A visual sensing weld tracking method for narrow gap welding

Abstract

The invention discloses a narrow gap welding visual sensing weld tracking method and system, the system includes: a narrow gap welding torch, a narrow gap welding control cabinet, an infrared camera, an image real-time transmission module, a computer image processing module and a data communication module; the method is that the infrared camera is used to trigger the image of the groove before welding or the global image of the welding area through the image real-time transmission module; after computer image processing, the left and right edges of the groove and the edge of the welding wire are extracted, and the groove center and the welding torch center are calculated through the pattern recognition anti-interference algorithm based on the coefficient of variation and the numerical filtering algorithm, and the difference between the two is used as the weld deviation; through the weld tracking controller and the data communication module, the obtained weld deviation adjustment amount is sent to the narrow gap welding control cabinet to correct the welding torch. The method has good environmental adaptability and high detection accuracy, and is suitable for the initial welding torch position guidance and weld tracking control of narrow gap welding.

Description

Technical Field

[0001] The invention belongs to the technical field of welding, and in particular relates to a narrow gap welding visual sensing weld tracking method and system. Background Art

[0002] In the process of narrow gap welding, due to the influence of groove processing error, assembly error, variable gap, welding thermal deformation, etc., the welding torch is easily caused to deviate from the groove center, resulting in weld deviation, making it difficult for the arc to effectively heat the groove side wall, affecting the fusion of the narrow gap welding groove side wall and the quality of joint formation. Therefore, weld tracking control is required. In order to achieve weld tracking control, sensing technology is the key. Among them, the passive visual sensing method directly uses ambient light or arc light as the background light source, without the need for additional auxiliary light source, with low cost, suitable for a variety of groove forms, and can realize sensing detection synchronized with the arc position. However, interference such as welding spatter, smoke and arc light will seriously affect the quality of welding images collected by the visual sensor, bringing challenges to the visual sensing detection of weld deviation and the precise detection of weld tracking control, thereby restricting the engineering application of narrow gap welding process.

[0003] In the process of realizing weld tracking control by using visual sensors, the detection with the position information of the weld groove edge as the sensing feature is the most common. The invention patent with Chinese patent number ZL201610517913.X and the name of "Welding groove edge position visual sensing detection method" discloses a welding groove edge position visual sensing detection method. It uses a visual sensor to collect the global image of the welding area, intercepts the groove edge image on the opposite side of the arc through the groove edge ROI window, extracts the groove edge line after image processing, and detects the straightest section on the groove edge line through the local pattern recognition small window, and uses the position detection value of the groove edge on the straightest section as the detection value of the weld groove edge position. Its disadvantages are: 1) This method uses standard deviation to compare and evaluate the discreteness and dispersion of each group of data. When the measurement scales of each group of data differ greatly, the standard deviation evaluation method is not applicable; 2) This method does not involve the detection of the center position of the welding torch and the weld tracking control.

[0004] The invention patent with Chinese patent number ZL201410741503.4 and the name of "Adaptive control method and device for narrow gap welding arc shaking" discloses an adaptive control method and device for narrow gap welding arc shaking. The method uses an infrared camera system to extract groove width information in real time, and realizes adaptive control of arc shaking angle according to the change of groove width; its disadvantages are: 1) the method does not use pattern recognition method to identify welding interference information, but obtains the left and right edge positions and groove width detection values ​​of the groove through median or mean filtering method; when the welding spatter is large, the detection accuracy of groove width and groove center is easily affected; 2) the method does not involve the initial welding torch position guidance function of welding.

[0005] The invention patent with Chinese patent number ZL201210325926.9 and titled "Narrow gap welding monitoring and weld deviation detection method based on infrared visual sensing" discloses a narrow gap welding monitoring and weld deviation detection method based on infrared visual sensing. The method uses an infrared CMOS camera to obtain welding images, and obtains the weld deviation value by calculating the distance from the left (or right) edge of the molten pool away from the arc side to the left (or right) boundary of the left (or right) interception window. Its disadvantages are: 1) This method does not involve the image real-time transmission module and the data communication module; 2) This method does not consider the influence of interference such as welding spatter on the groove center and weld deviation detection accuracy.

