JPH11282957A - Method for deciding recognition object area - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a new method capable of deciding a specified cell area on the image data of a slip whether it is a fixed form document or a fixed format document in which the position of the cell is shifted by printing slippage or the like. SOLUTION: In deciding one or plural specified cell areas to be a recognition object area from the image data of a slip provided with many cells divided by ruled lines, information for deciding the specified cell area of the slip is stored in a dictionary 21 beforehand. Then, the ruled line is detected for the image data of the slip, a table area turned to a closed area by the detected ruled line is specified and the specified cell area is decided, based on the information for deciding the specified cell area from the area divided by the ruled line in the table area.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for determining a recognition target area suitable for use in a character recognition device or the like.

[0002]

2. Description of the Related Art In the field of character recognition, the recognition of characters written on a tabular form is also performed. Typical examples of the tabular form include, for example, the following two types of forms. The first is
A form (hereinafter referred to as a fixed form form) in which the dimensions of the cells divided by vertical and horizontal ruled lines and the position of each cell with respect to the reference point of the form (for example, the upper left point of the image) are fixed as predetermined. . Second, although the dimensions of each cell and the absolute position of each cell with respect to the form reference point are not fixed, a form (hereinafter referred to as a form) in which the logical positional relationship of each cell, such as the order in which the cells are arranged, is predetermined. , Style fixed form).
Specifically, there is a △△ item column below the XX item column,
In addition, the form has a logical positional relationship as predetermined such that there is an entry column on the right side of each of these item columns.

In these tabular forms, one or a plurality of specific cell areas in the form are often used as character entry fields. Therefore, in order to perform character recognition, it is necessary to detect one or a plurality of specific cell areas from a form.

[0004] As a conventional method for detecting a specific cell area from a fixed format form, for example, reference 1 (“Overview of character recognition”,
Edited by Shinichiro Hashimoto, March 20, 1982, Ohmsha,
P203-206).

In this method, characters are cut out from image data of a form using read control data for various fixed forms, and character recognition is performed. Specifically, the position and the like of a specific cell from the upper left point of the form are prepared as read control data (204th document of Document 1).
Page), a specific cell area is specified based on this, and characters are cut out from the specified cell area to perform character recognition. One or more cell areas identified as above,
Since it is an area to be subjected to character recognition in subsequent processing, it is hereinafter referred to as a reading field. Note that, even when the area is not surrounded by a ruled line other than the cell, the area to be subjected to character recognition is called a read field.

However, in the above-described conventional character recognition method, for example, characters on a fixed form may not be read. As described above, the form fixed form only defines the logical positional relationship of each cell, but does not define the dimensions and the absolute positional relationship of each cell. Therefore, although the style is the same, the dimensions of each cell, the absolute position of the cell, the distance between the cells, and the like may be freely prepared. In such a case, the above-described conventional character recognition method is used. Even if the read control data is prepared, a specific cell which is a target area for character recognition cannot be specified.

Further, in the above-described conventional character recognition method, even in a fixed format form, if the absolute position of a cell is shifted due to, for example, a printing shift of a ruled line or a character, the character can be read. Sometimes it is impossible. However, in order to correct a printing shift or the like, there is a method in which a line mark is provided at one end of a line where a specific cell to be read is located, and the line to be read is specified by this mark (see Document 1). Page 205). However, in that case, an area for forming a line mark is required. Also, a form without a line mark cannot specify a specific cell to be read.

Therefore, even if the form is fixed form,
A new method is desired that can determine a specific cell area on image data of a form even in a fixed form where the cell position is shifted due to printing shift or the like.

[0009]

Means for Solving the Problems (1) In order to solve the above-mentioned problems, a first aspect of the present invention is to convert one or a plurality of recognition target areas from image data of a form including a large number of cells divided by ruled lines. In determining the specific cell area, the ruled line information or cell information of the form and the information for determining the specific recognition area of the form are stored in a dictionary in advance, and the ruled line is detected from the image data of the form, From the area divided by the ruled line in the table area, the specific cell area is determined, and further, based on information for determining the recognition area,
It is characterized in that a recognition area is determined.

(2) In the second invention, one or a plurality of specific cell areas to be recognized are determined from image data of one or a plurality of forms including a large number of cells divided by ruled lines. In doing so, a form in which a specific cell area has already been determined is used as a reference form, and for each table area closed by an intersecting ruled line in the form to be recognized,
Ruled line information or cell information and information for determining a specific cell area of the reference form are stored in the dictionary, and from image data of a form to be recognized, stored in the dictionary for each table area. The specific cell area is determined based on the ruled line information or cell information and information for determining a specific cell area of the reference form.

(3) The third invention is the first or second invention.
In the invention of the above, when there are a plurality of specific cell areas in the form to be recognized and there is an area to be integrated and recognized as the same recognition target, the ruled line information or the cell information and the reference form The information for determining the specific cell area and the information for determining the same recognition target area are stored in the dictionary, and the ruled line stored in the dictionary is obtained from image data of a form to be recognized. The specific cell area is determined based on information or cell information and information for determining a specific cell area of the reference form, and the identification is performed based on information for determining the same recognition target area. Are characterized by integrating the same recognition target areas from among the cell areas described above.

(4) Further, the fourth invention is the first or second invention.
In the invention, when only a part of the specific cell area included in the form to be recognized is the recognition target area, the ruled line information or the cell information and the specific cell area of the reference form are determined. And the internal area position information for determining the recognition target area in the specific cell area is stored in the dictionary, and the ruled line stored in the dictionary is obtained from image data of a form to be recognized. Based on information or cell information and information for determining a specific cell area of the reference form, determine the specific cell area, based on the internal area position information, included in the specific cell area It is characterized in that a recognition target area to be recognized is determined.

