JP2023023706A - System for reading engraved character and method for reading engraved character - Google Patents

Abstract

To provide a system for reading an engraved character and a method for reading an engraved character, capable of reading an engraved character with high accuracy.SOLUTION: A system 1 for reading an engraved character includes character position identification means 20 that identifies a position of an engraved character for each character in a photographic image obtained by imaging means 10, and character type determination means 30 that determines which of candidate characters a character type is in an engraved character image for each character at the position identified by the character position identification means 20. The character position identification means 20 identifies the position on the basis of a position learning model 21 obtained by learning a feature of an area where the engraved character is located for each character of each candidate character type by machine learning. The character type determination means 30 determines the character type on the basis of a position learning model 31 obtained by learning a feature of a shape for each character of each candidate character type by machine learning.SELECTED DRAWING: Figure 1

Description

The present invention relates to a stamped character reading system and a stamped character reading method for reading stamped characters.

In a car engine assembly line, for example, the number of metal assembly instructions stamped on an engine block is read, and the number is used to select the metal number. For example, numbers are read visually, but in recent years, automatic character reading technology has been developed. For example, in Patent Document 1, a number image is read by a CCD camera from a printed surface on which a string of stamped numbers is printed, transferred to a computer, and a slit narrower than the width of each stamped number is moved to obtain a slit image from the number image. A character recognition device is described which extracts characters and recognizes character types using a neural network. Further, Patent Document 2 describes an accounting processing system that performs character recognition from a read image by OCR processing to generate text data.

Japanese Patent No. 2972011 Japanese Patent No. 6528147

However, with these image analysis technologies, characters can be recognized in the case of printed or handwritten characters in which there is a color difference between the characters and the background, but in the case of engraved characters in which the characters and background are the same color, However, there is a problem that surface gloss, stains, etc. have an adverse effect on image analysis, and may not be recognized as characters. In particular, when characters are double-struck or misaligned, for example, as shown in FIG. As a result, the discrimination rate becomes low, making it difficult to recognize characters.

SUMMARY OF THE INVENTION The present invention is made based on such problems, and an object of the present invention is to provide a stamped character reading system and a stamped character reading method capable of reading stamped characters with high accuracy.

The stamped character reading system of the present invention reads which of a plurality of candidate character types a stamped character to be read belongs to, and uses a photographing means for photographing a stamped surface on which the stamped character is provided, and a photographing means. Character position specifying means for specifying the position of the stamped character for each character in the obtained photographed image, and a stamped character image for each character at the position specified by the character position specifying means in the photographed image obtained by the photographing means. character type determining means for determining which of the candidate character types the character type is, and the character position specifying means learns the characteristics of the area where the stamped character is located for each character for each candidate character type by machine learning The character type discrimination means discriminates based on the character type learning model in which the feature of the shape of the stamped character for each character is learned for each candidate character type by machine learning.

The stamped character reading method of the present invention reads which of a plurality of candidate character types a stamped character to be read belongs to. Character position specifying procedure for specifying the position of the stamped character for each character in the captured image, and a stamped character image for each character at the position specified by the character position specifying procedure in the captured image obtained by the photographing procedure , a character type determination procedure for determining which of the candidate character types the character type is, and in the character position specifying procedure, the position where the feature of the area where the stamped character is located for each character is learned for each candidate character type by machine learning The character type discrimination procedure uses a character type learning model in which the feature of the shape of the stamped character for each character is learned for each candidate character type by machine learning.

According to the present invention, using a position learning model in which the characteristics of the region where the stamped character is located for each character are learned for each candidate character type by machine learning, the position of the stamped character for each character is specified, and Using a character type learning model in which the feature of the shape of the stamped character is learned for each candidate character type by learning, it is determined which of the candidate character types the character type of the stamped character image at the specified position is. It can detect and learn features including surface gloss and dirt, and can read stamped characters with high accuracy.

1 is a diagram showing the configuration of a stamped character reading system according to an embodiment of the present invention; FIG. It is a figure showing one structural example of the hardware of a character position specification means and a character type discrimination|determination means. It is a figure showing the flow of the stamped character reading method which concerns on one embodiment of this invention. It represents an image obtained by binarizing the stamped characters.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 shows the configuration of a stamped character reading system 1 according to an embodiment of the present invention. This stamped character reading system 1 reads which of a plurality of candidate character types a stamped character to be read belongs to. Stamped characters are stamped characters formed by stamping. One character or a plurality of characters may be read. In the present embodiment, an example will be described in which stamped characters provided on a work M such as an engine block are read in a production line such as an engine assembly line.

