JPH09147114A - Pattern recognizing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent even a nondefective from being erroneously recognized as a defective by finding the matching degree of the both patterns of a binary expression standard unknown pattern and a binary expression expansion/ contraction pattern by comparing the patterns and recognizing the unknown pattern by defining the unknown pattern as the one matching with the binary expression pattern corresponding to the found highest matching degree. SOLUTION: An analog image is obtained by performing a raster scanning for a character string with a camera 1 and the analog image is made a digital image by an A/D converter 2. Each character is segmented as a unit pattern from the digital image by a pattern segment means 4, a binary expression standard pattern is generated from a standard pattern by a feature extraction means 5, the binary expression unknown pattern is generated from an unknown pattern and the matching degree of the both pattern is found. In the same way, the binary expression unknown pattern and a binary expression expansion or contraction pattern are compared, the matching degree is found by a rotation matching 6, the highest matching degree is defined as the binary expression pattern and the unknown pattern is recognized.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to patterns such as characters.
The present invention relates to a method for recognizing a character, and more particularly to a character recognition method for recognizing an arbitrary character by cutting out an arbitrary character from an image obtained by capturing a character string written on an IC or the like.

[0002]

2. Description of the Related Art Conventionally, a method of mechanically reading a character in a character string written on an IC or the like is such that a character is picked up by a TV camera, the image is taken into a computer, and a pattern matching method is used. Character recognition methods have been proposed.

In the pattern matching method, for each character line in the visual field taken in by the computer, the left and right ends of each character are detected by detecting the left and right ends of each character from the projection on the axis parallel to this character line. Characters are cut out, and the individual characters cut out while moving in the direction of the character string between the images are compared with all standard characters (hereinafter referred to as text) registered in advance for each character to be recognized, The matching degree of the character to be recognized is detected, and the character of the text having the highest matching degree is recognized as the character.

When the character lines in the visual field are not parallel to the scanning direction for recognition, they are made parallel by rotating the entire screen, or after the characters are cut out, the characters are cut out. It is corrected by rotating each time.

However, in such a character recognition method, an image obtained by raster-scanning an unknown character is stored as a bit map, or text is also stored in the bit map in advance, and the bit map of the unknown character is stored. And text are compared bit by bit to check the degree of coincidence.

In this method, since the number of bits is large, it takes time for comparison. Further, when the unknown character is rotated with respect to the scanning direction, even if the unknown character matches the text, the degree of matching is low. And the rotation angle is ± 5
There was a problem that the same character could not be recognized if it exceeded °.

Therefore, when the unknown character is rotated, it is necessary to rotate at least one of the unknown character and the text and compare them. However, in the conventional character recognition method, it takes time to rotate the character pattern by performing bitmap processing, which is an obstacle to improving the speed of character recognition. As described above, the conventional method of recognizing a pattern such as a character by pattern matching has a problem that the recognition speed is slow when the pattern is rotating in the scanning direction.

Therefore, in order to solve this problem, the inventors of the present invention, even if the pattern of characters or the like obtained by scanning with the TV camera is rotating in the scanning direction, A patent application has been filed for a pattern recognition method and device capable of quickly recognizing a pattern (JP-A-5-12491).

The pattern recognition method previously applied will be described below. As shown in FIGS. 5 to 8, this is the envelope of the standard pattern (the envelopes of A and B are shown in FIG. 3) and the center of gravity O of the area surrounded by this envelope. Ask for. Next, as shown in FIG. 5, among the end points of this standard pattern, a line segment connecting the end point farthest from the center of gravity O and the center of gravity O, and this line segment as a radius with the center of gravity O as the center. And ask for a circle 11 to do.

Further, a plurality of (for example, 16) diameter line segments 12 that are the length of the diameter of the circle 11 and are arranged at equal angles passing through the center of gravity O and that divide the circle 11 into equal parts are sequentially obtained.
As shown in FIG. 6, each diameter line segment 12 is equally divided into a plurality of line elements (for example, 32), and constituent elements (for example, character lines) of each standard pattern at positions corresponding to these line elements. Whether or not is represented by a binary value.

The center of gravity O of the area surrounded by the envelope of the pattern
The characteristics of the pattern are extracted depending on whether or not there is a constituent element of the pattern on the radiation centered at. The characteristics of the pattern extracted in this way are expressed in binary, and the pattern expressed in binary is used as a binary expression pattern, which is a binary value related to a standard pattern known in advance. The representation is determined and stored as a binary representation standard pattern (shown in FIG. 6). This binary representation standard pattern is created and stored in advance for each standard pattern.

