Character recognition method, character recognition system, and program

The character recognition method and system address the challenge of accurately recognizing irregularly shaped handwritten characters by employing extraction, correction, and database verification processes, enhancing accuracy and efficiency in steel product management.

JP2024168802AActive Publication Date: 2024-12-05JFE STEEL CORP +1
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
JP2023085754
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-24
Publication Date
2024-12-05
Estimated Expiration
2043-05-24

AI Technical Summary

Technical Problem

Existing character recognition methods in the steel industry struggle with accurately identifying irregularly shaped handwritten characters due to uneven spacing, leading to identification errors and inefficiencies in managing steel products.

Method used

A character recognition method and system that includes character area extraction, recognition, and correction processes to ensure characters adhere to predetermined rules, followed by database comparison to verify accuracy, especially for irregularly shaped characters.

Benefits of technology

Enhances the accuracy of character recognition for irregularly shaped characters, reducing identification errors and improving operational efficiency in steel product management.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024168802000001_ABST
    Figure 2024168802000001_ABST
Patent Text Reader

Abstract

To provide a character recognition method, a character recognition system, and a program for accurately recognizing an irregular character, such as a handwritten character.SOLUTION: A character recognition method for recognizing characters constituting a mark put on an identification target includes: an acquisition step of imaging the mark on the identification target, the mark on the identification target being a character string having a predetermined character string rule, to obtain captured image data; an extraction step of extracting a character region 21 of the mark from the captured image data; a character recognition step of recognizing characters from the extracted character region 21; a first determination step of determining whether the character string composed of the recognized characters satisfies the character string rule registered in advance on a dictionary; and a character correction step of extracting characters that do not satisfy the character string rule, regarding the character string determined to be negative in the first determination step, and correcting the extracted characters so as to satisfy the character string rule.SELECTED DRAWING: Figure 2
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present disclosure relates to a character recognition method, a character recognition system, and a program. [Background technology]

[0002] Conventionally, in the manufacturing industry, in order to identify and manage objects to be identified such as final products and intermediate products, identification strings such as identification numbers of objects to be identified have been handwritten, engraved, printed, or the like at the manufacturing site. In particular, in the steel industry, for example, the identification numbers of objects to be identified were handwritten on the surfaces of objects to be identified, and the identification numbers given to the objects to be identified were read visually, and then the read identification numbers were visually checked against a database in which information on objects to be identified for each identification number was accumulated. This resulted in a large workload for operators and occasional identification anomalies, so that a method of systematizing the process from reading the identification numbers of objects to checking information on objects to be identified has been developed.

[0003] For example, Patent Document 1 discloses a method for reading and recognizing the product markings marked on steel products and comparing the notations of each display item on the product marking with product information and manufacturing history information extracted from a database based on the notation of the production number among the notations.

[0004] Furthermore, Patent Document 2 discloses a slab matching method in which an image of a marking location of an actual slab number on an actual slab is captured, the actual slab number is read from the captured image data, character recognition is performed using a dictionary having multiple marking fonts built in, and the determined actual slab number is matched with the slab number in the rolling instruction data for the next process. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2010-61561 A [Patent Document 2] JP 2018-58114 A Summary of the Invention [Problem to be solved by the invention]

[0006] Incidentally, in the manufacturing process of steel products in the steel industry, when materials to be used in the next process are taken from an intermediate product, an identification object, the taking position may not be constant. In such cases, it is not possible to apply a mark for identifying the identification object (identification mark) to the identification object by online spraying or the like. There is also a method of attaching a label with an identification mark written on it to the identification object, but the label may peel off during the manufacturing process of steel products, and there are cases where the identification mark must be applied directly to the identification object by hand.

[0007] The method described in Patent Document 1 has a problem in that, in the case of handwritten characters, which are characters of an irregular shape, the identification mark cannot be accurately recognized because the character spacing is not uniform.

[0008] In addition, the method described in Patent Document 2 uses a dictionary that has multiple marking fonts built in, which improves the accuracy of character recognition, but the character recognition accuracy may still be insufficient for handwritten characters due to reasons such as uneven character spacing.

