JP3484446B2 - Optical character recognition device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical character recognition device, and more particularly, to an optical character recognition device used for registering and recognizing general fixed form documents such as a driver's license, an insurance card, and a student ID card. Regarding the configuration of.

[0002]

2. Description of the Related Art Conventionally, as an example of an optical character recognition device, a document processing device disclosed in Japanese Patent Application Laid-Open No. 6-52156 has been known. In this document processing device,
A document in which characters indicating information about personal attributes are left blank, a so-called sample document is digitized and registered, and the ruled line frame and the relationship between the ruled line frames are extracted from the image of the sample document. Characters such as item names written in are identified, and after checking against the given knowledge, the ruled line frame part and the image structure of the sample document consisting of the identified characters are understood and the Format information is being registered. When recognizing the input document, as in the case of registration, the image processing apparatus understands the image structure of the input document and determines which one of the sample documents has been registered in advance. A recognition process is to be executed based on the format information of the sample document.

[0003]

However, the document processing apparatus according to the conventional technique needs not only the sample document but also the means for realizing the function of understanding the image structure of the input document. In addition, it takes a lot of processing time to understand the image structure of the input document, which is inconvenient and impractical.

The present invention was devised in view of such inconveniences, and even if it is an input document in which characters indicating information regarding personal attributes have been entered, registration and recognition are possible, and the configuration is It is an object of the present invention to provide an optical character recognition device that is significantly simplified and that can shorten the processing time.

[0005]

An optical character recognition apparatus according to claim 1 of the present invention comprises a photoelectric conversion means for digitizing an input document, an image storage means for storing an image of the digitized input document, A ruled line frame extracting means for extracting a ruled line frame from the image of the input document, a ruled line frame coordinate sequence storage means for storing a coordinate sequence of the extracted grid points of the ruled line frame, a coordinate sequence of grid points of the ruled line frame and a registered name. A format setting means for setting the format of the input document represented by the attributes of the registration area delimited by the ruled line frame and external input information such as keywords, and the format of the input document is one of the registered formats. Format search / Search to determine if applicable /
An optical character recognition device comprising a judgment means and a format registration number storage means for storing the number of registered formats , wherein the format search is performed in the registration mode.
/ The determination means is set by the format setting means
Primary registration of the format in the format structure registration area
This primary registered format and registered
When mats are compared sequentially and format judgment is not possible,
The format setting means further includes a similar former.
Input and keyword comparison between input formats
Repeat and register for all recorded formats
Input format is different for all formats
If you want to register the format of the input document,
To do. The optical character recognition device according to claim 2 is the optical character recognition device according to claim 1.
The same format and input
If there is no difference in the keywords from the format,
The mat setting means displays that there is no difference, and either
One of the features is to reset the format
There is.

The optical character recognition device according to the third aspect of the present invention is an input device.
Photoelectric conversion means for digitizing documents and digitized input sentences
Image storage means for storing the image of the document and whether the image of the input document
Ruled line frame extraction means for extracting a ruled line frame from the extracted ruled line
Ruled line frame coordinate sequence storage means for storing a coordinate sequence of grid points of a frame
And the grid of coordinate points of the ruled line frame, separated by the registered name and ruled line frame.
Input from outside such as registered area attributes and keywords
A file that sets the format of the input document represented by information and
Format setting method and input document format are registered
Search for one of the supported formats
Format search / judgment means for judging
Format registration number storage means for storing the number of formats
Is an optical character recognition device equipped with
The format search / determination means,
Format the format set by the mat setting means.
The primary registration was made in the registration structure registration area, and this primary registration was performed.
Compare the format and the registered format one by one, and
If the format cannot be determined, set the above format.
The means are more similar to the input format
All formats registered for comparison by keywords
About the registered formats
If it succeeds in determining which of
The feature is that the necessary area is recognized.

The optical character recognition apparatus according to claim 4 is an input device
Photoelectric conversion means for digitizing documents and digitized input sentences
Image storage means for storing the image of the document and whether the image of the input document
Ruled line frame extraction means for extracting a ruled line frame from the extracted ruled line
Ruled line frame coordinate sequence storage means for storing a coordinate sequence of grid points of a frame
And the grid of coordinate points of the ruled line frame, separated by the registered name and ruled line frame.
Input from outside such as registered area attributes and keywords
A file that sets the format of the input document represented by information and
Format setting method and input document format are registered
Search for one of the supported formats
Format search / judgment means for judging
Format registration number storage means for storing the number of formats
Is an optical character recognition device equipped with
If the format search / determination means
Format the format set by the setting means
The primary registration is performed in the body registration area, and the primary registration is performed.
Format and the registered formats are sequentially compared, and the format is
If the result cannot be determined, at least one of the
If no keywords are registered in the mat,
Display information about formats that are not registered
Shows that keywords are registered in both formats.
Character recognition, the keyword registration name and
Characterized by comparing the character recognition results.

The optical character recognition apparatus according to claim 5 is an input device.
Photoelectric conversion means for digitizing documents and digitized input sentences
Image storage means for storing the image of the document and whether the image of the input document
Ruled line frame extraction means for extracting a ruled line frame from the extracted ruled line
Ruled line frame coordinate sequence storage means for storing a coordinate sequence of grid points of a frame
And the grid of coordinate points of the ruled line frame, separated by the registered name and ruled line frame.
Input from outside such as registered area attributes and keywords
A file that sets the format of the input document represented by information and
Format setting method and input document format are registered
Search for one of the supported formats
Format search / judgment means for judging
Format registration number storage means for storing the number of formats
An optical character recognition device equipped with
Check the relation between the attribute of the registered area and the character recognition result
However, if the character recognition result is incorrect,
The recognition result check means to show and the high level based on the input information.
Perform fast recognition or, alternatively, recognize at normal speed
Then, the character recognition result corresponding to the input information is obtained.
Whether the recognition result check means determines whether or not
It is equipped with a recognition mode selection means that the user selects.
To collect.