Then, when an unknown pattern in a tilted state (shown in FIG. 7) to be recognized is input, this unknown pattern is processed in the same manner as the method for obtaining the binary representation standard pattern. The binary expression pattern relating to the unknown pattern is obtained as the binary expression unknown pattern, as shown in FIG.

Therefore, while comparing at least one of the binary expression unknown pattern and the binary expression standard pattern by an equal angle in sequence, the two patterns of these binary expressions are compared, and both patterns are compared. -Determine the degree of coincidence.

Further, similarly, the other binary expression standard pattern and the binary expression unknown pattern are compared with each other to obtain the degree of coincidence, which corresponds to the binary expression standard pattern having the highest degree of coincidence. The unknown pattern is recognized as matching the standard pattern.

In the above method of the present inventors, since both patterns are expressed with a significantly smaller number of bits than a bit map, patterns of characters and the like obtained by a TV camera are scanned. Even if the pattern is rotating with respect to the direction, the pattern can be recognized quickly.

[0016]

However, in such a method, as shown in FIGS. 9 to 10, unknown patterns are generated.
If there is thickening or thinning of the lines that make up the characters in the standard pattern shown in FIG.
In the value expression unknown pattern, the number of points that do not match the binary expression standard pattern to be compared increases, and the degree of matching decreases.

If the degree of coincidence is 90% or more, it is recognized as a good product. Therefore, when there is a fatness or a thinness, although the patterns actually match, the degree of matching deteriorates, and there is a problem that the product is recognized as a defective product.

[0018]

The present invention provides an unknown pattern.
In this pattern recognition method for comparing an unknown pattern with a plurality of standard patterns by a pattern matching method, the presence / absence of the constituent elements of the plurality of standard patterns is checked.
The data obtained by expressing it as a value is obtained for each standard pattern as a binary expression standard pattern for each standard pattern, and this is registered as a comparison pattern. Each expression standard pattern is expanded a predetermined number of times to obtain a binary expression expansion pattern related to the binary expression standard pattern, and this is registered as a comparison pattern, and each binary expression standard pattern is registered. Each of the patterns is contracted a predetermined number of times to obtain a binary expression contraction pattern related to the binary expression standard pattern, and this is used for comparison.
The binary expression pattern for the unknown pattern is obtained as the binary expression unknown pattern in the same manner as the method for obtaining the binary expression standard pattern. And the binary expression standard pattern are compared to obtain the degree of coincidence between both patterns, and similarly, the binary expression unknown pattern and the binary expression 2
The degree of coincidence of both patterns is obtained by comparing with the value expression expansion pattern, and further, the degree of coincidence of both patterns by comparing the binary expression unknown pattern and the binary expression contraction pattern. Of the coincidences thus obtained, the unknown pattern is determined to match the pattern of the binary representation corresponding to the highest coincidence.
It is designed to recognize

Furthermore, the present invention is a binary representation standard pattern.
In addition to obtaining the envelope of each standard pattern, the center of gravity of the area surrounded by this envelope is obtained, and the end point of the standard pattern farthest from the center of gravity and the center of gravity are determined. The connecting line segment is obtained, the circle whose center is the center of gravity with this line segment as the radius is obtained, and a plurality of diametrical diameter line segments that equally divide this circle into equal angles are obtained through the center of gravity of this circle. The diameter line segment is equally divided into a plurality of line elements, and the presence or absence of the constituent elements of the standard pattern at the positions corresponding to the respective line elements is composed of binary data.

Further, in the present invention, the binary expression expansion pattern is obtained by expanding the data in each line element of the binary expression standard pattern by expanding the data in the longitudinal direction of the line element a predetermined number of times. It was made to become.

Further, according to the present invention, the binary expression contraction pattern is obtained from the data in each line element of the binary expression standard pattern which is contracted in the longitudinal direction of the line element a predetermined number of times. It was made to become.

Further, according to the present invention, the unknown pattern is recognized by the binary expression unknown pattern or the registered 2 pattern.
By comparing one of the two patterns of the value representation while rotating them by an equal angle sequentially, the degree of agreement with the binary representation unknown pattern is calculated, and the registered binary representation with the highest degree of agreement is obtained. The unknown pattern is recognized as the one corresponding to the pattern.