[0006] In the above patents, there is no effective solution to the problem of seam tracking control in oscillating arc narrow gap welding. However, when the center of the groove is inconsistent with the center of the welding torch during welding, it will affect the weld formation and the quality of the weld joint, so real-time seam tracking control is required, and the above methods are difficult to apply at this time. Summary of the invention

[0007] The purpose of the present invention is to provide a visual sensing weld tracking method and system for narrow gap welding with high detection accuracy and strong anti-interference ability, in view of the fact that the existing technology has not solved the problem of real-time weld tracking control of shaking arc narrow gap welding.

[0008] The present invention uses an infrared camera to obtain a pre-welding groove image or a global image of a welding area in real time, and according to the center position of the pre-welding groove image or the arc highest point position of the global image of the welding area, the groove edge ROI (Region of Interest) image and the welding wire ROI image are adaptively intercepted, and the left and right edges of the welding groove and the contour of the welding wire are respectively extracted through different image processing methods, and the original position point distribution of the groove center and the original position point distribution of the welding torch center are obtained, and through a pattern recognition anti-interference algorithm based on a coefficient of variation and a numerical filtering algorithm, the groove center position and the welding torch center position are respectively obtained, and the difference between the groove center position and the welding torch center position is calculated to obtain the weld deviation value; by developing a weld tracking controller, the controller parameters are automatically matched according to the size of the weld deviation value to obtain the weld deviation adjustment amount; through a data communication module, the weld deviation adjustment amount is sent to a narrow gap welding control cabinet, and the narrow gap welding torch is corrected, so as to realize the pre-welding position guidance of the welding torch and the real-time control of the weld tracking of the narrow gap welding.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] A narrow gap welding visual sensing weld tracking system, comprising a narrow gap welding torch, characterized in that: it also comprises a narrow gap welding control cabinet 14, an infrared camera 8, a filter system 7, a current sensor 13, a signal trigger 9, an image real-time transmission module 10, a computer image processing module 11 and a data communication module 12;

[0011] The infrared camera 8 is aimed at the welding groove 6 and is arranged at an angle θ with the horizontal plane. It is connected to the output end of the signal trigger 9 and the input end of the image real-time transmission module 10 through a video cable respectively; the input end of the signal trigger 9 is connected to the narrow gap welding control cabinet 14 and the current sensor 13 through a wire respectively, and the narrow gap welding control cabinet 14 is connected to the arc motion driver 3b of the narrow gap welding torch; one end of the welding power supply 3e in the narrow gap welding torch is connected to the arc motion driver 3b, and the other end of the connecting cable is connected to the workpiece where the welding groove 6 is located after passing through the detection ring of the current sensor 13 or connecting the current sensor 13 in series; the image real-time transmission module 10 includes an image acquisition card and image acquisition software, and is sequentially connected to a computer image processing module 11 including an edge extraction module, a weld deviation value obtaining module, and a weld deviation adjustment amount obtaining module, and a data communication module 12 including a communication protocol, communication software and a communication interface, and a narrow gap welding control cabinet 14 through a video cable. The filter system 7 is coaxially arranged at the front end of the infrared camera 8, and is composed of a filter, a dimming filter and a UV filter.

[0012] Furthermore, the infrared camera is a CMOS infrared camera.

[0013] Furthermore, the angle θ=10-80°.