(5) In a fifth aspect based on the fourth aspect, the internal area position information for determining the recognition target area in the specific cell area is specified by the specific cell and the recognition target area in the reference form. Is determined by comparing the position coordinates of
The recognition target region is determined by comparing the positional relationship between the determined internal region position information stored in the dictionary and the specified cell information.

(6) The sixth aspect of the present invention is directed to the first or second aspect.
In the invention, when the recognition target area exists outside the table in the form to be subjected to the recognition processing, the ruled line information or the cell information and the out-of-line recognition area for determining the recognition target area outside the table. The position information is stored in the dictionary, and based on the ruled line information or cell information stored in the dictionary and the position information of the out-of-line recognition area, based on the image data of the form to be recognized, the position information exists outside the table. It is characterized in that a recognition target region to be performed is determined.

(7) In a seventh aspect based on the sixth aspect, the table area is determined based on the ruled line information in the reference form, and the position coordinates of the table area are compared with the position coordinates of the recognition target area. Accordingly, if the recognition target area is outside the table area, the position coordinates of the recognition target area are stored in the dictionary as position information of the out-of-line recognition area for determining the recognition target area outside the table area, and stored. The recognition target area is determined by searching an area around the position coordinates from the position information of the out-of-line recognition area thus obtained.

(8) The eighth aspect of the present invention is directed to the first or second aspect.
In the invention according to the above, when the division of the form to be recognized by the ruled line of the table area is divided by a rule other than the ruled line connected to both ends in the horizontal direction or both ends in the vertical direction, it indicates that the ruled line is divided. The intersection additional information is stored in the dictionary, and a ruled line is detected from image data of a form to be recognized based on the intersection additional information stored in the dictionary.

(9) In a ninth aspect based on the eighth aspect, if intersection additional information is added to the dictionary,
A ruled line detecting process for detecting ruled lines other than ruled lines connected to both ends in the horizontal direction or both ends in the vertical direction from image data of a form to be recognized is performed.

[0018]

Embodiments of the present invention will be described below.

(Configuration of First Embodiment) In the first embodiment, a case where a plurality of table areas are included in a form to be determined (hereinafter simply referred to as a form) which is a form to be recognized. And recognition processing can be performed even when only one or a plurality of specific cell regions in each table region are read fields. In the first embodiment,
The description will be made assuming that a specific cell area is a recognition target area as it is.

FIG. 1 shows a first embodiment of the present invention and is a diagram showing a configuration of an apparatus for performing a method of determining a recognition target area. FIG. 1 shows a character recognition device 10 including a recognition target region determination device 20 that performs the recognition target region determination method of this embodiment.

The character recognition device 10 includes a recognition target area determination device 20, character recognition means 40, a CPU 50, and a memory 60.

The recognition target area determination device 20 specifies one or a plurality of table areas including a large number of cells divided into ruled lines from image data of a form, and specifies one or a plurality of table areas serving as recognition target areas such as characters. Determine a specific cell area.

The recognition target area determination device 20 includes a dictionary 2
1, a table-specific processing control unit 23, a ruled line detection unit 25, a correspondence determination unit 27, and a cell determination unit 29.

The dictionary 21 includes ruled line information and cell information of a table area in a form in which a specific cell area has already been determined (hereinafter referred to as a reference form), and information for determining a specific cell area to be a recognition target area. Is stored. When there are a plurality of table areas in the reference form, the table-specific processing control means 23 controls the ruled line detection means 25, the correspondence determination means 27 and the cell determination means 29, and To detect a ruled line or determine the correspondence between the ruled line and the cell area.
Details of the operation of the table-based processing control means 23 will be described later. The ruled line detecting means 25 detects a ruled line from the image data of the form and stores the ruled line data in a ruled line data memory 61.
To be stored. The correspondence determination unit 27 detects the ruled line detected by the ruled line detection unit 25, compares the ruled line stored in the ruled line data memory 61 with the ruled line information or cell information stored in the dictionary 21, and determines whether the detected ruled line is correct. And determine
The area divided by each ruled line is specified. In addition, the cell area determination means 29 stores, in the dictionary 21, from the cell areas on the image data which are classified by the ruled lines determined by the correspondence determination means 27 to be correct among the ruled lines detected by the ruled line detection means 25. A specific cell area is determined based on the ruled line information or cell information, and information on the position of that area is stored in the recognition area data memory 63.

The character recognizing means 40 recognizes characters on a form. In this case, a form image stored in a form image memory 73, which will be described later, is determined based on the information on the position of the specific cell area determined by the recognition target area determination device 20 and stored in the recognition area data memory 63. Characters and the like are cut out from the data and characters are recognized. The character recognition device 40 can be configured by known means. For example, a character is cut out from the character string in the recognition target area, and the characteristics of the cut out character are extracted. Then, the feature of the extracted character to be recognized is collated with the feature of each character in a dictionary prepared in advance to determine a candidate character.

The CPU 50 is a recognition target area determination device 2
0 and character recognition means 30 are controlled.

The memory 60 is constituted by any suitable storage medium such as a hard disk.

The character recognition device 10 is connected to an image data input device 70 for reading and storing image data from a form to be recognized. In the figure, reference numeral 71 denotes a scanner for reading a form image, and reference numeral 73 denotes a form image memory for storing the image read by the scanner 71.

(Regarding the operation of the first embodiment)
Method for determining recognition target area and dictionary 21 associated therewith
Will be described with reference to the form shown in FIG. FIG. 2 is a diagram showing the configuration of the reference form.

