JPH1069520A - Character recognition method and recording medium recording program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character recognition method capable of improving a recognition rate for an inclined character or character string. SOLUTION: The character string of a recognition object is surrounded by an inclined rectangular frame W of and a recognition range is specified. At the time, the character string of the recognition object is surrounded by the rectangular frame W inclined for the same angle as the inclination angle of the character string. Such a rectangular frame W is set by specifying three points P1 , P2 and P3 among the four vertexes. Then, by obtaining the angle formed by a horizontal direction and a straight line connecting the point P1 set first and the point P2 specified secondly among the three points P1 -P3 , the inclination angle θ of the rectangular frame W is calculated. Then, the inclination angle of the character string is decided based on the inclination angle θof the rectangular frame W, the inclination angle of the character string is taken into consideration and the characters are recognized.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition method for reading a drawing such as a map and recognizing a character or a character string entered in the drawing.

[0002]

2. Description of the Related Art A CAD system or the like has a character recognition function for reading a drawing such as a map and recognizing a character or a character string entered in the drawing. To perform such character recognition, first, an operator designates a recognition range by surrounding a character or a character string with a rectangular frame on the screen. Here, the rectangular frame has a horizontal side and a vertical side, and is set by designating two diagonally located points among the four vertices. When the recognition range is specified in this way, a plurality of feature amounts are extracted from the binary image data of the characters included in the recognition range, and a feature pattern is obtained. Then, character recognition is performed by comparing the characteristic pattern with a reference pattern that has already been registered.

[0003]

In general, characters or character strings written in drawings may be inclined. When recognizing such inclined characters or character strings,
In the conventional CAD system, an operator designates a character or a character string by enclosing it in a rectangular frame as described above, and inputs an inclination angle of the character or the character string from a keyboard. Then, character recognition is performed in consideration of the inclination angle of the character or character string. However, since the operator obtains the inclination angle of the character or the character string while watching the character on the screen, it is difficult to obtain an accurate inclination angle of the character or the character string. For this reason, there is a problem that a tilted character or a character string cannot be recognized in some cases, and the recognition rate is poor.

[0004] The present invention has been made based on the above circumstances, and has as its object to provide a character recognition method capable of improving the recognition rate of a tilted character or character string.

[0005]

According to the first aspect of the present invention, a character or a character string having an inclination with respect to a predetermined coordinate axis on the screen is recognized on the recognition screen. In the character recognition method, the character or the character string having the inclination is surrounded by a rectangular frame having an inclination substantially parallel to the inclination of the character or the character string, and the recognition range of the character or the character string is designated. Calculating the tilt angle of the rectangular frame with respect to the predetermined coordinate axis, determining the tilt angle of the character or character string based on the calculated tilt angle of the rectangular frame, and determining the tilt of the character or character string. The character or character string is recognized based on an angle.

According to a second aspect of the present invention, in the character recognition method according to the first aspect, the rectangular frame is
It is characterized in that three of the four vertices are designated and set. A character recognition method according to a third aspect of the present invention is the character recognition method according to the second aspect, wherein a straight line connecting a first designated point and a second designated point among the three points is defined by a horizontal direction. The tilt angle of the rectangular frame is calculated by obtaining an angle to be formed.

According to a fourth aspect of the present invention, in the character recognition method according to the third aspect, one angle is selected from a total of four angles that differ by 90 ° from the inclination angle of the rectangular frame. Thus, the inclination angle of the character or character string is determined. In the present invention, by specifying a recognition range by surrounding a character or character string inclined on the screen with a rectangular frame inclined in the same manner as the character or character string, by calculating the inclination angle of the rectangular frame, An accurate angle of inclination of the rectangular frame can be obtained.
For this reason, if the inclination angle of the character or the character string is determined based on the inclination angle of the rectangular frame, the inclination angle of the character or the character string can be accurately obtained. Therefore, even when the character or the character string is inclined, the recognition rate can be improved by recognizing the character or the character string in consideration of the inclination angle of the rectangular frame.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a schematic configuration diagram of a CAD system using a character recognition method according to an embodiment of the present invention, and FIG. 2 is a diagram illustrating a rectangular frame for specifying a recognition range in the character recognition method. The CAD system shown in FIG. 1 includes a scanner 12 as an image input device, a CRT display device 14, and a position input device (pointing device) 16.
, A central processing unit 18, an element file 22, an outline file 24, an area file 26, a recognition dictionary 28
And a recognition result storage unit 32. In the present embodiment, a case is considered in which a drawing is read and characters or character strings written in the drawing are recognized. In particular, here, it is assumed that only uppercase letters and numbers are recognized as characters. The characters written in the drawing may be printed characters or handwritten characters.