[0009] The present disclosure has been made in consideration of the above problems, and aims to provide a character recognition method, a character recognition system, and a program that can accurately recognize even irregular characters such as handwritten characters. [Means for solving the problem]

[0010] In order to solve the above-mentioned problems, the character recognition method of the present disclosure includes: [1] A character recognition method for recognizing characters constituting a mark attached to an object to be identified, comprising the steps of: The indication attached to the object to be identified is a character string having a predetermined character string rule, An acquisition step of acquiring captured image data of the marking attached to the identification object; an extraction step of extracting a character area of ​​the display from the captured image data; a character recognition step of recognizing characters in the extracted character region; a first determination step of determining whether or not a character string composed of recognized characters satisfies the character string rule registered in advance in a dictionary; a character correction step of extracting characters that do not satisfy the character string rule from the character string determined to be invalid in the first determination step, and correcting the extracted characters so that they satisfy the character string rule; The present invention is characterized by having the following.

[0011] In addition, the character recognition method of the present disclosure includes: [2] In the configuration described in [1] above, it is preferable that the display is an object identification string, and identification object information is associated with each object identification string and stored in a database, and the configuration further includes a second determination step of comparing a string determined in the first determination step to satisfy the string rule, or a string corrected in the character correction step for a string determined in the first determination step to not satisfy the string rule, with the object identification string stored in the database to determine whether or not a matching object identification string exists.

[0012] In addition, the character recognition method of the present disclosure includes: [3] In the configuration described in [2] above, it is preferable to further include a selection step of visually determining the character strings determined to be negative in the second determination step, and selecting an object identification character string from the database.

[0013] In addition, the character recognition method of the present disclosure includes: [4] In the configuration described in any one of [1] to [3] above, it is preferable that the display is a character string made up of characters with an irregular shape.

[0014] In addition, the character recognition method of the present disclosure includes: [5] In the configuration described in any one of [1] to [4] above, it is preferable that the character string rule is a rule that defines a character type for each character number indicating the order from the beginning of each character that constitutes the character string attached to the object to be identified.

[0015] In addition, the character recognition method of the present disclosure includes: [6] In the configuration described in any one of [1] to [5] above, it is preferable that the extraction of the character region is performed by detecting the center position of each character that constitutes the character string attached to the object to be identified, and the center position between characters.

[0016] In order to solve the above-mentioned problems, the character recognition system of the present disclosure includes: [7] A character recognition system that recognizes characters constituting a mark attached to an object to be identified, The indication attached to the object to be identified is a character string having a predetermined character string rule, an image acquisition unit that acquires captured image data of an indication attached to the identification object; a character region extraction unit that extracts a character region of the display from the captured image data; a character recognition unit that recognizes characters in the extracted character area; a first determination unit that determines whether or not a character string formed from recognized characters satisfies the character string rule that has been registered in advance; a character correction unit that extracts characters that do not satisfy the character string rule from the character string determined to be invalid by the first determination unit, and corrects the extracted characters so that they satisfy the character string rule; The present invention is characterized by having the following.

[0017] In addition, the character recognition system of the present disclosure includes: [8] In the configuration described in [7] above, it is preferable that the display is an object identification string, identification object information is associated with each object identification string and stored in a database, and the configuration further includes a second determination unit that compares a string determined by the first determination unit to satisfy the string rule or a string corrected by the character correction unit for a string determined by the first determination unit to not satisfy the string rule with the object identification string stored in the database, and determines whether a matching object identification string exists.

[0018] In addition, the character recognition system of the present disclosure includes: [9] In the configuration described in [8] above, it is preferable that the configuration further comprises a selection unit that performs a visual judgment on character strings judged as negative by the second judgment unit and selects an object identification character string from a database.

[0019] In addition, the character recognition system of the present disclosure includes:

[10] In the configuration described in any one of [7] to [9] above, it is preferable that the display is a character string made up of characters with an irregular shape.

[0020] In addition, the program of the present disclosure is

[11] A program for causing a computer to execute a character recognition method for recognizing characters constituting a mark attached to an object to be identified, The indication attached to the object to be identified is a character string having a predetermined character string rule, An acquisition step of acquiring captured image data of the marking attached to the identification object; an extraction step of extracting a character area of ​​the display from the captured image data; a character recognition step of recognizing characters in the extracted character region; a first determination step of determining whether or not a character string composed of recognized characters satisfies the character string rule registered in advance in a dictionary; a character correction step of extracting characters that do not satisfy the character string rule from the character string determined to be negative in the first determination step, and correcting the extracted characters so that they satisfy the character string rule; The present invention is characterized in that it is a program for executing the above. Effect of the Invention

[0021] According to the present disclosure, it is possible to provide a character recognition method, a character recognition system, and a program that are capable of accurately recognizing even characters with irregular shapes, such as handwritten characters. [Brief description of the drawings]