First, ruled line information, cell information, and information for determining a recognition area stored in the dictionary 21 will be described. Note that both the ruled line information and the cell information are stored in the dictionary 2
1 may be stored, but since the position of the cell area can be specified from either one, it will be described below that only one of them is stored in the dictionary 21.

The ruled line information is information on the positions of ruled lines, such as the number of ruled lines and the arrangement relationship of each ruled line. The cell information is information on the position of a cell, such as the number of cells and the arrangement of each cell. The ruled line information and the cell information can be information indicating an absolute position as in the related art, but in the present invention, they are information indicating a relative position.

That is, each cell region divided by a ruled line connecting both ends in the vertical direction or both ends in the horizontal direction of a table region formed by intersecting ruled lines in a form is referred to as a divided region of the first hierarchy. Is called the first level ruled line,
Each cell area divided by a ruled line connecting both ends in the vertical direction or both ends in the horizontal direction in each divided area of the first hierarchy is referred to as a divided area of the second hierarchy. The cell information and the ruled line information can be expressed as information of an n-layer structure by similarly referring to the divided region of the nth layer and the ruled line of the nth layer.

For example, FIG. 2 shows two table areas 210.
An example of a reference form including “a” and “210b” is shown.
In FIG. 2, table areas 210a and 210b are defined by ruled lines 220a to 220a.
The inside is hierarchically divided by i. As shown in FIG. 3, the ruled line information or cell information of such a reference form is, as shown in FIG.
1 is stored. FIG. 3 is a diagram showing the structure of ruled line information or cell information stored in the dictionary 21. In FIG. 3, Table 1 is information for specifying the position of the table area 210a, Table 2 is information for specifying the position of the table area 210b, T1
T232 is ruled line information relating to the ruled lines 220a to 220i or cell information of the cell regions 230a to 230c and other cell regions.

The information for determining a recognition area is, for example, information indicating whether or not a cell is a specific cell area to be a recognition target area for characters or the like (hereinafter, information for determining a specific cell area). Information), internal area position information described later, and out-of-line recognition area position information. Information for determining these recognition areas is stored in the dictionary 21 in association with ruled line information or cell information. The information for determining the recognition area is transmitted to the device by the operator by inputting numerical values or the like from an input means such as a keyboard (not shown) with reference to the reference form or by causing the apparatus to read the reference form. Is entered.

Next, the operation of the table-specific processing control means 23 will be described with reference to the flowchart of FIG. FIG. 4 is a flowchart showing the operation of the table-specific processing control means 23.

First, initialization is performed in step S401. That is, the variable i, which is stored in the table-specific processing control means 23 and indicates the number of the table that has already undergone the recognition processing (hereinafter referred to as the processing table number), is set to 1.

Next, in step S402, the number of tables stored in the dictionary 21 in advance (hereinafter referred to as the number of tables stored in the dictionary) is compared with the variable i indicating the processing table number. If it is larger than the number, the process ends. If not, the process moves to step S403. In the case of the example of FIG. 2, the number of tables stored in the dictionary is 2 and the current processing table number is 1, so the process proceeds to step S403.

In step S403, the position of Table i is specified, and one or more specific cell areas in Table i are determined. The processing is executed by the ruled line detecting means 25, the correspondence relation determining means 27 and the cell determining means 29. The ruled line detecting means 25 detects ruled lines from the image data of the form. Further, the correspondence determination unit 27 compares the ruled line detected by the ruled line detection unit 25 with ruled line information or cell information related to the table i stored in the dictionary 21, determines whether the detected ruled line is correct, and classifies the table i. Identify the ruled lines you want.

For example, the ruled line detecting means 25 detects a ruled line (ruled line of the first hierarchy) connecting both ends in the vertical direction or both ends in the horizontal direction of the table i.
The cell area (first-level divided area) divided by the one-level ruled line is compared with the divided area in the dictionary. Further, the ruled line detection means 25 detects a ruled line (ruled line of the second hierarchy) connecting both ends in the vertical direction or both ends in the horizontal direction of the divided area of the first hierarchy, and the correspondence determination means 27 determines a ruled line of the second hierarchy. Then, the cell area (second-level divided area) divided by the above is compared with the divided area in the dictionary. Hereinafter, similarly, the ruled line of the n-th layer is detected and the divided area of the n-th layer is collated.

After that, the cell area determining means 29 selects the correspondence determining means 2 from the ruled lines detected by the ruled line detecting means 25.
7 is determined based on the ruled line information or cell information stored in the dictionary 21 and the information for determining the specific cell region from each cell region of Table i divided by the ruled line determined to be correct. Is determined, and information on the position of the specific cell area is stored in the recognition area data memory 63.
When this process ends, the process moves to step S404.

In step S404, the variable i is increased by one in order to shift the processing to the next table. Then, the process proceeds to step S
Move to 402.

By repeating the same processing as described above, the dictionary 21
When the processing is completed for all the tables in the table, the table-specific processing control unit 23 ends the processing.

Thereafter, the character recognizing means 40 executes the form image memory 73 based on the information on the position of the specific cell area (or a part of the specific cell area described later) stored in the recognition area data memory 63. Characters and the like are cut out from the image data of the form stored in the document, and the characters are recognized.

(Effects of the First Embodiment) As described above, in the recognition processing of characters and the like, by using the method of determining the recognition target area of the first embodiment, a plurality of table areas can be included in a form. If it exists, the tablespace can be specified, and
A specific cell area to be a recognition target area can be determined for each table area. Therefore, it is possible to perform a process of recognizing a form having a complicated configuration with a device having a simple configuration.

