JPH08137992A - Character identification device - Google Patents

Abstract

PURPOSE: To provide a character identification device capable of improving the recognition rate of a character. CONSTITUTION: This device is the character identification device which obtains a binarized character in the character that is a recognition object and samples its contour and identifies the character from a feature point, and it is equipped with a directional component judging part 3 which blocks the moving direction of the next point for a remarked point in the contour of the character into prescribed directional components, a feature variable calculation part 4 which obtains the information of the directional components blocked by the directional component judging part 3 and calculates the feature variable of the character pattern by using an arithmetic equation corresponding to the information, and an identification part 5 which compares the feature variable calculated by the feature variable calculation part 4 with reference feature quantity decided in advance, respectively for plural characters, and recognizes the character in accordance with the reference feature variable with the minimum difference between them as the character in the character pattern.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character identification device for accurately identifying a character from a binarized character pattern of a read character.

[0002]

2. Description of the Related Art Conventionally, as a technique for identifying characters including alphanumeric characters and katakana, there is a character identification system disclosed in Japanese Patent Laid-Open No. 1-258086. In this character identification method, the outline of the character is extracted from the character pattern binarized by reading the character, and the relative position of the line segment forming the outline and the characteristic amount (Q Value) and compares this feature amount with a predetermined reference feature amount corresponding to a plurality of characters, and the character having the smallest difference is recognized as a character of the character pattern. are doing.

[0003]

However, such a character identification method has a problem that although it is possible to represent global characteristics of a character to be identified, it is difficult to express characteristics in a fine portion of the character. For example, when distinguishing the character “6” from the character “0”, it is difficult to make a feature difference from the character “0” when the loop of the character “6” becomes a little large, which may cause misrecognition. Has become one of.

[0004]

SUMMARY OF THE INVENTION The present invention is a character identification device made to solve such problems. That is, the character identification device of the present invention reads a character to be identified, obtains a binarized character pattern in the character, extracts the contour of the character from the character pattern, and identifies the character from the characteristic points. , Direction component determining means for segmenting the moving direction of the next point with respect to the target point among the pixels forming the outline of the character into predetermined direction components, and information of the segmented direction components from the direction component determining means. Characteristic amount calculation means for calculating a characteristic amount indicating a characteristic of a character pattern using an arithmetic expression corresponding to the characteristic amount, a characteristic amount calculated by the characteristic amount calculation means, and a reference characteristic amount predetermined for each of a plurality of characters. And a discriminating means for recognizing a character corresponding to the reference feature amount with which the difference between the two is compared with the character in the character pattern.

[0005]

According to the present invention, the direction component determining means divides the moving direction of the next point of the pixels forming the outline of the character extracted from the character pattern with respect to the target point into predetermined direction components. Further, the feature amount calculating means calculates the feature amount indicating the feature of the character pattern by using the arithmetic expression according to the divided direction component. As a result, it is possible to obtain the feature amount for each direction component, and it is possible to clearly show the difference in the feature amount according to the fine feature of the character pattern. Therefore, it is possible to accurately perform the character identification based on this difference. become able to.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the character identification device of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram for explaining the first embodiment of the character identification device of the present invention. (A) is a block diagram, (b) is a character pattern of "6", (c) is "0".
Character pattern, and (d) shows a direction component.

As shown in FIG. 1A, the character identifying apparatus in the first embodiment has a pattern register 1 for storing a binarized character pattern 1a, and a character pattern 1b output from the pattern register 1. From the contour extraction unit 2 for extracting the contour 2a of the character pattern 1b by a known method from the above, the moving direction of the next point of the pixels forming the contour 2a with respect to the target point is divided into, for example, four direction components, and A directional component determination unit 3 that adds information to the contour 2a and outputs a directional component-added contour 3a, and a feature amount that calculates the feature amount 4a of the character pattern 1b based on the directional component-added contour 3a obtained from the directional component determination unit 3. The calculation unit 4 compares the calculated feature amount 4a with a predetermined reference feature amount,
When it is detected that the difference between the two is the smallest, it is determined that the character corresponding to the reference feature amount is the character of the character pattern 1a, and the identification unit 5 that outputs the determination result 5a. .