[0022] [Figure 1] 1 is a block diagram showing a configuration of a character recognition system according to an embodiment of the present disclosure. [Diagram 2] 1 is a flowchart illustrating a procedure for implementing a character recognition method according to an embodiment of the present disclosure. [Diagram 3] This figure shows (a) an example of a marking written on an object to be identified, (b1) an example of a character region extracted from the marking, (b2) an example of an inter-character region extracted from the marking, and (c) an example of an extracted character string region. [Figure 4] FIG. 1A is a diagram showing an example of a marking written on an object to be identified, and FIG. 1B is a diagram showing an example of the result of character recognition of the marking. [Diagram 5] FIG. 13A is a diagram showing an example of character string rules and character recognition results for each character number, FIG. 13B is an example of correcting the number of characters to conform to the character string rules, and FIG. 13C is an example of correcting the character recognition results by applying a dictionary according to the character string rules. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0023] Hereinafter, a character recognition system 100 and a character recognition method according to an embodiment of the present disclosure will be illustrated in detail with reference to FIGS.

[0024] Fig. 1 is a block diagram showing the configuration of a character recognition system 100 according to this embodiment. As shown in Fig. 1, the character recognition system 100 includes an imaging device 30 (image acquisition unit) that captures an image of a mark attached to an object to be identified and acquires captured image data, a character recognition device 10 that recognizes a character string from the acquired captured image data, and a database 40 in which an object identification character string and information on the object to be identified are stored.

[0025] Character recognition device 10 includes an input unit 11 to which captured image data from imaging device 30 and material information from database 40 are input, an output unit 12 to which recognized character string data and the like are output, and a calculation unit 13 to perform character recognition, etc. Calculation unit 13 includes a character region extraction unit 14, a character recognition unit 15, a first determination unit 16, a character correction unit 17, and a second determination unit 18.

[0026] (Imaging device) The imaging device 30 captures an image of the marking attached to the object to be identified, and creates captured image data. The imaging device 30 is, for example, a tablet, smartphone, or digital camera having an imaging function, but is not limited to these as long as it is a device having a function of capturing an image of the marking attached to the object to be identified. In addition, when capturing an image of the marking on the object to be identified, it is preferable to fix the imaging device 30 and capture the image. By fixing the imaging device 30, camera shake and focus deviation during capture can be suppressed, and the quality of the captured image data can be improved.

[0027] The marking attached to the object to be identified may be a character string having a character string rule. The character string rule here is a rule that defines a character type for each character number indicating the order from the beginning of each character constituting the character string attached to the object to be identified. The character number indicates the position of the character from the beginning of the character string (the leftmost character "N" in the example of FIG. 4(a)). The character type is, for example, a number, an alphabetic character, a symbol, etc. The character string is, for example, a character string for identifying the object to be identified, such as an identification object number. The characters constituting the character string may be added by engraving, online spraying, printing, or handwriting, but the present disclosure is particularly suitable for recognizing characters with uneven character spacing or irregular shapes, such as handwritten characters.

[0028] (Input section) The input unit 11 is an input interface of the character recognition device 10 for the calculation unit 13 to acquire captured image data from the imaging device 30. In this embodiment, the input unit 11 acquires captured image data obtained by capturing an image of a mark attached to an object to be identified (step S101 in FIG. 2). The format of the captured image data may be a commonly used image data format such as JPEG, TIFF, or BMP.

[0029] Furthermore, the calculation unit 13 acquires the object identification character string and the identification object information in the database 40 through the input unit 11 .

[0030] 1, the calculation unit 13 is configured to only acquire information from the database 40 through the input unit 11, but the present invention is not limited to this embodiment, and the calculation unit 13 may be configured to output and store information to the database 40. The database 40 may be provided in the same PC (Personal Computer) as the calculation unit 13, etc., or may be provided in an external storage device connected to the PC. The database 40 may also be provided in a storage device connected via a network, such as the cloud.

[0031] (Output section) Output unit 12 is an output interface of character recognition device 10 that outputs the pass / fail judged by first judgement unit 16 and / or second judgement unit 18 and the character string data at the time of the judgement. Output unit 12 may be an output display of character recognition device 10, or may output the pass / fail result and the character string data at the time of the judgement as data to another device or the like.

[0032] (calculation section) The calculation unit 13 performs a calculation to identify the marking attached to the object to be recognized as a character string. In this embodiment, the calculation unit 13 includes a character region extraction unit 14, a character recognition unit 15, a first determination unit 16, a character correction unit 17, and a second determination unit 18.

