JP2014016304A - Image processing method and image processing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing method and an image processing system capable of recognizing the shape of the entire bead and enhancing the accuracy of image processing for recognizing the shape of the bead.SOLUTION: The image processing method for recognizing the shape of a bead of a welded part on the basis of an image obtained by photographing and acquiring the welded part with residual thermoluminescence occurring during beam system junction includes: setting a plurality of reference lines on an image; forming a contour point at a crossing place between a luminance boundary between a bright area in which the luminance of the residual thermoluminescence in the image is larger than a luminance threshold set so as to identify the contour line of the bead and a dark area in which the luminance of the residual thermoluminescence is smaller than the luminance threshold, and the reference lines; forming a backbone point at the middle of two contour points on the same reference line; determining an outline shape of the bead on the basis of the backbone point; and setting a corrected backbone point obtained by correcting the backbone point in accordance with an arc-shaped or linear shaped bead in a case where the outline shape of the bead is determined to be an arc shape or a linear shape. The image processing system uses the method.

Description

  The present invention relates to an image processing method for recognizing the shape of a bead formed in a welding area based on an image obtained by capturing and capturing a welding area in a state in which residual heat emission occurs during beam system joining, and The present invention relates to an image processing system using the method.

  Conventionally, as disclosed in Patent Document 1, a welded portion where light emission is generated during laser welding is photographed by a camera, an optical sensor, and the like, and an image of the photographed welded portion is captured and based on the captured image. Thus, image processing for recognizing the shape of the bead of the welded portion is performed. However, in the image processing of Patent Document 1, it is difficult to accurately recognize the bead shape of the welded part due to the influence of plasma light, sputtered particles, etc. generated during laser welding, which is a problem. ing.

  Therefore, as disclosed in Patent Document 2, in order to avoid the influence of plasma light, sputtered particles, and the like, thermoluminescence (hereinafter referred to as the heat emission) generated at the joining terminal portion of the bead within a short period of time in which heat generation of the metal remains immediately after laser welding. , “Residual heat emission”) is imaged by a camera, and image processing configured to capture a captured image of residual heat emission has been proposed. This image processing is configured to recognize the shape of the joint end portion of the bead based on the captured residual heat emission image. The image processing apparatus determines that a hole defect has occurred when a discontinuously separated bead shape is recognized in the captured image and when a concave shape is recognized at the weld end of the bead. It is configured as follows.

JP 2006-43741 A JP 2012-45610 A

  However, while the image processing of Patent Document 2 can accurately recognize the shape of the joining end portion of the bead in which residual heat emission occurs immediately after laser welding, this image processing is performed with the joining start end portion of the bead and the joining start end. Since the residual heat emission at the joint intermediate part of the bead located between the part and the joint end part is cooled and weakened immediately after laser welding, the shape of the bead joint start end part and the shape of the bead joint intermediate part are It has become difficult to recognize correctly. Therefore, it is desired to accurately recognize the shape of the entire laser welding bead, and further, it is desired to increase the accuracy in image processing for recognizing the shape of such a bead.

  The present invention has been made in view of such a situation, and an object of the present invention is to recognize the shape of the entire bead and to improve the accuracy of image processing for recognizing the shape of the bead. And providing an image processing system.

  In order to solve the problems, an image processing method according to an aspect of the present invention is formed on the welded portion based on an image obtained by capturing and capturing a welded portion in a state where residual heat emission occurs during beam system joining. An image processing method for recognizing the shape of a bead, wherein a plurality of reference lines are set on the image, and the luminance of residual heat emission in the image is set to identify the outline of the bead. A step of forming a contour point at a crossing point between a bright area larger than a luminance threshold and a luminance boundary between dark areas where the luminance of the residual heat emission is smaller than the luminance threshold and the reference line; and on the same reference line A step of forming a spine point in the middle between the two contour points; a step of determining a rough shape of the bead based on the spine point; and a case where the rough shape of the bead is determined to be an arc shape And a step of setting a corrected spine point obtained by correcting the spine point corresponding to the arc-shaped bead, and the step of setting the corrected spine point when the rough shape of the bead is an arc shape, Three or more representative spine points positioned so as to equally divide the length of the bead from the plurality of spine points are extracted, and the lines are arranged in the middle in the longitudinal direction of the line segment connecting the adjacent representative spine points. Calculating a bisector orthogonal to a minute, calculating an intersection of the plurality of bisectors, setting a plurality of correction reference lines extending radially from the intersection of the bisectors, and the luminance boundary Forming the corrected contour point at an intersection with the corrected reference line, and forming the corrected spine point in the middle between the two corrected contour points on the same corrected reference line.

  The image processing method according to an aspect of the present invention sets a corrected spine point obtained by correcting the spine point corresponding to the linear bead when it is determined that the approximate shape of the bead is a linear shape. A step of setting the correction spine point when the rough shape of the bead is a straight line, setting a plurality of correction reference lines orthogonal to a direction in which the bead extends, and the luminance boundary and the correction Forming a corrected contour point at an intersection with the reference line, and forming a corrected spine point in the middle between the two corrected contour points on the same corrected reference line.

  According to another aspect of the present invention, there is provided an image processing method in which a shape of a bead formed on a welded portion is formed based on an image obtained by capturing and capturing a welded portion in a state where residual heat emission occurs during beam system joining. An image processing method for recognizing a step of setting a plurality of reference lines on the image, and a brightness of residual heat emission in the image larger than a brightness threshold set to identify the contour line of the bead. Forming a contour point at the intersection of the reference line and a luminance boundary between a region and a dark region where the luminance of the residual heat emission is smaller than the luminance threshold, and the two contour points on the same reference line A step of forming a spine point in the middle, a step of determining a rough shape of the bead based on the spine point, and a straight shape when the rough shape of the bead is determined to be a straight shape Setting a corrected spine point in which the spine point is corrected corresponding to a bead, and the step of setting the corrected spine point when the general shape of the bead is a linear shape, in the extending direction of the bead A plurality of orthogonal correction reference lines are set, a correction contour point is formed at an intersection between the luminance boundary and the correction reference line, and correction is performed in the middle between two correction contour points on the same correction reference line. Including forming a spine point.

  In the image processing method according to one aspect or another aspect of the present invention, in the step of forming the contour point when the width of the bright region is equal to or smaller than a predetermined width threshold, the luminance boundary and the Contour points are formed at the intersections with the reference line.

  In order to solve the problem, an image processing system according to one aspect or another aspect of the present invention uses the above-described image processing method.

