JPH04280393A - Character/graphic recognizing device - Google Patents

Abstract

PURPOSE:To improve the rate of character recognition by reducing danger to erroneously extract a graphic as a character and reducing the integration error of character strings concerning the character/graphic recognizing device. CONSTITUTION:A first condition memory 22 stores the character conditions of respective characters making size conditions as the sizes of respective characters correspondent to category conditions as the types of respective characters. A character separation part 30 extracts characters by collating the respective areas of images with the size conditions of the respective characters, and a character recognition part 3 recognizes characters by collating the characters extracted by the character separation part 30 with the dictionary of the type instructed by the category condition corresponding to the size condition.

Description

[Detailed description of the invention]

0001

[Industrial application field] CAD/CAM, computer/
With the spread of mapping, there is an increasing need for automatic input of drawings as a means of converting existing maps, mechanical diagrams, circuit diagrams, etc. into digital data. The elements written in drawings can be broadly divided into graphic elements and text elements. In automatic input of drawings, graphic elements are required to be output as vector data, and character elements are required to be output as recognized character codes.

0002

2. Description of the Related Art FIG. 10 shows a flowchart of conventional character/graphic separation processing.

A drawing is read by a scanner, converted into raster data, and the raster data is thinned and approximated by broken lines to be converted into vector data as shown in FIG. 11(A), and the processing of the drawing is executed. Vector data is searched and mutually continuous vectors are grouped as surrounded by the broken line in FIG. 11(B) (step 10).

[0004] Taking advantage of the characteristics that characters have a relatively small area and are isolated from other figures and characters, step 1
In loop 1, the size of each group is checked, and character elements are extracted based on the size. Specifically, groups that match a predetermined character size are extracted as character elements (steps 12 and 13). Furthermore, elements other than character elements are treated as graphic elements (step 14).

[0005] After this, character strings are created by integrating neighboring character elements according to spacing and angle conditions (
Step 15). In FIG. 11(C), the central character element A
However, there are string-1 and string-2 within the range that can be integrated. The distance L1 from the nearest character of both character strings to A,
L2 is calculated, and the smaller the distance, the higher the evaluation value is given. Also, the difference θ1 between the slope of character string -1 and the slope when A is integrated into character string -1, the slope of character string -2, and the slope of character string -1
2 and A are integrated, and the smaller the difference, the higher the evaluation value is given. Based on these two evaluations, it is determined whether to integrate into character string-1, character string-2, or neither. Next, in a loop of steps 16 and 17, the character image is cut out according to the position, size, and inclination of the character, and the character is recognized.

On the other hand, primitives such as straight lines, circles, and arcs are recognized as graphic elements.

[0007]

As described above, the condition used to separate characters from figures is the size of the characters. Traditionally, this condition has taken the form of a minimum font size and a maximum font size.

[0008] However, in reality, characters of different sizes are often written in a mixture of characters in one drawing. In such cases, conventionally it was necessary to determine the maximum size according to the largest font and the minimum size according to the smallest font. This poses a problem in that the greater the difference between large and small characters, the more graphics other than characters with sizes between them will be classified as characters.

[0009] The present invention has been made in view of the above points.
It is an object of the present invention to provide a character/figure recognition device that reduces the risk of erroneously extracting figures as characters, reduces erroneous integration of character strings, and improves character recognition rate.

0010

[Means for Solving the Problems] The character/figure recognition device of the present invention reads an image in which characters and figures are mixed, and recognizes each of the characters and figures. A first condition memory stores character conditions for each character in which conditions are matched with a category condition, which is the character type of each character, and a character is extracted by comparing each region of the image with the size condition of each character. and a character recognition unit that performs character recognition by comparing the characters extracted by the character separation unit with a dictionary of character types specified by category conditions corresponding to size conditions.

[0011] Also, character/figure recognition reads an image containing a mixture of characters and figures, extracts character elements, integrates neighboring character elements to create a character string, and recognizes each character element of the character string. The device includes a second condition memory that stores rules for predetermining character string division conditions and recognition conditions according to the number of characters in the character string; It has a character string post-processing unit that changes elements to graphic elements.

0012

[Operation] In the invention as claimed in claim 1, since multiple character sizes can be specified in the character condition, there is less risk of erroneously extracting a figure as large as a character, and it is possible to specify the size of each character according to the size condition. The character recognition rate improves because the character type is checked against the dictionary.