(Regarding the Configuration of the Second Embodiment) In the recognition processing of characters and the like, it may be better to cut out characters and lines by integrating the same recognition target areas. For example, characters indicating money such as "$ 100" are "@",
Rather than performing divisional recognition processing such as "1", "0", and "0", it is more accurate and faster to perform recognition processing unified as the same recognition target such as "$ 100". can do. Further, characters described in the same entry column can be output collectively as one piece of information.

Therefore, the second embodiment determines the recognition target in the first embodiment so that regions to be integrated and recognized as the same recognition target (hereinafter, referred to as the same recognition target region) can be integrated. A cell integrating means 31 is provided at a stage subsequent to the cell determining means 29 of the recognition target determining apparatus provided with the method.

FIG. 5 shows a configuration diagram of the second embodiment. The recognition area data memory 63 stores information on a specific cell area that is a recognition target area determined by the cell determination unit 29. On the other hand, the dictionary 21 stores information indicating which area is the same recognition target area. The cell integrating means 31 refers to the information of the specific cell area in the recognition area data memory 63 and also refers to the dictionary 21 to determine whether the specific cell area to which the information indicating the same recognition target area is added is added. In some cases, the cell areas are integrated into one recognition target area, and information on the positions of those areas is associated with a number added to each recognition target area (hereinafter referred to as a recognition target area number). It is stored in the area data memory 63. If there is no specific cell area to which information indicating that the area is the same recognition target area is added, as in the first embodiment, the specific cell area remains as a single specific cell area. Is stored in the recognition area data memory 63 in association with the recognition target area number.

Thereafter, the character recognizing means 40 executes the form image memory 73 based on the information on the position of the specific cell area (or a part of the specific cell area described later) stored in the recognition area data memory 63. Characters and the like are cut out from the image data of the form stored in the document, and the characters are recognized. The processing contents of the other units are the same as in the first embodiment.

(Regarding the operation of the second embodiment)
FIG. 6 is a diagram showing the configuration of the form, FIG. 7 is a diagram showing the configuration of information stored in the dictionary 21, and FIG.
A method of integrating the same recognition target area will be described with reference to the flowchart showing the operation 1.

First, information stored in the dictionary 21 will be described. In the second embodiment, in addition to the position information of the table area, the ruled line information or the cell information, and the information for determining one or a plurality of specific cell areas, the specific cell area serving as the same recognition target area has As shown in FIG. 7, information indicating the same recognition target area is added and stored in the dictionary 21. The information indicating the same recognition target area is
For example, the information indicating the relative relationship between the cell regions in the above-described hierarchical structure and the information regarding the position of the field to be read, which are stored in the dictionary 21 independently, are compared with each other. The number is set for all the specific cell areas for each recognition target area, and the same number is assigned to the same recognition target areas at that time.

FIG. 7 shows that T2221 to T2225 surrounded by the same circle in the specific cell region are the same recognition target region. The number written on the left shoulder of the circle indicates the number of the recognition target area (hereinafter referred to as the recognition target area number).

Next, the operation of the cell integration means 31 will be described.

First, initialization is performed in step S701. That is, the variable i, which is stored in the cell integration means 31 and indicates the recognition target area number to be processed, is set to 1.

Next, in step S702, the largest one of all the recognition target region numbers is compared with the variable i indicating the recognition target region number to be processed, and if the variable i is larger than the largest of all the recognition target region numbers. , And the process ends. If not, the process moves to step S703. In the case of the example of FIG. 6, since the maximum of all the recognition target area numbers is 2, and the current processing recognition target area number is 1, step S70 is performed.
Move the process to 3.

In step S703, the recognition target area numbers of all the specific cell areas are checked, and the specific cell areas having the same number as the current processing recognition target area number i are integrated. In the case of the example of FIG. 6, since the specific cell region having the recognition target region number of 1 is only T11, the integration process is not performed, and the recognition target region is the T11 cell region itself.

Next, in step S704, the variable i is increased by one in order to integrate the next recognition target area. And
The process moves to step S702. In the case of the example of FIG.
It becomes 2.

Next, an end determination is made in step S702.
In this case, since the end determination described above does not apply, the process proceeds to step S703.

In step S703, the recognition target area numbers of all the specific cell areas are checked, and specific cell areas having the same number as the current processing recognition target area number i are integrated. In the case of the example of FIG. 6, the cell having the recognition target area number 2 is the five cells T2221 to T2225, and these cells are integrated. Then, it is stored in the recognition area data memory 63.

Next, in step S704, the variable i is increased by one in order to integrate the next recognition target area. Then, the process proceeds to step S702. In the case of the example of FIG. 6, i = 3
Becomes

Next, an end determination is made in step S702.
This case corresponds to the end determination described above. That is, the integration processing has been completed for all the recognition target areas. Therefore, the cell integration means 31 ends the operation.

Thereafter, the character recognizing means 40 executes the form image memory 73 based on the information on the position of the specific cell area (or a part of the specific cell area described later) stored in the recognition area data memory 63. Characters and the like are cut out from the image data of the form stored in the document, and the characters are recognized.

(Regarding the effects of the second embodiment) As described above, the same recognition target regions are integrated by using the method for determining the recognition target regions in the second embodiment in the recognition processing of characters and the like. And can be recognized accurately and quickly. Also, characters in the same recognition target area can be output as a recognition result in the same entry field. For example, even in a case where a numeral or the like is entered in a region where a numeral or the like is entered, cells corresponding to the amount column can be integrated with each other, and the recognition results can be output collectively.

(Regarding the Configuration of the Third Embodiment) In the process of recognizing a character or the like, only a part of a specific cell area may become a recognition target area. In the third embodiment, the cell determination means 29 of the recognition target determination apparatus provided with the recognition target determination method in the first embodiment can perform recognition of characters and the like described in such a recognition target area. A cell internal area determining means 33 is provided at the subsequent stage.

