JP4872930B2 - Character recognition device, fuel pump module assembling device equipped with such device, and character recognition method - Google Patents

Description

  The present invention relates to a character recognition device, a fuel pump module assembling device including such a device, and a character recognition method.

  Conventionally, in the manufacturing process of parts, etc., the part lot number or date of manufacture is stamped on the surface of the part, etc., and the manufacturing history etc. is managed by specifying the part etc. using this part lot number or date of manufacture. Has been done.

  For example, alphabetic characters, Arabic numerals or symbols are used as the characters used for the marking. And in a manufacturing process, the character stamped on the surface is read by the character recognition apparatus.

  In the manufacturing process, from the viewpoint of improving throughput, there is a limit to the time that can be allocated for reading the markings of parts and the like. In the manufacturing process, the stamped character may be scratched during the processing process, or the stamped character may be soiled with oil or dust, and the character may not be recognized correctly.

  Therefore, a character recognition device has been proposed that has been devised to accurately recognize characters.

  For example, in Patent Document 1, after binarizing an image obtained by imaging a plurality of recognized patterns, an image including each recognized pattern is divided into a plurality of partition portions, and partial pattern signals are created for each partition portion. Then, the recognized pattern is recognized by comparing the combination of the partial pattern signals of each section with the database, and for an unrecognizable image, the difference between the unrecognized partial pattern signal in the recognized pattern and the database A character recognition device is disclosed that corrects a threshold value when binarizing an image of a pattern to be recognized based on a signal, and re-checks a pattern to be recognized that cannot be recognized.

  Patent Document 2 discloses a pattern matching apparatus that performs image recognition using data satisfying both pattern matching using a coarse search template and pattern matching by a fine search processing unit.

  However, in the character recognition device described in Patent Document 1, if the threshold value for binarization is corrected, the accuracy with which the pattern to be recognized is collated may be changed.

  Further, in the pattern matching device described in Patent Document 2, since convergence calculation is used for the rough search template optimization process performed by the fine search processing unit, the calculation time is long, and the time required for character recognition may increase. is there.

  Further, in the conventional character recognition device using pattern matching, as shown in the example shown in FIG. 1, when characters engraved on parts or the like have a common character feature pattern, similar characters are displayed. There is a problem that cannot be distinguished or misrecognized.

  The example shown in FIG. 1 uses “I” (alphabet letter “eye”) as a character pattern template for Arabic numerals 1 to 9 and A to V alphabet letters engraved on parts. This is a result of pattern matching.

  As shown in FIG. 1, the Arabic numeral “1” and the alphabet letters “I” and “L” indicate a high matching score with a threshold value of 0.7 or more, and the conventional pattern In a character recognition device using matching, the Arabic numeral “1” and the alphabetic characters “I” and “L” may be indistinguishable or misrecognized.

Japanese Patent Laid-Open No. 5-5611 JP 2002-8028 A

  SUMMARY OF THE INVENTION An object of the present invention is to provide a character recognition device, a fuel pump module assembling device equipped with such a device, and a character recognition method. Another object of the present invention is to provide a character recognition device that can quickly and accurately recognize characters, a fuel pump module assembly device including such a device, and a character recognition method.

  In order to solve the above problems, a character recognition device according to the present invention includes an image acquisition unit that acquires an inspection image obtained by imaging an inspection object on which characters are written, a class having a plurality of character patterns, and a common character feature pattern. Storage means for storing non-common character feature information for identifying a character by examining a non-common part between characters, and for the inspection image, the plurality of read out from the storage means A class classification unit that examines the degree of coincidence with a character pattern and classifies the inspection image into the class having the common character feature pattern, and the inspection image classified into the class is associated with the class. Character specifying means for reading out the non-common character feature information from the storage means, and using the non-common character feature information to specify the character captured in the inspection image , Characterized in that it comprises.

  Thereby, a character can be recognized rapidly and correctly.

  Specifically, in the character recognition device of the present invention, the degree of coincidence with the plurality of character patterns is obtained by the class classification unit with respect to the inspection image, and the degree of coincidence is a predetermined threshold value or more. The group consisting of character patterns is the class having the common character feature pattern, and the class classification means classifies the inspection image into the class and assigns the character pattern having the highest matching degree to the class. The character specifying means uses the non-common character feature information associated with the character pattern selected as the representative character pattern to select a character captured in the inspection image. It is preferable to specify.

  In the character recognition device of the present invention, it is preferable that the non-common character feature information is a partial area of a character in a region at a predetermined position with respect to the center of gravity of the representative character pattern. In particular, in the character recognition device of the present invention, it is preferable that the partial area is an area of a region located above or below the center of gravity of the representative character pattern.

  Thereby, the process of character recognition can be performed more rapidly.

  In the character recognition device of the present invention, the inspection image in which a character is specified is binarized using the non-common character feature information, and the partial area is a pixel having a pixel value of 1. It is preferable that it is the number of these.

  Thereby, the process of character recognition can be performed more rapidly.

  In the character recognition device of the present invention, the class classification unit may determine the degree of coincidence with the plurality of character patterns for the inspection image based on the length of the contour that coincides with the character pattern. preferable.

  As a result, even if the characters written on the object to be inspected are scratched, or the characters are soiled by oil or dust, the characters can be recognized accurately.

  Further, in the character recognition device of the present invention, the class classification unit recognizes a character at a specific position included in a character string to be described on the inspection object as a reference character, and positions relative to the reference character Preferably, a search area is set based on the relationship, and the image of the search area in the inspection image is classified into the class having the common character feature pattern.

  Thereby, even when a plurality of characters are written on the inspection object, the target character can be recognized quickly and accurately.