[0033] The function of the calculation unit 13 can be realized as software processing by, for example, executing a predetermined program in a CPU (Central Processing Unit) or a DSP (Digital Signal Processor). However, the present invention is not limited to this embodiment, and the function of the calculation unit 13 may be configured to be realized as hardware processing by, for example, an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array). The above-mentioned predetermined program may be stored in, for example, a storage unit included in the character recognition device 10 or an external storage device.

[0034] (Character area extraction part) The character area extraction unit 14 extracts character areas 21 of each character constituting a character string in a marking attached to an object to be identified from the captured image data acquired from the imaging device 30 (step S102 in FIG. 2). It is preferable to use a method for extracting the character area 21 without preparing a fixed frame, for example, by detecting the center of the character and the space between characters. An example of the extraction of the character area 21 is shown in FIG. 3.

[0035] FIG. 3(a) shows captured image data acquired from the imaging device 30. The captured image data includes a character string "SG098-02T" handwritten on the object to be identified. In this embodiment, the character string is written in white in the captured image data, and is clearly distinguishable from the black color of the object to be identified on the image. The character region extraction unit 14 of the calculation unit 13 recognizes the center position of the character region 21 and the center position of the inter-character region 22 between characters from the captured image data by image processing. In FIG. 3(b1), the black part at the center of the substantially circular character region 21 is an area that the character region extraction unit 14 has determined to have a high probability of being the center position of the character. In FIG. 3(b2), the black part at the center of the substantially circular inter-character region 22 is an area that the character region extraction unit 14 has determined to have a high probability of being the center position of the inter-character region 22. The character region extraction unit 14 extracts character regions 21 for each character by, for example, recognizing the separation positions between characters from the center positions of inter-character regions 22 and further recognizing the size, shape, etc. of each character. Note that the colors of the character string and the object to be identified are not limited to the above-mentioned combinations, and may be other color combinations that are easy to distinguish even with a black-and-white camera, such as yellow and red, or white and blue.

[0036] As described above, character region extraction unit 14 recognizes character region 21 in which each character constituting a character string is written, by recognizing the center position of a character and the center position between characters. Character region extraction unit 14 further recognizes character string region 23 in which the entire character string is written, from recognized character region 21. Figure 3(c) shows an example in which character string region 23 in which the entire character string is written is extracted from character region 21 of each character extracted by character region extraction unit 14. In Figure 3(c), character string region 23 is shown surrounded by a square.

[0037] The operation of the character region extraction unit 14 may be performed as a part of a process in a program executed by the calculation unit 13, or may be performed by another processor, or the captured image data may be sent to a cloud to perform extraction processing of the character region 21. For extraction of the character region 21, open source CRAFT (Character Region Awareness For Text detection) or the like may be applied. By using the method of recognizing the center position of a character and the position between characters shown in FIG. 3, it is possible to accurately extract the character region 21 of each character and recognize the character even if the spacing between characters is not uniform, such as handwritten characters, or the character shape and size are not uniform.

[0038] It is preferable to apply image processing to the image data of the extracted character string area 23 (step S103 in FIG. 2). Any method such as white balance adjustment, size adjustment, noise removal, etc. may be applied as the image processing method. Image processing may be performed by combining several methods. Image processing may be applied individually to the character area 21 of each character, or image processing may be performed on the image data before the character area 21 is extracted.

[0039] (character recognition section) The character recognition unit 15 performs character recognition using image data of the character area 21 in the character string area 23 that has been image-processed (step S104 in FIG. 2). Each character that has been character-recognized is converted into character data, and character string data consisting of the converted characters is created. The function of the character recognition unit 15 may be, for example, OCR (Optical Character Recognition) as a part of processing in a program executed by the calculation unit 13, or processing that combines AI (artificial intelligence) technology with OCR. The above processing can also be realized by transmitting captured image data to a cloud and using a cloud service. The AI ​​used for character recognition may be configured to create a trained model in advance, for example, using a combination of a character image on an object to be identified and a correct character recognition result as teaching data, and to have the trained model read the image data of the character area 21 and output the character recognition result.

[0040] The character recognition unit 15 may output a character recognition result from among candidates that match a predetermined character string rule. For example, in the character string rule in the present disclosure, the first and second characters are alphabetic characters, so when performing character recognition of the first and second characters after extracting the character area 21, the character that is most likely to match the image data of the character area 21 may be selected from among the alphabetic character candidates. This can also be realized by performing character recognition using a learning model that has been trained using combinations of character images containing only alphabetic characters and correct character recognition results as training data.