  According to the present invention, the following effects can be obtained. An image processing method according to an aspect of the present invention recognizes the shape of a bead formed on a welded part based on an image obtained by capturing and capturing a welded part in a state where residual heat emission occurs during beam system joining. An image processing method comprising: setting a plurality of reference lines on the image; and a bright area having a luminance of residual heat emission in the image larger than a luminance threshold set so as to identify a contour line of the bead, And a step of forming a contour point at a crossing point between a luminance boundary between dark regions where the luminance of the residual heat emission is smaller than the luminance threshold and the reference line, and between the two contour points on the same reference line Forming a spine point in the middle; determining a rough shape of the bead based on the spine point; and determining that the rough shape of the bead is an arc shape; Setting a corrected spine point in which the spine point has been corrected corresponding to the position of the plurality of spine points when the rough shape of the bead is an arc shape. Three or more representative spine points located so as to equally divide the length of the bead are extracted from the middle, and the second orthogonal to the line segment in the middle in the longitudinal direction of the line segment connecting the adjacent representative spine points Calculating a branch line, calculating an intersection between the plurality of bisectors, setting a plurality of correction reference lines extending radially from the intersection of the bisectors, and the luminance boundary and the correction reference line Forming the corrected contour point at an intersection and forming the corrected spine point in the middle between the two corrected contour points on the same correction reference line. Further, the image processing method according to an aspect of the present invention provides a corrected spine point obtained by correcting the spine point corresponding to the linear bead when it is determined that the approximate shape of the bead is a linear shape. A step of setting the corrected spine point when the rough shape of the bead is a straight line, and setting a plurality of corrected reference lines orthogonal to the extending direction of the bead, and the luminance boundary. Forming a corrected contour point at an intersection with the corrected reference line, and forming a corrected spine point in the middle between the two corrected contour points on the same corrected reference line. According to another aspect of the present invention, there is provided an image processing method in which a shape of a bead formed on a welded portion is formed based on an image obtained by capturing and capturing a welded portion in a state where residual heat emission occurs during beam system joining. An image processing method for recognizing a step of setting a plurality of reference lines on the image, and a brightness of residual heat emission in the image larger than a brightness threshold set to identify the contour line of the bead. Forming a contour point at the intersection of the reference line and a luminance boundary between a region and a dark region where the luminance of the residual heat emission is smaller than the luminance threshold, and the two contour points on the same reference line A step of forming a spine point in the middle, a step of determining a rough shape of the bead based on the spine point, and a straight shape when the rough shape of the bead is determined to be a straight shape Setting a corrected spine point in which the spine point is corrected corresponding to a bead, and the step of setting the corrected spine point when the general shape of the bead is a linear shape, in the extending direction of the bead A plurality of orthogonal correction reference lines are set, a correction contour point is formed at an intersection between the luminance boundary and the correction reference line, and correction is performed in the middle between two correction contour points on the same correction reference line. Including forming a spine point. Therefore, a spine line serving as a reference for bead shape recognition is formed based on the corrected spine point calculated in consideration of the approximate shape of the bead determined by the provisionally defined spine point. As a result, the spine line can be accurately formed based on the correct corrected spine point, and the bead shape can be accurately recognized based on the accurate spine line. Therefore, the accuracy of image processing for recognizing the bead shape can be increased.

  In the image processing method according to one aspect or another aspect of the present invention, in the step of forming the contour point when the width of the bright region is equal to or smaller than a predetermined width threshold, the luminance boundary and the Since the contour point is formed at the intersection with the reference line, it is possible to remove the influence of plasma light and sputtered particles having a bright region having a width larger than a predetermined width threshold. As a result, in addition to the bead end portion, the bead start portion and the intermediate portion of the bead that are affected by the plasma light and the sputtered particles can be accurately recognized. Therefore, the shape of the entire bead can be recognized accurately, and the accuracy of image processing for recognizing the shape of the bead can be increased.

  An image processing system according to one aspect or another aspect of the present invention uses the above-described image processing method. Therefore, as described above, the shape of the entire bead can be recognized, and the accuracy of image processing for recognizing the shape of the bead can be improved.

It is a schematic diagram which shows the bead formed in a welding area | region. 1 is a block diagram illustrating an image processing system according to a first embodiment of the present invention. It is a schematic diagram which shows the laser irradiation part and monitor part of the image capture device in 1st Embodiment of this invention. It is a figure which shows the relationship between the radiation intensity of a black body radiation, and a wavelength. It is a figure which shows the relationship between the relative sensitivity of the image pick-up element produced from a silicon-type semiconductor, and the wavelength input into an image pick-up element. It is a figure which shows an example of the relationship between the time passage just after welding in the 1st Embodiment of this invention, and the brightness | luminance of residual heat emission. It is a block diagram which shows the image processing part of the image processing apparatus in 1st Embodiment of this invention. In 1st Embodiment of this invention, it is a schematic diagram which shows an example of the image which displayed the bead area | region corresponding to the actual bead. In 1st Embodiment of this invention, it is a schematic diagram which shows an example of the image which displayed the contour point and spine point formed on the reference line corresponding to the arc-shaped bead. In 1st Embodiment of this invention, it is a schematic diagram which shows an example of the image which displayed the contour point and spine point formed on the reference line corresponding to the linear bead. In 1st Embodiment of this invention, when the rough shape of a bead is circular arc shape, it is a schematic diagram which shows an example of the image which displayed the representative spine point, the bisector, and the intersection. In 1st Embodiment of this invention, when schematic shape of a bead is circular arc shape, it is a schematic diagram which shows an example of the image which displayed the correction reference line, the correction outline point, the correction spine point, the outline, and the spine line. is there. In 1st Embodiment of this invention, when the schematic shape of a bead is a linear shape, it is a schematic diagram which shows an example of the image which displayed the correction reference line, the correction outline point, the correction spine point, the outline, and the spine line. is there. 5 is a flowchart illustrating an overview of an image processing method for a captured image in the first embodiment of the present invention. It is a flowchart explaining the image processing method of the correction element for circular arcs in the flowchart of FIG. It is a flowchart explaining the image processing method of the correction element for straight lines in the flowchart of FIG.

  An image processing system according to a first embodiment of the present invention will be described.

  Initially, with reference to FIG. 1, the bead m formed in the welding area J of the welding member M which is the object which performs image processing is demonstrated. The bead m formed in a substantially arc shape includes a joining start end portion (hereinafter referred to as “start end portion”) m1 formed at a laser welding start point and a joining end portion (hereinafter referred to as “laser welding end point”). , M2 and a joining intermediate portion (hereinafter referred to as “intermediate portion”) m3 formed between the start end portion m1 and the end portion m2. In the present embodiment, as an example, the bead m is formed in a substantially arc shape, but may be another shape such as a straight shape, an L shape, or a wave shape.

  Referring to FIG. 2, the image processing system 1 includes an image capturing device 2. The image capturing device 2 can irradiate the welding region J of the welding member M with a laser beam L <b> 1 for welding. The welding region J where the residual heat emission occurs is photographed, and an image of the photographed welding region J is captured. The image processing system 1 also has an image processing device 3 connected to the image capturing device 2. The image processing device 3 includes an image processing unit 4, and the image processing unit 4 recognizes the shape of the bead m based on the image of the welding region J captured by the image capturing device 2. The image processing apparatus 3 includes a bead shape determination unit 5, and the bead shape determination unit 5 determines a defect occurring in the bead m based on the shape of the bead m recognized by the image processing unit 4. It has become.