  In the character recognition device of the present invention, the inspection image includes information of a plurality of characters, and the character specifying unit specifies the plurality of characters and the inspection of each of the specified characters. A character string information acquiring unit that acquires a position in the image, determines the order of the plurality of specified characters based on each of the positions acquired by the character specifying unit, and acquires character string information; When the column information is a character string that may be written on the inspection object, the character string information is determined to be correct, and the character string information is not likely to be displayed on the inspection object. In the case of a character string, it is preferable to include a determination unit that determines that the character string information is an error.

  As a result, it is determined whether the recognized character string is a character string that should not be written on the inspection object, and correct / incorrect determination is performed, so that the character recognition accuracy can be further improved.

  In addition, a fuel pump module assembling apparatus according to the present invention includes any one of the character recognition devices described above, and the object to be inspected is a component of the fuel pump module.

  Thereby, the fuel pump module can be assembled by managing the manufacturing history of the inspection object which is a part.

  Further, the character recognition method of the present invention obtains an inspection image obtained by imaging an inspection object on which characters are written, checks the degree of coincidence with a plurality of character patterns with respect to the inspection image, and obtains the inspection image. Categorized into a class having a common character feature pattern, and is associated with the class with respect to the inspection image classified into the class, and is non-common for identifying a character by examining a non-common part between characters The method includes a step of specifying a character captured in the inspection image using character feature information.

  Thereby, a character can be recognized rapidly and correctly.

  Specifically, in the character recognition method of the present invention, the group of character patterns whose degree of coincidence with the plurality of character patterns is a predetermined threshold or more with respect to the inspection image is the common character feature pattern. And classifying the inspection image into the class, selecting a character pattern having the highest degree of matching as a representative character pattern representing the class, and And the step of specifying specifies the character captured in the inspection image using the non-common character feature information associated with the character pattern selected as the representative character pattern. It is preferable to have.

  In the character recognition method of the present invention, it is preferable that the non-common character feature information is a partial area of a character in a region at a predetermined position with respect to the center of gravity of the representative character pattern.

  Thereby, the process of character recognition can be performed more rapidly.

  The reference numerals in parentheses attached to the above means are an example showing the correspondence with the specific means described in the embodiments described later.

  Hereinafter, the character recognition device of the present invention will be described with reference to FIGS.

  A character recognition device 10 (hereinafter, also simply referred to as this device 10) of the present embodiment is installed in a production line of a fuel pump module that supplies fuel to an in-vehicle engine as an example, and is a main component of the fuel pump module. Recognize a character string of a predetermined size and a predetermined font, such as a part number, date of manufacture, etc., engraved on the surface of the pump body (work) using pattern matching with a character pattern template prepared in advance. Is. The apparatus 10 also has a function of determining whether the recognized character string is a character string that may be imprinted on the workpiece and determining whether the workpiece is good or bad.

  FIG. 2 shows a system configuration diagram of the apparatus 10. The apparatus 10 shoots a workpiece 30 that is an inspection object on which characters are written, and captures an imaging unit 12 as an image acquisition unit that acquires an inspection image, and characters stamped on the surface of the workpiece 30 based on the inspection image. A processing unit 19 that recognizes a column and controls the character recognition device, and a character string determination information storage unit 20 that stores a manufacturing date and the like stamped on the work 30 are included.

  The character string determination information storage unit 20 includes a server used for production line management or control, a hard disk connected to the server, and the like, and stores manufacturing date information and the like of the work 30 flowing on the production line. The character string determination information storage unit 20 is connected to the processing unit 19 through a communication line, and can transmit and receive data to and from each other. In response to a request from the processing unit 19, the manufacturing date information of the workpiece 30 is transmitted to the processing unit 19. In addition, the character string determination information storage unit 20 receives the recognized character string, the correctness determination result of the recognized character string, and the quality determination result of the work 30 from the processing unit 19.

  Hereinafter, each component of this apparatus 10 is demonstrated in detail.

  The imaging unit 12 captures a character string stamped on the surface of the work 30 and acquires an inspection image. Further, the imaging unit 12 captures an image so that the entire character string is included in the inspection image and individual characters included in the character string can be identified. For this purpose, the imaging unit 12 includes a two-dimensional detector composed of a photoelectric converter such as a CCD or C-MOS sensor, and an imaging optical system that forms an image of the surface of the workpiece 30 on the two-dimensional detector. Have. From the viewpoint of preventing adverse effects due to irregular reflection from the workpiece 30, it is preferable to use dark illumination as the illumination used when photographing the inspection image.

  As shown in FIG. 3, a work 30 that is an object to be inspected of the apparatus 10 is a pump main body 30 that is one of the parts assembled to the fuel pump module 34. The fuel pump module 34 is configured by assembling a plurality of parts including the pump body 30. Specifically, the fuel pump module 34 includes an upper cover 31, a pump body 30, a fuel gauge 32, a module cover 33, and the like.

  Thus, since the fuel pump module 34 is configured by assembling a plurality of parts, the pump main body 30 which is a main part has a part number (assistant number) and information for identifying the pump main body 30 before assembling. The date of manufacture of the pump body 30 is recognized by the apparatus 10. When the apparatus 10 determines that the assembly number and the date of manufacture of the pump main body 30 are correct character string information, the pump main body 30 is determined to be non-defective, and the components of the fuel pump module 34 are determined. Assemble as. In addition, on the surface of the upper cover 31 of the assembled fuel pump module 34, the date of manufacture of the pump body 30 recognized by the present apparatus 10 is imprinted together with the date of manufacture of the fuel pump module 34. . In this way, the assembling of the fuel pump module 34 is prevented and the traceability is managed.

