JP2008083846A - Target area determination method and device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device determining a recognition area of a scanned business form image without correcting deviation of the recognition area from a recognition area of a registered image. <P>SOLUTION: This target area determination device has: a dictionary data storage means 21 previously storing each coordinates in a lateral direction and a longitudinal direction of a circumscribing rectangle of a table in a business form, and each coordinates in a lateral direction and a longitudinal direction of the recognition area in the circumscribing rectangle as dictionary data in each kind of the business form; a business form identification means 22 recognizing the corresponding business form from the business form image scanned by a scanner part 71, and retrieving the dictionary data of the business form from the dictionary data storage means 21; a table circumscribing rectangle detection means 23 detecting the circumscribing rectangle of the table from the scanned business form image; and a recognition area extraction means 24 recognizing the recognition area in the circumscribing rectangle from the detected circumscribing rectangle of the table and the retrieved dictionary data, and extracting it as data of an area to recognize a character. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a method and apparatus for determining a character recognition target region suitable for application to a character recognition device or the like.

  Conventionally, there are the following methods for determining a recognition target area from a form image. For example, two feature part coordinates from the left and right on the form image are detected, and the distance from the form edge reference is compared with the predefined feature part coordinates to determine the cause of the image deviation, and the image deviation factor is determined. An appropriate coordinate correction value from the form edge reference is calculated to determine a recognition target area such as a receipt column (see, for example, Patent Document 1).

Japanese Patent Laying-Open No. 2004-326672 (page 4, FIG. 2)

  In the conventional method described above, the factor is determined from the image shift caused by the comparison calculation with the predefined characteristic part coordinates, and then an appropriate coordinate correction value from the form edge reference according to the image shift factor is obtained. The calculation and the process until the recognition target area is determined are complicated.

  In the recognition target area determining method according to the present invention, the horizontal and vertical coordinates of the circumscribed rectangle of the table in the form, and the horizontal and vertical coordinates of the recognition area existing in the circumscribed rectangle are dictionary data. Is set for each form type, and the dictionary data of the corresponding form is retrieved from the form image scanned by the scanner unit, the circumscribed rectangle of the table is detected from the scanned form image, and the detected circumscribed rectangle of the table is detected. Then, the recognition area in the circumscribed rectangle is recognized from the retrieved dictionary data, and this is extracted as data of the character recognition target area.

  In the present invention, the dictionary data of the corresponding form is searched from the form image scanned by the scanner unit, the circumscribed rectangle of the table is detected from the scanned form image, and the detected circumscribed rectangle of the table is searched. Since the recognition area in the circumscribed rectangle is recognized from the data and this is extracted as the data of the character recognition target area, correction is made even if the recognition area is shifted from the recognition area on the dictionary data side. And the characters in the recognition area can be easily recognized.

FIG. 1 is a block diagram showing a schematic configuration of a character recognition device provided with a recognition target area determination device according to an embodiment of the present invention, and FIG. 2 is a diagram showing systematized dictionary data in the embodiment.
The character recognition device 10 shown in the figure is connected to an image data input device 70 for reading image data of a form, and a recognition target region determination device 20, a character recognition unit 40, a recognition target region determination device 20 and character recognition. A CPU 50 that controls each of the units 40 and a memory 60 including, for example, a hard disk are provided. The recognition target area determination device 20 includes a dictionary data storage unit 21, a form identification unit 22, a circumscribed rectangle detection unit 23, and a recognition region extraction unit 24. The image data input device 70 includes a scanner unit 71 that scans a form and a form image memory 72 that stores a form image scanned by the scanner unit 71.

  For example, as shown in FIG. 2, the dictionary data storage means 21 includes a form ID for identifying the form type, the table type for each form, the horizontal left and right coordinates of the circumscribed rectangle of each table, and the vertical The coordinates of the upper and lower ends in the direction, the left and right coordinates in the horizontal direction and the upper and lower coordinates in the vertical direction of the recognition area existing in the circumscribed rectangle are registered as dictionary data. The form identifying means 22 reads out the form image from the form image memory 72 and recognizes the form ID, and uses this as a search key to retrieve the dictionary data of the registered image (form) having the same form ID from the dictionary data storage means 21 and recognize it. The data is transferred to the area extracting unit 24 and the form image is transferred to the front circumscribed rectangle detecting unit 23.