The optical character recognition device according to claim 6
According to any one of claims 1 to 5,
Format search / judgment means
Which of the registered formats does Matt have?
A function to instruct the outside of the factors that made the judgment impossible
It is characterized by having. Optical sentence according to claim 7
The character recognition device is described in any one of claims 1 to 6.
The format search / determination means
Registered format corresponding to the document format
Has a function to let the outside know the registered name of
And are characterized.

An optical character recognizing device according to claim 8 claims
According to any one of claims 3 to 7,
The input document format is registered
Specify which is applicable based on instructions from the outside
Format specification method and the registered registered file
A format that directs the execution of matte-based character recognition
It is characterized in that it comprises a search / determination means. Claim
The optical character recognition device according to item 9 is described in item 8.
The format search / determination means is
Registered name of the format specified by the format specification method
It has a function to instruct the outside to
And

[0011]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of the optical character recognition device according to the present embodiment. The optical character recognition device according to the present embodiment is for registering and recognizing a fixed form document such as a driver's license. It is used when performing.

The optical character recognition device, as shown in FIG.
Along with registration / recognition mode switching means 1, photoelectric conversion means 2, RAM 3 as a temporary storage area, hard disk 4 as a data storage device, ruled line frame reading means 5 as ruled line frame extraction means, and format setting means 6. , A format designation means 7, a format search / determination means 8, a character recognition means 9, a character recognition result processing means 10, a recognition mode selection means 11 which is a character recognition procedure designation means, a recognition result check means 12, It comprises a recognition / word dictionary 13, a display 14 which is a means for displaying information to the outside, and a keyboard 15 and a mouse 16 which are means for inputting information from the outside. These various means are connected via a bus 17. Are connected to each other. Then, in the RAM 3, a document image storage area 3a, which is an image storage unit for temporarily storing the image of the input document digitized by the photoelectric conversion unit 2, and a grid of ruled line frames read by the ruled line frame reading unit 5. There is provided a ruled line frame coordinate sequence storage area 3b which is a ruled line coordinate sequence storage means for temporarily storing a coordinate sequence of points (hereinafter referred to as a ruled line coordinate sequence).

At this time, the hard disk 4 uses the ruled line frame coordinate sequence stored in the ruled line frame coordinate sequence storage area 3b of the RAM 3, the registered name and the attributes and keywords of the registered region delimited by the ruled line frame. A format structure registration area 4a as a format content registration means for registering input information given from the outside by a person and a format registration number storage means for storing the number of formats registered in the format structure registration area 4a. A certain number of format registration number storage area 4b and the recognition result /
An image storage area 4c is provided.

By the way, each of the various means constituting the optical character recognition device is as follows.
That is, the registration / recognition mode switching means 1 is a block for the user to determine whether to use the optical character recognition device in the input document registration mode or in the recognition mode, and specifically, a changeover switch or the like. Is. The photoelectric conversion unit 2 optically reads the input document and digitizes it, and then sends the digitized image of the input document to the document image storage area 3a of the RAM 3 to store the image. Reference numeral 5 is a block for extracting and reading a ruled line frame from the image of the input document stored in the document image storage area 3a, and storing the read ruled line frame coordinate sequence in the ruled line frame coordinate sequence storage region 3b for storage. .

The format setting means 6 is a RAM.
The ruled line frame coordinate sequence stored in the ruled line frame coordinate sequence storage region 3b of No. 3 and the input information externally given by the user such as the registered name and the attributes of the registered region delimited by the ruled line frame, and the keyword of the hard disk 4. It is a block that is registered and stored in the format structure registration area 4a, and is to be operated only in the registration mode. Then, the format setting means 6 determines whether or not the registered name given to the format of the input document corresponds to the registered name of the registered format, and only the format of the registered name that does not correspond to the registered name of the registered format. Function for registering the format structure registration area 4a with the keyboard 1
5 and the function of registering input information instructed from the outside by using the mouse 16 and determining whether or not a keyword for distinguishing the input document format from the registered format is necessary. It has the function of instructing the outside to input the necessary keywords.

Further, the format designating means 7 is a block for a user to externally designate and designate which of the formats registered in the format structure registration area 4a the format of the input document corresponds to. On the other hand, the format search / judgment unit 8 searches for a corresponding registered format when the format of the input document is directly specified, and automatically selects which of the registered formats when not specified. The format designating unit 7 and the format search / determination unit 8 operate only in the recognition mode. The format search / determination unit 8 searches for registered formats based on the fact that the ruled line frame coordinate sequence stored in the ruled line frame coordinate sequence storage area 3b is the same. If the coordinate sequence cannot be determined, the character recognition means 9 is operated and the determination based on the keyword is also executed. Then, the format search / determination unit 8
A factor that makes it impossible to determine which of the registered formats the input document format corresponds to, for example, a function of instructing the outside with a display on the display 14 of a keyword not input, and an input document format. Has a function of causing the outside to instruct the registered name of the corresponding registered format.