  Here, the details of the image capturing device 2 will be described with reference to FIGS. 2 and 3. Referring to FIG. 2, the image capturing device 2 includes a laser irradiation unit 6, and the laser irradiation unit 6 is configured to be able to irradiate the welding region J of the welding member M with the laser beam L <b> 1 for welding. Has been. The image capturing device 2 includes a monitor unit 7, and the monitor unit 7 is configured to continuously capture images of the welding region J irradiated with the laser. The image capturing device 2 includes a storage unit 8, and the storage unit 8 is configured to continuously store images taken by the monitor unit 7. The image capturing device 2 includes an image capturing unit 9, and the image capturing unit 9 is configured to capture an image stored in the storage unit 8.

  Referring to FIG. 3, the laser irradiation unit 6 includes a laser oscillator 6a, and the laser oscillator 6a is configured to be able to emit laser light L1. The laser irradiation unit 6 includes a substantially flat half mirror 6b, and the half mirror 6b is disposed on the optical path of the laser light L1 emitted from the laser oscillator 6a. The front surface 6b1 and the back surface 6b2 of the half mirror 6b are inclined at a predetermined angle θ1 with respect to the optical path of the laser light L1 emitted from the laser oscillator 6a. As an example, the predetermined angle θ1 is preferably 45 degrees. The laser irradiation unit 6 includes a welding condensing lens 6c, and the welding condensing lens 6c is disposed on the optical path of the laser light L1 that has passed through the half mirror 6b. Moreover, the condensing lens 6c for welding is comprised so that the laser beam L1 may be condensed on the welding member M. FIG. In such a laser irradiation unit 6, the laser beam L1 emitted from the laser oscillator 6a passes through the half mirror 6b and is collected by the welding condenser lens 6c, and then irradiated to the welding member M. It will be. Further, part of the laser light L1 irradiated to the welding member M is reflected, and this reflected light (hereinafter referred to as “monitoring light”) L2 passes through the welding condensing lens 6c and the half mirror 6b. After the angle is changed by reflecting on the back surface 6b2, the image is sent to the monitor unit 7.

  The monitor unit 7 includes a substantially flat mirror 7a. The mirror 7a is disposed on the optical path of the monitoring light L2 whose angle is changed by the half mirror 6c of the laser irradiation unit 6, and the reflection surface 7a1 of the mirror 7a. Is inclined at a predetermined angle θ2 with respect to the optical path of the monitoring light L2. The monitor unit 7 has a magnification adjustment lens 7b, and the magnification adjustment lens 7b is configured to adjust the photographing magnification. The magnification adjustment lens 7b is disposed on the optical path of the monitoring light L2 reflected by the mirror 7a. The monitor unit 7 includes a long-pass filter (hereinafter referred to as “Si long-pass filter”) 7c corresponding to an imaging element (hereinafter referred to as “Si-based imaging element”) 7f manufactured from a silicon-based semiconductor, which will be described later. The Si long pass filter 7c is configured to transmit light in a wavelength range of 800 nm or more. The Si long pass filter 7c is disposed on the optical path of the monitoring light L2 that has passed through the magnification adjusting lens 7b. The monitor unit 7 has an ND filter 7d, and the ND filter 7d is configured to reduce the amount of light passing therethrough. The ND filter 7d is disposed on the optical path of the monitoring light L2 that has passed through the Si long pass filter 7c. The monitor unit 7 includes an imaging condenser lens 7e, and the imaging condenser lens 7e is disposed on the optical path of the monitoring light L2 that has passed through the ND filter 7d. The imaging condenser lens 7e is configured to collect the monitoring light L2 on an imaging surface of an Si-based imaging element 7f described later. The monitor unit 7 includes a Si image sensor 7f, and the Si image sensor 7f is configured to convert the monitoring light L2 into an analog electric signal. The Si-based image sensor 7f has sensitivity to light in a wavelength range of 200 nm to 1100 nm. The monitor unit 7 has an amplifier 7g, and the amplifier 7g is configured to amplify an analog electric signal sent from the Si-based image sensor 7f. The monitor unit 7 includes an A / D converter 7h, and the A / D converter 7h is configured to convert the analog electric signal amplified by the amplifier 7g into a digital electric signal. The digital electrical signal converted by the A / D converter 7h is sent to the storage unit 8. The frame rate of the monitor unit 7 is 50 Hz or more and 1000 Hz or less.

  The conditions of the imaging surface of the Si-based imaging element 7f will be described. When the imaging surface of the Si imaging element 7f is significantly larger than the area where the entire bead m is arranged, the bead m displayed on the image is displayed with a small number of pixels. In this case, the shape of the bead m cannot be accurately recognized. Therefore, assuming that the imaging surface of the Si-based imaging element 7f is formed in a substantially square shape, the length of one side of the imaging surface is d, the focal length of the welding condenser lens 6c is f1, and the imaging surface When the focal length of the condenser lens 7e is f2, the magnification of the magnification adjusting lens 7b is s, and the maximum length of the region where the entire bead m is arranged is l, the length of one side of the imaging surface of the Si imaging element 7f. Is preferable to satisfy the condition of the following formula (1).

  The wavelength range of light passing through the Si long pass filter 7c will be described. Plasma light generated during laser welding contains many components of transition emission generated by orbital electrons of atoms, and the wavelength range of transition emission is typically in the range of 200 nm to 780 nm corresponding to the visible range. It has become. On the other hand, the relationship between the radiation intensity of black body radiation and the wavelength is as shown in the graph of FIG. 4 when the absolute temperature is calculated as 1200 K, 1000 K, and 800 K using the black body radiation equation. In the graph of FIG. 4, the calculation result when the absolute temperature is 1200K is indicated by a solid line A1, the calculation result when the absolute temperature is 1000K is indicated by a broken line A2, and the calculation result when the absolute temperature is 800K. This is indicated by a one-dot chain line A3. Referring to the graph of FIG. 4, the radiant intensity of blackbody radiation increases as the wavelength increases. Further, the radiation intensity of black body radiation increases as the temperature increases. Therefore, it can be confirmed that the radiation intensity of the black body radiation included in the residual heat emission having a high temperature increases in the long wavelength region.

  Further, the relationship between the relative sensitivity of the Si image sensor 7f and the wavelength of light input to the Si image sensor 7f is typically as shown in FIG. As shown in FIG. 5, the sensitivity of the Si-based image sensor 7f is lowered particularly in the wavelength range of 600 nm or more. Therefore, the Si-based image sensor 7f has a tendency to hardly detect low intensity light and to detect high intensity light in a wavelength range of 800 nm to 1100 nm.