The stamped character reading system 1 includes, for example, a photographing means 10 for photographing a stamped surface on which a stamped character to be read is provided, and for the photographed image obtained by the photographing means 10, the position of the stamped character is specified for each character. A character position identifying means 20, and a character type for determining which character type is a candidate character type for each stamped character image at a position identified by the character position identifying means 20 in the photographed image obtained by the photographing means 10. and determination means 30 .

The photographing means 10 is composed of, for example, a camera such as a CCD camera, and is fixedly arranged so as to photograph the stamped characters of the work M conveyed through the production line. It is preferable that the imaging range of the imaging means 10 is set so as to include, for example, the entire object to be read.

The character position specifying means 20 can be configured by a computer, for example, and is configured to function as the character position specifying means 20 by executing a program. The character position specifying means 20 is connected to, for example, the photographing means 10, specifies the position of the stamped character for each character from the photographed image obtained by the photographing means 10, and generates the stamped character image for each character at the specified position. It is preferably configured to be cut and saved. The position of the stamped character is specified, for example, based on a position learning model 21 that learns the features of the region where the stamped character is located for each candidate character type by machine learning.

Specifically, the character position specifying means 20 uses, for example, a position learning model 21 that has learned, for each candidate character type, the characteristics of the region where the stamped character is located for each character by machine learning, and the position learning model 21. It has a detection unit 22 that detects each character from a photographed image and cuts it as a stamped character image, and a storage unit 23 that stores the stamped character image cut by the detection unit 22 together with position information. In addition, the character position identifying means 20 preferably has position learning model generating means 24 for generating the position learning model 21, for example.

The position learning model 21 can be generated by the position learning model generation means 24 . For example, the position learning model generating means 24 performs position learning on an image within a region specified by enclosing each character with a frame or the like from a photographed image of a stamping surface on which stamped characters are provided for a plurality of candidate character types. It is configured to extract the feature amount of the position learning image by deep learning, and to learn it as the feature amount of the position where the stamped character of any one of the candidate character types is present. preferable. For example, if the candidate character types are seven types of numbers from 0 to 6, each character is surrounded by a frame or the like from a photographed image of a stamping surface on which seven types of stamped characters from 0 to 6 are provided. It is preferable to use the image in the area designated by as a position learning image, and to learn the position of any one stamped character. The stamped characters of each candidate character type may be provided on the same stamped surface and photographed, or may be provided on different stamped surfaces and photographed. It is preferable to use a convolutional neural network (CNN) for deep learning.

As a result, the position learning model 21 automatically detects and learns the feature amount of each candidate character type for the region in which a single stamped character is located, including glossiness and stains on the surface of the stamped surface. there is That is, the character position identifying means 20 is configured to be able to identify the positions of stamped characters of a plurality of candidate character types using one position learning model 21 . In addition, as the position learning image, there are various possible aspects of the stamped characters to be read, especially considering the range of variations in the parts flowing in the relevant production line, for example, an image of double-struck characters, an oblique It is preferable to include an image of characters whose orientation is rotated, such as upside down, or an image of characters with different brightness.

The character discrimination means 30 can be configured by a computer, for example, and is configured to function as the character discrimination means 30 by executing a program. The character discrimination means 30 is connected to, for example, the character position specifying means 20, and uses the stamped character image at the position specified by the character position specifying means 20 and machine learning to determine the feature of the shape of the stamped character for each candidate character type. Based on the learned character type learning model 31, the character type is determined.

Specifically, the character discrimination means 30 inputs a character type learning model 31 obtained by learning, for example, the feature of the shape of a stamped character for each candidate character type by machine learning, and a stamped character image to the character type learning model 31. and a determination unit 32 for determining which of the candidate character types the character type is based on the output value obtained thereby, and a display unit 33 such as a display for displaying the character type determined by the determination unit 32. . Further, the character determining means 30 preferably has character type learning model generating means 34 for generating the position type learning model 31, for example.