On the other hand, the character recognition means 9 is a block for executing character recognition, and the format search / determination unit 8
The result of the character recognition executed by the character recognition means 9 based on the instruction from is processed by the character recognition result processing means 10, and then stored and saved in the recognition result / image storage area 4c of the hard disk 4. Has become. The character recognition result at this time is to be displayed on the display 14, and the keyboard 15 and the mouse 1
The correction of the character recognition result by the user using 6 is also processed by the character recognition result processing means 10. The recognition mode selection means 11 is a block for designating an execution procedure of the character recognition in the character recognition means 9 based on the attribute of the registration area designated from the outside, and executes the high-speed recognition based on the input information. Alternatively, it is used when the user selects whether or not the recognition result check means 12 should perform whether or not the character recognition result corresponding to the input information is obtained after executing the recognition at the normal speed. .

Furthermore, this recognition result checking means 1
2 checks the relationship between the attribute of the registration area designated from the outside and the character recognition result, and if the character recognition result is inappropriate, displays on the display 14 that the character recognition result is inappropriate. It has a function to give instructions to the outside. The recognition / word dictionary 13 is a block used when the character recognition unit 9 executes character recognition, and the character recognition unit 9, the format search / determination unit 8, the character recognition result processing unit 10, and the recognition mode selection unit described above are used. 1
1, the recognition result checking means 12, and the recognition / word dictionary 13 are to operate only in the recognition mode. The display 14 displays information necessary for the user to work, for example, information regarding images and ruled line frames, and is used by the format setting means 6 and the character recognition result processing means 10. Has become. The keyboard 15 and mouse 16 are used when the user inputs an instruction or information from the outside, and are used in the format setting means 5 and the character recognition result processing means 10.

2A to 2D are explanatory views showing a format example of an input document registered and recognized by the optical character recognition apparatus according to the present embodiment, and as the input document at this time, A driver's license is common. First, the areas in which the characters “name”, “date of birth”, and “address” are described in these format examples are areas that have been printed in advance,
The area with the text "(photo)" is the area where the photo is pasted, while the area on the right side of the area with "name", "date of birth", and "address" is described. The generated character string is different for each of the four format examples.

The images of the respective input documents after photoelectric conversion are stored near the upper left corner in the document image storage area 3a of the RAM 3, as shown in FIG.
Coordinate points such as (0, 0) entered therein indicate coordinate positions when the upper left corner position in the document image storage area 3a is the origin. 2A and 2D, the coordinate points are slightly different, but such a difference in coordinate points is merely an error due to digitization, and these two input documents have the same format. Have Therefore, in this case, the input documents shown in FIGS. 2A and 2D have the same format, and the input documents shown in FIGS. 2B and 2C respectively have the same format as FIGS. It will have a different format than d).

FIG. 3 is a flow chart showing the operation procedure in the registration mode of the optical character recognition device according to the present embodiment. The operation procedure in the registration mode will be described below with reference to FIG.

First, the user sets the registration mode by using the registration / recognition mode switching means 1 (step 3-1), and the registration input is made to the photoelectric conversion means 2. After the document is digitized by binarization, the digitized image of the input document is stored in the document image storage area 3a of the RAM 3 (step 3-2). Next, the ruled line frame means 5 executes reading the ruled line frame from the image of the input document stored in the document image storage area 3a. Note that, here, the Hough transform is used in obtaining a straight line in image processing, and the ruled line frame is calculated by obtaining the logical sum of the ruled lines in both directions after obtaining the ruled lines in the vertical and horizontal directions.

Then, the read ruled line frame is represented as a ruled line frame coordinate sequence and stored in the ruled line frame coordinate sequence storage area 3b of the RAM 3 (step 3-3). That is, the order of the ruled line frame coordinate sequence at this time is the coordinate on the upper side and the coordinate on the left side in the same row, and the ruled line frame coordinate sequence in the document image in FIG. 2A is (0,
0), (30,0), (50,0), (120,0), (30,1
0), (50,10), (120,10), (30,20), (50,2)
0), (120,20), (0,30), (30,30), (50,3
It is represented as a ruled line frame coordinate sequence of length 14 such as 0) and (120, 30).

Subsequently, after the user sets the format of the input document by using the format setting means 6, the set format is registered in the format structure registration area 4a of the hard disk 4 ( Step 3-4). That is, the format here means a structure of information added to the input document, and together with the ruled line frame coordinate sequence obtained in step 3-3, the registration name or the registration separated by the ruled line frame. It contains external input information such as area attributes and keywords. The detailed procedure of the format setting operation will be described later with reference to FIG. Then, after the format setting of the input document is completed, it is judged whether or not there is an input document to be further registered (step 3-5), and if there is an input document to be registered, the step 3 is carried out. The process returns to step 2, and if there is no input document to be registered, the process according to the registration mode ends.

FIG. 4 is a flow chart showing the operation procedure when setting the format of the input document. In the format setting of the input document (step 3-4) in the registration mode of the optical character recognition device, the following procedure is followed. The action is taken.

First, a user sets a registered name (hereinafter also referred to as a label) of the format (step 4).
-1). The registered name or label here is a display indicating what document the input document is, such as "driver's license" or "XX university student card". Then, the format setting means 5 rejects the registration of the label which overlaps with the label of the registered format, and the label designated from the outside by using the keyboard 15 or the mouse 16 is stored in the format structure registration area 4a of the hard disk 4. If it overlaps with the label of the registered format, it is displayed on the display 14 to prompt the user to input a different label. Also, a list of labels in registered formats is displayed on the display 14 so that the user can easily input the labels.