[0023]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of the present invention is shown in FIGS.
It will be described in detail based on. FIG. 1 shows an embodiment of the present invention, and is a flow chart showing the procedure of pattern recognition, FIG. 2 is a diagram showing the main parts, FIG. 3 is a view showing the envelopes of characters A and B, FIG. 4 shows a characteristic pattern obtained by dividing the diameter line segment into 32 line elements, and expressing the presence or absence of the constituent elements of the pattern at the position corresponding to each line element in a binary value, and a comparison using this pattern. It is a figure explaining. The same parts as those in the conventional example have the same names and the same numbers, and the description thereof is omitted.

1 to 4, 1 is a TV camera,
A character string to be recognized is raster-scanned to obtain an analog image. An A / D converter 2 converts an analog image signal into a digital image signal represented by 1 or 0. A memory 3 stores a digital image. 4
Is a pattern cutting means for reading a digital image from the memory 3 and cutting each character from the image as a unit pattern.

Reference numeral 5 is a feature extraction means for extracting the features of the pattern of the character output from the pattern cut-out means 4, and a binary expression standard pattern is generated from the standard pattern to obtain an unknown characteristic. An unknown binary expression pattern is generated from the pattern. For example, if the character A is cut out by the pattern cutting means 4 as shown in FIG. 5 and input, the envelope of the character A is first obtained as shown in FIG. The envelope is a line formed by connecting the end points on the outer circumference of the character A, and is a line that does not bend inward.

Further, the feature extracting means 5 obtains the center of gravity O (FIGS. 5 and 7) of the area surrounded by the envelope and, as shown in FIGS. 5 and 7, among the end points of the character A. A circle 11 centered on the center of gravity O and 16 diameter line segments 12 that are the length of the diameter of the circle 11 and that equally divide the circle 11 through the center of gravity O are obtained. In addition, in FIG. 5 and FIG. 7, the diameter line segment 12 is a, b, c.
.., a ', b', c '...

After obtaining the diameter line segment 12 in this way, each diameter line segment 12 is further equally divided into 32 line elements,
The character A at the position corresponding to each of the 32 line elements is extracted, and the characteristic data of the character A expressed in binary is generated. This characteristic data is output as a binary expression unknown pattern for the character A. 6 and 8
In order to facilitate the explanation, the diameter line segment 12 is divided into nine line elements, and the line segments 12 are divided into positions corresponding to the line elements in FIGS.
The characteristic data of the character A is shown. The actual binary expression unknown pattern is composed of 512 (16.times.32) bits of data.

The text storage means 7 stores in advance a binary representation standard pattern (hereinafter referred to as text T ₁ ) created by extracting the characteristics of the standard pattern of each character. . This text T ₁ is displayed on a TV with the known characters as a standard pattern upright as shown in FIG.
An image is taken by the camera 1 and processed in the procedure described below. Then, the characteristic extraction means 5 generates a binary expression standard pattern (text T ₁ ) as characteristic data of the binary expression of the standard pattern and stores it in the text storage means 7. Therefore, the standard pattern - the text T ₁ of the character related emissions, binary representation unknown pattern - down as well as one character per 512 bits of the de - is represented by data.

Reference numeral 6 is a rotation matching means, which is a binary expression unknown pattern from the feature extracting means 5, a text T ₁ read from the text storing means 7 and a text T described later.
₂ and T ₃ are compared bit by bit. Since a plurality of characters are stored as texts in the text storage means 7, the rotation matching means 6 sequentially reads the texts one by one from the texts according to a predetermined order, and the rotation extraction means 6 reads the texts from the feature extraction means 5. It is compared with the binary expression unknown pattern.

Reference numeral 9 denotes an expansion / contraction processing means, which represents the data in each line element of the binary expression standard pattern (text T ₁ shown in FIG. 4) relating to the standard pattern feature-extracted by the feature extraction means 5. , This is for expanding or contracting a predetermined number of times in the longitudinal direction of this line element, and as shown in FIG. 4, the expanded binary representation standard pattern T ₁ is
It is stored and registered in the text storage means 7 as a binary expression expansion pattern (text T ₂ ).

Similarly, as shown in FIG. 4, the contracted binary expression standard pattern T ₁ is stored and registered in the text storage means 7 as a binary expression contracted pattern (text T ₃ ). It

An arithmetic unit 8 identifies an unknown pattern on the basis of an output from the rotation matching means 6, and a bitwise logical product and an exclusive OR of both patterns from the rotation matching means 6. (XOR) is calculated to obtain the degree of coincidence between both patterns.

