JPH10240869A - Device for preparing character recognition dictionary and method therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and device for preparing a character recognition dictionary in which plural many sample character data for recognition can be automatically prepared, and the labor of a user can be saved. SOLUTION: In a character recognition dictionary preparing part 1, an image data inputting part 10 fetches the image data of a character being a basic from the outside, and a segmenting part 11 operates a work for segmenting the part of a character from the image data fetched by the image data inputting part 10, and obtains the basic image data, and a disturbance processing part 12 operates a prescribed picture processing for simulating various kinds of disturbance to the basic image data, and generates sample character data for recognition. Thus, the many sample character data for recognition can be automatically generated from a few basic image data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition dictionary creation device and a character recognition dictionary creation method for creating a character recognition dictionary used for character recognition by a computer or the like.

[0002]

2. Description of the Related Art Factory automation (FA)
In some fields, a computer or the like may automatically recognize characters indicating a model number, a product number, or the like, which are printed or stamped on parts or products flowing on a production line. In that case, the computer performs a pattern matching process between the image data of the character read by the television camera or the like and the sample character data for recognition of each character stored in the storage means in advance. Even if the same character type is the same, various disturbances are added when reading it on a production line or the like.To improve the accuracy of character recognition, a large number of recognition sample character data are required for one character type. It is necessary to prepare. In a character recognition operation on an actual production line or the like, when a certain character recognition operation is performed, the pattern of the character is registered and becomes sample character data for recognition for the next and subsequent pattern matching, so the recognition operation is repeated. As a result, the sample character data for recognition gradually increases. However, when the character recognition device is installed, a certain number of sample character data for recognition must be prepared in advance. In the recognition of printed or engraved characters in the general FA field, several tens to several hundreds of sample character data for recognition are prepared in advance for one character type, and the target of the sample character data for recognition is The entirety of such prepared sample character data for recognition for each of a large number of character types is provided as a character recognition dictionary.

[0003]

In order to improve the recognition accuracy in character recognition, it is necessary to increase the number of sample character data for recognition per character type when the character recognition device is installed. However, conventionally, it has been necessary for a human to create the recognition sample character data included in the character recognition dictionary one by one, or take it as image data from an actual product or the like and store it in a computer. For this reason, if an attempt is made to increase the number of recognition sample character data per character type in order to improve recognition accuracy, considerable labor is required, and a heavy burden is imposed on the user.

[0004] The present invention has been made based on the above circumstances, and is capable of automatically generating a large number of sample character data for recognition, thereby reducing the user's labor. And a method for creating a character recognition dictionary.

[0005]

In order to solve the above-mentioned problems, a character recognition dictionary creating apparatus according to the present invention comprises: a storage unit for storing a character recognition dictionary; and a sample character for recognition registered in the character recognition dictionary. Character input means for inputting character data as a source of data, and image processing simulating a disturbance applied to a character to be recognized is applied to the character data input by the character input means, and the character data is input by the character input means. And image processing means for generating sample character data for recognition from the extracted character data and registering the sample character data in a character recognition dictionary of the storage means.

The character recognition dictionary creating method according to the present invention is characterized in that image processing for simulating a disturbance applied to a character to be recognized is performed on character data which is a source of sample character data for recognition, and the character input means performs the processing. A character recognition dictionary is created by generating sample character data for recognition from input character data and registering them in a character recognition dictionary in a storage unit.

According to the present invention, a recognition sample character is obtained by subjecting character data, which is a source of the recognition sample character data registered in the character recognition dictionary, to image processing that simulates a disturbance applied to the character to be recognized. Since the character data is created, it is possible to automate the creation of the recognition sample character data. Therefore, a large number of recognition sample character data can be created quickly and in large quantities from a small number of original character data.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a schematic block diagram of a character recognition device including a character recognition dictionary creation device according to one embodiment of the present invention. The character recognition device of FIG.
It comprises a character recognition dictionary creation unit 1 and a character recognition unit 2. The image data input unit 10 receives image data of a basic character from outside. As the image data input unit 10, a television camera, an image scanner, or the like can be used. The image signal captured by the image data input unit 10 is sent to the character recognition dictionary creation unit 1.