Next, the correction / registration of the ruled line frame coordinate sequence will be executed. Here, the ruled line frame coordinate sequence temporarily stored in the ruled line frame coordinate sequence storage area 3b of the RAM 3 is used while utilizing the ruled line frame coordinate sequence. Create an image and display 14
While visually checking the ruled line frame image displayed above, the user corrects the ruled line frame coordinate sequence using the keyboard 15 or the mouse 16, and the corrected ruled line frame coordinate sequence is stored in the format structure registration area 4a of the hard disk 4. The registration is performed for that (step 4-2). Subsequently, the attribute of the registration area is set / registered for each of the fields represented as the registration area, which is a rectangular small area divided by the ruled line frame, and the ruled line frame coordinate sequence storage area 3b of the RAM 3 is set. The user executes the attribute registration for each field by using the coordinates and the number of fields extracted from the ruled line coordinate sequence temporarily stored therein.

As the flow of processing at this time, first, the field number counter CN is initialized (CN =
After 0) (step 4-3), attribute setting / registration is repeatedly performed for the field of the number AN of fields (step 4-4 to step 4-6). Note that the fields for which attributes are set / registered are in no particular order, and the coordinates of the fields at this time are the upper left coordinates and the lower right coordinates that are automatically registered, and the lower left coordinates and the upper right coordinates are the upper left coordinates and the lower right coordinates. After being calculated from
The user prompted to register the attribute of the current field by the display on the display 14 sets the attribute using the keyboard 15 or the mouse 16. By the way, the attributes of a field are a character registration area, an image registration area,
It means the standard to show the difference from other registration areas. For the character registration area, "address", "name",
Detailed attributes such as "date of birth" are set.

Further, when the field number counter CN becomes equal to or more than the field number AN, it means that the setting / registration of the attribute for each field is completed, and it is selected whether or not to register the keyword (step 4-7). When the user selects the keyword registration, the keyword registration is subsequently executed (step 4-8). The keyword is set so that even an image of an input document having a similar ruled line frame can be distinguished by recognizing one field. That is, the second
When the input document shown in FIG. 2A and the input document shown in FIG. 2C are registered, both have the same format of ruled line frame coordinates, and the determination cannot be made by this alone. Keywords are registered in order to avoid.

At the time of registering a keyword, a keyword name registration for registering a word and a keyword attribute registration for registering a keyword attribute are adopted.
In the keyword name registration, it is registered that the word "name" is described in the area between the upper left coordinates (30,0) and the lower right coordinates (50,10) in FIG. 2 (a). On the other hand, upper left coordinates (30,0) and lower right coordinates (50,10) in Fig. 2 (c).
It is registered that the word "birth date" is described in the area between and. Further, in keyword attribute registration, the upper left coordinates (50, 0) and the lower right coordinates (120,
The area between the upper left coordinates (50,0) and the lower right coordinates (120,10) in Fig. 2 (c) is registered as the "year of birth". It is registered as an attribute of "month day". Further, after the keyword registration is executed or when it is selected not to register the keyword in step 4-7, the process relating to the format setting of the input document is finished.

FIG. 5 is an explanatory diagram illustrating the format of the input document registered in the format structure registration area 4a of the hard disk 4. This format is set according to the operation procedure of the flowchart shown in FIG. 4 is an example of a format registered in the format structure registration area 4a in step 3-4 of the flowchart shown in FIG.

The format at this time is a structure of various information added to the input document, and the first 32
An area 5a for bytes is reserved for the label set in step 4-1 and an area 5b for the next 2 bytes is reserved for the length of the ruled line frame coordinate sequence registered in step 4-2. In addition, the area 5c for the ruled line frame coordinate sequence is secured for each 4 bytes by the length of the ruled line coordinate sequence. Then, in these areas 5c, the first 2 bytes are for the X coordinate and the remaining 2 bytes are for the Y coordinate. Subsequently, a 2-byte area 5d indicating the total number of fields is secured, and a numerical value indicating the total number of fields registered in either the character registration area or the image registration area is stored in this area 5d. Has been done.
At this time, fields that are not used for character registration or image registration are omitted.

Further, a plurality of areas 5e of 10 bytes each indicating the coordinates and attributes of each field set in steps 4-3 and 4-4 are secured by the number corresponding to the total number of fields. 2 bytes each are used for upper left coordinate (X coordinate), upper left coordinate (Y coordinate), lower right coordinate (X coordinate), lower right coordinate (Y coordinate), and field attribute. The 2 bytes for the field attribute indicate whether it is a character registration area or an image registration area, and if it is a character registration area, it is "name" or "address". Numerical values will be stored. Subsequently, a 2-byte area 5f indicating the presence / absence of keyword registration is secured, and when there is keyword registration, a 2-byte area 5g indicating a keyword attribute and 3 indicating a keyword name.
An area 5h for 2 bytes is secured respectively, and an area 5 for each 2 bytes showing the coordinates of the keyword.
i is secured.

In this case, each of 2 bytes in the area 5i is for upper left coordinate (X coordinate), upper left coordinate (Y coordinate), lower right coordinate (X coordinate), lower right coordinate (Y
Coordinates) and field attributes. It should be noted that while a numerical value of 0 is stored in the area 5f when there is no keyword registration, a numerical value of 1 is stored in the area 5f when there is keyword registration. Numerical values indicating keyword attributes are stored in the area 5g. Further, a meaningless numerical value is stored in the area 5g indicating the keyword attribute when the keyword name is registered, and the keyword name is stored in the area 5h. Therefore, FIG.
The registered name in the format example of the input document shown in (a) is "XX University Student ID Card", and the upper left coordinates (30,0) and lower right coordinates (5
If the area written as “name” between 0 and 10) is registered as keyword name = “name”, the following format will be obtained.