The character type learning model 31 can be generated by the character type learning model generating means 34 . For example, the character type learning model generating means 34 learns the character type from the photographed image of the stamped surface on which the stamped characters are provided for a plurality of candidate character types, and the image within the area specified by enclosing the stamped characters with a frame or the like for each character. It is preferable that the feature amount of the character type learning image is extracted for each candidate character type by deep learning, and the feature amount is associated with the candidate character type of the correct character type for learning. For example, if the candidate character types are seven types of numbers from 0 to 6, each character is surrounded by a frame or the like from a photographed image of a stamping surface on which seven types of stamped characters from 0 to 6 are provided. It is preferable that the image in the area specified by 2 is used as the character type learning image, and is associated with the candidate character type of the correct character type for learning. It is preferable to use the same image as the position learning image for the character type learning image. A convolutional neural network is preferably used for deep learning.

As a result, the character type learning model 31 detects and learns by itself the feature values of the stamped character shape for each of a plurality of types of candidate character types, including glossiness and dirt on the surface of the stamped surface. That is, the character discrimination means 30 is configured to be able to discriminate which of the candidate character types the character type is using one character type learning model 31 . Note that the character type learning image, like the position learning image, can be a variety of possible aspects of the stamped character to be read. It is preferable to include an image of a character that is rotated, an image of a character that is rotated such as obliquely or upside down, and an image of a character that is different in brightness.

The determination unit 32 is connected to the storage unit 24, the character type learning model 31, and the display unit 33, for example. For example, the discrimination unit 32 inputs the stamped character image stored in the storage unit 24 to the character type learning model 31, and the degree of matching is determined based on the output value obtained from the character type learning model 31, for example, the score value of each candidate character type. It is preferable that the character type is determined to be the highest candidate character type. If there is no candidate character type with a high degree of matching, or if there are two or more candidate character types with a similar degree of matching, the character type may be determined to be unknown.

The display unit 33 is preferably configured to display the stamped characters on the stamped surface based on, for example, the character type of the stamped character image determined by the determination unit 32 and the position information of the stamped character image.

FIG. 2 shows an example of the hardware configuration of the character position identifying means 20 and the character type determining means 30. As shown in FIG. The character position specifying means 20 and the character type determining means 30 include, for example, a CPU (Center Processing Unit) 41, a ROM (Read Only Memory) 42, a RAM (Random Access Memory) 43, an HDD (Hard Disk Drive) 44, and an operation interface (operation I/F) 45 . The CPU 41 executes various processes according to various programs recorded in the ROM 42 or various programs loaded from the HDD 44 to the RAM 43 . The RAM 43 also stores data necessary for the CPU 41 to execute various processes. Various data are stored in the HDD 44 .

This stamped character reading system 1 is used, for example, as follows. FIG. 3 shows the flow of the stamped character reading method using the stamped character reading system 1 . First, as a preparatory procedure, the position learning model generating means 24 generates the position learning model 21, and the character type learning model generating means 34 generates the character type learning model 31 (preparatory procedure; step S110).

Specifically, for example, a stamping surface on which stamped characters are provided for a plurality of types of candidate character types is photographed by the photographing means 10, and the stamped characters are enclosed in a frame or the like for each character from the obtained photographed image, and within a designated area is used as a position-learning image, the feature quantity of the position-learning image is extracted by deep learning, and a position-learning model 21 learned as the feature quantity of the position where any one of the complementary character types is present is generated. Further, for example, using the same image as the position learning image as the character type learning image, the feature amount of the character type learning image is extracted for each candidate character type by deep learning, and the feature amount is associated with the candidate character type of the correct character type for learning. A character type learning model 31 is generated. For example, if the candidate character types are seven types of numbers from 0 to 6, each character is surrounded by a frame or the like from a photographed image of a stamping surface on which seven types of stamped characters from 0 to 6 are provided. Prepare images in the area designated by as a character position learning image and a character type learning image.