In the character recognition dictionary creating section 1, a cutout section 11 performs a work of cutting out each character portion from the image data taken in by the image data input section 10. This operation is performed by the operation of the operator. Hereinafter, the character image data cut out in this manner is referred to as “basic image data”. The disturbance processing unit 12 performs predetermined image processing that simulates various disturbances on the basic image data thus cut out, and generates recognition sample character data. Thereby, a large number of sample character data for recognition is automatically generated from a small number of basic image data. Further, in order to simulate a disturbance based on a plurality of causes, a plurality of image processes can be performed on one piece of basic image data as described later. The recognition sample character data generated in this manner is subjected to predetermined processing by the dictionary creation processing unit 13 so as to be usable as sample data for predetermined character recognition,
It is registered in the character recognition dictionary 14 as a database.

The dictionary created in this way is used in the actual character recognition work in the character recognition unit 2. For example, image data of characters printed or stamped on a product or the like is taken in from an image data input unit 20 provided near the production line, and a recognition unit 21 sequentially reads sample character data for recognition from the character recognition dictionary 14, and The type of character is recognized by performing processing such as pattern matching with the recognition target character fetched from the image data input unit 20. The recognized characters are output and displayed on the CRT, for example, in the character output unit 22 as necessary. The character recognized in this way is fed back as new sample character data for recognition and registered in the character recognition dictionary 14. Therefore, the number of recognition sample character data is gradually increased by repeatedly using the present apparatus, and the recognition rate is gradually improved by such a learning effect. Note that the image data input unit 20
It can be shared with the image data input unit 10 described above.

Here, the processing in the disturbance processing unit 12 of FIG. 1 will be described in detail. The left column of the table of FIG. 2 lists some possible disturbances when the image data input unit 20 of FIG. 1 captures image data of a character to be recognized. Each item indicates the cause of the disturbance shown on the left side. For example, the disturbance "image blur"
This may be caused by a defocus of the camera of the image data input unit 20 or bleeding of ink. Disturbances such as “thickening of the line” and “thinning of the line” can be caused by a shift in the binarization level. Disturbance called "dirt" can be caused by adhesion of dirt and the like, signal noise, and the like. "Character enlargement and reduction"
Can be caused by a change in the distance between the object and the camera. And the disturbance called "tilt of characters"
It can be caused by the tilt of the object or camera.

Each of the image processes listed in the right column of FIG.
This is the type of image processing for simulating the disturbances listed in the left column of each row. For example, when an image of a character to be recognized is actually captured, disturbance such as image blurring that may be caused by a camera out of focus or ink bleeding in the image data input unit 20 is performed by performing a smoothing process on the basic image data. Can be simulated by Similarly, the thickening of the line is performed by dilation processing on the basic image data, the thinning of the line is performed by shrinking processing, the dirt generates pseudo noise on the basic image data, and the enlargement / reduction of the character is performed on the image based on the basic image data. Can be simulated by the enlargement / reduction processing, and the inclination of the character can be simulated by the rotation processing on the basic image data.

FIG. 3 shows the disturbance processing unit 12 shown in FIG.
9 is a table summarizing a part of image processing actually performed on basic image data. In the table, "No disturbance"
Is a case where the basic image data is output without any processing. Also, the difference between “smoothing” and “smoothing” is that the former is a process of calculating the average of each pixel at its adjacent nine pixels, while the latter is the process of calculating the average of each pixel at the adjacent 16 pixels. The point is that it is the required process. The difference between “fat” and “fat” is that the former is 4
In contrast to the thickening process in the direction, the latter is a thickening process in eight directions. The difference between "narrow" and "narrow"
The former is a thinning process in four directions, while the latter is a thinning process in eight directions. When the ratio of “pseudo noise” is “0.02”, it means that the image is changed at a ratio of 2%. “Enlarge” enlarges the image by α times in aspect ratio. The value of α can be freely set by the user. "Shrink"
Is reduced to 1 / β times in aspect ratio. The value of β can be freely set by the user. "Rotation" and "Rotation"
Is the process of rotating the clockwise angle θ.
The latter is a process of rotating counterclockwise by the angle θ.
Note that FIG. 3 is only an example of image processing capable of simulating a disturbance, and more image processing can be actually prepared.