The scanner 12 optically scans the drawing,
This is input as binary image data. The CRT display device 14 displays the input result. The position input device 16 moves, for example, a crosshair cursor on the screen of the CRT display device 14 and indicates the position coordinates of the recognition range to specify the recognition range.
A mouse, trackball, or the like is used. In the present embodiment, as shown in FIG. 2, a recognition range is specified by surrounding a character or a character string to be recognized with a slanted rectangular frame W. In fact, if the character or character string to be recognized is tilted,
A rectangular frame W tilted by the same angle as the tilt angle of the character or character string (hereinafter also simply referred to as the tilt angle of the character)
To surround the character or character string to be recognized. The rectangular frame W is set by designating three points among the four vertices.

Here, as the binary image data input from the scanner 12, raster data based on a run length format used in facsimile and the like is adopted. This raster data has, for example, a starting point at which the pixel value changes from “0” (white) to “1” (black) on the scanning line, and a last coordinate at which the pixel value changes from “1” to “0”. Is the end point.

The element file 22 stores feature amounts (element data) extracted for each graphic element after raster data is subjected to labeling processing. Here, the center of gravity, area, circumscribed rectangle, moment, and the like are used as element data. The contour file 24 stores contour data obtained by extracting the edges of the raster data for each graphic element. The outline is approximated by a polygon and is composed of many line segments, and one outline is a closed polygonal line. For this reason, the contour line data has, as vectorized data, the line segments constituting each contour line. The element data is linked by the contour data and the file pointer or the address pointer on the memory.

The area file 26 stores area data indicating which area of the entire drawing read on the screen contains each line segment forming the contour line. The area data is managed in a tree structure (root, branch node, leaf). In data management using such a tree structure, each branch node represents an area in the middle of division on the drawing, and a leaf represents an area of the minimum division unit. The leaves and the contour data included in the region represented by the leaves are associated with each other by a pointer.
Is stored in the main memory 19. Therefore, when a certain region is specified, a leaf corresponding to the specified region is searched, and a line segment of a contour line included in the specified region can be determined based on a pointer associated with the leaf. it can.

The central processing unit 18 performs pre-labeling processing, feature quantity extraction processing, contour line extraction processing, etc., based on the input raster data in order to create element data, contour line data, and area data. And executes character recognition processing. In the recognition dictionary unit 28, reference patterns obtained from sample characters are registered in advance. The recognition result storage unit 32 stores the result recognized by the central processing unit 18.

Next, an operation procedure for enclosing a character or a character string to be recognized in a rectangular frame in the CAD system of the present embodiment will be described. FIG. 3 is a diagram showing an example of a procedure for designating the rectangular frame. Here, a case will be considered where a character string “ABCDEFG” inclined as shown in FIG. 2 is recognized. The operator first uses the mouse to
As shown in FIG. 3 (a), specifying the first-th point P ₁ at the lower left position in the string. Thus, as the first-th origin P ₁ point, xy coordinate system is set. This x
As shown in FIG. 4A, the y-coordinate system is a rectangular coordinate system in which the x-axis is in a horizontal direction and the y-axis is in an up-down direction. Is facing upwards.
Next, as shown in FIG. 3 (b), on a straight line along the arrangement direction of the first-th point character from P _1, specifies the second th point P ₂ at the lower right position of the string . 1st point P
Straight line connecting the _first and and the second point P ₂ is a side of the frame W. Thereafter, as shown in FIG. 3C, the second point P
Through _2, on a line perpendicular to a line connecting the first-th point P ₁ and a second th point P _2, which specifies a third-th point P ₃ at the upper right position in the string. By the third point P ₃ Sadamari height from one side of the frame W, thereby,
An inclined rectangular frame W is determined.