[0013] Furthermore, in the invention of claim 2, since constituent elements of a character string can be specified as rules, character extraction errors and character integration errors can be suppressed, and furthermore, character positions within a character string can be used as recognition targets. Narrow down categories and improve recognition rate.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows a block diagram of a first embodiment of a character/figure recognition device according to the present invention. In the figure, host computer 2
0, the drawing reading conditions and recognition conditions are specified and a drawing recognition request is made to the character/figure recognition device 21. The reading/recognition conditions are stored in a condition memory 22 within the device.

Based on the reading conditions, the drawing is read by the scanner 23 and the image is stored in the frame memory 24. At this time, the scanner control section 26 within the main control section 25 controls the scanner. The image in the frame memory 24 is thinned by the image processing section 27. A vector generation unit 28 generates vector data for the thinned image and stores it in a vector memory 29. A character separation unit 30 groups the vector data and extracts character elements. At this time, the characters stored in the condition memory 22 are referred to.

The separated character elements are recognized by a character recognition unit 31 by cutting out an image from the frame memory character by character. The recognized results are stored in code memory 32.

On the other hand, the vector calculation unit 33 recognizes graphic primitives (straight lines, circles, arcs, etc.) according to the recognition conditions stored in the condition memory 22. The results are stored in vector memory 29. The vector data and character data thus obtained are output via the communication control unit 34 to a host computer having a CAD or mapping system.

When reading, for example, the image shown in FIG. 2 by the scanner 23, the following character conditions are given.

Character conditions ■ Size conditions: In Figure 3,
Corresponds to the large letter (PID) in the upper right corner
Category conditions: Alphabetic characters Character conditions■ Size conditions:
Corresponds to the characters near the symbol (TB, PDT)

Category conditions: Alphabetic characters Character conditions■ Size conditions: Characters near the line (204-104-1)
to
equivalent size
Category conditions: Numbers and symbols font conditions (font) shall be the same.

The raster data of the image shown in FIG. 2 is subjected to line thinning and broken line approximation, and then the graphic character separation process shown in FIG. 3 is executed.

In FIG. 3, in step 40 the vector
Search the data and group mutually consecutive vectors together. In the loop of step 41, vertical and
Measure the horizontal size. In step 43 within the loop of step 42, it is determined whether all vector groups meet all character size conditions. If any of the conditions is met, a code indicating the met condition is given (step 44), and if it is not met, it is designated as a graphic element (step 45).

Next, in step 46, the distance between adjacent character elements (the length of the straight line connecting the center points of the circumscribed rectangle of the character element) and the angle (the length of the straight line connecting the center points of the circumscribed rectangle of the character element, A character string is created by integrating character elements that are adjacent to each other depending on the angle between the main scanning line and the main scanning line. At this time, the angle of the character string (the angle between the main scanning point and the straight line connecting the center points of the first character and the last character of the character string) is given as the angle of the character elements constituting the character string. As a result, the character string integration result shown in FIG. 4 is obtained from the image shown in FIG. 2. In FIG. 4, the integrated character string is surrounded by a broken line, and the character condition that satisfies the character string is indicated by a circled number on the left shoulder of the dashed line.

In the loop of step 47, font conditions and category conditions of character conditions that are compatible with all character strings are set in the character recognition unit 31.

In step 48, an image is cut out for each character in each character string according to the character position and character string angle, and the characters are recognized. At this time, matching is performed only with the target dictionary using the font condition and category condition specified in the character condition. Since string-1 in FIG. 4 was extracted under character condition 1, only the alphabetical dictionary is compared. Similarly, strings-2, 3, and 4 are checked against only an alphabetic dictionary according to character condition 2, and strings-5 and 6 are checked against only a numeric dictionary according to character condition 3.

As described above, according to the first embodiment, since a plurality of character conditions can be specified, the possibility of erroneously extracting relatively small figures as characters is reduced, and only characters of the same size are extracted as characters. To target column integration,
String integration errors can be reduced. Furthermore, since it is possible to specify the dictionary font and category to be checked for each character condition, it also contributes to improving the character recognition rate.

FIG. 5 shows a block diagram of a second embodiment of the device according to the invention. In this figure, the same parts as in FIG. 2 are designated by the same reference numerals, and their explanations will be omitted.

In FIG. 5, a character string post-processing section 50 reads rule parameters, which are rules for character strings written in a drawing, from a condition memory 52, analyzes them, and reorganizes the character strings in accordance with the rules. Here, the data of the element whose character attribute has changed to a graphic attribute is moved from the code memory 32 to the vector memory 29.