[0041] (First judgment section) The first determination unit 16 determines whether the character string data recognized and created by the character recognition unit 15 satisfies a character string rule registered in advance in a dictionary (step S105 in FIG. 2). In this embodiment, the character string written on the object to be identified has a character string rule in which the first, second, and ninth characters are alphabetic characters, the third, fifth, and seventh and eighth characters are numbers, and the sixth character is a hyphen, which is a symbol (see the second line in FIG. 5(a)).

[0042] Assume that the character region extraction unit 14 extracts the character region 21 from the display of the object to be identified having such a character string rule, "NF185-01B" (see FIG. 4(a)), and the character recognition unit 15 recognizes the character string consisting of each character in the character region 21 as "NF185-0I81" (see FIG. 4(b)). As shown in FIG. 5(a), the first to seventh characters in the character string recognition result satisfy the predetermined character string rule, whereas the eighth to tenth characters do not satisfy the predetermined character string rule (the recognition result that does not satisfy the character string rule is displayed as a white-out character in the third line of FIG. 5(a)). When the first determination unit 16 determines that the character string recognized by the character recognition unit 15 does not satisfy the character string rule (No in step S105 of FIG. 2), it causes the character correction unit 17 to perform a process of correcting the characters. Furthermore, when the first judgment unit 16 judges that the character string recognized by the character recognition unit 15 satisfies the character string rule (Yes in step S105 of FIG. 2), it causes the second judgment unit 18 to compare the recognized character string with the database 40 (step S108 of FIG. 2).

[0043] (Character correction department) If the recognized character string does not satisfy the character string rules, the character correction unit 17 extracts the eighth to tenth characters (shown as white characters in FIG. 5) that do not satisfy the character string rules from the character string data (step S106 in FIG. 2). Next, the character correction unit 17 corrects the characters extracted in step S106 so that they satisfy the character string rules.

[0044] The character correction unit 17 corrects the number of characters so that the character string recognized by the character recognition unit 15 has the number of characters defined in a predetermined character string rule (step S107 in FIG. 2). In the example shown in FIG. 5, the character recognition result is corrected to delete the 10th character that exceeds the number of characters defined in the character string rule (the character string after correction is shown in FIG. 5(b)). However, the present invention is not limited to this embodiment, and characters in the character string recognition result may be associated with the character string rule so that as many characters as possible in the character string recognition result satisfy the character string rule, and then characters to be deleted may be selected from among the characters that do not satisfy the character string rule. The character string recognition result may be directly compared with the object identification character string stored in the database 40, and characters that do not exist in the object identification character string may be deleted. Alternatively, a trained model may be created in advance using the character image on the identification object and the correct recognition result as teaching data, and the trained model may be configured to read the image data of the character area 21 and delete the recognized character with the smallest probability of being a character.

[0045] Next, the character correction unit 17 applies a dictionary according to the character string rule to the character string whose number of characters has been corrected, and corrects the character string so that the character string satisfies the predetermined character string rule (step S107 in FIG. 2). The character correction unit 17 can apply, for example, the dictionary shown in Table 1 below.

[0046] [Table 1]

[0047] In Table 1, the "character number" indicates the number of the character from the left in the character string recognized from the captured image data. The "character to be converted" and "converted character" mean that when a "character to be converted" appears in the character string recognized from the captured image data, it is converted to the "converted character" to the right. A checked box in Table 1 indicates that the conversion from the "character to be converted" to the "converted character" applies only to the character with the checked "character number."

[0048] For example, the third line of Table 1 indicates that if "character to be converted: D" appears in "character number: 9," it is converted to "character after conversion: B." Since "character number: 9" does not use the character "D" in the identification string of the object to be identified in this disclosure, a dictionary is provided that converts "character number: 9" to the character "B," which has a similar external shape to the character "D," when it is recognized as the character "D."

[0049] For "character numbers: 3 to 8" in the third row of Table 1, the character string rule is "number" or "symbol (hyphen)" as shown in FIG. 5(a), so the conversion to "converted character: B" is not applied and is displayed in gray (meaning that the character string rule is not followed). Also, for "character numbers: 1 to 2", the character string rule is "alphabetical character" (see FIG. 5(a)), so the conversion to "converted character: B" can be applied according to the character string rule. However, since the character "B" is not used in "character numbers: 1 to 2" in the identification character string of the object of identification of the present disclosure, the conversion from "converted character: D" to "converted character: B" is not applied to "character numbers: 1 to 2". Therefore, in Table 1, it is displayed as an unchecked white square (meaning that the character string rule is followed but such conversion is not performed).