  FIG. 4 shows an example of an inspection image acquired by the imaging unit 12. As shown in FIG. 4, a plurality of characters are engraved on the surface of the work 30. On the surface of the work 30, an Arabic numeral character string “31” is written in a large font, and an Arabic numeral and alphabet character string “6IP” is written in a small font. In the present embodiment, “31” is an assembly number of the work 30 that is a pump body, and this character string is a symbol unique to the work 30, and the same assembly number is written on all the works. On the other hand, “6IP” means the date of manufacture of the workpiece 30 and is a character string that varies depending on the workpiece. The letters on the left side indicate the year of manufacture, and 10 Arabic numerals 0 to 9 are used to specify the year in which the work 30 was manufactured among the 10 years. The middle letter means the manufacturing month, and the alphabet in which the workpiece 30 is manufactured is specified by using 12 alphabetic characters A to L. The characters on the right side indicate the date of manufacture, and the date on which the work 30 was manufactured is the day of the month using 9 Arabic numerals 1 to 9 and 22 alphabetic characters A to V. Is identified. The acquired inspection image is transmitted to the processing unit 19 described below.

  The processing unit 19 analyzes the inspection image acquired by the imaging unit 12 and recognizes a character string stamped on the surface of the workpiece 30. And the recognition result is compared with the character string determination information which should be imprinted on the workpiece | work 30 acquired from the character string determination information storage part 20, and it is determined whether the recognized character string is correct.

  Hereinafter, the processing unit 19 will be described in detail. FIG. 5 shows a functional block diagram of the processing unit 19. As illustrated in FIG. 5, the processing unit 19 includes a control unit 11, a communication unit 18, a storage unit 13, a class classification unit 14, a character identification unit 15, a character string information acquisition unit 16, and a determination unit 17. The processing unit 19 includes, for example, a personal computer (PC), its peripheral devices, and a computer program that operates on the PC. The processing unit 19 includes a display unit (not shown) such as a liquid crystal display and an input unit such as a keyboard.

  The control means 11 is composed of a central processing unit (CPU) of a PC and a semiconductor memory such as a read-only memory (ROM) or a random access memory (RAM), and operates according to a computer program read into the CPU. Each unit of the unit 12 and the processing unit 19 is controlled.

  The communication unit 18 is an input / output interface that transmits and receives control signals, image data, or data signals between the processing unit 19 and the imaging unit 12 and the character string determination information storage unit 20, and includes USB, SCSI, RS232C, Ethernet ( It comprises various communication interface circuits such as registered trademark) and driver software for driving them. The processing unit 19 receives the inspection image from the imaging unit 12 through the communication unit 18. On the other hand, the control signal generated by the control unit 11 is transmitted to the imaging unit 12 through the communication unit 18. Further, the communication unit 18 outputs the character string recognized based on the inspection image, the correctness / incorrectness determination result and the quality determination result of the work 30 to an external device such as the display unit or the character string determination information storage unit 20. To do.

  The storage unit 13 includes a random access memory (RAM) or a nonvolatile memory such as a magnetic disk, an optical disk, or a flash memory, and temporarily stores an inspection image received from the imaging unit 12. The storage unit 13 stores a program for controlling the processing unit 19.

  The storage unit 13 stores a plurality of character pattern templates corresponding to each character (for example, alphabetic characters, Arabic numerals, symbols) that may be imprinted on the work 30.

  In the present apparatus 10, as shown in FIGS. 3 and 4, the work 30 as an object to be inspected has a cylindrical shape, and the engraved characters are written on a curved surface. Accordingly, the shape of the captured character may differ depending on the position of the character written on the work 30 and the direction in which the inspection image is captured by the image capturing unit 12. Therefore, the storage means 13 stores three types of corresponding character pattern templates for each character. That is, for each character, a template corresponding to the case where the character imprinted on the cylindrical pump body is imaged from the front, a template corresponding to the case where the imprinted character is imaged slightly from the right, and the imprinted character There are three types of templates corresponding to the case where the image is taken from the left side.

  In each character pattern template, the position of the center of gravity is defined as indicated by the crosses in the drawings of FIGS. The position of the center of gravity is stored in the storage unit 13 in association with the character pattern template. The position of the center of gravity associated with the character pattern template is used by the character specifying means 15 described later.

  The storage unit 13 stores identification information for identifying a character pattern template and a character code representing a character corresponding to the character pattern template in association with each character pattern template.

  In addition, the storage unit 13 stores non-common character feature information that is associated with a class having a common character feature pattern, and that identifies a character by examining a non-common part between characters. Specifically, the non-common character feature information is stored in association with each character pattern template that is selected and becomes a representative character pattern template representing the class. The non-common character feature information is also stored in the storage unit 13 in association with a character code representing a character corresponding to the associated character pattern template. This non-common character feature information will be described later in detail.

  Next, the class classification unit 14, the character identification unit 15, the character string information acquisition unit 16, and the determination unit 17 described below are implemented as program modules executed on the CPU, for example. Alternatively, each of these units may be mounted as a dedicated processing board including an image processing processor provided separately from the CPU.

  Next, the class classification means 14 will be described below. The class classification unit 14 examines the degree of coincidence of the inspection image with a plurality of character patterns read from the storage unit 13, and classifies the inspection image into a class having a common character feature pattern. The class classification unit 14 first performs preprocessing before classifying the inspection image stored in the storage unit 13 into a class having a common character feature pattern. As the preprocessing, if the inspection image is taken under dark illumination, for example, the dynamic range is widened to increase the luminance dispersion. Further, as preprocessing, unevenness correction processing, noise removal processing, or the like may be performed.