Furthermore, a target category is set for each character in the code memory 32 as a recognition condition. The separated character elements are recognized by the character recognition section 31 by cutting out an image from the frame memory character by character. At this time, matching is performed only with the dictionary of the category set as the recognition condition.

The drawing rules for the drawing shown in FIG. 6 are as follows.

1. There are a 1-character string and a 3-character string.

2. A string of one character is an alphabetic character.

[0032] The first character string of 3.3 characters is an alphabetic character,
The last two characters are numbers.

From this convention, rules are set. Here, the rule grammar is defined as follows.

Sentence: IF (condition) T
HEN
process
ENDIF condition :={
Condition for the number of characters in a string | Condition for the number of vectors in a character string
String composition Number of characters condition: = {Number of characters in string composition =
=N}
Number of character constituent vectors condition: = {Number of character constituent vectors Comparison operator N} Comparison operator:
＝｛＜｜＞｜≦｜≧｜==｜! =} Processing
:= {Separation condition | Recognition condition | Attribute condition |
Basic sentence pattern} Separation condition := Separation condition = (
{N character string | (separation condition + N character separation string) |
(Separation condition or separation condition)}
) Recognition condition := Recognition condition = (
{Category | (Category condition/Category condition}
) Category condition := {Category | (Category condition or category)} Attribute condition := {Shape attribute} Category := {Alphabet | Number | Symbol} Here, ==
means that the left and right sides are equal! = indicates that the left side and right side are not equal, / indicates a separation between the Nth character and the N+1th character, | indicates an OR condition, and N indicates a positive integer value.

When the drawing convention of FIG. 6 is expressed as a rule, it is as follows.

(a) If a character string of one character exists, the number of vectors constituting the character is checked, and if the number of vectors is two or less, it is determined as a figure. If it is 3 or more, it is recognized as an alphabetic character. (b) If a string of two characters exists, separate each character into two strings.

(c) If a string of three characters exists, the first character is recognized as an alphabetic character and the remaining two characters are recognized as numbers.

(d) If a 4-character string exists, it is separated into a 3-character string and a 1-character string.

(e) If a 5-character string exists, either separate it into a 3-character string and a 2-character string, or create a 3-character string and a 1-character string x 2. Separate into columns. (f) If a string of 6 characters exists, a string of 3 characters ×
Separate into 2.

When the rule in FIG. 6 is described according to the above rule grammar, it is as follows.

(a) IF (number of characters==1) THENIF
(Number of character constituent vectors ≦ 2) THEN ENDIF as figure attribute Recognition condition = (English characters) ENDIF (b) IF (Number of characters = = 2) THEN Separation condition = (1 character string + 1 character string) ENDIF (c) IF (Number of characters = = 3) THEN Recognition condition = (Alphabets/Numbers/Numbers) ENDIF (d) IF (Number of characters = = 4) THEN Separation condition = (1 character string + 3 character string) ENDIF (e) IF (Number of characters = =5) THEN Separation condition = (2 character string + 3 character string) or (3 character string + 1 character string + 1 character string) ENDIF (f) IF (number of characters = = 6) THEN Separation condition = (3 characters Character string + 3 character character string) ENDIF By applying thinning and broken line approximation to the raster data of the image in FIG. The string integration result shown in FIG. 8 is obtained. In FIG. 8, the integrated character string is surrounded by a broken line, and the number of the character string is indicated by a circled number on the left shoulder.

Next, in step 62 of the loop for all character strings in step 61 of FIG. 7, the character strings are classified into cases.

[0043] Rules (b), (d), (e),
In the case where the separation condition (f) exists, in step 63, characters that can be separated according to the separation condition are searched for, an evaluation value between each character is determined, and the character string is separated according to this. In the case where the character string has the recognition conditions of rules (a) and (c), in step 64, the category specified by the above recognition condition is set in the category field on the code memory of each character composing the character string, and this Mark the string as processed. Rule (a) in string
In the case where there is a condition (attribute condition) for making the character a graphic attribute, it is determined in step 65 whether or not this character satisfies the attribute condition. If the character is satisfied, the character is deleted from the code memory and registered in the vector memory in step 66, and if it is not satisfied, the recognition condition is set in the category term of the character on the code memory in step 67, and the processing is completed.

Next, within the loop for all character strings at step 68, a loop for all characters at step 69 is performed. Here, an image for one character is cut out according to the character position and character string inclination in the code memory, the category condition specified in the category item of the code memory is set in the character recognition section, and the character image is recognized.

Note that in step 60, step 4 in FIG.
Steps 10 to 15 in Figure 10 instead of processing steps 0 to 45
You may also perform the following processing.