==== Registered Name (32byte) ○○ University Student ID Coordinate Sequence Length (2byte) 14 Coordinates (4 × 14byte) (0,0), (30,0), (50,0), ( 120,0), (30,10) (50,10), (120,10), (30,20), (50,20), (120,20) (0,30), (30,30) , (50,30), (120,30) Number of registrations (2byte) 4 1st (10byte) (0,0), (30,30) ... Numerical value representing image registration 2nd (10byte) (50,0) , (120,10)… 3rd numerical value representing “name” (10byte) (50,10), (120,20)… 4th numerical value representing “date of birth” (10byte) (50,20), ( 120, 30)… Presence / absence of numeric keyword indicating “address” (2byte) 1 attribute (2byte) Numeric keyword name not assigned to attribute (32byte) Name coordinate (4 × 2byte) (30,0), (50,10) ) ==== FIG. 6 shows optical character recognition according to the present embodiment. A flow chart showing an operation procedure in the recognition mode of location, the following, an operation procedure at the time of recognition mode will be described with reference to FIG.

First, after the user has set the recognition mode using the registration / recognition mode switching means 1 (step 6-1), when the user specifies the format, the keyboard 15 or The registered format of the label designated from the outside by using the mouse 16 is directly set by the format designation means 7 (step 6-2). And the recognition mode selection means 1
When the user selects the recognition mode with 1,
A high-speed mode that executes high-speed recognition based on input information,
A check mode for checking whether or not a character input result corresponding to the input information is obtained after performing recognition at the normal speed is selected based on an instruction from the outside (step 6-3). After that, the registration document input to the photoelectric conversion unit 2 is binarized and digitized, and the image of the digitized input document is stored in the document image storage area 3a of the RAM 3.
(Step 6-4).

Next, the ruled line frame is read from the image of the input document stored in the document image storage area 3a, and the read ruled line frame is converted into a ruled line frame coordinate sequence and then R
It is stored in the ruled line frame coordinate sequence storage area 3b of AM3 (step 6-5). Then, it is determined whether or not the format is designated (step 6-6),
If the format is designated in step 6-2, the format search / determination means 8 searches for and identifies the designated registered format, and then sends the designated registered format to the character recognition means 9. Is performed (steps 6-7). Further, the format search / determination means 8 when the format is not specified is
A determination is made after searching which of the registered formats corresponds to the format temporarily stored in the ruled line frame coordinate sequence storage area 3b (step 6-8). It should be noted that when the format determination based on only the ruled line frame coordinate sequence is not possible, the keyword recognition is also executed, but a detailed description of the format determination operation will be given later with reference to FIGS. 7 and 8.

Then, after it is determined in step 6-9 whether or not the format determination is successful, if the determination is unsuccessful, the format is redesignated and the process returns to step 6-7. The user determines whether to continue the character recognition or to end the processing in the recognition mode (step 6-10). If the format determination is successful, the character based on the determined or designated format is used as in the case where the format is directly designated based on the instruction from the outside (step 6-7). The recognition will be executed by the character recognition means 9. Prior to executing the character recognition, whether the high speed mode or the check mode is in step 6-
The operation differs depending on whether it is selected in step 3 (step 6-11). In the high speed mode, character recognition based on the determined or designated format is executed by the character recognition means 9 (step 6-12). When the check mode is selected, the character recognition means 9 performs character recognition at normal speed, and
The recognition result checking means 12 checks whether or not a character recognition result corresponding to the input information has been obtained (step 6-13).

That is, in the character recognition at this time,
After calculating the registration area of the input image corresponding to the area that requires character recognition in the determined or specified format, character recognition is performed in the registration chain area of this input image. Is a character recognition according to the attribute of the registration area, for example, in the field having the attribute of "name", the recognition / character recognition without using the arithmetic digits or symbols of the word dictionary 13 is executed, while in the check mode. Is character recognition unrelated to the attributes of the registration area, for example,
Even in the field having the attribute of "name", the character recognition is executed while using all the dictionaries of the recognition / word dictionary 13 including the numerals and symbols. And
In the check mode, the character recognition result and the field attribute are subsequently transferred to the recognition result checking means 12, and the recognition result checking means 12 causes a contradiction between the character recognition result and the field attribute. Checking is done. Note that, in this check, for example, it is determined that the character recognition result cannot be trusted if the numeral recognition character or the symbol exists in the character recognition result of the field having the attribute of "name". If it is determined that the character recognition result is untrustworthy, the fact that the character recognition result is untrustworthy is transferred to the character recognition result processing means 10 together with the character recognition result.

Further, subsequently, the character recognition result processing means 10 processes the character recognition result, and the user recognizes the character recognition result displayed on the display 14 while viewing the character recognition result. Correction and the like are executed using the keyboard 15 and the mouse 16 (step 6-14). In the field where it is determined that the character recognition result is unreliable, the user's attention is called by the reverse display and the corrected character recognition result is stored in the recognition result / image storage area 4c of the hard disk 4. Saved. By the way, when the attribute of the field at this time is the image registration area, the image in the image registration area is based on the image registration area calculated from the specified or determined registered format, and the image in the image registration area is the recognition result / image. It will be saved by the storage area 4c. Then, when the recognition process as described above is completed, it is determined whether or not there is an input document to be recognized (step 6-1).
5) If there is an input document to be recognized, the process returns to step 6-4, and if not, the process according to the recognition mode ends.