Next, a method of creating a text will be described. In the present invention, according to the procedure shown in FIG.
As shown in FIG. 9, for example, in addition to the standard pattern of the letter A, this standard pattern is expanded and contracted, and as shown in FIGS. 9 and 10, a thick and thin pattern of about ± 20%, respectively. −
Are created and respectively adopted as expansion pattern and contraction pattern, and a total of three basic standard comparison patterns are created and registered in the text storage means 7 as text. .

Therefore, first, regarding the registration of the text,
This will be described with reference to FIGS. An analog image of the character A (FIG. 3) as a standard pattern photographed by the TV camera 1 is converted into a digital image by the A / D converter 2 and stored in the memory 3. This digital image is read from the memory 3 by the pattern cutting means 4 (step P1), an arbitrary character, for example, the character A is cut out from this digital image (step P2), and is input to the feature extracting means 5. It

In the feature extracting means 5, as shown in FIG. 3, the envelope of the character A is obtained by the same procedure as described in the conventional example (step P3), and the area surrounded by the envelope is determined. The center of gravity O (FIGS. 5 and 7) is obtained (step P
4). Next, as shown in FIGS. 5 and 7, a circle 11 centered on the center of gravity O is obtained (step P5).

Next, 16 diameter line segments 1 of the diameter of the circle 11 and passing through the center of gravity O and equally dividing the circle 11
2 is obtained, and further, each diameter line segment 12 is equally divided into 32 line elements, the character A at the position corresponding to each of these 32 line elements is extracted, and the characteristic of the character A expressed in binary Data is generated (step P6) and is stored in the text storage means 7 as a binary representation standard pattern for the character A based on this characteristic data. In this way, a binary expression standard pattern for a plurality of standard patterns is created and registered in the text storage means 7 as text (step P9).

Next, when the text of the expanded pattern is created, for the binary expression standard pattern generated for the standard pattern in step P6, each line element of this binary expression standard pattern. After the expansion process is performed on the data in the line element a predetermined number of times in the longitudinal direction of the line element (step P).
8), text storage means 7 as a binary expression expansion pattern
Is registered (step P9).

Similarly, when the text of the contraction pattern is created, the data in each line element of the binary expression standard pattern relating to the standard pattern from the feature extracting means 5 is expanded and contracted. In 9, the contraction process is performed a predetermined number of times in the longitudinal direction of the line element (step P8), and then registered in the text storage means 7 as a binary expression contraction pattern (step P9).

The number of times of expansion processing or contraction processing is determined depending on the expansion pattern or contraction pattern to be used as a text up to what percentage of the standard pattern, thick or thin.

Next, a method of actually recognizing an unknown pattern will be described. Similar to the case of registering a standard pattern or the like, an analog image of an unknown pattern taken by the TV camera 1 is converted into a digital image by the A / D converter 2 and stored in the memory 3. This digital image is read from the memory 3 by the pattern cutting means 4 (step P1), an arbitrary character, for example, the character A is cut out from this digital image (step P2) and sent to the feature extracting means 5. It

In the feature extracting means 5, as in the case described above, the character A has an envelope curve as shown in FIG. 3 (step P3), and the center of gravity O of the area surrounded by this envelope curve (see FIG. 5, FIG. 7) is obtained (step P4), and as shown in FIGS. 5 and 7, a circle 11 centered on the center of gravity O is obtained (step P5), and 16 circles are equally divided through the center of gravity O. Diameter line segment 12 and each diameter line segment 12 are equally divided into 32 line elements, and the letter A is extracted according to this and the letter A is expressed in binary.
The binary expression unknown pattern relating to the unknown pattern is created as the characteristic data of (step P6) and is output to the rotation matching means 6.

The rotation matching means 6 compares the binary expression unknown pattern with the text read from the text storage means 7 (step P7). At this time, the text storage means 7 stores a plurality of characters as standard patterns and texts T ₁ , T ₂ and T ₃ relating to their weight and thinness. The text is read character by character according to the order in which it is compared with the binary expression unknown pattern.

For example, in order to simplify the explanation, the patterns of FIGS. 6 and 8 will be explained. It is assumed that the feature extraction means 5 outputs the binary expression unknown pattern shown in FIG. 8 and the text storage means 7 reads out the text shown in FIG. Then, in the rotation matching means 6, a pattern constructed by sequentially arranging data of rows a, b, c, d shown in FIG. 6 and a ', b', c '. The pattern formed by arranging the data of the row of d'in order is compared for each bit, and the logical product and the exclusive logical product (XO
R) is obtained, sent to the calculation unit 8, and output respectively. At this time, the number of coincident bits is represented by the number 1 of the logical product, and the number of non-coincident bits is represented by the number 1 of the XOR.