  Further, the class classification unit 14 may binarize the inspection image as preprocessing. The binarization of the inspection image is preferably performed at the latest before the character specifying unit 15 processes the inspection image. When pattern matching is performed using a binarized image, the character pattern template is preferably a binarized pattern so that the character portion and the background portion have different values.

  The class classification unit 14 recognizes a character at a specific position included in the character string to be written on the work 30 as a reference character with respect to the pre-processed inspection image, and positions relative to the reference character A search area is set based on the relationship. In the present apparatus 10, the assembly number of the workpiece 30 is used as the reference character. The reason why the assembly number is used as the reference character is that, as described above, the assembly number is a character string that is written in common to the workpieces that are all pump bodies.

  The class classification unit 14 performs pattern matching on the inspection image using the character pattern template corresponding to the assembly number, which is the Arabic numeral “31” in the apparatus 10.

  As shown in FIG. 6, when classifying means 14 recognizes the character string of the assembly number in the inspection image, it obtains coordinates 101g in inspection image 101 corresponding to the barycentric position of the character pattern template corresponding to the assembly number. .

  The class classification unit 14 sets the search area 102 based on the relative positional relationship with the coordinates 101g in the inspection image corresponding to the barycentric position. In the example of FIG. 6, the search area 102 is set to a horizontally long rectangle at a position close to the reference character. Since the position of the marking may vary greatly when a laser marking is used, the search area 102 is preferably set to be larger than the expected variation range of the marking position. If the character “31” as the assembly number is not recognized, the class classification unit 14 determines that the work 30 is a defective product, and transmits the determination result to the determination unit 17 for subsequent processing. Cancel.

  Next, the class classification unit 14 obtains the degree of coincidence with a plurality of character pattern templates read from the storage unit 13 in the search area 102 set in the inspection image, and the degree of coincidence is equal to or higher than a predetermined threshold value. The group of a certain character pattern template is set as a class having a common character feature pattern, and the image of the search area 102 in the inspection image is classified into the class. Images with different common character feature patterns are classified into different classes.

  The class classification unit 14 obtains the degree of coincidence by pattern matching while changing the relative position and angle between the search area 102 and each character pattern template within the search area. Then, the class classification unit 14 specifies the character pattern template that maximizes the matching degree, and obtains the position of the character pattern template when the matching degree is maximized.

  Here, as shown in FIG. 6, when a plurality of characters are imaged in the search area 102, each character image is classified into a class. Further, the classification into each class is the portion of the character in the inspection image corresponding to the character pattern template, in which the degree of coincidence with the character pattern template shows a value equal to or higher than the predetermined threshold in the inspection image. is there.

  Next, processing for classifying the inspection image into classes by the class classification unit 14 will be described below. FIG. 7 shows an example of a result of checking the degree of coincidence by performing pattern matching using a plurality of character pattern templates on an image obtained by capturing each character imprinted on the work 30 with the apparatus 10. For example, when pattern matching is performed between an alphabetic character “I (I)” engraved on the work 30 and each of a plurality of character pattern templates, an Arabic numeral “1” and an alphabetic character “I” ) ”And“ L ”showed a high degree of coincidence of 0.8 or more. Therefore, in the present apparatus 10, the Arabic numeral “1” and the alphabetic characters “I” and “L” are commonly used as a group of character pattern templates having a matching degree of 0.8 or more. We decided to classify into the same class with character feature patterns. That is, it is considered that the inspection image classified into this class includes the Arabic numeral “1” and the alphabetic characters “I” and “L”.

  Further, when pattern matching is performed between the alphabet character “O (O)” as the character engraved on the workpiece 30 and each of the plurality of character pattern templates, the alphabet characters “O (O)” and “Q (queue) ) "Character pattern template showed a high degree of coincidence of 0.9 or more. Therefore, in the present apparatus 10, the alphabet characters “O (o)” and “Q (queue)” are the same class having a common character feature pattern as a group of character pattern templates having a matching degree of 0.9 or more. It was decided to classify.

  Furthermore, when pattern matching is performed between “P (pea)” of the alphabet character as the character engraved on the work 30 and each of the plurality of character pattern templates, the alphabet characters “P (pea)” and “R (R) ) "Showed a high degree of coincidence of 0.7 or more. Therefore, alphabet letters “P” and “R” are also classified into the same class having a common character feature pattern as a group of character pattern templates having a matching degree of 0.7 or more. did.

  As a result of the above consideration, in the present apparatus 10, the threshold value of the degree of coincidence classified into the same class is set to 0.7.

  Therefore, when the class classification unit 14 obtains the degree of coincidence with the alphabet character “I” as the character pattern template in the search area 102 set in the inspection image, the alphabet displayed in the search area 102 is displayed. A degree of coincidence of 0.7 or more is obtained for the character “I (eye)”. Similarly, the class classification means 14 uses the search area 102 set in the examination image to read the Arabic numeral “1” or the alphabetic letter “L” as the character pattern template read from the storage means 13. When the degree of coincidence is obtained, a degree of coincidence of 0.7 or more is obtained for the alphabet letter “I” or the letter “L” written in the search area 102.

  Therefore, the class classification means 14 obtains 0.7 or more as the degree of coincidence with the Arabic numeral “1” and the alphabetic character “I” or “L” as the character pattern template. The character portions in the inspection image corresponding to each character pattern template are classified into the same class.

  Similarly, the class classification unit 14 determines that each character pattern template has a matching score of 0.7 or more as “O (o)” or “Q (queue)” of the alphabetic character that is the character pattern template. The character portion in the corresponding inspection image is classified into another class. Similarly, the class classification means 14 determines that each character pattern template has a matching score of 0.7 or higher as “P (Piece)” or “R (R)” of the alphabetic characters that are character pattern templates. Classify the character part in the corresponding inspection image into another class

  In FIG. 7, the alphabet characters “I”, “O” and “P” are used as character images engraved on the work 30, and patterns with a plurality of character pattern templates are shown. Although the matching degree when matching is shown, characters having a matching degree of 0.7 or more are matched with the same class by pattern matching with a character pattern template using an image of another character stamped on the work 30. Can be classified.