FIG. 7 is a flow chart showing the operation procedure at the time of format judgment. In the format judgment (step 6-8) in the recognition mode of the optical character recognition device, the operation according to the following procedure is executed.

First, initialization is performed by setting the format counter CN to 0 and setting the distance values FD1 and FD2 to infinity (step 7-1). In this case, it is not necessary to set the initial values of FN1 and FN2 indicating the number of registered formats and the number of registered formats, and the number of registered formats stored in the format registration number storage area 4b. Is 1 or more (FN ≧
1). Next, the distance dis (CN) between the ruled line frame coordinate sequence temporarily stored in the ruled line frame coordinate sequence storage area 3b of the RAM 3 and the ruled line frame coordinate sequence of the CNth format.
Is calculated (step 7-2). By the way, as a method of calculating the distance in this case, DP-matching (dynamic programming) for obtaining the distance of time-series data is used, and the distance between each coordinate is expressed by using the absolute value difference distance. At the same time, the error of several dots is absorbed. Therefore, when applied to the input document of FIG.
The ruled line frame coordinate sequences of the format examples shown in (a), (c), and (d) are all the same and the distance is 0, while
Unlike the others, only the ruled line frame coordinate sequence of the ruled line frame coordinate sequence of the format example shown in FIG. 2B is different from the others, and the distance is not zero. The numerical value of the distance depends on how the DP-matching is configured.

Subsequently, FD1 is compared with dis (CN) (step 7-3). If FD1 <dis (CN), F
D2 = FD1 and FN2 = FN1 are set, and FD1 and FN1 are newly set to FD1 = dis (CN) and FN1 = CN (step 7-4). Also, FD1 <dis
If it is not (CN), it is judged whether or not dis (CN) <FD2 (step 7-5), and dis (C).
If N) <FD2, then FD2 = dis (CN), FN2 = C
It is reset to N (step 7-6). Next, the value of the format counter CN is incremented by 1 (step 7-7), it is determined whether CN ≧ AN (step 7-8), and steps 7-2 to 7 are repeated until CN ≧ AN. After repeating 7-7, CN
If it is determined that ≧ AN, FD1 ≠ FD2
Is determined (step 7-9). When FD1 ≠ FD2, FD1 is the minimum value and FD2 is the second smallest value, so it is determined that the format of the input document is the closest to the FD1 registered format. As a result, the value of FN1 as the number of the registered format is transmitted to the character recognizing means 9, and the determination of the format is successful, and the format determination process is terminated (step 7-12).

On the other hand, when FD1 = FD2 is determined in step 7-9, the distance between the FD1th registered format and the FD2th registered format is equal, and the format of the input document is determined only by the ruled line frame coordinate sequence. Since it is not possible to specify the registered format that most closely resembles the above, the determination is performed using the keyword. Therefore, subsequently, it is determined using a keyword, that is, which of the FD1st and FD2th registered formats is close to the format of the input document is determined using the keyword ( In step 7-10), it is determined whether or not the determination using the keyword is successful (step 7-11). If the determination using the keyword is successful, the number of the specified registered format is used as the character recognizing means 9 Then, the format determination process is terminated (step 7-12).

If it is determined in step 7-11 that the determination using the keyword has failed, the fact that the determination using the keyword has failed is displayed on the display 14 and the format determination processing ends (( Steps 7-13). The operation at the time of keyword determination will be described later with reference to FIG. By the way, in the present embodiment, the keyword determination is executed when FD1 = FD2. However, the required threshold value TH is set in advance, and the value of FD1 is smaller than the value obtained by subtracting the threshold value TH from the value of FD2. It is also conceivable to execute the keyword determination only when is large (FD1> FD2-TH). Further, in the present embodiment, the approximate format is obtained from the two registered formats that form the minimum distance from the ruled line frame coordinate sequence in the format of the input document, but the approximate format is obtained from the more registered formats. It is also possible to make an extension that asks for something.

FIG. 8 is a flow chart showing the operation procedure at the time of keyword determination, and this keyword determination is adopted when the format determination cannot be performed only by the ruled line frame coordinate sequence.

First, it is determined whether or not keywords have been registered for both the FN first format and the FN second format (step 8-1). If the keyword is not registered in at least one of the formats, the number of the format in which the keyword is not registered, the registered name, etc. are displayed on the display 14, and the keyword determination fails for the user. Notify that (step 8-5). If keywords are registered in both formats, the character recognition means 9 executes character recognition for the image in the document image storage area 3a surrounded by the coordinates of the keywords, and
The character recognition results for both formats are sent to the format search / determination means 9 (step 8-2). Next, based on both recognition results, it is determined whether only one of them is the correct answer (step 8-3). Note that in this determination, when the keyword registration performed in step 4-8 of FIG. 4 is executed with the keyword registration name, the keyword registration name and the character recognition result are the same, and thus the correct answer is obtained, and the keyword attribute is set. When the registration is being performed, the recognition result check means 12 checks whether or not the character recognition result is appropriate, and if the result is correct, it is determined as a correct answer.

If only one of them is correct, the judgment is successful, the number of the correct registered format is transmitted to the sentence recognition means 9 and the process is terminated (step 8-4). . If neither is the correct answer, a message such as “wrong answer of registered format of either (FN1 and FN2)” is displayed on the display 14 to inform the user that the keyword determination has failed. Notify (step 8-5). Although only one keyword is registered in this embodiment, it goes without saying that a plurality of keywords may be registered, and the keyword registration name and character in at least one of the registered formats may be registered. If the condition that the recognition result is equal is too strict, it is also possible to make the determination based on the condition that the character recognition result is closer to the keyword registration name.