Next, the lines are shifted by one line, and
A pattern formed by sequentially arranging data of rows b, c, and d
And b ′, c ′. A pattern formed by arranging the data of the rows d'and a'in order is compared bit by bit, processed in the same manner, and the logical product and XOR are obtained respectively. Sent out. In the same manner, the patterns formed by arranging the lines in FIG. 8 while shifting them line by line and the patterns of FIG. 6 are compared one after another.

In this manner, the binary expression unknown pattern shown in FIG. 8 is shifted line by line, that is, compared with the text while rotating. When the comparison is completed for one text, the next text is read and compared in the same procedure. In this way, the binary representation unknown pattern is compared for all texts.

FIG. 4 shows an embodiment of the present invention in which a diameter line segment is divided into 32 line elements, and the presence or absence of pattern constituent elements at positions corresponding to each line element is represented by a binary pattern.
That is, data of 32 bits per line is combined into 16 lines to form one pattern. Therefore, in FIG. 4, the text T ₁ (binary representation standard pattern) relating to the standard pattern is shown.
2) and a binary expression expansion pattern (text T ₂ ) and a binary expression expansion pattern (text T ₃ ) created by expanding and contracting these patterns, respectively.

The comparison between these texts and the binary expression unknown pattern M is performed by ANDing and XO for 512 bits.
It is performed by taking R. However, in order to facilitate processing,
Both patterns are divided into 32 bits for pattern comparison. Here, M is a bit number and only the first digit is described.

As is clear from this figure, the text T ₁
And the logical product of the binary expression unknown pattern M is 7 bits "
1 ”(S = 7) and 10 bits of XOR are“ 1 ”(F = 10). Similarly, the logical product of the text T ₂ and the binary expression unknown pattern M is 8 bits. It is "1" (S = 8), and 8 bits of XOR are "1" (F =
8), the logical product of the text T ₃ and the binary expression unknown pattern M is “1” in 4 bits (S = 4), and the XOR is 8
The bit is "1" (F = 8).

In this way, in the rotation matching means 6, the logical product obtained by comparing both patterns bit by bit and the value of XOR are input to the arithmetic unit 8. Therefore, in the computing section 8, the degree of coincidence R = ΣS / (ΣS + ΣF) of the characters is calculated from the logical product and the XOR value for each pattern to be compared, and the unknown pattern is determined (step P7). ). Here, S is the number of matching bits and F is the number of non-matching bits.
The degree of character mismatch R ′ _S = ΣF / (ΣS + ΣF) is also obtained. However, since the coincidence bit is obtained from the logical product, the coincidence bit number S represents only the number where the number of bits of both patterns to be compared is "1", and both are "
It does not include numbers that are 0 ".

[0050] As described above, 'seek _S (or degree of matching R), similarly inconsistency R for other text' character of inconsistency R _S is obtained. Such 'the minimum value of _S inconsistency R of the text' inconsistency R obtained in the adopted as _S. Similarly, one after another seek inconsistency R _'S for other text, selects the text corresponding to the minimum inconsistency R's of all the inconsistency.

Even if the dissimilarity R's is 0.5 or less, when the second smallest disagreement R _'S is less than half, the character of the text is recognized as the character of the unknown pattern. However, in other cases, it is determined to be unrecognizable.

[0052]

According to the present invention, in a pattern recognition method for comparing an unknown pattern with a plurality of standard patterns by a pattern matching method and recognizing the unknown pattern, a plurality of standard patterns are recognized. The data obtained by expressing the presence or absence of the constituent elements of the binary in binary is obtained as a binary expression standard pattern for each standard pattern for each standard pattern, and this is used for comparison. The binary expression standard pattern is expanded a predetermined number of times to obtain a binary expression expansion pattern for the binary expression standard pattern, and this is used as a comparison pattern. The binary expression standard pattern is contracted a predetermined number of times to obtain the binary expression contraction pattern for the binary expression standard pattern, and this is registered as a comparison pattern. Binary representation standard In the same way as the method for obtaining the pattern, the binary expression pattern relating to the unknown pattern is obtained as the binary expression unknown pattern, and the binary expression unknown pattern and the binary expression standard pattern are obtained. The degree of coincidence between the two patterns is calculated by comparing the two patterns, and similarly, the unknown pattern of the binary expression and the expanded pattern of the binary expression are compared to obtain the degree of coincidence between the two patterns. The expression unknown pattern and the binary expression contraction pattern are compared to obtain the degree of coincidence between both patterns, and the binary value corresponding to the highest degree of coincidence among the degrees of coincidence thus obtained. Since the unknown pattern is recognized to match the pattern of the expression, the unknown pattern such as characters obtained by the TV camera is inclined with respect to the scanning direction, or the standard pattern. Even if the pattern is thicker or thinner than the Come. Therefore, even a non-defective product will not be erroneously recognized as a defective product.

