position as the black pixel of the registered blank form image, from the alignment result information (positional shift information), and making the pixel a white pixel. The pre-print erased image thus obtained is stored in the pre-print erased image storage unit 69, and character recognition is performed on the pre-print erased image.

As described above, by removing the pre-printed portion from the image of the form to be recognized, it is possible to remove noises such as ruled lines and printed characters which have a bad influence on the cut-out and recognition of the described characters. .

Next, a seventh embodiment will be described. In the seventh embodiment, a background image of a blank form is registered in advance, so that the background image is erased from the image of the manually recognized recognition target form, and the recognition target character is cut out and the accuracy of character recognition is improved. Then, the schematic configuration of the processing apparatus is shown in FIG. 11, which will be described below.

First, the form to be recognized is set in the reading unit,
A scanner 71 serving as an image input unit captures a grayscale image of the form and stores it in the recognition target image storage unit 72. The background image of the blank form is stored in advance in the registered background image storage unit 73. Therefore, the registration processing unit 74 uses the target mark of each image stored in the two image storage units 72 and 73 as a reference to determine the target mark.
The two images are aligned, and the positional deviation information is stored in the alignment result storage unit 75.

As shown in FIG. 12, the background image erasing unit 76 uses the registered background image stored in the registered background image storage unit 73 with reference to the positional deviation information stored in the alignment result storage unit 75. The background image previously printed on the form is removed from the recognition target form image stored in the recognition target image storage unit 72, and only the character image to be recognized (background image erased image) is extracted.

The processing of the background image erasing unit 76 will be described in more detail. By using the alignment result information (positional deviation information), the density of a pixel in the registered background image and the pixel in which the position of the recognition target image coincides with that of the pixel. If the density difference is not significant and the background image textures match if there is not much difference in density, the pixels in the recognition target image are set to white pixels, and if the density difference is large, there is a possibility that text is written here. Is high,
The pixel has the same density as the recognition target image. By doing so, even if the recognition target character is in close contact with the background image, it is possible to reliably remove only the background image. The background image erased image thus obtained is stored in the background image erased image storage unit 77, and character recognition is performed on the background image erased image.

Next, an eighth embodiment will be described. In the form of shipping industry (delivery slips, etc.), the character "sai" is often used as a unit of weight and capacity.
By recognizing this, the accuracy of weight recognition can be improved. "Sai" represents a unit of weight or volume,
Usually, 1 year old = 10 kg.

FIG. 13 shows an example of a character string including the character "sai". In addition, FIG. 14 shows a procedure for recognizing the weight column accompanied by the character "sai", which will be described below. The input form image undergoes binarization processing, character cutting processing, and weight recognition processing. First, the cut-out character is recognized (step S1). Next, the recognition result of the number portion of the weight is obtained based on the recognition result (step S2). After that, the recognition result of the unit of the weight column is obtained (step S3). As a result of recognizing the unit, if the unit is blank or "Kg", the result of the numeral recognition is output as it is (step S4). When the result of unit recognition is "age", the result of numeral recognition is multiplied by 10 (step S5), and the result is output by aligning the units.

Next, a ninth embodiment will be described. FIG. 13 shows an example of a character string containing the characters "age". Fig. 13 (a) represents "1 year old" and Fig. 13 (b) represents "3 year old". As can be seen from FIG. 13, there are many variations in the writing of “age”. Therefore, even if an attempt is made to recognize "age" by the conventional character recognition technology, the recognition rate becomes low.

In the weight / capacity column of the form, only the character types of the numbers "0" to "9" and "age" usually appear. Therefore, even if "character" cannot be recognized in the character recognition process, if it is found that all the character candidates "0" to "9" are not recognized, the character may be regarded as "character".

FIG. 15 shows a flow for determining the recognition of the character "sai". According to FIG. 15, when the character recognizing unit cannot say “senior”, whether the character is the character is judged by the 10 discriminating units from the “0” discriminating unit to the “9” discriminating unit.
Determine if it is not the letter. As a result of this determination, if it is found that all the categories from “0” to “9” are not found, the recognition result of the character is set to “age”. In this way, the category of characters to be entered is decided,
When there is a large variety of written transformations in a certain category and the recognition rate is low, the recognition rate of that character can be increased by determining that it is not another character.