  In addition, the class classification unit 14 classifies the inspection image into classes, and selects a character pattern template having the highest matching score as a representative character pattern template representing the class. This representative character pattern template is associated with the class. Specifically, in the example of FIG. 7, when the character engraved on the work 30 is an alphabetic character “I (eye)”, the character pattern template “I (eye)” with the highest degree of matching is obtained. Are selected as representative character pattern templates. However, when the work 30 is inspected, the characters stamped on the work 30 are actually unknown.

  Similarly, when the character engraved on the work 30 is an alphabetic character “O”, the character pattern template “O” is selected as the representative character pattern template. When the character engraved on the work 30 is an alphabetic character “P”, the character pattern template “P” is selected as the representative character pattern template.

  In this way, in the example of FIG. 6, the class classification unit 14 selects three representative character pattern templates corresponding to the three characters in the search area 102. For example, for an alphabetic character “I” in the search area 102, the character pattern template “I” is selected as the representative character pattern template.

  Further, the class classification unit 14 acquires the position in the inspection image of the character recognized by each of the selected representative character patterns. As described above, since each character pattern template and the position of its center of gravity are associated and stored in the storage means 3, the class classification means 14 stores the character pattern template selected as the representative character pattern template. The position of the center of gravity is read from the storage unit 13. Then, the class classification unit 14 acquires the coordinates in the inspection image corresponding to the position of the center of gravity of the character pattern template selected as the representative character pattern template, and the character recognized by each of the selected representative character patterns. Obtain the position in the inspection image.

  Moreover, in this apparatus 10, the class classification means 14 calculates | requires the coincidence degree with a some character pattern template with respect to a test | inspection image based on the length of the outline matched with the character pattern template.

  The characters stamped on the workpiece may be scratched or may be soiled with oil or dust. In this case, when pattern matching is performed based on the length of the contour, only the pixel of the character existing on the contour of the engraved character has an adverse effect on the matching degree. Relatively less. Accordingly, it becomes easy to accurately detect characters.

  When class classification means 14 finishes recognizing all characters in search area 102, identification information for identifying the character pattern template selected as the representative character pattern template in search area 102, the value of coincidence, and the selected value The coordinates of the center of gravity as the position in the inspection image of the character recognized for each representative character pattern are stored in the storage means 13.

  Next, the character specifying means 15 will be described below. The character specifying unit 15 reads out the non-common character feature information associated with the character pattern template selected as the representative character pattern template associated with the class from the storage unit 13, and uses the non-common character feature information. The character captured in the inspection image is specified.

  In the present apparatus 10, the inspection image in which the character specifying unit 15 specifies a character using the non-common character feature information is an image binarized by the class classification unit 14, and the pixel value of the portion corresponding to the character is 1 The pixel value is 0 for the other portion. Here, a pixel value of 1 can be white, and a pixel value of 0 can be black.

  In the present apparatus 10, the non-common character feature information is a partial area of a character in a region at a predetermined position with respect to the center of gravity of the representative character pattern template, as shown in FIGS. . That is, the partial area is the number of pixels having a pixel value of 1.

  Hereinafter, further description will be given using the example of FIG. FIG. 8A shows an image D obtained by cutting out the image portion of the search area 102 in which the representative character pattern template shows the best match. In FIG. 6, in the search area 102, an area to be cut out for each character is surrounded by a square.

  Assume that the representative character pattern template “I” is selected for the characters in the search area 102. First, the character specifying unit 15 reads from the storage unit 13 the position of the center of gravity of the character pattern template “I” selected as the representative character pattern template and the non-common character feature information associated with the character pattern template “I”.

  Next, the character specifying unit 15 cuts out the image portion of the search area 102 in which the representative character pattern template “I” shows the best match, and specifies the position of the center of gravity Tig of the representative character pattern template in the cut out image D. . Next, the character specifying means 15 checks the partial area of the character in the region Tiu at the upper position and the region Tid at the lower position with respect to the center of gravity Tig of the representative character pattern as non-common character feature information.

  In the present apparatus 10, the non-common character feature information is the character partial area of the region Tiu in the upper position and the region Tid in the lower position with respect to the center of gravity Tig of the representative character pattern. This non-common character feature information is preferably a character portion other than a feature portion common to character patterns used for classification, and a portion having a large amount of character information.

  Now, since the representative character pattern template “I (eye)” is selected for the inspection image, the characters included in the inspection image are the Arabic numeral “1”, the alphabetic character “I (eye)” or “L”.

  The character specifying means 15 uses the partial area to determine whether the character included in the inspection image is an Arabic numeral “1”, an alphabetic character “I” or “L”. Identify.

  As shown in FIG. 8A, if the character engraved on the workpiece is an alphabetic character “I”, the region Tiu at the upper position and the region at the lower position with respect to the center of gravity Tig. Both of the partial areas of the characters of Tid are values below a predetermined threshold value. Since the character having such a partial area is a character corresponding to the alphabet character “I” of the character pattern template classified into the class, the character specifying means 15 determines that the cut image D is an alphabet character. “I (eye)”. Then, the character specifying unit 15 acquires a character code representing the character corresponding to the character pattern template of the alphabet character “I”.