By the way, a specific example of the registration and recognition of the input document described above is shown in FIG. 2 as follows. First, when FIGS. 2A and 2B are registered, even if no keyword is registered, FIG.
It is determined that (d) is the same as FIG. 2 (a). However, when FIGS. 2A and 2C are registered without keyword registration, the determination based on the ruled line coordinate sequence at the time of recognition is performed as shown in FIG.
2D will be determined to be the same as FIGS. 2A and 2C, and since the keyword is not registered, a message indicating that the keyword is required is displayed on the display 14. Become. 2 (a) and 2 (c), in FIG. 2 (a), the word "name" is written in the area of the upper left coordinates (30,0) and the lower right coordinates (50,10). Registered, and in Fig. 2 (c), the upper left coordinate (30,0)
If you register that the word "date of birth" is written in the area of the lower right coordinates (50,10),
(D) is determined to have the same format as that of FIG. 2 (a) by the keyword determination.

Next, a modified example in which the optical character recognition device according to the present embodiment is developed will be described with reference to FIG.
That is, in the optical character recognition device described above,
A keyword is used to determine the format of an input document having a similar ruled line frame, and it was notified that the keyword registration was inadequate at the time of recognition. It is characterized by having a notification function. Since the configuration of the optical character recognition device according to this modification is basically the same as that of the embodiment, detailed description thereof will be omitted here.

FIG. 9 is a flow chart showing the operation procedure in the registration mode of the optical character recognition apparatus according to the modification. The procedure for registering the format of the input document (one sheet) will be described below. Note that the symbol CN used in FIG. 9 is a counter, and AN indicates the number of registered registered formats stored in the format registration number storage area 4b of the hard disk 4.

First, the user externally sets the format for the currently registered input document through the format setting means 6 (step 9-1). The operation procedure of the format setting is exactly the same as that of the embodiment. Therefore, the set format will be temporarily registered in the format structure registration area 4a, but the number of registered formats registered in the format registration number storage area 4b will not increase. At this time, the format search / determination means 8 distinguishes between the newly temporarily registered format and other formats. Next, after initialization of the counter CN = 0 is executed (step 9-2) and it is judged whether or not CN ≧ AN (step 9-3), if CN ≧ AN is not satisfied. It is determined whether or not the currently registered format is different from the CNth format (step 9-
4). Note that the determination at this time is whether the ruled line frame coordinate sequence is different, that is, the distance by DP-matching is 0.
Or if the distance is 0, it is supposed to be based on whether the keywords are different,
The detailed procedure is the same as in the embodiment.

When it is determined in step 9-4 that the formats are different, the value of the counter CN is incremented by 1 (step 9-5) and compared with the next registered format. Will be repeatedly executed, and if a difference between the currently registered format and the CNth format cannot be found, step 9-
After proceeding to step 1, the fact that no difference can be found is displayed on the display 14 as in the embodiment. Therefore, the user will again set the format for both the new format and the CNth format, or for either one. Furthermore, if it is determined that the format is different from all the registered formats, the steps 9-3 to 9
As the process proceeds to -6, the format being temporarily registered is actually registered. As a result, the number of registered formats also increases by 1, and the registration process ends.

[0055]

As described above, according to the first aspect of the present invention.
According to the optical character recognizing device of the present invention, since the sample document is not required and the structure of the input document is not required to be understood, even if the input document in which the characters indicating the information regarding the personal attributes are written is registered. And the recognition is possible, the configuration is greatly simplified, and the processing time is shortened.
Yes, moreover, the judgment cannot be made only with the ruled line coordinate sequence.
When making a decision, it is possible to make a reliable judgment using keywords.
Become.

Further, according to the configuration of claim 6 , the factor indicating that the corresponding format cannot be determined is indicated to the user, and according to the configuration of claim 8 , the registration is performed by the user's designation. A corresponding format can be selected from the already-used formats.

[0057]

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an optical character recognition device according to an embodiment.

FIG. 2 is an explanatory diagram showing a format example of an input document registered and recognized in the optical character recognition device according to the present embodiment.

FIG. 3 is a flowchart showing an operation procedure in the registration mode of the optical character recognition device according to the present embodiment.

FIG. 4 is a flowchart showing an operation procedure when setting a format.

FIG. 5 is an explanatory diagram illustrating a format of an input document registered in a format structure registration area.

FIG. 6 is a flowchart showing an operation procedure in the recognition mode of the optical character recognition device according to the present embodiment.

FIG. 7 is a flowchart showing an operation procedure at the time of format determination.

FIG. 8 is a flowchart showing an operation procedure at the time of keyword determination.

FIG. 9 is a flowchart showing an operation procedure in the registration mode of the optical character recognition device in the modified example.