In the character identifying device of the first embodiment, for example, a character pattern of "6" as shown in FIG.
The difference between the feature amount and the character pattern of "0" as shown in FIG. 1C is clearly shown to identify the difference, and the operation will be described below.

First, a known character pattern input means (for example, a digitizer) reads a character to be identified, for example, a character "6" on a notepad, and the binarized data is associated with the xy coordinate position. Composed character pattern 1
Store a in pattern register 1. Then, the pattern register 1 outputs the character pattern 1b to the contour extraction unit 2.

The contour extracting unit 2 obtains the character pattern 1b output from the pattern register 1 and outputs the character pattern 1b.
The process of extracting the contour at is performed. To extract the contour, a known method is used, for example, contour tracing is performed using a pixel located at the upper end of the character pattern 1b as a starting point, and the contour information is extracted as the contour 2a. Note that the contour extraction method may be other than this.

The direction component determination unit 3 obtains the contour 2a extracted by the contour extraction unit 2, and a line segment is, for example, 4 by the moving direction of the next point with respect to the target point among the pixels forming the contour 2a.
Decompose into one direction component. The four direction components are shown in Fig. 1.
As shown in (d), there are a horizontal component, a vertical component, a + 45 ° component, and a −45 ° component. As a method of decomposing into four directional components, first, dx and dy are calculated using the equation (1), and it is determined depending on which of the equations (2) the dx and dy correspond.

[0012]

[Equation 1]

[0013]

[Equation 2]

In the equation (1), xi, yi are the coordinate values of the target point, and xi + 1, yi + 1 are the coordinate values of the next point. This also applies to the following equations.

In the direction component judging unit 3, the line segment in the contour 2a is decomposed in this way into four direction components, and then a predetermined code corresponding to the four direction components is added to the contour 2a to add the direction component. The contour 3a is output.

The feature amount calculation unit 4 receives the contour 3a with a direction component output from the direction component determination unit 3, and performs a feature amount calculation process using an arithmetic expression according to the added direction component. For example, if the horizontal component (dy = 0),
Only q0, q1, q8 and q9 shown in the equation (3) are calculated.

[0017]

(Equation 3)

Note that HX and HY shown in the equation (3) represent the character width of the character pattern 1a in the x and y directions, respectively. This also applies to the following equations.

The feature quantities q0 and q1 calculated here are xy with the lower left of the circumscribed rectangular area 10 (rectangular area circumscribing the character pattern) shown in FIGS. 1B and 1C as the origin.
It is a value calculated with respect to the coordinate system, and the closer to the origin, the smaller the value. The feature quantity q0 is calculated for the x coordinate value (xi + xi + 1) on the line segment, and the feature quantity q1 is calculated for the y coordinate value (yi + yi + 1) on the line segment. That is, the feature quantity q0 becomes larger as the horizontal component (dy = 0) in the character pattern 1a increases on the right side of the circumscribed rectangular area 10, and the feature quantity q1 becomes larger than the horizontal component (dy).
= 0), the larger the value in the upper side of the circumscribed rectangular area 10, the larger the value.

Therefore, it is difficult to distinguish between the case where a long horizontal component exists on the left side and the case where a short horizontal component exists on the right side in the circumscribed rectangular area 10 in the feature quantity q0. Therefore, by using the feature quantity q8 similarly calculated with the upper right of the circumscribed rectangular area 10 as the origin, the distinction is made. That is, when there is a long horizontal component on the left side of the circumscribed rectangular area 10, the feature quantity q0 is small and the feature quantity q8 is large, and when there is a short horizontal component on the right side of the circumscribed rectangular area 10, the feature quantity q0 is small. , Feature quantity q8
Distinguishes by being small. Note that the feature quantity q9 is also calculated in the same way, and the same calculation as the feature quantity q1 is performed with the upper right corner of the circumscribed rectangular area 10 as the origin.