  As shown in FIG. 8B, if the character engraved on the workpiece is the Arabic numeral “1”, the partial area of the character in the region Tiu located above the center of gravity Tig is predetermined. The partial area of the character in the region Tid that is larger than the threshold value and in the lower position is a value equal to or smaller than the predetermined threshold value. Since the character having such a partial area is a character corresponding to the Arabic numeral “1” in the character pattern template classified into the class, the character specifying unit 15 uses the Arabic image “1” as the cut image D. ". Then, the character specifying unit 15 acquires a character code representing the character corresponding to the character pattern template “1” of the Arabic numeral.

  Further, as shown in FIG. 8C, if the character engraved on the workpiece is an alphabetic character “L”, the partial area of the character in the region Tiu located above the center of gravity Tig. Is a value equal to or smaller than a predetermined threshold value, and the character partial area of the region Tid in the lower position is larger than the predetermined threshold value. Since the character having such a partial area is a character corresponding to the alphabet character “L” of the character pattern template classified into the class, the character specifying unit 15 determines that the cut image D is an alphabet character. It is specified as “L”. Then, the character specifying means 15 acquires a character code representing the character corresponding to the character pattern template “L” of the alphabet character.

  As described above, the character specifying unit 15 uses the partial area of the character in the region Tiu in the upper position and the region Tid in the lower position as the non-common character feature information with respect to the center of gravity Tig of the representative character pattern. The cut out image D is examined, and the character captured in the cut out image D is specified.

  As the region Tiu in the upper position and the region Tid in the lower position with respect to the center of gravity Tig of the representative character pattern, for example, the region in the upper position in the representative character pattern divided into three equal parts, and It can be a region in a lower position.

  In the search area 102, the character specifying unit 15 stores a character code representing the specified character in the storage unit 13 together with the coordinates of the center of gravity in the inspection image of the character.

  9 and 10 show non-common character feature information when the alphabet characters “O” and “P” are selected as the representative character pattern template.

  In the example of FIGS. 9A and 9B, the representative character pattern template “O (O)” is selected for the inspection image, and therefore, the characters included in the inspection image are “O ( O) "or" Q (queue) ".

  As shown in FIG. 9A, if the character engraved on the workpiece is an alphabetic character “O”, the partial area of the character in the region Tod located below the center of gravity Tog is It becomes a value below a predetermined threshold value. Since the character having such a partial area is a character corresponding to the alphabet character “O (O)” of the character pattern template classified into the class, the character specifying unit 15 determines that the cut image D is an alphabet character. “O (O)”.

  Also, as shown in FIG. 9B, if the character engraved on the workpiece is an alphabetic character “Q (queue)”, the partial area of the character in the region Tod located below the center of gravity Tog. Becomes a value larger than a predetermined threshold value. Since the character having such a partial area is a character corresponding to “Q (cue)” of the alphabetic character of the character pattern template classified into the class, the character specifying means 15 determines that the cut image D is an alphabetic character. "Q (queue)".

  As described above, the character specifying unit 15 checks the partial area of the character in the region Tod at the lower position with respect to the center of gravity Tog of the representative character pattern in the cut image D as non-common character feature information, Characters captured in the cut out image D are specified.

  In the example of FIGS. 10A and 10B, the representative character pattern template “P (Pi)” is selected for the inspection image. Therefore, the characters included in the inspection image are the alphabet characters “P”. (P) "or" R (R) ".

  Similarly, as shown in FIGS. 10A and 10B, the character specifying unit 15 includes a character portion of a region Tpd located below the center of gravity Tpg of the representative character pattern in the cut out image D. The area is examined using the non-common character feature information, and the character captured in the cut out image D is specified.

  In the example of FIG. 9 or 10, the region at the lower position with respect to the center of gravity of the representative character pattern is, for example, the region at the lower position of the representative character pattern divided in half in the vertical direction. Can do.

  The character string information acquisition unit 16 determines the order of the plurality of specified characters based on each character position acquired by the character specifying unit 15 and acquires the character string information. Specifically, the character string information acquiring unit 16 determines the character arrangement order based on the positional relationship of all the characters recognized by the character specifying unit 15. And the character string information acquisition means 16 acquires the character string information showing a manufacture date based on the determined arrangement order of the characters.

  In the present apparatus 10, as shown in FIG. 4, a character string specifying the date of manufacture is written on the work 30. Therefore, the character string information acquisition unit 16 refers to the position information of the center of gravity of the character stored in the storage unit 13 as the character recognized in the search area 102, and the x direction (lateral direction) stored in the storage unit 13. Are compared and arranged in order from the leftmost character. Thereafter, the character string information acquisition unit 16 acquires character string information with reference to a character code representing the specified character. The character string information acquisition unit 16 stores the acquired character string information in the storage unit 13.

  The determination unit 17 determines that the character string information is correct when the character string information is a character string that may be written on the work 30, and the character string information is written on the work 30. If the character string is not likely to be detected, it is determined that the character string information is incorrect.

  Specifically, the determination unit 17 determines whether the character string information is correct by checking whether the character string information acquired by the character string information acquisition unit 16 is not an impossible character string. . Moreover, the quality of the workpiece | work 30 is determined based on the determination result. For this purpose, the determination unit 17 acquires character string determination information serving as a determination reference from the character string determination information storage unit 20 through the communication unit 18. The character string determination information is, for example, a manufacturing date range of the workpiece 30 currently flowing on the line. Then, the recognized character string information is compared with a matching condition set based on the character string determination information.

  For example, the determination unit 17 determines that the date of manufacture included in the recognized character string information is a date that is not actually possible (for example, 19th, 38th, etc.), or the recognized date of manufacture. When the corresponding work 30 is out of the manufacturing date range flowing on the line (for example, only the parts manufactured in October should be on the line, but the recognized manufacturing date is September. It is determined that the character string information recognized as the manufacturing date information is incorrect in the case of before or after November.