FIG. 4 is a block diagram of the character recognition dictionary creating apparatus shown in FIG.
FIG. 3 is a diagram showing a concept of a process performed by a disturbance processing unit 12;
And multiple stages (in this case, 3
Disturbance simulation is performed by the image processing section
Indicates that it can be done. In FIG. 4, the control unit 30
And the first-stage image processing unit 31 and the second-stage image processing unit 31
₁~ 31_N, The third stage image processing unit 31₁₁~ 31_1N, ...
・, 31_N1~ 31_NNConsists of First-stage image processing unit 31
Has N outputs and each output has a switch S₁~ S_N
Is provided. Each of the N outputs is connected to a switch S₁~ S
_NThrough the second stage image processing unit 31₁~ 31_NConnected to
Have been. Each image processing unit in the second stage also has the first stage
There are N outputs, each with a switch S₁₁~
S_1N, ..., S_N1~ S_NNIs provided. Second stage
Each image processing unit 31₁~ 31_NIs the switch S₁₁~ S_1N,
... S_N1~ S_NNThrough the third-stage image processing unit 31
₁₁~ 31_1N, ..., 31_N1~ 31_NNConnected to
The N outputs of each image processing unit in the third stage
S₁₁₁~ S_11N, ..., S_NN1~ S_NNNIs connected
I have. Switch S of each stage₁~ S_N, S₁₁~ S_1N, ...
・, S_N1~ S_NN, S₁₁₁~ S_11N, ..., S_NN1~ S
_NNNAre all controlled by the control unit 30.

Each of the three stages of image processing units shown in FIG.
The image processing illustrated in FIG. 3 and other image processing that simulates a disturbance are performed. However, the parameters shown in the “ratio” in the table of FIG. That is, for example, pseudo noise is set to 0.02 in the first-stage image processing unit, 0.08 in the second-stage image processing unit, and 0.06 in the third-stage image processing unit. Α = 2 in the image processing unit, α = 4 in the second image processing unit, and α = 8 in the third image processing unit.
For the reduction, in the first stage image processing unit, β =
2, β = 4 in the second stage image processing unit, β = 8 in the third stage image processing unit, and rotation = θ = 5 ° in the first stage image processing unit; Then θ = 10 °,
In the third-stage image processing unit, θ is set to 15 °. Then, only the signal of the image processing output whose switch is turned on by the control unit 30 is sent to the next image processing unit. Therefore, when all the switches are turned on by the control unit 30, if the type of processing of one image processing unit is N, the number of sample character data for recognition obtained by three-stage image processing is N is the third power (N ³ ).

The control unit 30 shown in FIG. 4 can turn off the switch so as not to perform image processing for simulating such disturbance when there is a disturbance that cannot occur in an actual production line or the like. . For example, when it is known from the situation of the production line that disturbance such as image rotation cannot occur, a predetermined switch is turned off so as not to perform image processing such as rotation. Thus, generation of useless recognition sample character data is reduced, and the efficiency of the recognition operation in the recognition unit 21 in FIG. 1 can be improved.

FIG. 5 shows a character recognition dictionary (conventional dictionary) created by a conventional method, that is, a method of reading and registering recognition sample character data one by one, and a character recognition dictionary created by the apparatus of the present embodiment. 9 is a table showing results of an experiment of character recognition using (dictionary of the present embodiment). The character to be recognized is a character obtained by capturing the engraved character as an image from a video and performing image processing for simulating the disturbance shown in FIG. as a result,
The recognition rate of the conventional dictionary was 88.67%, whereas the dictionary of the present embodiment achieved a high recognition rate of 99.46%.