[0014] In order to achieve the above object, the present invention adopts another technical solution as follows:

[0015] A tracking method for a narrow gap welding visual sensing weld tracking system comprises the following steps:

[0016] 1) Real-time image acquisition: The arc current base value signal i detected by the current sensor 13 is b and arc position signal P arc , a single channel or a dual channel is sent to a signal trigger 9, the infrared camera 8 is controlled to start collecting a pre-weld groove image with a width and height of h1 (h1≤1024) and h2 (h2≤1024) pixels respectively or a global image of the welding area synchronized with the arc position 15, and an image acquisition program is developed through C or C++ language and image processing software to form an image real-time transmission module 10 with real-time image acquisition, storage and output functions, so as to realize real-time transmission of images and provide real-time images for pre-weld torch position guidance and weld tracking control of narrow gap welding;

[0017] 2) Extract the groove edge and obtain the groove center position x g : including obtaining the groove center position x of the groove image before welding or the global image of the welding area 15 g ; The groove center position x of the groove image before welding g The acquisition is based on a single frame of the pre-weld groove image collected in real time, and adaptively locates the left and right edge ROI windows of the groove with a height of h5 (5≤h5≤h2) pixels by the midpoint method or the midline method, and adaptively intercepts the left and right edge ROI images of the groove, and extracts the left edge 4 and the right edge 5 of the groove before welding by the pre-weld groove edge ROI image processing method in the computer image processing module 11, and obtains the original position point distribution of the groove center by calculating the mean value or the fitting value along the height direction of the groove edge ROI window, and calculates the groove center position x in the pre-weld groove image by the mean method. g ;

[0018] The groove center position x of the global image 15 of the welding area g The method for obtaining the arc 2 is to extract the position of the arc 2 in the global image 15 of the welding area through the global image processing method in the computer image processing module 11, and to adaptively set the groove edge ROI window 16 on the opposite side of the arc 2 through the position of the arc 2, and to intercept the groove edge ROI image with a height of h3 (5≤h3≤h2) pixels, and to extract the groove left edge 4 or the groove right edge 5 on the opposite side of the arc 2 through the groove edge ROI image processing method in the computer image processing module 11; according to the groove left edge 4 and the groove right edge 5 extracted from the two adjacent frames of the global image 15 of the welding area, the original position point distribution of the groove center is obtained by calculating the mean value or the fitting value along the height direction of the groove edge ROI window; the discreteness of the distribution of the original position point of the groove center is analyzed through the pattern recognition anti-interference algorithm based on the coefficient of variation, and after eliminating the abnormal original position point of the groove center, the groove center position x of the global image 15 of the welding area is obtained from the retained effective position points of the groove center through the method of calculating the mean value or the fitting value. g , providing a basis for correcting the position of the narrow gap welding torch;

[0019] 3) Extract the welding wire contour and obtain the welding torch center position x e : Including obtaining the welding torch center position x of the groove image before welding and the global image of the welding area 15 e ; The center position of the welding torch before welding groove image x eThe image before welding groove is obtained, including the image before welding groove when the welding wire is stopped and the image before welding groove when the welding wire is shaken; for the image before welding groove when the welding wire is stopped or the welding wire is shaken, the image before welding wire ROI window is intercepted with a height of h6 (5≤h6≤h2) pixels, and the contour of the welding wire is extracted by the image processing method before welding wire ROI, and the center position x of the welding torch of the image before welding groove is obtained. e , initial torch position guidance for narrow gap welding;

[0020] The center position x of the welding torch in the global image 15 of the welding area e The acquisition is to intercept a welding wire ROI image with a height of h4 (5≤h4≤h2) pixels through the welding wire ROI window 17 on the same side of the arc 2, and extract the welding wire edge and the welding wire center through the welding wire ROI image processing method; according to the welding wire left contour (including the welding wire left edge and the welding wire right edge), the welding wire left center or the welding wire right contour (including the welding wire left edge and the welding wire right edge), and the welding wire right center extracted from the two adjacent frames of the welding area global image 15, the welding torch center position sampling value is obtained through the welding wire single-side edge detection method or the welding wire double-side edge detection method; the obtained welding torch center position sampling value is obtained through a numerical filtering algorithm to obtain the welding torch center position x of the welding area global image 15. e ;