As described above, according to the above embodiment, the positions of ruled lines and entry fields on the form are automatically extracted,
The information on the extracted ruled lines and entry fields can be registered as it is, or the user can simply register the information. In addition, the attributes of the entry fields are automatically recognized by using the character recognition technology to recognize the characters printed near the entry fields and the attributes that the user has previously handwritten on the form, and using the results. You can register. This allows
The time and effort required for the user to register the form is greatly reduced,
Further, even an inexperienced user can easily perform form registration.

Further, the blank form image is registered as it is, and the image is used at the time of form recognition to remove the preprinted portion from the image of the form to be recognized. It is possible to remove noises such as ruled lines and printed characters, which have a bad influence on clipping and recognition of characters, and improve the character recognition accuracy. Furthermore, by registering the background image of the blank form in advance,
Erase the background image from the image of the recognition target form that was manually operated,
It is possible to improve the accuracy of character segmentation and character recognition.

[0054]

As described above in detail, according to the present invention, a form form registration method is possible in which the user's time and effort for registering a form is significantly reduced, and even an inexperienced user can easily perform form registration. Can be provided. Also,
According to the present invention, it is possible to provide a form recognition method capable of improving character recognition accuracy.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of a ruled line extraction device according to a first embodiment.

FIG. 2 is a diagram showing a display example of a ruled line extraction result.

FIG. 3 is a block diagram showing a schematic configuration of an entry field extraction device according to a second embodiment.

FIG. 4 is a diagram showing various image examples used for explaining an entry column extraction process.

FIG. 5 is a diagram illustrating an example of entry field attribute detection according to the third embodiment.

FIG. 6 is a diagram showing an example of a correspondence table of numbers and attribute information that a user fills in a form in the third embodiment.

FIG. 7 is a block diagram showing a schematic configuration of a processing device according to a fourth embodiment.

FIG. 8 is a block diagram showing a schematic configuration of a processing device according to a fifth embodiment.

FIG. 9 is a block diagram showing a schematic configuration of a processing device according to a sixth embodiment.

FIG. 10 is a diagram illustrating a form recognition process using a blank form image according to the sixth embodiment.

FIG. 11 is a block diagram showing a schematic configuration of a processing device according to a seventh embodiment.

FIG. 12 is a diagram illustrating a form recognition process using a texture image in the seventh embodiment.

FIG. 13 is a diagram showing an example of a character string including a character of “age” to be processed in the eighth embodiment.

FIG. 14 is a flowchart showing a procedure for recognizing a weight column accompanied by the characters “sai” in the eighth embodiment.

FIG. 15 is a diagram showing a flow of recognition determination of a character “sai” in the eighth embodiment.

FIG. 16 is a diagram showing an example of a recognition target form.

[Explanation of symbols]

11 ... Scanner (image input means), 12 ... Gray image storage unit, 13 ... Binarization unit, 14 ... Binary image storage unit, 15 ...
Vertical projection unit, 16 ... Horizontal projection unit, 17 ... Vertical projection result storage unit, 18 ... Horizontal projection result storage unit, 19 ... Ruled line extraction unit.

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI theme code (reference) G06T 7/60 180 G06T 7/60 180A 200 200K (72) Inventor Hideo Horiuchi 70 Yanagicho, Yuki-ku, Kawasaki-shi, Kanagawa Address: Toshiba Yanagicho Office (72) Inventor Naoki Natori 70 Yanagicho, Saiwai-ku, Kawasaki City, Kanagawa Prefecture In-house Toshiba Yanagimachi Office (72): Akihiko Nakao 70, Yanagicho, Saiwai-ku, Kawasaki City, Kanagawa Toshiba Yanagimachi Business In-house (72) Yasuhiro Aoki 70 Yanagi-cho, Saiwai-ku, Kawasaki-shi, Kanagawa Prefecture, Yanagi-cho, Toshiba Office (72) Inventor Noriyuki Hamamura 70, Yanagi-cho, Sai-ku, Kawasaki-shi, Kanagawa F-Term, Toshiba Yanagi-cho, Ltd. (reference) 5B029 BB02 CC26 CC29 EE08 5B057 AA11 BA29 CA02 CA08 CA12 CB02 CB06 CB12 CE12 CH01 CH11 DA08 DB02 DB05 DB 09 DC16 5B064 AA01 BA01 EA20 5L096 BA20 EA37 FA03 FA16 FA46 FA73 LA05