[0050] Furthermore, the fifth line of Table 1 indicates that if "character to be converted: i" appears in "character numbers: 3 to 5 and 7 to 8", it is converted to "character after conversion: 1". Since "character numbers: 3 to 5 and 7 to 8" in the identification string of the object to be identified in this disclosure must be "numbers" according to the string rules, a dictionary is provided that converts "character numbers: 3 to 5 and 7 to 8" into the number "1" which has a similar external shape to "i" when it is recognized as "i", which is not a "number".

[0051] It should be understood that Table 1 merely illustrates a portion of a dictionary for correcting characters that do not conform to the character string rules of the present disclosure, and that other characters that should be subject to conversion but are not included in Table 1 are also included in this dictionary.

[0052] The character correction unit 17 corrects the character numbers 8 and 9 in FIG. 5(a) that do not satisfy the character string rule, using the conversion dictionary in Table 1. In Table 1, a dictionary is registered for converting "character to be converted: I" to "converted character: 1" when "character numbers 3 to 5, 7 to 8" appear. The character correction unit 17 applies the conversion from "character to be converted: I" to "converted character: 1" to "character number: 8". In addition, in Table 1, a dictionary is registered for converting "character to be converted: 8" to "converted character: B" when "character number: 9" appears. The character correction unit 17 applies the conversion from "character to be converted: 8" to "converted character: B" to "character number: 9". FIG. 5(c) shows the character string after conversion using the dictionary in Table 1, and it can be seen that it matches the display on the identification object shown in FIG. 4(a).

[0053] A character string composed of characters corrected by the character correction unit 17 is judged by the second judgment unit 18.

[0054] (Second Judgment Section) The second determination unit 18 compares the character string recognized by the character recognition unit 15 and determined by the first determination unit 16 to satisfy the character string rule or the character string corrected by the character correction unit 17 with the object identification character string registered in the database 40. The second determination unit 18 determines whether the recognized character string or the corrected character string exists as an object identification character string in the database 40 (step S108 in FIG. 2). If the object identification character string corresponding to the character string recognized from the identification object or the corrected character string exists in the database 40 (Yes in step S108), the second determination unit 18 acquires the identification object information corresponding to the object identification character string stored in the database 40 and outputs it from the output unit 12. On the other hand, if the object identification character string corresponding to the recognized character string or the corrected character string does not exist in the database 40, it is assumed that a problem has occurred in any step when recognizing the character string from the captured image data. Therefore, the calculation unit 13 returns to step S101 and acquires the captured image data from the imaging device 30 again. Alternatively, an operator may visually recognize the character string indicated on the object to be identified, and obtain the corresponding object information from the database 40 .

[0055] The object identification string is a unique string set for each object to be identified, and may be an object number to be identified, or may be one or more combinations of numbers, letters, symbols, etc. indicating the type name of the object to be identified. The object information to be identified includes, for example, information on the name, specifications, and dimensions of the product from which the object to be identified was collected, information on the collection date of the object to be identified, and the number of objects to be identified collected, and is stored in database 40 in association with each object identification string.

[0056] The character recognition system 100 of the present disclosure is also capable of simultaneously capturing images of marks attached to multiple objects to be identified and recognizing characters. The multiple objects to be identified may be limited to only objects having the same object identification character string, or multiple objects to be identified having different object identification character strings may be simultaneously captured.

[0057] As described above, the character recognition method according to the present embodiment is a character recognition method for recognizing characters constituting a mark attached to an object to be identified, the mark attached to the object to be identified is a character string having a predetermined character string rule, and includes an acquisition step of capturing an image of the mark attached to the object to be identified and acquiring captured image data, an extraction step of extracting a character area 21 of the mark from the captured image data, a character recognition step of recognizing characters in the extracted character area 21, a first determination step of determining whether a character string composed of the recognized characters satisfies a character string rule registered in advance in a dictionary, and a character correction step of extracting characters that do not satisfy the character string rule from the character string determined to be negative in the first determination step and correcting the extracted characters so that they satisfy the character string rule. By adopting such a configuration, the character area 21 of each character constituting the character string is extracted from the captured image data of the mark on the object to be identified and character recognition is performed, so that even characters with an indefinite shape such as handwritten characters can be recognized with high accuracy. In addition, by correcting the character recognition result so that the predetermined character string rule is satisfied, errors in character recognition can be efficiently corrected to improve the accuracy of character recognition.