  If the determination unit 17 determines that the character string information is correct, the determination unit 17 determines the work 30 as a non-defective product. On the other hand, when it is determined that the character string information is incorrect, the work 30 is determined as a defective product. Further, since the assembly number is incorrect, the determination unit 17 also determines that the workpiece 30 is defective even when the class classification unit 14 determines that the workpiece 30 is defective.

  The determination unit 17 transmits the recognized character string information and the pass / fail determination result of the work 30 to the display unit through the communication unit 18 and notifies the user of the result. Alternatively, the recognized character string information and the quality determination result of the work 30 are transmitted to the character string determination information storage unit 20 or a computer (not shown) managing the production line, and the work 30 determined to be defective is removed from the line. You may make it do.

  The apparatus 10 described above may be incorporated in a part of the assembly apparatus of the fuel pump module. The fuel pump module assembling apparatus includes the present apparatus 10 and can determine whether or not the fuel pump module is a non-defective product before assembling the work body 30 as a component into the fuel pump module. Therefore, the fuel pump module can be assembled by accurately managing the manufacturing history of the workpiece 30 as a part. Moreover, about the assembly | attachment method of the components containing the workpiece | work 30, a conventional normal method can be used, for example.

  Next, the operation of the character recognition device 10 of the present embodiment described above will be described. As described above, the apparatus 10 recognizes characters captured in the inspection image using a two-step procedure.

  First, the apparatus 10 examines the degree of coincidence of a test image with a plurality of character pattern templates, and classifies the test image into a class having a common character feature pattern. At this stage, for example, using the least common divisor character information in the character pattern template, the inspection images are roughly classified into groups for which the character pattern template features are common.

  Next, the apparatus 10 uses the non-common character feature information for identifying the character by examining the non-common part between the characters with respect to the inspection image classified into the class. The character captured in the inspection image is specified. At this stage, individual characters are specified using the character information of the portion excluding the least common divisor character information between characters classified into the same class. The non-common character feature information is preferably, for example, a character portion other than a feature portion common to the character pattern templates used for the classification, and a portion having a large amount of character information.

  There are no particular restrictions on the characters written on the object to be inspected which is the inspection target of the character recognition device 10. Examples of characters that are to be recognized include alphabet characters, Arabic numerals, hiragana, kanji, symbols, and the like.

  Next, an example of the operation of the character recognition device 10 of the present embodiment described above will be described below with reference to FIGS. The operation of the character recognition device 10 is controlled by the control means 11 of the processing unit 19.

  First, in step S10, the imaging unit 12 acquires an inspection image obtained by imaging the work 30 on which a plurality of characters are written. The storage unit 13 stores the captured inspection image.

  Next, in step S11, the class classification unit 14 performs preprocessing on the inspection image.

  Next, in step S12, the class classification unit 14 recognizes an assembly number, which is a character at a specific position included in the character string to be written on the work 30, as a reference character, and a relative position with respect to the reference character. Based on the relationship, a search area is set in the inspection image.

  Next, in step S13, the class classification unit 14 obtains a degree of coincidence with a plurality of character pattern templates for the search region of the inspection image, and from the character pattern template whose degree of coincidence is a predetermined threshold value or more. A group having a common character feature pattern is classified into this class.

  Next, in step S14, the class classification unit 14 selects the character pattern template having the highest matching score as the representative character pattern template representing the class, and is recognized as the selected representative character pattern. Get the position of the character in the inspection image.

  Next, in step S15, the character specifying unit 15 specifies the character imaged in the inspection image using the non-common character feature information associated with the character pattern template selected as the representative character pattern template. Specifically, the character specifying means 15 uses the partial area of the character in a region at a predetermined position with respect to the center of gravity of the representative character pattern template as non-common character feature information.

  Next, the character specifying unit 15 determines whether or not all the characters in the search area have been specified. If the identification of all characters in the search area has not been completed, the process returns to S15. On the other hand, if all the characters in the search area have been specified, the process proceeds to step S17.

  Next, in step S <b> 17, the character string information acquisition unit 16 identifies the plurality of identified plural characters based on the positions in the inspection image of the characters recognized by the selected representative character patterns acquired by the class classification unit 14. Determine the order of characters and get string information.

  Next, in step S <b> 18, the determination unit 17 acquires character string determination information from the character string determination information storage unit 20.

  Next, in step S <b> 19, the determination unit 17 determines whether or not the character string information is a character string that may be written on the work 30. If the character string information is a character string that may be written on the work 30, the process proceeds to step S20. On the other hand, if the character string information is not a character string that is not likely to be written on the work 30, the process proceeds to step S21.

  In step S <b> 20, the determination unit 17 determines that the character string information is correct, and outputs a determination result that the work 30 is non-defective.

  On the other hand, in step S21, the determination unit 17 determines that the character string information is incorrect, and the work 30 outputs a determination result that the work 30 is defective.

  According to the character recognition device 10 of the present embodiment described above, characters can be recognized quickly and accurately. In particular, the apparatus 10 can accurately recognize a character even if the character written on the object to be inspected is scratched or dirty due to oil or dust.

  In addition, the apparatus 10 uses a partial area of a character in a region located at a predetermined position with respect to the center of gravity of the representative character pattern template as non-common character feature information, and this partial area is binarized. Since the number of pixels in the image is 1, the character recognition process can be performed more quickly.

  Further, according to the present apparatus 10, a character at a specific position included in a character string to be written on the work 30 is recognized as a reference character, and a search area is set based on a relative positional relationship with the reference character. Thus, even when a plurality of characters are written, the target character can be recognized quickly and accurately.