[Brief description of the drawings]

FIG. 1 is a flow chart showing a pattern recognition procedure according to an embodiment of the present invention.

FIG. 2 is a configuration diagram of a main part of the present invention.

FIG. 3 shows an embodiment of the present invention, and is a diagram showing envelope curves for characters A and B.

FIG. 4 shows an embodiment of the present invention, in which the diameter line segment is 3
It is divided into two line elements and the pattern at the position corresponding to each line element.
Feature pattern obtained by expressing the presence or absence of the components of
FIG. 3 is a diagram for explaining comparison between the pattern and this pattern.

FIG. 5 is a diagram illustrating a circle and a diameter line segment related to a character A.

FIG. 6 is a binary expression pattern extracted in the embodiment of FIG.
FIG.

FIG. 7 is a diagram illustrating a circle and a diameter line segment related to a character A rotated by 45 ° from the posture of FIG. 5;

FIG. 8 is a binary expression pattern extracted in the embodiment of FIG.
FIG.

FIG. 9 is a diagram showing the thickening of a character A which is a standard pattern.

FIG. 10 is a diagram showing the thinning of a character A which is a standard pattern.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 TV camera 2 A / D converter 3 Memory 4 Pattern cutout means 5 Feature extraction means 6 Rotation matching means 7 Text storage means 8 Arithmetic unit 9 Expansion / contraction processing means

Claims (5)

[Claims] 1. A pattern recognition method for recognizing an unknown pattern by comparing an unknown pattern with a plurality of standard patterns by a pattern matching method. The data obtained by expressing the presence or absence of the constituent elements in a binary value is a binary expression standard pattern for each of the standard patterns.
Each of the standard patterns is obtained as a pattern, and is registered as a comparison pattern, and each of the binary expression standard patterns is expanded a predetermined number of times to obtain the binary expression standard pattern. Binary expression expansion pattern for
The binary expression contraction pattern for the binary expression standard pattern is obtained by registering the binary expression standard pattern for a predetermined number of times and registering it as a comparison pattern. And the pattern is registered as a comparison pattern, and the binary expression pattern relating to the unknown pattern is converted into a binary expression unknown pattern in the same manner as in the method for obtaining the binary expression standard pattern. The binary expression unknown pattern is compared with the binary expression standard pattern to obtain the degree of coincidence between the two patterns. Similarly, the binary expression unknown pattern and the binary expression unknown pattern are obtained. The degree of coincidence between the two patterns is obtained by comparing the binary expression expansion pattern, and further, the binary expression unknown pattern and the binary expression contraction pattern are compared to both patterns. The degree of coincidence of Pattern and recognizes the down - - emissions recognition method the unknown pattern as to match the emission - binary pattern representation corresponding to have the degree of coincidence. 2. The binary representation standard pattern obtains an envelope of each of the standard patterns, obtains a center of gravity of a region surrounded by the envelope, and selects one of the end points of the standard pattern. Find the line segment connecting the end point at the farthest position from the center of gravity with the center of gravity, find the circle centered on the center of gravity with this line segment as the radius, and pass this circle through the center of gravity to form an equal angle To obtain a plurality of diametrical diametrical line segments that are equally divided into 2. The pattern registration method according to claim 1, wherein the pattern registration data is composed of binary data. 3. The binary expression expansion pattern comprises data obtained by expanding the data in each of the line elements of the binary expression standard pattern a predetermined number of times in the longitudinal direction of the line element. The pattern registration method according to claim 1, wherein: 4. The binary representation contraction pattern comprises data obtained by subjecting the data of each of the line elements of the binary representation standard pattern to a contraction process a predetermined number of times in the longitudinal direction of the line element. The pattern registration method according to claim 1, wherein: 5. The unknown pattern is recognized by comparing either the binary expression unknown pattern or the registered binary expression pattern while sequentially rotating them by equal angles. The degree of coincidence with the binary expression unknown pattern is obtained, and the registered degree of coincidence is highest.
The pattern registration method according to claim 1, wherein an unknown pattern is recognized as a pattern corresponding to the value expression pattern.