  Therefore, in consideration of the upper limit 780 nm of the wavelength range of transition emission included in the plasma light and the variation in wavelength, the wavelength range of the light transmitted through the Si long pass filter 7c is set to 800 nm or more. Therefore, light is transmitted to the Si image sensor 7f in a wavelength range of 800 nm or more, and the Si image sensor 7f senses input light in a wavelength range of 800 nm to 1100 nm. In such a wavelength range, the Si-based imaging element 7f is difficult to detect plasma light including a transition light emission component having a low radiation intensity, but includes a large amount of a light emission component due to black body radiation having a high radiation intensity. Residual heat emission is easily detected. Note that the Si long-pass filter 7c is preferably configured to block light having a wavelength shorter than 800 nm. However, the Si imaging element 7f only needs to be able to sense light in the wavelength range of 800 nm to 1100 nm. For example, the Si long-pass filter 7c is configured to block light having a wavelength smaller than 800 nm, 900 nm, or 1000 nm. It may be.

  The range of the frame rate of the monitor unit 7 will be described. Since the generated plasma occupies a large area near the start of welding, the bead m that has already been welded cannot often be observed. Therefore, it is effective to store an image in the storage unit 8 after a certain time has elapsed since the start of welding. Further, in order to set the entire bead m as a target of image processing, it is only necessary to use a part of a plurality of images acquired in succession. Therefore, a part of the plurality of images may be acquired. Also, the images used for image processing need not be completely continuous. In the terminal portion m2 of the bead m, when a plurality of images are not captured at a high speed, the metal cools down. Therefore, it is necessary to capture images at a high sampling rate. On the other hand, when an image is captured at a high sampling rate during observation from the start of welding to the end of welding, a plurality of images with almost no change are acquired. Since it is useless to calculate similar characteristics in such a plurality of images, it is preferable to determine the frame rate of the monitor unit 7 in consideration of shortening the calculation processing time, saving the capacity of the storage unit 8, and the like. . Therefore, as the upper limit of the frame rate increases, the number of images to be captured increases, so that the accuracy of recognizing the shape of the bead m increases, while the number of images to be captured per second increases. When the frame rate is greater than 1000 frames (that is, when the frame rate is greater than 1000 Hz), the upper limit value of the frame rate is set to 1000 Hz in consideration of an increase in calculation load of image processing.

  On the other hand, when the lower limit value of the frame rate is low, the welding region J of the bead m to be photographed moves from the timing of photographing one image to the timing of photographing the next image, and the monitor unit There is a risk of deviating from the shooting range captured by 7. In this case, the bead m cannot be accurately photographed. Therefore, the lower limit value of the frame rate needs to satisfy the following conditions.

  When the frame rate is f, the moving speed of the laser beam L1 used for welding is Vm, and the shortest distance from the center of the shooting range to the boundary between the shooting range and the shooting range is Rm, the frame rate Preferably satisfies the condition of the following expression (2).

  Under typical conditions, Vm is 0.133 m / sec (8 m / min) and Rm is 0.002 m (2 mm). Therefore, it is necessary to make f larger than about 67 Hz from the equation (2). Furthermore, the lower limit value of the frame rate is set to 50 Hz in consideration of variations in the moving speed of the laser light L1.

  Referring to FIG. 6, the image capturing unit 9 is configured to capture an image of residual heat emission that occurs within a predetermined time immediately after the end of welding. Specifically, the image capturing unit 9 starts capturing an image when the average luminance G of the image is equal to or lower than a predetermined capture start (inspection start) luminance threshold g1, and captures a predetermined number of images. Is completed, or when the average luminance G of the image is equal to or less than a predetermined capture end luminance threshold value g2, it is recognized that the residual heat emission has ended, and the image capture is terminated.

  Here, an example of a preferable form of the image capturing device 2 will be described. The residual heat emission is attenuated after being generated for several tens of milliseconds (milliseconds) immediately after the end of welding. In this case, a high-speed camera with a frame rate of about 500 Hz (cycle 2 ms) is preferably used for the monitor unit 7 and about 10 images (= 20 ms ÷ 2 ms) are captured. Moreover, it is preferable that a high dynamic range camera is used for the monitor unit 7, and since the signal detection dynamic range is wide and the measurable luminance range is wide, it is possible to take images during and immediately after welding. . In the image capturing unit 9, it is preferable that the start luminance threshold value g <b> 1 for starting image capturing is an intermediate value of the average luminance G during and immediately after welding. The end luminance threshold value g2 for completing the image capture is preferably smaller than the start luminance threshold value g1 and larger than the average luminance G of the image that has become so dark that residual heat emission cannot be confirmed.

  Next, details of the image processing unit 4 of the image processing apparatus 3 will be described with reference to FIGS. Referring to FIG. 7, the image processing unit 4 includes a bead recognition unit 10, which recognizes a bead area candidate based on the image captured by the image capturing device 2. Yes. Referring to FIG. 7, the image processing unit 4 includes a bead discriminating unit 11, and when the bead discriminating unit 11 recognizes a plurality of bead area candidates, as shown in FIG. The bead area E corresponding to the actual bead is discriminated from the area candidates. The bead region E includes a start end portion E1 corresponding to the start end portion m1 of the bead m, an intermediate portion E2 corresponding to the end portion m2 of the bead m, and a terminal end portion E3 corresponding to the intermediate portion m3 of the bead m. Yes. Referring again to FIG. 7, the image processing unit 4 includes a reference line setting unit 12, and the reference line setting unit 12 includes a plurality of reference lines in the captured image as illustrated in FIGS. 9 and 10. L is set in a grid pattern. Referring to FIG. 7, the image processing unit 4 includes a contour point forming unit 13, and the contour point forming unit 13 contours the bead region E on the reference line L as shown in FIGS. 9 and 10. A point p is formed. Referring to FIG. 7, the image processing unit 4 has a spine point forming unit 14, and the spine point forming unit 14 is formed on the same reference line L as shown in FIGS. 9 and 10. A spine point q located in the middle between the two contour points p is formed.

  Referring to FIG. 7, the image processing unit 4 includes a rough shape determining unit 15, and the rough shape determining unit 15 determines a rough shape of the bead m based on the spine point q. Referring to FIG. 7, the image processing unit 4 has an arc correcting element 16, and the arc correcting element 16 is configured so that the approximate shape determining means 15 converts the approximate shape of the bead m into an arc shape as shown in FIG. 9. 11 and 12, as shown in FIGS. 11 and 12, a corrected reference line L ′, a corrected contour point p ′, and a corrected spine point q ′ obtained by correcting the reference line L, the contour point p, and the spine point q, respectively. Is calculated. Referring to FIG. 2, the image processing unit 4 includes a straight line correction element 17. As shown in FIG. 10, the straight line correction element 17 has a rough shape determining means 15 that changes the rough shape of the bead m to a straight line shape. 13, the corrected reference line L ′, the corrected contour point p ′, and the corrected spine point q ′ obtained by correcting the reference line L, the contour point p, and the spine point q, respectively, are calculated as shown in FIG. It is like that. Referring to FIG. 7, the image processing unit 4 includes a contour line forming unit 18, which forms a contour line P based on the contour point p, or FIGS. 12 and 13. As shown in FIG. 8, a contour line P is formed based on a plurality of corrected contour points p ′ calculated by the arc correcting element 16 and the straight correcting element 17. Referring to FIG. 7, the image processing unit 4 includes a spine line forming unit 19, which forms the spine line Q based on the spine point q, or FIGS. 12 and 13. As shown in FIG. 4, a spine line Q is formed based on a plurality of corrected spine points q ′ calculated by the arc correcting element 16 and the straight correcting element 17.