  The front circumscribing rectangle detecting means 23 detects all ruled lines in the horizontal and vertical directions from the form image, groups the intersecting rectangles to detect the circumscribed rectangle of the front, and the left end of the circumscribed rectangle in the horizontal direction. And the coordinates at the right end and the coordinates at the upper and lower ends in the vertical direction are recognized and sent to the recognition area extracting means 24 at the subsequent stage.

  The recognition area extracting unit 24 includes the dictionary data of the registered image extracted by the form identifying unit 22, the horizontal left and right coordinates of the circumscribed rectangle detected by the front circumscribed rectangle detecting unit 23, and the upper end of the vertical direction. From the coordinates of the lower end, the horizontal left and right coordinates and the vertical upper and lower coordinates of the recognition area existing in the circumscribed rectangle of the form image are calculated, extracted as recognition area data, and extracted from the memory 60. To store.

  The character recognition unit 40 recognizes characters entered in the recognition area data determined by the recognition target area determination device 20 and stored in the memory 60. For character recognition, for example, a character is cut out from a character string in a recognition area, and a feature of the cut out character is extracted. Then, the extracted character features of the recognition target are collated with the character features registered in a dictionary (not shown) prepared in advance to determine candidate characters.

Next, the operation of this embodiment will be described with reference to FIGS. FIG. 3 is a flowchart showing the operation of the character recognition apparatus according to the embodiment, and FIG. 4 is an example for explaining the extraction of the recognition area of the form.
First, the form identification unit 22 of the recognition target area determination device 20 reads out the form image stored in the form image memory 72, recognizes the form ID, and uses this as a search key to have the same form ID from the dictionary data storage unit 21. A form is identified (301). Then, the dictionary data of the registered image of the form is taken out and transferred to the recognition area extracting unit 24, and the form image is transferred to the outer circumscribed rectangle detecting unit 23. When the form image is transferred, the front circumscribing rectangle detecting unit 23 detects all the horizontal and vertical ruled lines from the form image, groups the intersecting lines and detects the circumscribed rectangle of the table, In addition, the horizontal left and right coordinates and the vertical top and bottom coordinates of the circumscribed rectangle are recognized and sent to the recognition area extracting means 24 in the subsequent stage (302). The recognition area extracting unit 24 extracts a recognition area existing in the circumscribed rectangle from the dictionary data of the registered image extracted by the form identifying unit 22 and the circumscribed rectangle of the table detected by the outer circumscribed rectangle detecting unit 23 ( 303).

  Here, extraction of the recognition area in the circumscribed rectangle will be described with reference to FIG. In the figure, (a) shows an example of a registered image of a form based on the dictionary data searched by the form identifying means 22. This registered image includes the horizontal left and right coordinates iXa and iXd and the vertical top and bottom coordinates iYa and iYd of the circumscribed rectangle of Table 1, which is the dictionary data, and a recognition area existing in the circumscribed rectangle. This is based on the left and right coordinates iXb and iXc in the horizontal direction and the coordinates iYb and iYc in the upper and lower ends in the vertical direction.

  Further, (b) shows the dictionary data extracted by the form identifying means 22, the horizontal left and right coordinates oXa, oXd and the vertical edges of the circumscribed rectangle in Table 1 detected by the outer circumscribed rectangle detecting means 23. The figure shows the recognition area existing in the circumscribed rectangle of the form image from the coordinates oYa and oYd of the upper end and the lower end of the direction. The recognition area is determined by the following arithmetic expression. The calculation of the recognition area is performed by the recognition area extraction unit 24.