Such characteristic quantities q0, q1, q8, q9
By calculating, it is possible to represent at which position the horizontal component is distributed in the character pattern 1a.

Further, when the contour 3a with direction component from the direction component judgment unit 3 indicates a vertical component, only the feature quantities q2, q3, q10, q11 are calculated using the equation (4).
Further, the contour 3a with direction component from the direction component determination unit 3 is +
When the 45 ° component is indicated, only the feature quantities q4, q5, q12 and q13 are calculated using the equation (5). further,
The contour 3a with the direction component from the direction component determination unit 3 is -45 °.
When the component is shown, the feature quantity q is calculated using the equation (6).
Only 6, q7, q14 and q15 are calculated.

[0023]

[Equation 4]

[0024]

(Equation 5)

[0025]

(Equation 6)

Characteristic quantities q0 to q for the character pattern 1a
By performing the calculation for calculating 15, it becomes possible to represent in detail where in the circumscribed rectangular area 10 the data forming the contour is distributed. Moreover, since it is divided into the direction components and the calculation is performed using only the calculation formulas corresponding thereto, it is possible to perform the processing in a short time without performing the unnecessary calculation of the feature amount. When the character to be identified has a plurality of contours, the feature quantities q0 to q15 are calculated for all the contours.

Next, the identifying section 5 identifies the character.
In the identification unit 5, reference feature amounts for a plurality of characters are stored in advance. In order to identify a character, first, these stored reference feature amounts and the feature amounts q0 to q15 of the character pattern 1a derived from the contour 2a are sequentially compared to determine the difference between them, that is, the matching distance. Are derived respectively. Then, among the derived matching distances, it is assumed that the character corresponding to the reference feature amount having the minimum value is the character to be identified, and this is output as the determination result 5a.

As described above, in the character identifying apparatus according to the first embodiment, the direction component of the line segment is divided according to the moving direction of the next point with respect to the point of interest of the contour 2a, and the feature amount according to the division is calculated. Therefore, it is possible to clearly represent more detailed features in the character pattern 1a. That is, there is a case where the global features are similar, such as a character pattern of "6" as shown in FIG. 1B and a character pattern of "0" as shown in FIG. 1C. However, different features (slight directivity of line segments and difference in distribution) can be accurately identified within the broken line frame in the figure, and the recognition rate can be significantly improved. In the first embodiment, the example in which the directional component is divided into four has been shown, but other divisions may be performed.

Next, the second character recognition apparatus of the present invention
An embodiment will be described with reference to FIG. 2A and 2B are views for explaining a character identification device according to the second embodiment. FIG. 2A is a block diagram, FIG. 2B is a character pattern of "6", and FIG. 2C is a character pattern of "0". . As shown in FIG. 2A, the character identifying apparatus according to the second embodiment includes a pattern register 1, a contour extracting unit 2, a direction component determining unit 3, a feature amount calculating unit 4, and an identifying unit 5. Similar to the character recognition apparatus shown in the first embodiment, except that the contour extraction unit 2 outputs the contour 2
The difference is that an area determination unit 6 that obtains a and outputs a predetermined area code 6a to the feature amount calculation unit 4 is provided.

That is, the operation of each component of the pattern register 1, the contour extraction unit 2, the direction component determination unit 3, the feature amount calculation unit 4, and the identification unit 5 is the same as that of the first embodiment, and therefore the following is different. The operation of the area determination unit 6 will be mainly described.

As shown in FIGS. 2B and 2C, the area determination unit 6 divides the circumscribed rectangular area 10 into, for example, four areas, and coordinates ((xi + xi + 1) / 2, (yi + i). +1
) / 2) is output to the feature amount calculation unit 4 in which area. That is, the area determination unit 6 first obtains the contour 2a from the contour extraction unit 2, and then the area determination unit 6 in FIGS.
As shown in, the circumscribed rectangular area 10 is divided into four equal parts with HX / 2 and HY / 2 as boundaries, and the process is divided into areas. Then, the coordinates of the contour 2a ((xi + x
When (i + 1) / 2, (yi + i + 1) / 2) is set to (x, y), it is determined which of the expressions (7) is applicable, and the corresponding area code 6a is set as the feature quantity. It is output to the calculation unit 4.