Next, for example, in a production line such as an engine assembly line, the stamped characters provided on the work M such as an engine block are read as follows. First, the photographing means 10 photographs the stamped surface of the workpiece M on which the stamped characters, for example, the number "32232" is provided (photographing procedure; step S121). Next, for example, the character position identifying means 20 uses the position learning model 21 to characterize the position of the stamped character for each character in the photographed image obtained by the photographing means 10 (character position identifying procedure; step S122). . Specifically, for example, the detection unit 22 detects each character of the stamped character from the captured image using the position learning model 21, cuts it as a stamped character image, and stores the cut stamped character image together with the position information in the storage unit 23. Save to

Subsequently, for example, the character discriminating means 30 uses the character type learning model 31 to discriminate which of the candidate character types the stamped character image of each character at the position specified by the character position specifying means 20 is ( Character type discrimination procedure; step S123). Specifically, for example, the determination unit 32 inputs the stamped character image to the character type learning model 31, and based on the output value obtained from the character type learning model 31, determines that the character type has the highest degree of matching among the candidate character types. and displayed on the display unit 33. For example, the score value of each candidate character type obtained from the character type learning model 31 is 0% for "0", 0% for "1", 98% for "2", 1% for "3", " 4” is 0%, “5” is 1%, and “6” is 0%, it is determined to be the number 2 with the highest score value. This is performed for each stamped character image, and the determined character type is displayed on the display unit 33 as, for example, "32232" based on the position information of each stamped character image.

In the engine assembly line, the engraved characters provided on the engine block were read using the engraved character reading system 1 and OCR. The read rate by System 1 was 100%.

As described above, according to the present embodiment, the position learning model 21, in which the feature of the region where the stamped character is located for each character is learned for each candidate character type by machine learning, is used to determine the position of the stamped character for each character. Which of the candidate character types is the character type of the stamped character image at the specified position using a character type learning model 31 that identifies and learns the features of the shape of the stamped character for each character by machine learning for each candidate character type? can be learned by detecting characteristics including surface gloss and dirt, and the stamped characters can be read with high accuracy.

Although the present invention has been described above with reference to the embodiments, the present invention is not limited to the above embodiments and can be variously modified. For example, in the above embodiments, each component was specifically described, but the specific structure and shape of each component may be different, and all of the components described above may not be provided. , may comprise other components.

Further, in the above embodiment, the case of reading the stamped characters provided on the engine block in the engine assembly line was specifically described, but the present invention can also be applied to the case of reading the stamped characters in other production lines. can. Furthermore, in the above-described embodiment, the case of numerals as candidate character types has been specifically described, but character types other than numerals can also be applied.

REFERENCE SIGNS LIST 1 engraved character reading system 10 photographing means 20 character position specifying means 21 position learning model 22 detection unit 23 storage unit 24 position learning model generation means 30 character type discrimination means 31 Character type learning model 31, 32 Determining unit 33 Display unit 34 Character type learning model generating means 41 CPU 42 ROM 43 RAM 44 HDD 45 Operation interface

Claims (2)

A stamped character reading system for reading which of a plurality of candidate character types a stamped character to be read is,
a photographing means for photographing the stamped surface on which the stamped characters are provided;
Character position specifying means for specifying the position of the stamped character for each character in the photographed image obtained by the photographing means;
Character type determination means for determining which of the candidate character types the character type is for the stamped character image of each character at the position specified by the character position specifying means in the photographed image obtained by the photographing means,
The character position specifying means specifies the characteristics of the region where the stamped character is located for each character based on a position learning model learned for each of the candidate character types by machine learning,
A stamped character reading system, wherein the character type discriminating means discriminates the feature of the shape of the stamped character for each character by machine learning based on a character type learning model that is learned for each of the candidate character types. A stamped character reading method for reading which of a plurality of candidate character types a stamped character to be read is,
a photographing procedure for photographing the stamped surface on which the stamped characters are provided;
A character position identification procedure for identifying the position of the stamped character for each character in the photographed image obtained by the photographing means;
a character type determination procedure for determining which of the candidate character types the character type is for the stamped character image of each character at the position specified by the character position specifying procedure in the photographed image obtained by the photographing procedure,
In the character position specifying procedure, the feature of the region where the stamped character is located for each character is specified by machine learning using a position learning model learned for each of the candidate character types,
In the character type discrimination procedure, a character type learning model in which the feature of the shape of the engraved character is learned for each of the candidate character types by machine learning is used to discriminate using a character type learning model.

Images