[0021] 4) Obtain the weld deviation value △x i : Based on the groove center position and the welding torch center position obtained respectively from the groove image before welding or the global image of the welding area 15, the weld deviation detection module in the computer image processing module 11 is used to detect the weld deviation according to the formula △x i =(x g -x e ) respectively calculate the deviation value of the welding torch relative to the groove center before welding or during welding, obtain the weld deviation of the groove image before welding or the global image 15 of the welding area, and provide a basis for the welding torch position guidance and weld tracking control before welding;

[0022] 5) Obtain the weld deviation adjustment amount: for the obtained weld deviation value △x i , through proportional, integral, differential controllers, or separable combinations, intelligent controllers, and composite controllers integrated with proportional, integral, and differential controllers, develop narrow gap weld tracking controllers to form weld tracking controllers that automatically match control parameters according to weld deviations, thereby obtaining weld deviation adjustment amounts;

[0023] 6) Correction operation: for the obtained weld deviation adjustment amount, data communication software and communication interface are developed through computer language to form a data communication module 12 with data storage and output functions; the weld deviation adjustment amount is sent to the narrow gap welding control cabinet 14 through the data communication module 12 to guide the welding torch before welding and correct the deviation during welding;

[0024] 7) Repeat the above steps 1) to 6) until the welding process is completed.

[0025] Further, in step 2), the midpoint method or the midline method includes: by searching the midpoint coordinates (h1 / 2, h2 / 2) of the pre-weld groove image, or finding the midline segment position x=h1 / 2 or y=h2 / 2 of the pre-weld groove image, based on the initial groove width value, the midpoint or midline segment position of the pre-weld groove image, after translating up and down and left and right for a certain distance, automatically finding the groove left and right edge ROI window positions of the pre-weld groove image, and adaptively intercepting the pre-weld groove edge ROI image;

[0026] Further, in step 2), the pre-weld groove edge ROI image processing method includes: median filtering, morphological operation, adaptive threshold and edge extraction algorithm;

[0027] Further, in step 2), the steps of the global image processing method are: performing median filtering, histogram analysis, global threshold and morphological operations on the global image 15 of the welding area, extracting the morphology of the arc 2, obtaining the position of the highest point of the arc by grayscale recognition method, and numerically filtering the ordinate values ​​of the obtained continuous multiple highest point positions of the arc, so as to reduce the influence of welding interference on the highest point position of the arc;

[0028] Further, in step 2), the steps of the groove edge ROI image processing method are: median filtering for denoising, contrast stretching to improve image contrast, four-partition threshold processing of the ROI image, morphological operation to eliminate isolated points, and then Canny edge operator to extract the groove edge.

[0029] Further, in step 2), the steps of the pattern recognition anti-interference algorithm based on the coefficient of variation are: partitioning the obtained h3 groove center original position points into n equal parts, calculating the coefficient of variation of the groove center original position points in each equal partition through the coefficient of variation (Coefficient of Variation) formula (CV = standard deviation / mean), retaining the groove center original position point in the equal part corresponding to the minimum coefficient of variation as the groove center true position point; globally matching the groove center true position point with the h3 groove center original position points, retaining the groove center original position point with the same numerical value as the groove center true position point in situ as the groove center effective position point; obtaining the groove center position sampling value by linear fitting or calculating the mean of the groove center effective position points; filtering the groove center position sampling values ​​for the obtained multiple consecutive groove center position sampling values ​​by a numerical filtering algorithm to obtain the groove center position value.

[0030] Further, in step 3), the steps of the welding wire ROI image processing method before welding are: performing median filtering, local adaptive threshold processing, morphological operation, and Canny edge operator on the intercepted welding wire ROI image before welding to extract the welding wire contour; the steps of the welding wire ROI image processing method are: performing median filtering denoising, local adaptive threshold processing, morphological operation, and Canny edge operator on the intercepted welding wire ROI image to extract the welding wire contour.