It is to be noted that the same frame W can be obtained by a method other than the method of specifying the three points as described above. FIG. 4 shows some ways of specifying three points to obtain the same frame W. here,
FIG. 4A shows how to specify the three points shown in FIG. There quadruplicate on whether to any of FIRST its four vertices P ₁ point in the frame W, of the two vertices the second th point P ₂ adjacent to the first-th point P ₁ Because there are two ways, there are a total of eight ways to specify three points. The operator can set the frame W by specifying three points in a manner that is easy to specify.

When the frame W is designated in this way, the central processing unit 18 determines that one side of the rectangular frame W, here a straight line connecting the origin (first point P ₁ ) and the second point P ₂ , An angle θ (0 ° ≦ θ <360 °) with the positive direction of the x-axis is calculated as the inclination angle of the frame W. By the way, as shown in FIG. 4, the calculated inclination angle θ of the frame W differs depending on how the three points are specified. Now, ₀ the inclination angle theta of the frame W theta calculated when the specified manner three points shown in FIG. 4 (a)
(0 ° ≦ θ ₀ <90 °), the inclination angle θ of the frame W calculated when the above eight ways are specified is
θ ₀ , θ ₀ + 90 °, θ ₀ + 180 °, θ ₀ + 270 °
Is one of For example, if the operator selects FIG.
When the three points are designated in the manner shown in (1), the central processing unit 18 calculates the inclination angle θ of the frame W as θ ₀ + 90 °. In the present embodiment, the inclination angle of the character is determined based on the inclination angle θ of the frame W. Since the frame W is set to be tilted in the same manner as the character, the tilt angle of the character (considered in the range of 0 ° to 360 °) is 9 from the calculated tilt angle θ of the frame.
One of the above four angles that differ by 0 °. Therefore, in the present embodiment, the command line is displayed on the screen of the CRT display device 14, and the operator recognizes the inclination angle of the character among the above four angles regardless of the designation method of any three points. You can choose.

FIG. 5 is a diagram showing an example of this command line. FIG. 5 shows a case where θ _{0 is set} to 30 ° and the three points shown in FIG. 4B are designated. On the command line, in addition to the inclination angle “120 °” of the frame W calculated by the central processing unit 18, angles “30 °”, “210 °”, and “300 °” different from the inclination angle by ± 90 ° and + 180 ° are displayed. Is done. To select the inclination angle “120 °” of the frame W, the right button of the mouse is pressed or the “W” key is pressed. To select another angle, press the key indicated on the command line. The character string "ABC shown in FIG.
In "DEFG", "30" is the correct inclination angle of the character, so the "Q" key is pressed. Thus, the inclination angle of the character is determined from the inclination angle θ of the frame W by the interactive processing.

Next, a processing procedure for character recognition in the character recognition method of the present embodiment will be described. FIG. 6 is a diagram for explaining a processing procedure for character recognition. Here, the character string “ABCDEF” which is inclined as shown in FIG.
Consider the case of recognizing "G". First, the drawing is read by the scanner 12 to obtain raster data (step
12). After performing a labeling process based on the raster data, the central processing unit 18 extracts a feature amount such as a circumscribed rectangle for each of the labeled graphic elements, and stores this in the element file 22. In addition, the edge of the raster data is extracted for each graphic element to obtain contour data, which is stored in the contour file 24. Next, area data is created based on the outline data in order to manage information on an area including each line segment constituting the outline in a tree structure. Then, the area data is stored in the area file 26. The above operation is automatically executed by the central processing unit 18 after the drawing is read by the scanner 12.