  Details of the bead recognition means 10 will be described. The bead area candidate recognized by the bead recognizing means 10 is an area where the residual heat emission luminance H is higher than the predetermined contour luminance threshold value h, and the contour luminance threshold value h can identify the contour line of the bead m. Is set to The bead region E is composed of a plurality of bead region points e. As an example, the bead area point e is represented by one or more pixels on the image, and the plurality of bead area points e may be arranged adjacent to each other or arranged at intervals of a pixel unit. .

  With reference to FIG. 7, the detail of the bead discrimination | determination means 11 is demonstrated. The bead discriminating means 11 recognizes that such a bead area candidate corresponds to the actual bead m when the width B of the bead area candidate is equal to or smaller than a predetermined width threshold value b. As an example, since the width of the bead m formed in the welding region J of the welding member M is unlikely to be larger than twice the predetermined value due to variation, the width threshold value b is equal to the predetermined width of the bead m. The width threshold b is preferably 1.5 times the predetermined width of the bead m in order to improve the accuracy. Further, the bead discriminating unit 11 calculates the aspect ratio (aspect ratio) of each bead area candidate, and selects a bead area E corresponding to the actual bead m from the plurality of bead area candidates based on the aspect ratio. It is preferable that For example, when the aspect ratio is larger than a value corresponding to a horizontally long shape of a general bead m, it may be set to recognize that such a bead area candidate corresponds to an actual bead m. When the bead area candidates corresponding to the actual beads m and the sputtered particles scattered during welding are recognized, the bead area candidates corresponding to the sputtered particles are formed in an isotropic shape, The bead m is formed in a horizontally long shape. Therefore, the bead area E corresponding to the actual bead m is accurately selected. Further, in consideration of the case where the bead area candidates corresponding to the sputtered particles are not formed in an isotropic shape, the bead discrimination means 11 moves in an image in which a plurality of bead area candidates are captured before and after time. It is preferable to determine whether the bead area is stopped or stopped and to recognize that the stopped bead area candidate is the bead area E corresponding to the actual bead m.

  Details of the reference line setting means 12 will be described with reference to FIGS. The plurality of reference lines L set by the reference line setting means 12 are arranged at intervals of pixel units. The interval between the plurality of reference lines L is preferably larger than the interval between the plurality of bead region points e.

  Details of the contour point forming means 13 will be described with reference to FIGS. The contour point p is formed at the intersection of the boundary corresponding to the contour of the bead m and the reference line L based on the image captured by the image capturing device 2. The boundary corresponding to the contour of the bead m has a bright region in which the residual heat emission luminance H is larger than the contour luminance threshold h set so as to identify the contour line of the bead m, and the residual heat emission luminance H is the contour luminance. It is located at a luminance boundary with a dark area smaller than the threshold value h. That is, the contour luminance threshold value h is smaller than the residual heat emission luminance H corresponding to the inner region of the bead m contour and corresponds to the outer region of the bead m contour so as to correspond to the bead m contour. It is set to a value larger than the luminance H of light emission. As an example, the contour luminance threshold value h may be set by obtaining in advance the relationship between the contour of the bead m and the luminance H of the residual heat emission.

  Details of the approximate shape determining means 15 will be described with reference to FIGS. Based on the positions of the plurality of spine points q, the approximate shape determining means 15 determines whether the shape of the bead m is an arc shape, a linear shape, or a shape other than the arc shape and the linear shape. .

  Details of the arc correcting element 16 will be described with reference to FIGS. 7, 11, and 12. Referring to FIG. 7, the arc correcting element 16 has a representative spine point extracting means 20, and the representative spine point extracting unit 20, as shown in FIG. 11, extracts the length of the bead region E from a plurality of spine points q. Three or more representative spine points q0 located so as to be equally divided are extracted. Referring to FIG. 7, the arc correcting element 16 has a bisector calculation means 21, and the bisector calculation means 21 is a line connecting adjacent representative spine points q0 as shown in FIG. A bisector D perpendicular to the line segment is calculated at the middle in the longitudinal direction of the minute. Referring to FIG. 7, the arc correcting element 16 has an intersection calculation unit 22, and the intersection calculation unit 22 calculates an intersection c between a plurality of bisectors D as shown in FIG. 11. It has become. Referring to FIG. 7, the arc correction element 16 has an arc correction reference line setting means 23, and the arc correction reference point setting means 23 is the intersection of the bisectors D as shown in FIG. A plurality of correction reference lines L ′ extending radially around c are set. Referring to FIG. 7, the arc correcting element 16 has an arc correcting contour point forming means 24, and the arc correcting contour point forming means 24 corresponds to the contour of the bead m as shown in FIG. A corrected contour point p ′ is formed at the intersection of the luminance boundary and the corrected reference line L ′. Referring to FIG. 7, the arc correcting element 16 has an arc correcting spine point forming means 26. As shown in FIG. 12, the arc correcting spine point forming means 25 has the same correction reference line L ′. A corrected spine point q ′ is formed in the middle between the above two corrected contour points p ′.

  With reference to FIG.7 and FIG.13, the detail of the correction element 17 for straight lines is demonstrated. Referring to FIG. 7, the straight line correction element 17 has a straight line correction reference line setting unit 26, and the straight line correction reference line forming unit 26 extends in the direction in which the bead region E extends as shown in FIG. 13. A plurality of orthogonal correction reference lines L ′ are set. Referring to FIG. 7, the straight line correcting means 16 has a straight line corrected contour point forming means 27, and the straight line corrected contour point forming means 27 corresponds to the contour of the bead m as shown in FIG. A corrected contour point p ′ is formed at the intersection of the luminance boundary and the corrected reference line L ′. Referring to FIG. 7, the straight line correcting means 16 has straight straight spine point forming means 28, and the straight straight spine point forming means 28 has the same correction reference line L ′ as shown in FIG. A corrected spine point q ′ is formed between the above two corrected contour points p ′.

With reference to FIG.11 and FIG.13, the detail of the correction outline formation means 18 and the correction spine line formation means 19 is demonstrated. The corrected contour line forming means 18 forms a contour line P based on a plurality of corrected contour points p ′ using the fitting function fn1. The modified spine line forming means 19 forms a spine line Q based on a plurality of corrected spine points q ′ using the fitting function fn2. As an example, at least one of the fitting function fn1 and the fitting function fn2 determines the coordinates of each modified contour point p ′ or each modified spine point q ′ as (X _J , Y _J ) (J = 1, 2, 3,... , N, N + 1) may be defined by a polygonal line based on the following formula (3).