It should be noted that the present invention is not limited to the above embodiment, and various changes can be made within the scope of the invention. For example, in the above-described embodiment, “no disturbance”, “smoothing”,
The case where “thick”, “thin”, “pseudo noise”, “enlarge”, “shrink”, and “rotate” are performed has been described, but these are merely examples, and image processing such as “differential processing” is performed as described above. May be performed together with each image processing. Further, in the above embodiment, the characters printed or engraved on a product or the like flowing on the production line are recognized.
It is also possible to apply to handwritten characters.

[0019]

As described above, according to the present invention,
The recognition sample character data is created by applying image processing that simulates the disturbance applied to the recognition target character to the character data that is the basis of the recognition sample character data registered in the character recognition dictionary. Since the creation of character data can be automated, and a large number of sample character data for recognition can be created quickly and in large quantities from a small number of base character data, the sample character data for recognition The burden on the user is greatly reduced as compared to the case where the data is created by the user one by one. In addition, by selecting an appropriate image processing as described above, it is possible to better simulate the disturbance applied to the character to be recognized. Therefore, the recognition work is performed using the recognition sample character data created by the image processing. By performing the above, it is possible to provide a character recognition dictionary creation device and a character recognition dictionary creation method that improve recognition accuracy.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram of an embodiment of a character recognition dictionary creation device according to the present invention.

FIG. 2 is a table listing types of disturbances that can occur with respect to a recognition target character and causes thereof.

FIG. 3 is a table summarizing a part of image processing actually performed on basic image data in a disturbance processing unit of the embodiment.

FIG. 4 is a diagram illustrating a concept of a process performed by a disturbance processing unit included in the character recognition dictionary creation device of the present embodiment.

FIG. 5 is a table showing the results of an experiment of character recognition using a character recognition dictionary created by a conventional method and a character recognition dictionary created by the apparatus of the present embodiment.

[Explanation of symbols]

Reference Signs List 1 character recognition dictionary creation unit 2 character recognition unit 10, 20 image data input unit 11 cutout unit 12 disturbance processing unit 13 dictionary creation processing unit 14 character recognition dictionary 21 recognition unit 22 character output unit 30 control unit 31 first stage image processing unit 31 _{1 to} 31 _N Second-stage image processing units 31 _{11 to} 31 _1N ,..., 31 _{N1 to} 31 _NN Third-stage image processing units S _{1 to} S _N First-stage switches S _{11 to} S _1N ,. .., S _{N1 to} S _NN second-stage switches S _{111 to} S _11N ,..., S _{NN1 to} S _NNN third-stage switches

 ──────────────────────────────────────────────────続 き Continuation of front page (72) Inventor Ichiro Kurihara 2-6-3 Otemachi, Chiyoda-ku, Tokyo New Nippon Steel Corporation

Claims (6)

[Claims] 1. A storage unit for storing a character recognition dictionary, a character input unit for inputting character data which is a source of recognition sample character data registered in the character recognition dictionary, and a disturbance applied to a character to be recognized is simulated. Applying the image processing to the character data input by the character input means, generating sample character data for recognition from the character data input by the character input means, and registering it in a character recognition dictionary of the storage means. A character recognition dictionary creation device, comprising: processing means. 2. The character recognition dictionary creating apparatus according to claim 1, wherein said image processing means performs a plurality of different image processes on one character data input by said character input means. 3. The character recognition according to claim 1, wherein the image processing means performs the same type of image processing with different parameters a plurality of times on one character data input by the character input means. Dictionary creation device. 4. An image processing which simulates a disturbance applied to a character to be recognized is applied to character data which is a source of the sample character data for recognition. And generating a character recognition dictionary by registering them in a character recognition dictionary in a storage unit. 5. The image processing according to claim 4, wherein a plurality of different image processes are performed on the character data that is the basis of the recognition sample character data registered in the character recognition dictionary.
How to create the described character recognition dictionary. 6. The image processing according to claim 4, wherein the same type of image processing with different parameters is performed a plurality of times on the character data serving as the base of the recognition sample character data registered in the character recognition dictionary. 5. A method for creating a character recognition dictionary according to 5.