[Explanation of symbols]

2 photoelectric conversion means 3a document image storage area (image storage means) 3b ruled line frame coordinate sequence storage area (ruled frame coordinate sequence storage means) 4a format structure registration area (format content registration means) 4b format registration number storage area (format registration Number storage means 5 Ruled line frame reading means (ruled line frame extraction means) 6 Format setting means 7 Format designating means 8 Format search / determination means 9 Character recognition means 10 Character recognition result processing means 11 Recognition mode selection means (Character recognition procedure designating means) ) 12 Recognition result check means

Continuation of the front page (56) Reference JP-A-7-152859 (JP, A) JP-A-7-114616 (JP, A) JP-A-3-268084 (JP, A) JP-A-7-282193 (JP , A) JP 61-54575 (JP, A) JP 63-155386 (JP, A) JP 4-23185 (JP, A) JP 8-221510 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) G06K 9/00-9/82

Claims (9)

(57) [Claims] 1. A photoelectric conversion means for digitizing an input document,
An image storage unit that stores an electronic image of an input document, a ruled line frame extraction unit that extracts a ruled line frame from the image of the input document, and a ruled line frame coordinate sequence that stores a coordinate sequence of grid points of the extracted ruled line frame Format setting means for setting the format of the input document represented by the storage means, the coordinate sequence of the grid points of the ruled line frame, the registered name, the attributes of the registration area delimited by the ruled line frame, and external input information such as keywords And an optical character including a format search / determination means for searching and determining whether the format of the input document corresponds to any of the registered formats and a format registration number storage means for storing the number of registered formats. In the case of the recognition device and in the registration mode, the format search / determination means is
Format the format set by the setting means
The primary registration is performed in the body registration area, and the primary registration is performed.
Format and the registered formats are sequentially compared, and the format is
If it cannot be determined, the format setting means
In addition, the key word between the similar format and the input format
For all formats that have been registered for comparison
Repeat for all registered formats.
If the format is different, enter the input document format.
Optical character recognition device characterized by recording . 2. The optical character recognition device according to claim 1 , wherein the keywords of the similar format and the input format are used.
If there is no difference, the format setting means has no difference.
Message is displayed and either format is reset.
An optical character recognition device characterized by performing . 3. A photoelectric conversion means for digitizing an input document,
Image storage means for storing an electronic image of an input document
And a ruled line frame extractor that extracts the ruled line frame from the image of the input document.
A rule that stores the columns and the coordinate sequence of the extracted grid points of the ruled line frame
Line frame coordinate sequence storage means and registration of coordinate sequence of grid points of ruled line frame
Names, attributes of registered areas separated by ruled lines, keywords, etc.
Input document format represented by which external input information
Formatting means to set matte and input document
The format is one of the registered formats
Format search / judgment
Means and a former that stores the number of registered formats
It is an optical character recognition device equipped with
In the recognition mode, the format search / determination means is
Format the format set by the setting means
The primary registration is performed in the body registration area, and the primary registration is performed.
Format and the registered formats are sequentially compared, and the format is
If the judgment is impossible, the format setting means,
Furthermore, the key word of the similar format and the input format
For all formats that have been registered for comparison
Repeatedly, any of the registered formats
If it succeeds in determining whether or not
An optical character recognition device characterized by recognizing a region to be marked . 4. A photoelectric conversion means for digitizing an input document,
Image storage means for storing an electronic image of an input document
And a ruled line frame extractor that extracts the ruled line frame from the image of the input document.
A rule that stores the columns and the coordinate sequence of the extracted grid points of the ruled line frame
Line frame coordinate sequence storage means and registration of coordinate sequence of grid points of ruled line frame
Names, attributes of registered areas separated by ruled lines, keywords, etc.
Input document format represented by which external input information
Formatting means to set matte and input document
The format is one of the registered formats
Format search / judgment
Means and a former that stores the number of registered formats
It is an optical character recognition device equipped with
In the recognition mode, the format search / determination means sets the format setting
The format set by
Primary registration in the recording area, and the format in which this primary registration was made
And the registered formats are sequentially compared, and the format
If it cannot be determined, at least one of the formers
If the keyword is not registered in the
Display information about formats that are not registered
And keywords are registered in both formats
In this case, character recognition is executed and the keyword registration name and character
An optical character recognition device characterized by comparing recognition results . 5. A photoelectric conversion means for digitizing an input document,
Image storage means for storing an electronic image of an input document
And a ruled line frame extractor that extracts the ruled line frame from the image of the input document.
A rule that stores the columns and the coordinate sequence of the extracted grid points of the ruled line frame
Line frame coordinate sequence storage means and registration of coordinate sequence of grid points of ruled line frame
Names, attributes of registered areas separated by ruled lines, keywords, etc.
Input document format represented by which external input information
Formatting means to set matte and input document
The format is one of the registered formats
Format search / judgment
Means and a former that stores the number of registered formats
It is an optical character recognition device equipped with
Of the registered area attributes and character recognition results
Check the relationship and confirm that the character recognition result is incorrect.
Recognition result checking means to instruct the outside and input information
Performs fast, information-based recognition, or
A sentence corresponding to the input information after performing recognition at normal speed
Whether the character recognition result is obtained or not
The recognition mode selection means by which the user selects whether to perform
An optical character recognition device that is equipped with . 6. The optical character recognition device according to claim 1, wherein the format search / determination means is a format of the input document.
The registered format corresponds to one of the registered formats
It has a function to instruct the outside of the reason why
Optical character recognition apparatus characterized by being. 7. The method according to any one of claims 1 to 6.
In the optical character recognition device described above, the format search / determination means is a format of the input document.
The registered name of the registered format corresponding to
Optics characterized by having the function of instructing
Character recognizer. 8. The method according to any one of claims 3 to 7.
The optical character recognition device installed in the
This is a function to specify whether this applies based on instructions from the outside.
Format specification method and registered registered format
Format search that directs character recognition based on text
/ Optical character recognition characterized by having a determination means
Intelligence device. 9. The optical character recognition device according to claim 8.
The format search / determination means is a format designation means.
Externally instruct the registered name of the format specified in
Optical character recognition characterized by having the function of
apparatus.