A _XJ , B _XJ , A _YJ , B _YJ , and t _J in the formula (3) are defined by the following formulas (4) to (8), respectively.

  Details of the bead shape determination unit 5 will be described. The bead shape determination unit 5 is generated in the bead m based on the relationship between the contour line P and the spine line Q recognized by the image processing unit 4 and the information on the luminance of the residual heat emission along the spine line Q. Defects are judged.

  Here, a method for recognizing the shape of the bead m formed in the welding region J using the image processing system 1 of the present embodiment will be described. First, an image capturing method in the image capturing device 2 will be described. In the laser irradiation unit 6 of the image capturing device 2, a laser oscillator 6a emits a laser beam L1. The laser beam L1 emitted from the laser oscillator 6a passes through the half mirror 6b and is collected by the welding condensing mirror 6c, and then irradiated to the welding region J of the welding member M. Thereafter, residual heat emission is emitted from the portion where the bead m is formed by the laser light L1 irradiated to the welding region J of the welding member M, and becomes the monitoring light L2. The monitoring light L2 passes through the welding condenser lens 6c and is sent to the monitor unit 7 after the angle is changed by the half mirror 6b.

  In the monitor unit 7 of the image capturing device 2, the monitoring light L2 sent from the laser irradiation unit 6 is sent to the magnification adjusting lens 7b after the angle is changed by the mirror 7a. When the monitoring light L2 passes through the magnification adjusting lens 6b, the magnification is adjusted. The monitoring light L2 whose magnification is adjusted by the magnification adjusting lens 7b is sent to the Si long pass filter 7c. In the Si long pass filter 7c, a wavelength component smaller than 800 nm of the monitoring light L2 is blocked, and a wavelength component larger than 800 nm of the monitoring light L2 passes through the Si long pass filter 7c. The monitoring light L2 that has passed through the Si long pass filter 7c is sent to the ND filter 7d. In the ND filter 7d, the amount of the monitoring light L2 decreases so that the shape of the bead m can be clearly recognized by the image processing unit 4 of the image processing apparatus 3. The monitoring light L2 that has passed through the ND filter 7d is sent to the imaging condenser lens 7e. The monitoring light L2 is collected in the imaging condenser lens 7e. The monitoring light L2 collected by the photographing condenser lens 7e is irradiated to the Si-based image sensor 7f. In the Si-based imaging device 7f, the component in the wavelength range of 800 nm to 1100 nm in the monitoring light L2 is sensed, the sensed monitoring light L2 is converted into an analog electric signal, and this analog electric signal is sent to the amplifier 7g. In the amplifier 7g, the analog electric signal is amplified, and the amplified analog electric signal is sent to the A / D converter 7h. In the A / D converter 7h, the analog electric signal is converted into a digital electric signal, and this digital electric signal is sent to the storage unit 8. In the storage unit 8, the digital electric signal is stored as one image. In this way, the Si-based image sensor 7f is irradiated with the monitoring light L2, the Si-based image sensor 7f converts the monitoring light L2 into an analog electric signal, and the A / D converter 7h converts the analog electric signal into a digital electric signal. The operation of storing the digital electric signal as one image is continuously performed based on a frame rate larger than 50 Hz and smaller than 1000 Hz. Further, the image stored in the storage unit 8 is captured by the image capturing unit 9.

  Next, an image processing method in the image processing apparatus 3 will be described with reference to the flowchart of FIG. Based on the image captured by the image capturing unit 9 of the image capturing device 2, a bead region candidate having a residual heat emission luminance H greater than a predetermined contour luminance threshold value h is recognized, and a plurality of bead region candidates are recognized. If one is found, one of a plurality of bead area candidates is read (S1). It is determined whether or not the read width B of the bead area candidate is equal to or smaller than a predetermined width threshold value b (S2).

  When it is determined that the width B of the bead area candidate is equal to or smaller than the predetermined width threshold value b (YES), a plurality of reference lines L are set in a lattice shape (S3), and a luminance boundary corresponding to the contour of the bead m A contour point p is formed at the intersection with the reference line L (S4), and a spine point q located in the middle between two contour points p formed on the same reference line L is formed (S5). It is determined whether or not the approximate shape of the bead m is an arc shape based on the spine point q (S6).

  If it is determined that the approximate shape of the bead m is an arc shape (YES), the reference line L, the contour point p, and the spine point q are corrected by the arc correction element 16, respectively. A point p ′ and a corrected spine point q ′ are calculated (S7). Further, a contour line P is formed based on the plurality of corrected contour points p '(S8), and a spine line Q is formed based on the plurality of corrected spine points q' (S9). Thereafter, it is determined whether or not all bead area candidates have been confirmed (S10). When all the bead area candidates are confirmed (YES), the image processing is terminated. If all of the bead area candidates have not been confirmed (NO), another one of the plurality of bead area candidates is read (S11), and the width B of the read bead area candidate is equal to or less than a predetermined width threshold value b. It is determined whether or not there is (S2).

  When it is determined that the approximate shape of the bead m is not an arc shape (NO), it is determined whether or not the approximate shape of the bead m is a linear shape (S12). When it is determined that the approximate shape of the bead m is a straight line shape (YES), the reference line L, the contour point p, and the spine point q are corrected by the straight line correction element 17, respectively. A point p ′ and a corrected spine point q ′ are calculated (S13). Further, a contour line P is formed based on the plurality of corrected contour points p '(S8), and a spine line Q is formed based on the plurality of corrected spine points q' (S9). Thereafter, it is determined whether or not all bead area candidates have been confirmed (S10). When all the bead area candidates are confirmed (YES), the image processing is terminated. If all of the bead area candidates have not been confirmed (NO), another one of the plurality of bead area candidates is read (S11), and the width B of the read bead area candidate is equal to or less than a predetermined width threshold value b. It is determined whether or not there is (S2).

  When it is determined that the approximate shape of the bead m is not a straight line shape (NO), the contour line P is formed based on the plurality of contour points p (S14), and the spine line Q is formed based on the plurality of spine points q. (S15). Thereafter, it is determined whether or not all bead area candidates have been confirmed (S10). When all the bead area candidates are confirmed (YES), the image processing is terminated. If all of the bead area candidates have not been confirmed (NO), another one of the plurality of bead area candidates is read (S11), and the width B of the read bead area candidate is equal to or less than a predetermined width threshold value b. It is determined whether or not there is (S2).

  On the other hand, if it is determined that the width B of the bead region candidate is larger than the predetermined width threshold value b (NO), that is, if it is determined that the bead region candidate is plasma light or sputtered particles, the bead region candidate It is determined whether all of the above have been confirmed (S10). When all the bead area candidates are confirmed (YES), the image processing is terminated. If all of the bead area candidates have not been confirmed (NO), another one of the plurality of bead area candidates is read (S11), and the width B of the read bead area candidate is equal to or less than a predetermined width threshold value b. It is determined whether or not there is (S2).