The coordinates oXb, oXc of the left and right edges in the horizontal direction of the recognition area are
oXb = oXa + (iXb−iXa) × (oXd−oXa) / (iXd−iXa)
oXc = oXa + (iXc−iXa) × (oXd−oXa) / (iXd−iXa)
Sought from
In addition, the coordinates oYb and oYc of the upper end and the lower end in the vertical direction of the recognition area are
oYb = oYa + (iYb−iYa) × (oYd−oYa) / (iYd−iYa)
oYc = oYa + (iYc−iYa) × (oYd−oYa) / (iYd−iYa)
It is requested from. Of the above arithmetic expressions, (oXd−oXa) / (iXd−iXa) and (oYd−oYa) / (iYd−iYa) are the size (horizontal) of the recognition area on the form image side relative to the recognition area on the registered image side. This is an equation for correcting the direction and the vertical direction.

  The horizontal left and right coordinates oXb and oXc and the vertical top and bottom coordinates oYb and oYc of the recognition area obtained from the above arithmetic expression are stored in the memory 60 as recognition area data. On the other hand, the character recognition unit 40 recognizes the recognition area on the form image based on the recognition area data (oXb, oXc, oYb, oYc) stored in the memory 60. Then, the characters entered in this recognition area are cut out as described above, the features of the cut-out characters are extracted, and the features are compared with the features of each character registered in the dictionary to determine candidate characters. In Table 2, the recognition area is determined using the above arithmetic expression.

  As described above, according to the embodiment, all the horizontal and vertical ruled lines are detected from the form image, the intersections are grouped to detect the circumscribed rectangle of the table, and the form image is supported. Since the recognition area existing in the circumscribed rectangle is extracted from the dictionary data of the registered image to be registered, it is not corrected even if the recognition area of the scanned form image is deviated from the recognition area on the registered image side. The character in the recognition area can be easily recognized.

It is a block diagram which shows schematic structure of the character recognition apparatus which shows embodiment of this invention. It is a figure which systematizes and shows the dictionary data in embodiment. It is a flowchart which shows operation | movement of the character recognition apparatus which concerns on embodiment. It is an example for demonstrating extraction of the recognition area | region of a form.

Explanation of symbols

  10 character recognition device, 20 recognition target area determination device, 60 memory, 70 image data input device.

Claims (4)

The horizontal and vertical coordinates of the circumscribed rectangle of the table in the form in advance, and the horizontal and vertical coordinates of the recognition area existing in the circumscribed rectangle are set as dictionary data for each type of form, respectively. Search the corresponding form dictionary data from the form image scanned by the department,
Detect the bounding rectangle of the table from the scanned form image,
A recognition target area determination method, comprising: recognizing a recognition area in a circumscribed rectangle from the detected circumscribed rectangle of the table and searched dictionary data, and extracting the recognition area as data of a character recognition target area.   2. The recognition target area determining method according to claim 1, wherein the circumscribed rectangle of the table is detected by recognizing horizontal and vertical ruled lines from the scanned form image and grouping those that intersect. A dictionary in which the horizontal and vertical coordinates of the circumscribed rectangle of the table in the form and the horizontal and vertical coordinates of the recognition area existing in the circumscribed rectangle are stored as dictionary data for each type of form. Data storage means;
A form identifying means for recognizing a corresponding form from the form image scanned by the scanner unit, and retrieving dictionary data of the form from the dictionary data storage means;
A circumscribed rectangle detecting means for detecting a circumscribed rectangle of the table from the scanned form image;
The recognition area in the circumscribed rectangle is recognized from the circumscribed rectangle of the table detected by the outer circumscribed rectangle detecting means and the dictionary data searched by the form identifying means, and this is extracted as data of the character recognition target area. A recognition target area determining apparatus comprising a recognition area extracting unit.   4. The front circumscribing rectangle detecting means recognizes ruled lines in the horizontal direction and the vertical direction from the scanned form image, and groups the intersecting ones to detect a circumscribing rectangle of the front. Recognition target area determination device.

Images