[0032]

(Equation 7)

As in the first embodiment, the feature quantity calculation unit 4 receives the contour 3a with direction component from the direction component determination unit 3 to calculate the feature quantities q0 to q15, and the area determination unit 6 According to the output area code 6a, the feature amount for each corresponding area is calculated and output. In this case, for example, the feature amounts in all areas as described in the first embodiment are set to q0,0 to q0,15, and the feature amounts in the areas, areas, and areas are set to q1,0 to q, respectively.
1,15, q2,0 to q2,15, q3,0 to q3,15, q4,0 to q4,1
Expressed as 5.

The identification unit 5 receives the feature amount 4a output from the feature amount calculation unit 4, performs matching with the reference feature amount of the dictionary registered in advance using the equation (8), and the distance value dk thereof. The character corresponding to the minimum value of is output.

[0035]

(Equation 8)

In the equation (8), dk is the matching result of the character k, wi is the weighting coefficient of the area i, and q
i and j indicate the feature amount of the input character, and q′k, i, j indicate the reference feature amount of the character k stored in the dictionary. (Weighting factors wi are, for example, w0 = 1.0, w1 = 0.5, w2 = 0.5, w3
= 0.5, w4 = 0.5. )

For example, as shown in FIGS. 2B and 2C, when attention is paid only to the upper right area when the circumscribed rectangular area 10 is divided into four equal parts, the character "0" indicates the upper and right sides of the area. Although it is distributed in the area, the character "6"
It can be seen that is distributed in the upper part and the central part of the area. As a result, in the character pattern of "6" and the character pattern of "0", the feature amount q in the area
The values of 4,0 to q4,15 are clearly different.

As a result, for the character pattern of "6" and the character pattern of "0", the overall feature quantities q0,0 to q
There is a large difference in the values of the feature quantities q4,0 to q4,15 in 0, 15 and the area, and based on this, it is possible to perform more reliable character identification. In the second embodiment, the circumscribed rectangular region 10 is divided into four equal parts, but this is an example, and the same is true even if the number is divided into other parts. By increasing the number of divisions, a higher recognition rate can be obtained.

Next, the third character recognition apparatus of the present invention will be described.
An embodiment will be described with reference to FIG. 3A and 3B are views for explaining a character identification device according to the third embodiment. FIG. 3A is a block diagram, FIG. 3B is a character pattern of "6", and FIG. 3C is a character pattern of "0". . As shown in FIG. 3A, in the character identifying apparatus according to the third embodiment, the pattern register 1, the contour extracting unit 2, the direction component determining unit 3, the feature amount calculating unit 4, the identifying unit 5, and the area determining unit 6 are included. It is similar to the character identifying apparatus according to the second embodiment in that it is provided with a barycenter calculation unit 7 that obtains the contour 2a from the contour extraction unit 2 and outputs the barycentric position 7a of the character pattern to the area determination unit 6. It's different. Hereinafter, the operation of the character center in the third embodiment will be sequentially described from the operation of the gravity center calculating unit 7.

The center of gravity calculating unit 7 receives the outline 2a from the outline extracting unit 2, and calculates the center of gravity (CGX, CGY) of the character pattern by the equation (9).

[0041]

[Equation 9]

Next, the area determination unit 6 receives the contour 2a from the contour extraction unit 2 and the barycenter position 7a of the character pattern from the barycenter calculation unit 7, and as shown in FIGS. 3 (b) and 3 (c). The circumscribed rectangular area 10 is divided into four with the position of the center of gravity of the character pattern as a boundary. Then, which area of the area to the area the contour exists in is determined using the equation (10), and the area code 6a is output to the feature amount calculation unit 4.