  With reference to the flowchart of FIG. 15, details of a method for calculating the correction reference line L ′, the correction contour point p ′, and the correction spine point q ′ by the arc correction element 16 will be described. Three or more representative spine points q0 located so as to equally divide the length of the bead region E from a plurality of spine points q are extracted (S21). A bisector D perpendicular to the line segment in the middle of the line segment connecting the adjacent representative spine points q0 is calculated (S22). An intersection c between the plurality of bisectors D is calculated (S23). A plurality of correction reference lines L 'extending radially about the intersection c of the bisector D are set (S24). A corrected contour point p 'is formed at the intersection of the luminance boundary corresponding to the contour of the bead m and the correction reference line L' (S25). A corrected spine point q 'is formed in the middle between two corrected contour points p' on the same corrected reference line L '(S26).

  The details of a method for calculating the correction reference line L ′, the correction contour point p ′, and the correction spine point q ′ by the straight line correction element 17 will be described with reference to the flowchart of FIG. 16. A plurality of correction reference lines L 'perpendicular to the extending direction of the bead area E are set (S31). A corrected contour point p 'is formed at the intersection of the luminance boundary corresponding to the contour of the bead m and the corrected reference line L' (S32). A corrected spine point q 'is formed between two corrected contour points p' on the same corrected reference line L '(S33).

  Next, a bead shape determination method in the image processing apparatus 3 will be described. In the bead shape determination unit 5 of the image processing device 3, the position and shape of the contour line P and the spine line Q, the relationship between the contour line P and the spine line Q, and the information on the luminance of the residual heat emission along the spine line Q Based on the above, a defect occurring in the bead m is determined.

  As described above, according to the present embodiment, based on the corrected spine point q ′ calculated in consideration of the approximate shape of the bead m recognized based on the provisionally defined spine point q, A spine line Q serving as a reference for shape recognition is formed. As a result, the spine line Q can be accurately formed based on the correct corrected spine point q ', and the shape of the bead m can be accurately recognized based on the accurate spine line Q. Therefore, the accuracy of image processing for recognizing the shape of the bead m can be improved.

  According to this embodiment, when the width B of the bright region of the bead region candidate is equal to or smaller than the predetermined width threshold value b, the contour point p is formed at the intersection of the luminance boundary and the reference line L, and then Since the process proceeds to image processing for recognizing the shape of the bead m, it is possible to remove the influence of plasma light and sputtered particles having a bright region having a width B larger than a predetermined width threshold value b. As a result, in addition to the bonding end portion m2 of the bead m, the bonding start end portion m1 and the bonding intermediate portion m3 of the bead m affected by the plasma light and the sputtered particles can be accurately recognized. Therefore, the shape of the entire bead m can be recognized accurately, and the accuracy of image processing for recognizing the shape of the bead m can be increased.

[Second Embodiment]
An image processing system according to the second embodiment of the present invention will be described below. The second embodiment is basically the same as the first embodiment. Elements similar to those in the first embodiment will be described using the same symbols and names as those in the first embodiment. Here, a configuration different from the first embodiment will be described.

  In this embodiment, instead of the Si long-pass filter 6c in the first embodiment, an InGaAs long-pass corresponding to an image sensor (hereinafter referred to as “InGaAs-based image sensor”) manufactured from an indium gallium arsenide-based semiconductor described later. A filter is provided. The long pass filter for InGaAs is configured to transmit light in a wavelength range of 1200 nm or more. In the present embodiment, the image pickup device 6f has an InGaAs image pickup device instead of the Si image pickup device 6f1 of the first embodiment. The InGaAs imaging device has sensitivity to light in a wavelength range of 800 nm to 2000 nm. The InGaAs-based image sensor is configured to convert the monitoring light L2 sent to the imager 6f into an analog electric signal. Further, the conditions of the imaging surface of the InGaAs imaging device are the same as the imaging surface conditions of the Si imaging device 6f1 in the first embodiment.

  The wavelength range of light transmitted through the long pass filter for InGaAs will be described. As described above, plasma light generated during laser welding contains many components of transition emission generated by orbital electrons of atoms, and the wavelength range of this transition emission typically corresponds to the visible range. It is in the range of 200 nm to 780 nm. On the other hand, the radiation intensity of blackbody radiation increases as the wavelength increases. Further, the radiation intensity of black body radiation increases as the temperature increases. Therefore, it can be confirmed that the radiation intensity of the black body radiation included in the residual heat emission having a high temperature is increased. In addition, since light emitted from a laser or the like includes a wavelength component around 1060 nm, it is preferable that the light input to the InGaAs-based image pickup device avoid a wavelength range around 1060 nm.

  Therefore, in consideration of the upper limit 780 nm of the wavelength range of transition emission included in plasma light, avoiding light having a wavelength near 1060 nm corresponding to the emission wavelength of laser, etc., and variations in emission wavelength of laser, etc. The wavelength range of light passing through the InGaAs long pass filter is set to 1200 nm or more. Therefore, light is sent to the InGaAs-based image sensor in a wavelength range of 1200 nm or more, and the InGaAs-based image sensor senses input light in the wavelength range of 1200 nm to 2000 nm. The InGaAs long-pass filter is preferably configured to block light having a wavelength shorter than 1200 nm. However, since the InGaAs-based imaging device only needs to be able to sense light in the wavelength range of 800 nm to 2000 nm, for example, a long pass filter for InGaAs emits light having a wavelength smaller than 1300 nm, 1400 nm, 1500 nm, 1600 nm, 1700 nm, 1800 nm, or 1900 nm. You may be comprised so that it may interrupt | block.

  The method for recognizing the shape of the bead m formed in the welding region using the image processing system 1 of the present embodiment is the same as that of the first embodiment except for the configuration described above of the present embodiment. .

  As described above, according to the present embodiment, the same effects as those of the first embodiment can be obtained.

[Third Embodiment]
An image processing system according to the third embodiment of the present invention will be described below. The third embodiment is basically the same as the first embodiment. Elements similar to those in the first embodiment will be described using the same symbols and names as those in the first embodiment. Here, a configuration different from the first embodiment will be described.

  In the present embodiment, a Si band-pass filter is provided instead of the Si long-pass filter 6c in the first embodiment. The band pass filter for Si is configured to transmit light in a wavelength range of 800 nm or more and 1050 nm or less.

  The wavelength range of light passing through the Si bandpass filter will be described. The lower limit value of the wavelength range of the light transmitted through the Si bandpass filter is based on the same reason that the lower limit value of the wavelength range of the light transmitted through the Si longpass filter 6c in the first embodiment is set to 800 nm. It is set to 800 nm or more. In addition, in consideration of avoiding light having a wavelength near 1060 nm, which corresponds to the emission wavelength of a laser, etc., and variations in the emission wavelength of the laser, etc., the upper limit value of the wavelength range of the light passing through the Si bandpass filter is It is set to 1050 nm.