Here, a case will be described in which the post-processing shown in FIG. 7 is applied to the character string integration result shown in FIG. 8. Note that the characters in the character string are called the first character and the second character in order from left to right in the figure. Further, the evaluation value of the connection state between the Nth character and the N+1th character is assumed to be E(N).

(1) Since the character string (■) consists of one character, it is recognized as an alphabetic character according to rule (a).

(2) Since the character string of becomes. The right side is recognized as an alphabetic character.

(3) Since the character string ■ consists of three characters, the first character is an alphabetic character according to rule (c), and the remaining two
Recognize characters as numbers.

(4) Since the character string ■ is composed of 4 characters, rule (d) determines the connected state of the first character and the second character and the connected state of the third character and the fourth character based on the distance and angle. evaluate. In this example, since E(1)<E(3), the first character and the second character are separated. separated first
Characters are recognized as alphabetic characters by applying (a). 2nd to 4th
Rule (c) is applied to the characters as an example of three characters, and they are recognized as letters, numbers, and numbers.

(5) Since the character string (■) is composed of five characters, it can be separated into two or three characters according to rule (d).

To separate into two character strings, 1) Separate between the second and third characters.

2) Separate between the third and fourth characters.

To separate into three character strings, 3) Separate between the first and second characters, and between the fourth and fifth characters. Therefore, the connection state between the four characters is evaluated.

E(2), E(3) and (E(1)+E(4)
)) Compare ÷2.

In this example, since the last one has a high evaluation, the characters are separated between the first and second characters, and between the fourth and fifth characters.

The first character is made into a figure by applying rule (a).

[0058] Applying rule (c) to the second to fourth characters,
Recognize as letters, numbers, and numbers.

The fifth character is recognized as an alphabetic character by applying rule (a).

(6) Since the character string ■ consists of 6 characters, the third and fourth characters are separated according to rule (e) and rule (c) is applied to each character to form alphabetic characters. Recognize as numbers/digits.

By the above post-processing, the reorganized character string integration result shown in FIG. 9 is obtained.

As described above, according to the second embodiment, since the structural requirements of a character string can be specified as a rule, it is possible to prevent character extraction errors and character string integration errors, and to determine the recognition target based on the order of characters in the character string. Since the categories can be narrowed down, it also contributes to improving the recognition rate.

[0063]

[Effects of the Invention] As described above, according to the character/figure recognition device of the present invention, it is possible to reduce the possibility of erroneously extracting a figure as a character, reduce erroneous integration of character strings, and improve character recognition. It is extremely useful in practice.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a first embodiment of the device of the present invention.

FIG. 2 is a diagram showing a read image.

FIG. 3 is a flowchart of a first embodiment of character/figure separation processing of the apparatus of the present invention.

FIG. 4 is a diagram showing the result of character string integration.

FIG. 5 is a block diagram of a second embodiment of the device of the present invention.

FIG. 6 is a diagram showing a read image.

FIG. 7 is a flowchart of a second embodiment of character/figure separation processing of the apparatus of the present invention.

FIG. 8 is a diagram showing the result of character string integration.

FIG. 9 is a diagram showing the result of reorganized character string integration.

FIG. 10 is a flowchart of an example of character/figure separation processing of a conventional device.

FIG. 11 is a diagram for explaining character/figure separation processing.

[Explanation of symbols]

22, 52 Condition memory 28 Vector generation section 29 Vector memory 30 Character separation section 31 Character recognition section 32 Code memory 50 String post-processing section

Claims (2)

[Claims] [Claim 1] A character/figure recognition device that reads an image containing a mixture of characters and figures and recognizes each of the characters and figures, which has a size condition, which is the size of each character, and a category condition, which is the type of each character. a first condition memory (22) that stores character conditions for each of the associated characters; a character separation unit (30) that extracts characters by comparing each region of the image with the size condition of each character; The character recognition unit (31) performs character recognition by comparing the characters extracted by the character separation unit (30) with a dictionary of character types specified by the category condition corresponding to the size condition. Character/figure recognition device. [Claim 2] Character and figure recognition that reads an image containing a mixture of characters and figures, extracts character elements, integrates neighboring character elements to create a character string, and recognizes each character element of the character string. The device includes a second condition memory (52) that stores rules for predetermining division conditions and recognition conditions for a character string according to the number of characters in the character string, and a second condition memory (52) for recursively dividing the character string according to the rules. A character/graphic recognition device comprising a character string post-processing unit (50) that changes character elements into graphic elements according to recognition conditions.