FIG. 9 shows a configuration diagram of the third embodiment. In the third embodiment, the internal area position information is described as being added to a specific cell area where the recognition target area is only a part, but is added to all the specific cell areas. It is also possible.

The recognition area data memory 63 stores information on a specific cell area determined by the cell determination means 29. On the other hand, the dictionary 21 stores information for determining a recognition target area in a specific cell area for a specific cell area whose recognition target area is only a part. The information for determining the recognition target area in the specific cell area is information indicating where in the specific cell area the recognition target area is, and hereinafter, is referred to as internal area position information. The cell internal area determining means 33 refers to the information on the position of the specific cell area stored in the recognition area data memory 63 and refers to the dictionary 21 to add the internal area position information to the specific cell area. If there is, only the area determined by the internal area position information is set as a recognition target, and information on the position of the recognition target area is stored in the recognition area data memory 63. When the internal area position information is not added to the specific cell area, information on the position of the specific cell area which is the recognition target area is stored in the recognition area data memory 63.

Thereafter, the character recognizing means 40 converts the image data of the form stored in the form image memory 73 into characters and the like based on the information on the position of the recognition target area stored in the recognition area data memory 63. Cut out
Recognize characters. The processing contents of the other units are the first.
This is the same as the embodiment.

(Regarding the Operation of the Third Embodiment)
A part of a specific cell area is shown by using the figure showing the structure of the form shown in FIG. 10, the figure showing the structure of information stored in the dictionary 21 of FIG. 11, and the flowchart showing the operation of the cell internal area determining means of FIG. Is determined as the recognition target area,
A method of performing the recognition processing will be described.

First, information stored in the dictionary 21 will be described. In the third embodiment, in addition to the position information of the table area, the ruled line information or the cell information, and the information for determining one or a plurality of specific cell areas, the internal area position information is specified for each specific cell area. The data is added and stored in the dictionary 21. For example, the internal area position information compares the information indicating the relative relationship between the cell areas in the hierarchical structure and the information indicating the position of the read field, which are stored in the dictionary 21 independently of each other. Then, it is set by specifying a cell area overlapping with the reading field and adding information indicating at which position in the specific cell area the recognition target area exists.

FIG. 11 shows that specific cell regions T11 and T222 surrounded by a circle have internal region position information added thereto.

Next, the operation of the cell internal area determining means 31 will be described.

First, initialization is performed in step S1001. That is, the variable i indicating the number of the specific cell area to be processed, stored in the cell internal area determining means 33, is set to 1
To

Next, in step S1002, the number of specific cell areas determined in advance is compared with the variable i indicating the number of the specific cell area to be processed, and the variable i is the number of specific cell areas determined in advance. If it is larger, the process ends. If not, the process moves to step S1003. In the case of the example of FIG. 10, since the number of the specific cell areas determined in advance is 2 and the number of the specific cell area to be processed at present is 1, the process proceeds to step S1003.

Next, in step S1003, for a specific cell area to be processed, an area to be recognized inside the cell is determined. Specifically, if the specific cell area to be processed does not have the internal area position information, the specific cell area itself is determined as the recognition target area, and the information on the position of the specific cell area is stored in the recognition area data memory 63. Output to If the specific cell area to be processed has internal area position information, based on the internal area position information,
A recognition target area in a specific cell area is determined, and information on the position of the area is stored in a recognition area data memory 63.
Output to In the case of the example of FIG. 10, the specific cell area in which the number of the specific cell area to be processed is 1 is T11,
Since the internal area position information is “upper cell half”, the upper half of the specific cell area T11 is determined as the recognition target area.

Next, in step S1004, the variable i is increased by one in order to determine a recognition target area for a specific cell area to be processed next. Then, the process proceeds to step S
Move to 1002. In the example of FIG. 10, i = 2.

The above processing is performed on all the specific cell areas, and when the recognition target area in all the specific cell areas is determined, the operation of the cell internal area determining means 33 ends. In the case of the example of FIG. 10, when the recognition target area is determined for T11 and T222, the cell internal area determination unit 33 ends the operation.

Thereafter, the character recognizing means 40 determines the character or the like from the image data of the form stored in the form image memory 73 based on the information on the position of the recognition target area stored in the recognition area data memory 63. Cut out
Recognize characters.

In the above embodiment, the internal area position information is expressed as a ratio with respect to the cell area such as “upper half of the cell”. However, it is expressed by the coordinates of the cell area and the coordinates of the internal recognition target area. May be.

(Effects of the Third Embodiment) As described above, in the recognition processing of characters and the like, the recognition target area determination method in the third embodiment is used, so that only the area to be recognized inside the cell is to be recognized. Since recognition processing can be performed, unnecessary processing can be omitted. As a result, the processing time can be made more efficient. In addition, it is possible not to output a recognition result of an unnecessary portion that is not a recognition target inside the cell.

(Regarding the Configuration of the Fourth Embodiment) In recognition processing of characters and the like, it is desired that recognition processing can be performed even on a recognition target area existing in an area outside the table area.

Therefore, in the fourth embodiment, the cell determination of the recognition object determination apparatus provided with the recognition object determination method in the first embodiment is performed so that a recognition target area existing outside the table area can be extracted. An out-of-line area determining means 35 is provided at the subsequent stage of the means 29.

FIG. 13 shows the configuration of the fourth embodiment. No.
In the fourth embodiment, the cell determination means 29 determines the recognition target area for the area inside the table of the form, stores information on the position of the recognition target area in the recognition area data memory 63, and stores the information in the form table. For the outside area, the out-of-line area determination means 35 determines the recognition target area, and stores information on the position of the recognition target area in the recognition area data memory 63.