  A method for recognizing the shape of the bead m formed in the welding region using the image processing system 1 of the present embodiment is the same as that of the first embodiment except for the configuration described above of the present embodiment.

  As described above, according to the present embodiment, the same effects as those of the first embodiment can be obtained.

[Fourth Embodiment]
An image processing system according to the fourth embodiment of the present invention will be described below. The fourth embodiment is basically the same as the second embodiment. The same elements as those of the second embodiment will be described using the same symbols and names as those of the second embodiment. Here, a configuration different from the second embodiment will be described.

  In this embodiment, an InGaAs band-pass filter is provided instead of the InGaAs long-pass filter in the second embodiment. The bandpass filter for InGaAs is configured to transmit light in a wavelength range of 1200 nm or more and 2000 nm or less.

  The wavelength range of light that passes through the bandpass filter for InGaAs will be described. The lower limit value of the wavelength range of light passing through the InGaAs bandpass filter is 1200 nm based on the same reason that the lower limit value of the wavelength range of light passing through the InGaAs longpass filter in the second embodiment is set to 1200 nm. It is stipulated in. In addition, the upper limit of sensitivity to light in the long wavelength high sensitivity type InGaAs imaging device is 2600 nm, and the upper limit of sensitivity to light in the short wavelength high sensitivity type InGaAs imaging device is 1700 nm. Further, the short wavelength high sensitivity type InGaAs imaging device has a constant sensitivity (response performance) even in the range of 1700 nm to 2000 nm because its band gap gradually decreases. Therefore, the upper limit of the wavelength range of the light transmitted through the InGaAs band-pass filter is set to 2000 nm so as to correspond to all InGaAs-based image pickup devices such as the long wavelength high sensitivity type and the short wavelength high sensitivity type.

  As described above, according to the present embodiment, the same effects as those of the first embodiment can be obtained.

  Although the embodiments of the present invention have been described so far, the present invention is not limited to the above-described embodiments, and various modifications and changes can be made based on the technical idea of the present invention.

  For example, as a modified example of the present invention, a welding region J in a state where residual heat emission occurs during beam system joining other than laser welding is formed in the welding region J based on an image captured and captured by the monitor unit 7. The shape of the bead m may be recognized. In this case, the beam system bonding may be laser brazing, laser cladding, plasma welding, electron beam welding, or the like.

DESCRIPTION OF SYMBOLS 1 Image processing system 3 Image processing apparatus 4 Image processing part 10 Bead recognition means 11 Bead discrimination means 12 Reference line setting means 13 Contour point formation means 14 Spine point formation means 15 Outline shape determination means 16 Arc correction element 17 Straight line correction element 18 contour line forming means 19 spine line forming means 20 representative spine point extracting means 21 bisector calculating means 22 intersection calculating means 23 arc correction reference line setting means 24 arc correcting contour point forming means 25 arc correcting spine point forming Means 26 Straight line correction reference line setting means 27 Straight line correction contour point forming means 28 Straight line correction spine point forming means M Welding member J Welding area m Bead m1 Joint start end portion (start end portion)
m2 Junction termination part (termination part)
m3 Joining middle part (middle part)
L1 Laser light L2 Monitoring light θ1, θ2 Angle G Average luminance g1 Start luminance threshold g2 End luminance threshold E Bead region e Bead region point L Reference line L ′ Corrected reference line P Contour line p Contour point p ′ Corrected contour point Q Spine line q spine point q0 representative spine point q 'corrected spine point D bisector c intersection point A1 solid line A2 broken line A3 alternate long and short dash lines S1-S13, S21-S26, S31-S33 Steps

Claims (5)

An image processing method for recognizing the shape of a bead formed in the welded part based on an image obtained by capturing and capturing a welded part in a state where residual heat emission occurs during beam system joining,
Setting a plurality of reference lines on the image;
A luminance boundary between a bright area in which the luminance of the residual heat emission in the image is larger than a luminance threshold set so as to identify a contour line of the bead, and a dark boundary between the dark areas in which the luminance of the residual heat emission is smaller than the luminance threshold Forming a contour point at the intersection with the reference line;
Forming a spine point midway between the two contour points on the same reference line;
Determining a rough shape of the bead based on the spine point;
When it is determined that the rough shape of the bead is an arc shape, a corrected spine point is corrected by correcting the spine point corresponding to the arc-shaped bead, and
The step of setting the corrected spine point when the rough shape of the bead is an arc shape,
Extracting three or more representative spine points positioned so as to evenly divide the length of the bead from the plurality of spine points;
Calculate a bisector perpendicular to the line segment in the middle in the longitudinal direction of the line segment connecting the representative spine points adjacent to each other;
Calculating the intersection of the plurality of bisectors,
A plurality of correction reference lines extending radially from the intersection of the bisectors,
Forming the corrected contour point at the intersection of the luminance boundary and the corrected reference line, and forming the corrected spine point in the middle between the two corrected contour points on the same corrected reference line. Image processing method. When the rough shape of the bead is determined to be a linear shape, the method further includes a step of setting a corrected spine point obtained by correcting the spine point corresponding to the linear shape bead,
The step of setting the corrected spine point when the approximate shape of the bead is a linear shape,
Setting a plurality of correction reference lines orthogonal to the extending direction of the bead;
Forming a corrected contour point at an intersection between the luminance boundary and the corrected reference line, and forming a corrected spine point in the middle between the two corrected contour points on the same corrected reference line. 2. The image processing method according to 1. An image processing method for recognizing the shape of a bead formed in the welded part based on an image obtained by capturing and capturing a welded part in a state where residual heat emission occurs during beam system joining,
Setting a plurality of reference lines on the image;
A brightness boundary between a bright area in which the luminance of the residual heat emission in the image is larger than a luminance threshold set so as to identify a contour line of the bead, and a luminance boundary between dark areas in which the luminance of the residual heat emission is smaller than the luminance threshold; Forming a contour point at the intersection with the reference line;
Forming a spine point midway between the two contour points on the same reference line;
Determining a rough shape of the bead based on the spine point;
When a rough shape of the bead is determined to be a straight shape, a corrected spine point is corrected by correcting the spine point corresponding to the linear shape bead, and
The step of setting the corrected spine point when the approximate shape of the bead is a linear shape,
Setting a plurality of correction reference lines orthogonal to the extending direction of the bead;
Forming a corrected contour point at an intersection between the luminance boundary and the corrected reference line, and forming a corrected spine point in the middle between the two corrected contour points on the same corrected reference line. Method.   The contour point is formed at the intersection of the luminance boundary and the reference line in the step of forming the contour point when the width of the bright region is equal to or smaller than a predetermined width threshold. The image processing method according to any one of the above.   An image processing system using the image processing method according to claim 1.

Images