[0031] Further, in step 3), the single-side edge detection method of the welding wire is to use the mean or fitting value of the single-side edge position point distribution of the welding wire as the sampling value of the center position of the welding torch; including the inner edge detection method of the welding wire and the outer edge detection method of the welding wire; the inner edge detection method of the welding wire refers to obtaining the sampling value of the center position of the welding torch by calculating the mean or fitting value according to the distribution of the edge points of the welding wire that are closer to the groove edge on the same side where the welding wire stops; the outer edge detection method of the welding wire refers to obtaining the sampling value of the center position of the welding torch by calculating the mean or fitting value according to the distribution of the edge points of the welding wire that are farther from the groove edge on the same side where the welding wire stops; the double-side edge detection method of the welding wire is to first obtain the distribution of the left or right center line position points of the single-side welding wire according to the extracted left and right edges of the welding wire, and then obtain the sampling value of the center position of the welding torch of the current frame by calculating the mean or fitting value of the left and right center line position points of the welding wire in two adjacent frames of welding images; the numerical filtering algorithm includes: median filtering, mean filtering, anti-pulse mean or median filtering, and developed numerical filtering algorithm.

[0032] Further, in step 4), the expression △x i =(x g -x e ), also expressed as △x i =(x e -xg ).

[0033] Compared with the prior art, the advantages and beneficial effects of the present invention are mainly:

[0034] 1) By developing the real-time image transmission module and data communication module, the real-time acquisition, storage and output of the pre-weld groove image and the global image of the welding area are realized, the real-time communication of the weld deviation adjustment amount is completed, and a complete narrow gap weld tracking control system is established;

[0035] 2) Through the pattern recognition anti-interference algorithm based on the coefficient of variation, after analyzing the coefficient of variation of the original position point of the groove center, the simulated welding spatter with a diameter of less than 1.7 mm can be removed, thereby improving the detection accuracy of the groove center position. The algorithm is easy to implement and has good environmental adaptability;

[0036] 3) By developing a weld tracking controller, the parameters of the weld tracking controller can be automatically matched according to the size of the weld deviation value, thereby improving the stability of the narrow gap weld tracking control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a schematic diagram of the structure of the narrow gap welding visual sensing weld tracking system;

[0038] Figure 2 To use Figure 1 The system shown is a schematic diagram of a global image of a welding area where the arc is located on the right side wall of the groove (taking an I-type groove as an example);

[0039] Figure 3 This is the global image processing flow chart of the welding area;

[0040] Figure 4 It is the flow chart of the anti-interference algorithm of pattern recognition based on coefficient of variation;

[0041] Figure 5 Get examples of the wire edge and wire centerline for a global image of a narrow gap welding area; Figure 5 (a) and (b) are the global images of the welding area including the left and right ROI windows of the welding wire, respectively. Figure 5 (c) and (d) are the left and right ROI images of the intercepted welding wire, respectively. Figure 5 (e) and (f) are the grayscale histograms of the left and right ROI images of the welding wire, respectively. Figure 5 (g) and (h) are the left and right edges of the welding wire and the left and right center lines of the welding wire extracted by the welding wire ROI image processing method;

[0042] Figure 6 This is an example of global image processing in a narrow gap welding area when the welding torch is tilted to the left; Figure 6(a) and (b) are the global images of the welding area including the left and right edge ROI windows of the groove and the right and left ROI windows of the welding wire, respectively. Figure 6 (c) and (d) are the groove edge ROI images containing the extracted left and right edge lines of the groove, respectively. Figure 6 (e) and (f) are the welding wire ROI images containing the extracted right and left contours of the welding wire and the right and left center lines of the welding wire, respectively;

[0043] Figure 7 Examples of groove edge and centerline distribution, weld deviation detection based on coefficient of variation, wire center and torch centerline distribution; Figure 7 (a) is the left and right edges and center line of the groove transformed from the horizontal coordinate to the global image of the welding area. Figure 7 (b) is the distribution of the original position points of the groove center. Figure 7 (c) is the effective position point of the groove center obtained by the pattern recognition anti-interference algorithm based on the coefficient of variation. Figure 7 (d) The left and right center lines of the welding wire and the center line of the welding torch are transformed from the horizontal coordinate to the global image of the welding area;