Next, the operator designates a recognition range by surrounding the character string to be recognized with a rectangular frame W using a mouse (step 14). Then, the command line shown in FIG. 5 is displayed on the screen, and the operator displays the character string “A” on the screen.
With reference to “BCDEFG”, a character inclination angle θ ₀ = 30 ° is selected. Next, the central processing unit 18 searches the area management tree to find a leaf corresponding to the area surrounded by the frame W. Based on the pointer stored in the main memory 19 in relation to the leaf, the outline data included in the frame W can be known, whereby the graphic element included in the frame W can be known. Here, in the present embodiment, only uppercase letters and numerals are handled as characters, and all the characters are composed of one component, so each graphic element represents one character. Next, information on the circumscribed rectangle of the graphic element is extracted from the element file 22 among the element data linked with the outline data. Thereafter, raster data is cut out using the information on the circumscribed rectangle (step 16). The circumscribed rectangle has a horizontal side and a vertical side with respect to the screen.

When characters are inclined and adjacent characters are close to each other, a part of the adjacent characters may be included in the range cut out by the circumscribed rectangle. Some of such characters become noise and cause a reduction in recognition accuracy. Therefore, in order to remove a part of such characters, labeling processing is performed in each of the cut-out ranges, and the largest raster data in the range is obtained.
Only the raster data is cut out again (step 18).
In general, when handling characters such as symbol marks and kanji, it is necessary to cut out raster data representing one character in consideration of the relative positional relationship between components.

Next, considering the inclination angle θ _{0 of the} character,
Character recognition processing is performed on the raster data cut out again in step 18 (step 22). That is, each raster data, by rotating the tilt angle theta ₀ only clockwise obtained characters step 14, after the inclination angle of the character to zero, newly extracts a feature quantity for each raster data. After obtaining a characteristic pattern based on these characteristic amounts, character recognition is performed by comparing the characteristic pattern with a reference pattern registered in the recognition dictionary unit 28.

Thereafter, a character string conversion process for arranging the recognized characters is performed (step 24). In this processing, xy
The arrangement order of the characters is determined by comparing the positions of the centers of gravity of the characters in the coordinate system. At this time, the way of determining the order differs depending on the inclination angle θ ₀ of the character. This is divided into the following four cases. Case (1) 0 ° ≦ θ <90 ° Case (2) 90 ° ≦ θ <180 ° Case (3) 180 ° ≦ θ <270 ° Case (4) 270 ° ≦ θ <360 ° In Case (1), When the x-coordinate and y-coordinate of the center of gravity of one character are both larger than the x-coordinate and y-coordinate of the center of gravity of the other character, insert one character to the right of the other character;
When the x-coordinate and y-coordinate of the center of gravity of one character are both smaller than the x-coordinate and y-coordinate of the center of gravity of another character, one character is inserted to the left of the other character. In Case (2), when the x coordinate of the center of gravity of one character is smaller than the x coordinate of the center of gravity of another character, and the y coordinate of the center of gravity of one character is larger than the y coordinate of the center of gravity of another character, One character is inserted to the right of the other character, and the x coordinate of the center of gravity of one character is larger than the x coordinate of the center of gravity of another character, and the y coordinate of the center of gravity of one character is the center of gravity of the other character. Inserts one character to the left of another when it is less than the y coordinate of. In Case (3), when the x-coordinate and y-coordinate of the center of gravity of one character are both smaller than the x-coordinate and y-coordinate of the center of gravity of another character, one character is inserted to the right of the other character, Also, the x coordinate of the center of gravity of one character, y
When each of the coordinates is larger than the x and y coordinates of the center of gravity of the other character, one character is inserted to the left of the other character. In Case (4), when the x coordinate of the center of gravity of one character is larger than the x coordinate of the center of gravity of another character, and the y coordinate of the center of gravity of one character is smaller than the y coordinate of the center of gravity of another character , One character is inserted to the right of the other characters, the x coordinate of the center of gravity of one character is smaller than the x coordinate of the center of gravity of another character, and the y coordinate of the center of gravity of one character is Inserts one character to the left of another character when it is greater than the y coordinate of the center of gravity of. In this way, each recognized character is arranged,
The final recognition result is displayed in a dialog box on the screen of the CRT display device 18.