  Further, according to the present apparatus 10, since it is determined whether or not the recognized character string is a character string that should not be written on the work 30, correct / incorrect determination is performed, so that the character recognition accuracy can be further improved.

  The character recognition device of the present invention, the assembly device of the fuel pump module equipped with such a device, and the character recognition method are not limited to the above-described embodiments, and can be changed as appropriate without departing from the spirit of the present invention. is there.

  For example, in the character recognition device of the present embodiment described above, the non-common character feature information is a partial area of a character in a region at a predetermined position with respect to the center of gravity of the representative character pattern. The information may be, for example, a character aspect ratio, circularity, or the like.

  In the character recognition device of the present embodiment described above, the partial area is an area of a region located above or below the center of gravity of the representative character pattern. However, the partial area includes the center of gravity. It may be the area of the central region.

It is the result of having performed pattern matching by the character recognition apparatus of the conventional example. It is a system configuration figure of one embodiment of the character recognition device of the present invention. It is a disassembled perspective view explaining the workpiece | work which the embodiment of FIG. 2 inspects. It is a figure which shows an example of the test | inspection image which imaged the workpiece | work of embodiment of FIG. FIG. 3 is a functional block diagram of a processing unit in FIG. 2. It is a figure which shows the test | inspection image to which the search area | region was set. It is a figure explaining the class which has a common character feature pattern. It is a figure explaining the process which specifies the character imaged by the test | inspection image using non-common character feature information. It is a figure which shows the other example of FIG. It is a figure which shows the other example of FIG. It is a flowchart explaining the operation | movement of embodiment of FIG. It is a flowchart explaining the operation | movement of embodiment of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Character recognition apparatus 11 Control means 12 Imaging part (image acquisition means)
DESCRIPTION OF SYMBOLS 13 Memory | storage means 14 Class classification means 15 Character identification means 16 Character string information acquisition means 17 Determination means 18 Communication means 19 Processing part 20 Character string determination information storage part 30 Workpiece (inspection object)

Claims (11)

Image acquisition means for acquiring an inspection image obtained by imaging an inspection object on which characters are written;
Storage means for storing a plurality of character patterns and non-common character feature information for identifying a character by examining a non-common part between characters, which is associated with a class having a common character feature pattern;
A class classification unit that examines the degree of coincidence with the plurality of character patterns read from the storage unit with respect to the inspection image, and classifies the inspection image into the class having the common character feature pattern;
For the inspection images classified into the class, the non-common character feature information associated with the class is read out from the storage unit, and the inspection image is captured using the non-common character feature information. A character identification means for identifying a character;
Equipped with a,
The character classifying device according to claim 1, wherein the class classification unit obtains a degree of coincidence with the plurality of character patterns for the inspection image based on a length of a contour coincident with the character pattern . The class classification means obtains the degree of coincidence with the plurality of character patterns with respect to the inspection image, and a group of character patterns having the degree of coincidence is a predetermined threshold value or more includes the common character feature pattern Said class having
The class classification means classifies the inspection image into the class, and selects a character pattern having the highest matching score as a representative character pattern representing the class,
The character recognition according to claim 1, wherein the character specifying unit specifies a character imaged in the inspection image using the non-common character feature information associated with the character pattern selected as the representative character pattern. apparatus.   The character recognition device according to claim 2, wherein the non-common character feature information is a partial area of a character in a region located at a predetermined position with respect to the center of gravity of the representative character pattern.   The character recognition device according to claim 3, wherein the partial area is an area of a region located above or below the center of gravity of the representative character pattern.   5. The inspection image in which a character is specified using the non-common character feature information is a binarized image, and the partial area is the number of pixels having a pixel value of 1. The character recognition device described. The class classification means recognizes a character at a specific position included in a character string to be written on the inspection object as a reference character, sets a search area based on a relative positional relationship with the reference character, character recognition apparatus according to the image of the search area, to any one of claim 1 to 5 for classifying the class with the common character feature pattern in the inspection image. The inspection image includes information of a plurality of characters,
The character specifying means specifies the plurality of characters and obtains positions of the specified characters in the inspection image,
Character string information acquisition means for determining the order of the plurality of specified characters based on each of the positions acquired by the character specification means, and acquiring character string information;
When the character string information is a character string that may be written on the inspection object, it is determined that the character string information is correct, and the character string information may be written on the inspection object If it is no character string, the character string information is character recognition apparatus according to any one of claim 1 to 6, and a, a determination unit determines that an error. An assembly apparatus for a fuel pump module, comprising the character recognition device according to any one of claims 1 to 7, wherein the object to be inspected is a component of the fuel pump module. Obtain an inspection image that captures the inspected object with characters on it,
Check the degree of coincidence with a plurality of character patterns for the inspection image, classify the inspection image into a class having a common character feature pattern,
The inspection image classified into the class is associated with the class, and is imaged in the inspection image using non-common character feature information for identifying a character by examining a non-common part between characters. Identify the characters
Has steps ,
The step of classifying into the class is to obtain a degree of coincidence with the plurality of character patterns for the inspection image based on a length of a contour that coincides with the character pattern . A group consisting of character patterns having a matching degree with the plurality of character patterns equal to or higher than a predetermined threshold with respect to the inspection image is the class having the common character feature pattern,
The classifying step includes
Classifying the inspection image into the class;
Selecting a character pattern having the highest matching score as a representative character pattern representing the class;
Have
The identifying step includes:
Using the non-common character feature information associated with the character pattern selected as the representative character pattern, identifying the character imaged in the inspection image,
The character recognition method according to claim 9 . The character recognition method according to claim 10 , wherein the non-common character feature information is a partial area of a character in a region located at a predetermined position with respect to the center of gravity of the representative character pattern.

Images