[0058] In this embodiment, the display is an object identification string, and identification object information is associated with each object identification string and stored in the database 40, and the method further includes a second determination step of comparing a string determined in the first determination step to satisfy the string rule, or a string corrected in a character correction step for a string determined not to satisfy the string rule in the first determination step, with the object identification string stored in the database 40 to determine whether or not a matching object identification string exists. By employing such a configuration, it is possible to efficiently obtain identification object information associated with the object identification string in the database 40 from the recognized object identification string.

[0059] Moreover, in this embodiment, the method is configured to further include a selection step of visually determining the character string determined to be negative in the second determination step and selecting an object identification character string from the database 40. By adopting such a configuration, it is possible to quickly obtain identification object information for an exceptional object identification character string that could not be automatically recognized without re-imaging the object.

[0060] In the present embodiment, the marking is a character string composed of irregular characters. By adopting such a configuration, it is possible to prevent problems such as peeling off of a label bearing an identification mark when the label is attached to an object to be identified during the manufacturing process of a steel product by writing the handwritten characters.

[0061] In this embodiment, the character string rule is configured to be a rule that defines a character type for each character number indicating the order from the beginning of each character that constitutes the character string attached to the object to be identified. By adopting such a configuration, it is possible to optimize the correction dictionary applied for each character number of the recognized character string, thereby improving the accuracy of correcting the recognized character string.

[0062] In this embodiment, the character region 21 is extracted by detecting the center position of each character that composes the character string attached to the object to be recognized, and the center position of the space between characters. By adopting such a configuration, it is possible to recognize the boundary positions between characters from the center positions of the space between characters 22, and further to recognize the size, shape, etc. of each character, thereby accurately extracting the character region 21 of each character.

[0063] The present embodiment is a character recognition system 100 for recognizing characters constituting a mark attached to an object to be identified, the mark attached to the object to be identified is a character string having a predetermined character string rule, and includes an image acquisition unit for capturing an image of the mark attached to the object to be identified and acquiring captured image data, a character region extraction unit 14 for extracting a character region 21 of the mark from the captured image data, a character recognition unit 15 for recognizing characters in the extracted character region 21, a first determination unit 16 for determining whether a character string composed of the recognized characters satisfies a preregistered character string rule, and a character correction unit 17 for extracting characters that do not satisfy the character string rule from a character string determined to be negative by the first determination unit 16 and correcting the extracted characters so that they satisfy the character string rule. By adopting such a configuration, the character region 21 of each character constituting the character string is extracted from the captured image data of the mark on the object to be identified and character recognition is performed, so that even characters with an irregular shape such as handwritten characters can be recognized with high accuracy. Furthermore, by correcting the character recognition results so that they satisfy predetermined character string rules, errors in character recognition can be efficiently corrected and the accuracy of character recognition can be improved.

[0064] In addition, the present embodiment is configured to be a program that causes a computer to execute a character recognition method for recognizing characters constituting a mark attached to an object to be identified, the mark attached to the object to be identified being a character string having a predetermined character string rule, an acquisition step of capturing an image of the mark attached to the object to be identified and acquiring captured image data, an extraction step of extracting a character area 21 of the mark from the captured image data, a character recognition step of recognizing characters in the extracted character area 21, a first determination step of determining whether a character string composed of the recognized characters satisfies a character string rule registered in advance in a dictionary, and a character correction step of extracting characters that do not satisfy the character string rule from the character string determined to be negative in the first determination step and correcting the extracted characters so that they satisfy the character string rule. By adopting such a configuration, the character area 21 of each character constituting the character string is extracted from the captured image data of the mark on the object to be identified and character recognition is performed, so that even characters with an indefinite shape such as handwritten characters can be recognized with high accuracy. In addition, by correcting the character recognition result so that it satisfies the predetermined character string rule, character recognition errors can be efficiently corrected to improve the accuracy of character recognition.

[0065] Although the present disclosure has been described based on the drawings and embodiments, it should be noted that those skilled in the art can easily make various modifications and corrections based on the present disclosure. Therefore, it should be noted that these modifications and corrections are included in the scope of the present invention. For example, the functions included in each component can be rearranged so as not to cause logical inconsistency, and multiple components can be combined into one or divided. It should be understood that these are also included in the scope of the present invention.