In the character recognition method of the present embodiment, a recognition range is specified by surrounding a character or character string inclined on the screen with a rectangular frame inclined in the same manner as the character or character string, and specifying the recognition range. By automatically calculating the tilt angle, the tilt angle of the rectangular frame can be accurately obtained. Then, the operator looks at the characters or character strings on the screen,
Since the inclination angle of the character or the character string can be selected from four angles that differ by 90 ° from the inclination angle of the rectangular frame, the accurate inclination angle of the character or the character string can be easily determined by a simple operation. be able to. Therefore, when the character or the character string is inclined, the recognition rate can be improved by recognizing the character or the character string in consideration of the inclination angle of the character or the character string. Note that the program for executing the above-described operation is entirely or partially recorded or stored in a portable medium such as a floppy disk, a CD-ROM, or a storage device such as a hard disk. The program is read by the computer, and all or part of the operation is executed.

It should be noted that the present invention is not limited to the above embodiment, and various modifications can be made within the scope of the invention. In particular, in the present embodiment, a rectangular frame was used to specify the recognition range. However, the present invention is not limited to this. The column inclination angle may be calculated.

[0025]

As described above, according to the present invention, a recognition range is specified by surrounding a character or character string inclined on the screen with a rectangular frame inclined in the same manner as the character or character string,
By calculating the inclination angle of the rectangular frame, an accurate inclination angle of the rectangular frame can be obtained. Therefore, if the inclination angle of the character or character string is determined based on the inclination angle of the rectangular frame, Since the inclination angle of a character or a character string can also be accurately obtained, it is possible to provide a character recognition method capable of improving the recognition rate of a tilted character or a character string.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram of a CAD system using a character recognition method according to an embodiment of the present invention.

FIG. 2 is a diagram showing a rectangular frame for specifying a recognition range in the character recognition method.

FIG. 3 is a diagram showing an example of a procedure for specifying the rectangular frame.

FIG. 4 is a diagram showing an example of how to specify three points to obtain the same rectangular frame.

FIG. 5 is a diagram illustrating an example of a command line for selecting a tilt angle of a character.

FIG. 6 is a diagram for explaining a processing procedure when character recognition is performed in the character recognition method according to the embodiment.

[Explanation of symbols]

 Reference Signs List 12 scanner 14 CRT display device 16 position input device 18 central processing unit 19 main memory 22 element file 24 contour file 26 area file 28 recognition dictionary unit 32 recognition result storage unit

Claims (6)

[Claims] 1. A character recognition method for recognizing a character or a character string having a tilt with respect to a predetermined coordinate axis on the screen on a recognition screen, wherein the character or the character string having the tilt is converted to the character or the character. The recognition range of the character or the character string is designated by enclosing it in a rectangular frame having a tilt substantially parallel to the column tilt, and the tilt angle of the specified rectangular frame with respect to the predetermined coordinate axis is calculated. Determining the inclination angle of the character or the character string based on the inclination angle of the rectangular frame, and recognizing the character or the character string based on the inclination angle of the character or the character string. Method. 2. The character recognition method according to claim 1, wherein the rectangular frame is set by designating three of the four vertices. 3. An inclination angle of the rectangular frame is calculated by obtaining an angle between a horizontal line and a straight line connecting a first designated point and a second designated point among the three points. 3. The character recognition method according to claim 2, wherein: 4. The tilt angle of the character or character string is determined by selecting one angle from a total of four angles that differ by 90 ° from the tilt angle of the rectangular frame. Item 3. The character recognition method according to Item 3. 5. A recording medium on which a program for recognizing a character or a character string having a tilt with respect to a predetermined coordinate axis on the screen on a recognition screen is recorded. A procedure of specifying the recognition range of the character or the character string by enclosing it in a rectangular frame having an inclination substantially parallel to the inclination of the character or the character string, and calculating an inclination angle of the specified rectangular frame with respect to the predetermined coordinate axis And determining the inclination angle of the character or character string based on the calculated inclination angle of the rectangular frame. The character or character string based on the inclination angle of the character or character string. A recording medium storing a procedure for recognizing and a program for executing the procedure. 6. The recording medium according to claim 1, further comprising a step of setting the rectangular frame by designating three of the four vertices.