Thereafter, the character recognizing means 40 converts the character data or the like from the image data of the form stored in the form image memory 73 based on the information on the position of the recognition target area stored in the recognition area data memory 63. Cut out
Recognize characters. The processing contents of the other units are the first.
This is the same as the embodiment.

(Regarding the Operation of the Fourth Embodiment)
A method of determining a recognition target region existing in a region outside the table and performing a recognition process will be described with reference to the diagram showing the configuration of the form in FIG. 14 and the diagram showing the configuration of information stored in the dictionary 21 in FIG.

First, information stored in the dictionary 21 will be described. In the fourth embodiment, the position information and the ruled line information or the cell information of the table area and the information for determining one or a plurality of specific cell areas necessary for determining the specific cell area inside the table are determined. In addition, information indicating the position of the recognition target area existing outside the table (hereinafter, position information of the out-of-line recognition area) is added. Note that the recognition target areas existing outside the table are, for example, areas indicated by 240a and 240b in FIG.

FIG. 15 shows the structure of information stored in such a dictionary 21. FIG. 15 shows that information for determining a specific cell area inside the table is stored in the left system, and that positional information of the out-of-line recognition area is stored in the right system. Is shown. Since the form shown in FIG. 14 has two recognition target areas outside the table, two pieces of information R1 and R2 corresponding to the two areas are stored in the system on the right side of FIG.

As described above, with respect to the area inside the table of the form, the cell determining means 29 determines the recognition target area and stores the information on the position of the recognition target area in the recognition area data memory 63.
The area outside the table of the form is determined by the out-of-line area determination means 35, and information on the position of the recognition target area is stored in the recognition area data memory 63. Since the operation of the cell determining means 29 is the same as that of the first embodiment, the operation of the out-of-line area determining means 35 will be described below.

The out-of-line area determining means 35 specifies the recognition target area based on the position information of the out-of-line recognition area stored in the dictionary 21 in association with each of the recognition target areas existing outside the table. Then, a character area described in the recognition target area is detected, and information on the position of the area is stored in the recognition area data memory 63. The detection of the character area can be performed by a known recognition processing technique. For example,
It is possible to obtain a lump of black pixels in the vicinity of each recognition target area and extract a character area based on the information, and also examine a white pixel distribution around each recognition target area,
It is also possible to extract a character region by using a position where white pixels continue at a certain threshold or more as a boundary of the character region. Furthermore, a character area can be extracted from position information of a ruled line obtained when a specific cell area inside a table is detected, using a relative positional relationship to the position.

When the above processing is performed on the position information of all the out-of-line recognition areas, the out-of-line area determination means 35 ends the operation.

Thereafter, the character recognizing means 40 uses the information on the position of the recognition target area stored in the recognition area data memory 63 to extract the characters and the like from the image data of the form stored in the form image memory 73. Cut out
Recognize characters.

(Effects of the Fourth Embodiment) As described above, in the recognition processing of characters and the like, the recognition target area is determined outside the table by using the method of determining the recognition target area in the fourth embodiment. Even in some cases, the area can be determined, and recognition processing can be performed. As a result, it is possible to perform recognition processing on characters and the like written in all places regardless of the inside and outside of the table area of the form.

(Regarding the Configuration of the Fifth Embodiment) In the process of recognizing characters and the like, for example, ruled lines in a table area are divided by ruled lines other than ruled lines connecting both ends in the vertical direction or both ends in the horizontal direction of the table area. There is a case where a specific cell region to be a recognition target region is included in the cells formed by the above. In the fifth embodiment, information indicating that a ruled line has been divided (hereinafter referred to as “intersecting ruled line additional information”) is stored in the dictionary 21 so that such a specific cell area can be determined. Has a function of detecting the ruled line to be divided.

(Operation of Fifth Embodiment)
Ruled lines in the table area other than ruled lines connecting both ends in the vertical direction or both ends in the horizontal direction (hereinafter referred to as ruled lines) using the diagram showing the configuration of the form in FIG. 16 and the diagram showing the configuration of information stored in the dictionary 21 in FIG. The following describes a method of detecting the vertical division rule and the horizontal division rule in the case of division by the vertical division rule and the horizontal division rule, and determining the recognition target area.

First, information stored in the dictionary 21 will be described. In the fifth embodiment, corresponding to the table 210 in the reference form shown in FIG. 16, ruled line information on a ruled line connecting both ends in the vertical direction or both ends in the horizontal direction of the table area, or a cell formed by this ruled line The cell information relating to the area is hierarchically stored in the dictionary 21 and, in FIG.
As shown by b and c, the rule information on the vertical divided ruled lines and the horizontal divided ruled lines, or the cell information on the cell area formed by the vertical divided ruled lines and the horizontal divided ruled lines is associated with the intersection additional information and hierarchically associated with the dictionary 21. Is stored in

FIG. 17 shows that the cell area T2 has vertical dividing ruled lines,
It is shown that the cell region T21 and the specific cell region T22 are formed by being divided by the horizontal dividing ruled lines.

When the ruled line information or cell information stored in the dictionary 21 includes not only the number but also the position information, the vertical divided ruled line, the horizontal divided ruled line, or the cell area formed by these ruled lines is obtained from the position information. Therefore, it is possible not to add the intersection ruled line additional information.

Next, a method for determining a specific cell area will be described. Except for the ruled line detecting means 25, the processing contents of each means are the same as in the first embodiment of the present invention.