[0043]

[Equation 10]

As in the second embodiment, the feature quantity calculation unit 4 receives the contour 3a with direction component from the direction component determination unit 3 to calculate the feature quantities q0 to q15, and the area determination unit 6 According to the output area code 6a, the feature amount for each corresponding area is calculated and output. In this case, for example, the feature amounts in all areas described in the first embodiment are set to q0,0 to q0,15, and the area, area,
Area and feature quantity in the area are q1,0 to q1,15,
Represented as q2,0 to q2,15, q3,0 to q3,15, q4,0 to q4,15.

After that, as in the second embodiment, the identifying unit 5 receives the feature amount 4a output from the feature amount calculating unit 4,
The expression (8) is used to perform matching with the reference feature amount of the dictionary registered in advance, and the character corresponding to the minimum value of the distance value dk is output. In this way, the circumscribed rectangular area 10 is divided with the position of the center of gravity of the character pattern as a boundary, and the feature amount is calculated for each area, so that the written characters are less affected than in the case of simply dividing the area. It becomes possible to perform stable feature amount calculation. This makes it possible to perform character recognition with higher reliability.

Next, the fourth character recognition apparatus of the present invention
An embodiment will be described with reference to FIG. 4A and 4B are views for explaining a character identification device in the second embodiment. FIG. 4A is a block diagram, FIG. 4B is a projection amount of a character pattern, FIG. 4C is a character pattern “6”, and FIG. This shows a character pattern of "0". As shown in FIG. 4A, in the character identifying device according to the fourth embodiment, the pattern register 1, the contour extracting unit 2, the direction component determining unit 3, the feature amount calculating unit 4, the identifying unit 5, and the area determining unit 6 are included. The character identification device is similar to the character identification device according to the second embodiment in that the character pattern 1a is obtained from the pattern register 1 to calculate the projection amount of the character pattern, and the divided position of the circumscribed rectangular area 10 corresponding to the calculated projection amount. The difference is that a division position calculation unit 8 that outputs 8a to the area determination unit 6 is provided. Hereinafter, the operation of the character identification device in the fourth embodiment will be described in order centering on the operation of the division position calculation unit 8.

The division position calculator 8 receives the character pattern 1a from the pattern register 1 and first calculates the projection amount of the character pattern as shown in FIG. 4 (b). That is,
The circumscribed rectangular area 10 is scanned in the x direction, the number of black dots at that time is counted, and this is set as the projection amount. Although not shown, the number of black dots when the circumscribed rectangular area 10 is scanned in the y direction is also counted.

Next, the division position is determined by using the equation (11) based on the counted projection amounts in the x and y directions so that the division position can be divided into, for example, four areas.

[0049]

[Equation 11]

In order to determine the division position using the equation (11), the y coordinate values (division candidate positions) of the block whose projection amount is equal to or larger than the threshold value are obtained in the ascending order, and they are set to y1, y2 ... Select the division position from. For example, in FIG.
In the example shown in (b), the projection amount is equal to or more than the threshold value in three of y1, y2, and y3. In this case, the middle y2 is the division position in the x direction. Similarly, y
The division position in the direction is also obtained, and the division position 8a is output to the area determination unit 6.

The area determination unit 6 receives the division position 8a from the division position calculation unit 8 and determines the circumscribed rectangular area 10 as shown in FIG.
While dividing into four as shown in (c) and (d), the contour 2a is received from the contour extraction unit 2 and it is determined by the equation (12) which area among the areas the area exists. The area code 6a is output to the feature amount calculation unit 4.

[0052]

(Equation 12)

As in the second and third embodiments, the feature quantity calculating unit 4 receives the contour 3a with direction component from the direction component determining unit 3 to calculate the feature quantities q0 to q15, and According to the area code 6a, which is the output from the determination unit 6, the feature amount for each corresponding area is calculated and output. In this case, for example, the feature amounts in all areas described in the first embodiment are set to q0,0 to q0,15, and the feature amounts in each area, area, and area are set to q.
1,0 to q1,15, q2,0 to q2,15, q3,0 to q3,15, q4,
Expressed as 0 to q4,15.