[0066] For example, in the present embodiment, character recognition system 100 has been described as including imaging device 30 for capturing an image of a mark attached to an object to be identified, but is not limited to this embodiment. Character recognition system 100 may not include imaging device 30, and may only include an image acquisition unit that acquires captured image data captured outside character recognition system 100. In this case, input unit 11 in FIG. 1 serves as the image acquisition unit. [Explanation of symbols]

[0067] 10 Character recognition device 11 Input section 12 Output section 13 Arithmetic section 14 Character area extraction part 15 Character recognition section 16 First Judgment Section 17 Character correction section 18 Second Judgment Section 21 Character area 22 Inter-character area 23 String area 30 Imaging device (image acquisition unit) 40 Databases 100 Character Recognition System

Claims

1. A character recognition method for recognizing characters constituting a mark attached to an object to be identified, comprising the steps of: The indication attached to the identification object is a character string having a predetermined character string rule, An acquisition step of acquiring captured image data of the marking attached to the identification object; an extraction step of extracting a character area of ​​the display from the captured image data; a character recognition step of recognizing characters in the extracted character region; a first determination step of determining whether or not a character string composed of recognized characters satisfies the character string rule registered in advance in a dictionary; a character correction step of extracting characters that do not satisfy the character string rule from the character string determined to be negative in the first determination step, and correcting the extracted characters so that they satisfy the character string rule; A character recognition method comprising the steps of:

2. The display is an object identification character string, and identification object information is associated with each object identification character string and stored in a database; 2. The character recognition method according to claim 1, further comprising a second determination step of comparing a character string determined in the first determination step as satisfying the character string rule or a character string corrected in the character correction step for a character string determined in the first determination step as not satisfying the character string rule with object identification character strings stored in the database to determine whether or not a matching object identification character string exists.

3. 3. The character recognition method according to claim 2, further comprising a selection step of visually determining the character string determined to be negative in said second determination step, and selecting an object identifying character string from said database.

4. The character recognition method according to claim 1 , wherein the display is a character string composed of characters with an irregular shape.

5. The character recognition method according to claim 1 , wherein the character string rule is a rule that defines a character type for each character number indicating an order from the beginning of each character that constitutes the character string attached to the object to be identified.

6. The character recognition method according to claim 1 , wherein the extraction of the character region is performed by detecting a center position of each character constituting the character string attached to the object to be identified and a center position between characters.

7. A character recognition system that recognizes characters constituting a mark attached to an object to be identified, The indication attached to the identification object is a character string having a predetermined character string rule, an image acquisition unit that acquires captured image data of an indication attached to the identification object; a character region extraction unit that extracts a character region of the display from the captured image data; a character recognition unit that recognizes characters in the extracted character area; a first determination unit that determines whether or not a character string formed from recognized characters satisfies the character string rule that has been registered in advance; a character correction unit that extracts characters that do not satisfy the character string rule from the character string determined to be invalid by the first determination unit, and corrects the extracted characters so that they satisfy the character string rule; A character recognition system having the following structure:

8. The display is an object identification character string, and identification object information is associated with each object identification character string and stored in a database; 8. The character recognition system according to claim 7, further comprising a second determination unit that compares a character string determined by the first determination unit to satisfy the character string rule or a character string corrected by the character correction unit for a character string determined by the first determination unit to not satisfy the character string rule with object identification character strings stored in the database, and determines whether a matching object identification character string exists.

9. 9. The character recognition system according to claim 8, further comprising a selection unit that performs a visual judgment on the character string judged as not correct by the second judgment unit and selects an object identifying character string from a database.

10. The character recognition system according to claim 7 , wherein the display is a character string made up of characters with an irregular shape.

11. A program for causing a computer to execute a character recognition method for recognizing characters constituting a mark attached to an object to be identified, The indication attached to the object to be identified is a character string having a predetermined character string rule, An acquisition step of acquiring captured image data of the marking attached to the identification object; an extraction step of extracting a character area of ​​the display from the captured image data; a character recognition step of recognizing characters in the extracted character region; a first determination step of determining whether or not a character string composed of recognized characters satisfies the character string rule registered in advance in a dictionary; a character correction step of extracting characters that do not satisfy the character string rule from the character string determined to be negative in the first determination step, and correcting the extracted characters so that they satisfy the character string rule; A program that executes the following.

Citation Information

Patent Citations

  • OCR result error correction method and equipment based on verification rule

    CN113128504A

  • Device and method for post-character recognition processing

    JP1994162275A

  • Product display confirmation method for steel product

    JP2010061561A

  • Slab collation method and slab collation system

    JP2018058114A