When the additional ruled line addition information is not added to the dictionary 21, the ruled line detection means 25 detects a ruled line connecting both ends in the vertical direction or both ends in the horizontal direction of the table area. If the intersection ruled line additional information is added, an area in which the ruled line to be detected exists is specified based on the intersection ruled line additional information, and, for example, continuous in the horizontal direction and the vertical direction are determined from the image data of the area. Rule line candidates are detected from the number of black pixels and the like. Then, of those ruled lines, a ruled line in which both horizontal and vertical ruled lines intersect each other at an end point is detected,
Identify the cell area.

If this cell area is a specific cell area, the cell determining means 29 determines that cell area as a specific cell area and stores the data in the recognition area data memory 63 together with the number of the area to be recognized. I do.

If the specific cell region is located further below the cell region, the process is shifted to the detection of a ruled line of a lower hierarchy by a similar method until the specific cell region is determined.

(Effects of the Fifth Embodiment) As described above, in the recognition processing of characters and the like, by using the method of determining a recognition target area in the fifth embodiment, a vertical division rule line or a horizontal division rule line is used. Even if there are configured cell areas, these ruled lines can be detected based on the ruled line intersection additional information, and a specific cell area can be accurately determined. As a result, it is possible to accurately perform recognition processing of characters and the like in those specific cell regions.

(Regarding the Embodiment)
Although up to four have been described individually, the respective embodiments may be combined. FIG. 18 shows the configuration.

FIG. 18 is a diagram showing a configuration obtained by combining the embodiments. In the example of FIG. 18, the function of detecting the vertical division ruled line and the horizontal division ruled line is added to the ruled line detection means 25 to the configuration of the first embodiment, and the cell integration means 31 and the cell An internal area determining means 33 and an out-of-line area determining means 35 are provided. The dictionary 21 and the recognition area data memory 63 store the recognition target area number, the position information of the out-of-line recognition area, and other information. The processing of each means is the same as that described in the first to fifth embodiments.

Such a character recognition device can cut out characters from the form shown in FIG. 19 and recognize the characters.
FIG. 19 is a diagram showing the configuration of a form. In a form, there are a plurality of table areas, an area to be integrated and recognized as the same recognition target exists, and a specific cell area. This shows a form in which there is an area in which only a part is recognized, an area to be recognized exists outside the table area, and an area divided by a vertical division rule or a horizontal division rule.

The character recognition device shown in FIG.
First, the ruled line detection means 25
A ruled line connecting both ends in the horizontal direction or the vertical direction of the table area, a vertical divided ruled line and a horizontal divided ruled line are detected, and the cell determining means 29 detects the ruled line based on the information stored in the dictionary 21.
Identify the cell area.

The dictionary 21 stores information as shown in FIG. 20, which is a diagram showing the structure of information stored in the dictionary. FIG. 20 shows information about two table areas, Table 1 and Table 2, and information about the outside of the table, which indicates an area where characters and the like existing outside the table area are written.
For example, T1 to T222 are shown as information relating to Table 1, in which T221 which is a cell area of the third hierarchy is shown.
Indicates a specific cell area. Table 2
T1 to T233 are shown as related information, in which T222, T232, and T233, which are cell areas in the third hierarchy, are specific cell areas, and T222 further includes "cell left part 3" as internal area position information. / 4 "is added, and information indicating that the recognition target areas are the same is added to T232 to T233. Also, R1 is shown as information relating to out-of-line.

Next, the cell integration means 31 integrates the areas to be integrated and recognized as the same recognition target, and the cell internal area determination means 33 determines the recognition target area in the specific cell area, and The area determining means 35 determines a recognition target area outside the table area. Information indicating the position of the recognition target area determined by these is stored in the recognition area data memory 63.

Thereafter, the character recognizing means 40 converts the character data from the image data of the form stored in the form image memory 73 based on the information indicating the position of the recognition target area stored in the recognition area data memory 63. And character recognition.

Note that the present invention is not limited to the above-described embodiment.
It is possible to change information stored in a dictionary or the like, or to change the configuration or operation.

[0109]

The present invention described above has the following effects.

In the recognition processing of characters and the like, by using the first or second invention described in the first embodiment, a table area can be specified even when a plurality of table areas exist in a form. In addition, a specific cell area to be a recognition target area can be determined for each table area. Therefore, it is possible to perform a process of recognizing a form having a complicated configuration with a device having a simple configuration.

In the recognition processing of characters and the like, by using the third invention described in the second embodiment, the same recognition target areas can be integrated and recognized, so that And can be identified quickly. Also,
Characters in the same recognition target area can be output as a recognition result in the same entry field. For example, even in a case where a numeral or the like is entered in a region where a numeral or the like is entered, cells corresponding to the amount column can be integrated with each other, and the recognition results can be output collectively.

Further, in the recognition processing of characters and the like, by using the fourth or fifth invention described in the third embodiment, it is possible to perform the recognition processing only on the area in the cell where recognition is to be performed. Unnecessary processing can be omitted. As a result, the processing time can be made more efficient. Also,
Recognition results of unnecessary parts that are not to be recognized inside the cell can also be prevented from being output.

Further, by using the sixth or seventh invention described in the fourth embodiment in the recognition processing of characters and the like, even if a recognition target area exists outside the table, the area can be determined. , A recognition process can be performed. As a result, it is possible to perform recognition processing on characters and the like written in all places regardless of the inside and outside of the table area of the form.

Further, in the recognition processing of characters and the like, by using the eighth or ninth invention described in the fifth embodiment, a cell area constituted by vertical division ruled lines or horizontal division ruled lines exists. However, these ruled lines can be detected based on the ruled line intersection additional information, and a specific cell region can be accurately determined. As a result, it is possible to accurately perform recognition processing of characters and the like in those specific cell regions.

[Brief description of the drawings]

FIG. 1 is a processing configuration diagram of a first embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of a form.