Claims (12)

[Claims] 1. A form on a form is registered in advance,
In recognizing the characters written on the form based on this registered form, an image of a blank form on which nothing is written is captured using the image input means, and the position of the ruled line from the captured image,
A form form registration method comprising extracting thickness, line type, etc., and registering the extracted ruled line information as form information. 2. The form on the form is registered in advance,
In recognizing the characters written on the form based on this registered form, an image of a blank form with nothing written is captured by using the image input means, and the captured image is surrounded by a solid line or a broken line. A form form registration method, characterized in that the extracted area is extracted as an entry field, and the information in the extracted entry field is registered as form information. 3. The image captured by the image input means is
3. The form form registration method according to claim 2, wherein the binarized image is binarized, the black pixels of the binarized image are regarded as the background to extract the connected component, and the entry field is extracted using the extracted information. 4. The form registration according to claim 2, wherein the item name printed near the extracted entry field is recognized, and the recognition result is registered in the form information as an attribute of the entry field. Method. 5. The character or mark written by the user in the extracted entry field or in the vicinity thereof is recognized, and the recognition result is registered in the form information as an attribute of the entry field. Report form registration method. 6. A form on a form is registered in advance,
In the one that recognizes the characters written on the form based on this registered form, each image of a plurality of forms is captured by using the image input means, and the plurality of captured images are preprinted on all the forms. The form form registration method is characterized in that the extracted information is extracted and the extracted information is registered as form information. 7. The total value of pixels obtained by adding the density values of pixels corresponding to the same position in a plurality of captured images and dividing the pixel at each position into a pixel having a high sum and an average value and a pixel having a low value. 7. The form form registration method according to claim 6, wherein ruled lines, characters, background patterns, and the like that are printed in advance on the sheet are extracted. 8. A form on a form is registered in advance,
In the case of recognizing characters written on a form based on this registered form, each image of a plurality of completed forms is captured using image input means, and character recognition is performed on the captured multiple images. The form form registration method characterized in that the attribute of the entry field is determined based on the character recognition result of the same entry field of the plurality of forms, and the determined attribute is registered in the form information. 9. The form on the form is registered in advance,
When recognizing the characters written on the form based on this registered form, at the time of form registration, an image of a blank form that has no description is captured using the image input means, and at least one of the captured images is captured. The part is registered together with the form information, and at the time of character recognition, the ruled lines and characters etc. printed in advance on the form are input from the image of the input form to be recognized using the blank form image previously registered. A form recognition method, which comprises removing and then performing character recognition on the form. 10. A form recognition method for recognizing a character described in an image on a form, wherein an image on a form in which no character is described is registered in advance, and a recognition target form input at the time of character recognition. A form recognition method, characterized in that the image on the form is removed from the image using the previously registered image, and then character recognition is performed on the form. 11. A form recognition method for recognizing a character string written in a weight column of a form used in the transportation industry, wherein a character string in the weight column is recognized, and a weight unit portion of the recognition result indicates a weight unit. In the case of the character "zai" shown, the recognition result of the numeral which is the weight part of the recognition result is multiplied by a predetermined numerical value and output, and the weight unit part of the recognition result is other than the character "zai". In this case, the form recognition method is characterized in that the recognition result of the numeral which is the weight portion of the recognition result is output as it is. 12. A form recognition method for recognizing a specific character written in a specific entry field of a form when a limited number of character types appear in the particular entry field, When recognizing a specified specific character, if it can be determined that the specified character is not all the character types except one of the possible character types, the recognition result of the specific character is regarded as the remaining one character. A form recognition method characterized by the above.