The identifying section 5 is the same as that of the second embodiment and the third embodiment.
As described in the embodiment, the feature amount 4a output from the feature amount calculation unit 4 is received, the reference feature amount of the dictionary registered in advance is matched using the formula (8), and the distance value thereof is calculated. The character corresponding to the minimum value of dk is output. In this way, by dividing the circumscribed rectangular area 10 according to the projection amount of the character pattern and calculating the feature amount for each area, it is possible to clearly cause the feature amount difference of the character pattern in a finer portion. Therefore, it is possible to accurately identify characters.

Although the circumscribed rectangular area 10 is divided into four in the fourth embodiment, this is an example.
You may divide by another number. By increasing the number of divisions, a higher recognition rate can be obtained. Also, in the first to fourth examples, "6"
The character pattern of “0” and the character pattern of “0” have been described as an example, but the present invention is not limited to this, and the same applies when identifying other similar character patterns.

[0056]

As described above, the character identification device of the present invention has the following effects. In other words, since the line segments forming the outline of a character are focused on and divided into predetermined direction components, and the feature amount is calculated by an arithmetic expression according to each, the characters are generally similar. However, it is possible to clearly express the feature difference in the detailed portion, and it is possible to perform character recognition with a high recognition rate.

[Brief description of drawings]

FIG. 1 is a diagram illustrating a first embodiment, (a) is a block diagram, (b) is a character pattern of “6”, (c) is a character pattern of “0”, and (d) is a direction component. Shows.

2A and 2B are diagrams illustrating a second embodiment, FIG. 2A is a block diagram, FIG. 2B is a character pattern of “6”, and FIG. 2C is a character pattern of “0”.

3A and 3B are diagrams illustrating a third embodiment, FIG. 3A is a block diagram, FIG. 3B is a character pattern of “6”, and FIG. 3C is a character pattern of “0”.

FIG. 4 is a diagram for explaining a fourth embodiment, (a) is a block diagram, (b) is a projection amount of a character pattern, (c) is a character pattern of “6”, and (d) is “0”. It shows the character pattern.

[Explanation of symbols]

 1 pattern register 2 contour extraction unit 3 direction component determination unit 4 feature amount calculation unit 5 identification unit 6 area determination unit 7 center of gravity calculation unit 8 division position calculation unit 10 circumscribed rectangular area

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Shizuo Nagata 1-7-12 Toranomon, Minato-ku, Tokyo Oki Electric Industry Co., Ltd.

Claims (4)

[Claims] 1. A character identification which reads a character to be identified, obtains a binarized character pattern in the character, extracts the contour of the character from the character pattern, and identifies the character from the feature points. A device, wherein the direction component judging means divides the moving direction of the next point with respect to the target point among the pixels forming the outline of the character into predetermined direction components, and the direction component divided from the direction component judging means. Information, and a feature amount calculation unit that calculates a feature amount indicating the feature of the character pattern using an arithmetic expression corresponding to the information, the feature amount calculated by the feature amount calculation unit, and a plurality of characters On the other hand, it is characterized in that it is provided with an identification means for performing a comparison with a predetermined reference feature amount and recognizing a character corresponding to the reference feature amount having the minimum difference as a character in the character pattern. Character recognition device for collection. 2. A division area determination means for dividing a rectangular area circumscribing the character pattern into predetermined small areas and causing the characteristic amount calculation unit to calculate the characteristic amount in units of the small regions. The character identification device according to claim 1, wherein: 3. A center-of-gravity position calculation means for calculating the center-of-gravity position of the character pattern and causing the divided area determination means to divide the circumscribed rectangular area centered on the center-of-gravity position is provided. The character identification device according to claim 2, wherein the character identification device is a character identification device. 4. The projection information obtained by adding the number of pixels of each character pattern along the vertical and horizontal directions to the circumscribed rectangular area is obtained, and the circumscribed rectangular area is divided based on the projection information. 3. The character identification device according to claim 2, further comprising division position calculation means for causing the division area determination means.