FIG. 3 is a diagram showing a configuration of information stored in a dictionary.

FIG. 4 is a flowchart illustrating an operation of a table-specific process control unit.

FIG. 5 is a processing configuration diagram of a second embodiment of the present invention.

FIG. 6 is a diagram showing a configuration of a form.

FIG. 7 is a diagram showing a configuration of information stored in a dictionary.

FIG. 8 is a flowchart of the cell integration means.

FIG. 9 is a processing configuration diagram of a third embodiment of the present invention.

FIG. 10 is a diagram showing a configuration of a form.

FIG. 11 is a diagram showing a configuration of information stored in a dictionary.

FIG. 12 is a flowchart of a cell internal area determining unit.

FIG. 13 is a processing configuration diagram of a fourth embodiment of the present invention.

FIG. 14 is a diagram showing a configuration of a form.

FIG. 15 is a diagram showing a configuration of information stored in a dictionary.

FIG. 16 is a diagram showing a configuration of a form.

FIG. 17 is a diagram showing a configuration of information stored in a dictionary.

FIG. 18 is a processing configuration diagram in the embodiment.

FIG. 19 is a diagram showing a configuration of a form.

FIG. 20 is a diagram showing a configuration of information stored in a dictionary.

[Explanation of symbols]

 REFERENCE SIGNS LIST 10 character recognition device 20 recognition target area determination device 21 dictionary 23 table processing control means 25 ruled line detection means 27 correspondence determination means 29 cell determination means 40 character recognition means 50 CPU 60 memory 61 ruled line data memory 63 recognition area data memory 70 image Data input device 71 Scanner 73 Form image memory

Claims (9)

[Claims] When determining one or a plurality of specific cell areas to be a recognition target area from image data of a form including a large number of cells partitioned by ruled lines, the ruled line information or cell information of the form and the form information of the form are determined. Information for determining a specific recognition area is stored in a dictionary in advance, a ruled line is detected from the image data of the form, and the specific cell area is determined from an area divided by the ruled line in the table area. A method for determining a recognition target region, comprising determining a recognition region based on information for determining and further determining a recognition region. 2. A method for determining one or a plurality of specific cell areas to be a recognition target area from image data of one or a plurality of forms including a large number of cells divided by ruled lines, wherein a specific cell area has already been determined. The determined form is used as a reference form, and ruled line information or cell information and a specific cell area of the reference form are determined for each table area closed by an intersecting ruled line in the form to be recognized. Is stored in the dictionary, and from the image data of the form to be recognized, the ruled line information or cell information stored in the dictionary and a specific cell area of the reference form are determined for each table area. Determining the specific cell region based on the information for determining the region to be recognized. 3. The method for determining a recognition target area according to claim 1, wherein the form to be recognized has a plurality of specific cell areas, and the specific cell areas are to be integrated and recognized as the same recognition target. When there is an area, the ruled line information or cell information, information for determining a specific cell area of the reference form, and information for determining the same recognition target area are stored in the dictionary. From the image data of the form to be processed, based on the ruled line information or cell information stored in the dictionary and information for determining a specific cell area of the reference form, determine the specific cell area, A method for determining a recognition target area, comprising: integrating the same recognition target areas from the specific cell area based on information for determining the same recognition target area to determine a recognition target area. 4. The method according to claim 1 or 2, wherein only a part of the specific cell area included in the form to be recognized is a recognition target area. Information or cell information, information for determining a specific cell area of the reference form, and internal area position information for determining a recognition target area in the specific cell area are stored in the dictionary; From the image data of the form to be processed, based on the ruled line information or cell information stored in the dictionary and information for determining a specific cell area of the reference form, determine the specific cell area, A method for determining a recognition target area, comprising: determining a recognition target area included in the specific cell area based on internal area position information. 5. The method for determining a recognition target area according to claim 4, wherein the internal area position information for determining the recognition target area in the specific cell area is determined based on the specific cell and the recognition target area in the reference form. A recognition target area determined by comparing position coordinates, and comparing the positional relationship between the determined internal area position information stored in the dictionary and the specified cell information to determine a recognition target area; How to determine. 6. The method for determining a recognition target area according to claim 1, wherein the ruled line information or the cell information is used when a recognition target area exists outside a table area in a form to be recognized. The position information of the out-of-line recognition area for determining the recognition target area outside the table area is stored in the dictionary, and from the image data of the form to be recognized, the ruled line information or cell information stored in the dictionary is stored. And determining a recognition target area existing outside the table area based on the position information of the outside table recognition area. 7. The method for determining a recognition target area according to claim 6, wherein a table area is determined based on ruled line information in the reference form,
By comparing the position coordinates of the table region with the position coordinates of the recognition target region, if the recognition target region is outside the table region, the position coordinates of the recognition target region are determined as the recognition target region outside the table region. Recognition in a dictionary as position information of an out-of-line recognition area for performing a search, and determining a recognition target area by searching for a region around the position coordinates from the stored position information of the out-of-line recognition area. How to determine the target area. 8. The method for determining a recognition target area according to claim 1, wherein the division of the form to be recognized by the ruled line of the table area is performed by:
If it is divided by a rule other than a ruled line connected to both ends in the horizontal direction or both ends in the vertical direction, the intersection additional information indicating that the ruled line is divided is stored in the dictionary, and the image data of the form to be recognized is stored. Detecting a ruled line based on the additional intersection information stored in the dictionary. 9. The method for determining a recognition target area according to claim 8, wherein if intersection additional information is added to the dictionary, both ends in the horizontal direction or in the vertical direction are extracted from the image data of the form to be recognized. A method for determining a recognition target area, comprising performing ruled line detection processing for detecting ruled lines other than ruled lines connected to both ends.