[0044] Figure 8 An example of guiding the initial position of the welding torch before welding; Figure 8 (a) is the image of the groove before welding, the ROI image and its edge extraction result when the welding wire is stopped. Figure 8 (a1) is the image of the groove before welding when the welding wire is stopped, which contains two groove edge ROI windows with a height of h5 pixels and one welding wire ROI window with a height of h6 pixels. Figure 8 (a2) is the ROI image of the left edge of the groove before welding and its edge extraction result. Figure 8 (a3) is the ROI image of the welding wire before welding and the extraction results of the left and right edges of the welding wire. Figure 8 (a4) is the ROI image of the right edge of the groove before welding and its edge extraction result; Figure 8 (b) is the image of the groove before welding and the extraction result of the welding wire edge when the welding wire is shaking. Figure 8 (b1) is the image of the groove before welding when the welding wire including the welding wire ROI window is shaken to the left and right sides of the groove respectively. Figure 8 (b2) is the intercepted left ROI image of the welding wire before welding. Figure 8 (b3) is the intercepted right ROI image of the welding wire. Figure 8 (b4) is the extracted left contour of the welding wire before welding (including the left and right edges of the welding wire), Figure 8 (b5) is the extracted right profile of the welding wire before welding (including the left and right edges of the welding wire);

[0045] Description of reference numerals: Figure 1In the figure, 1a—left wall of welding groove; 1b—right wall of welding groove; 2—arc; 3a—wire feeder; 3b—arc motion driver; 3c—conductive rod mechanism; 3d—welding wire; 3e—welding power source; 4—left edge of groove; 5—right edge of groove; 6—welding groove; 7—light filtering system; 8—infrared camera; 9—signal trigger; 10—real-time image transmission module; 11—computer image processing module; 12—data communication module; 13—current sensor; 14—narrow gap welding control cabinet; V w — welding speed; θ—shooting angle of infrared camera 8; P arc —Arc position signal; i b —Arc base current signal;

[0046] Figure 2 In the figure, 15 is the global image of the welding area, which shows the situation when the arc is located at the right wall of the groove; 16 is the ROI window of the groove edge; 17 is the ROI window of the welding wire; 18 is the tail edge of the welding pool; 19 is the front edge of the welding pool. i —The origin of the global image of the welding area in the current frame, which is also the rectangular coordinate system x i -O i -y i h1 represents the width of the global image 15 of the welding area, h2 represents the height of the global image 15 of the welding area, h3 represents the height of the groove edge ROI window 16, and h4 represents the height of the welding wire ROI window 17. i Indicates the arc highest point position of the current frame image, obtained through global welding image processing. 1j Indicates the x coordinate value of the left edge of the groove at the j position in the vertical direction of the groove edge ROI window, variable j = 1, 2, ..., h3; x 2j It represents the x-coordinate value of the right center point of the welding wire at position j in the vertical direction of the welding wire ROI window, and the variable j = 1, 2, ..., h4. DETAILED DESCRIPTION

[0047] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited to the following embodiments, and all technical solutions obtained by equivalent replacement or equivalent transformation are within the protection scope of the present invention.

[0048] like Figure 1As shown, it is a schematic diagram of the structure of the narrow gap welding visual sensing weld tracking system of the present invention, which mainly includes: a narrow gap welding torch, a visual sensing system, and a narrow gap welding control cabinet 14. Among them, the narrow gap welding torch includes an arc motion driver 3b and a conductive rod mechanism 3c, wherein the arc motion driver 3b includes a driving mechanism and a welding torch controller, and the conductive rod mechanism 3c is a bent conductive rod mechanism composed of a bent conductive rod and a straight conductive nozzle connected thereto, or an eccentric conductive rod mechanism composed of a straight conductive rod and an eccentric conductive nozzle connected thereto; the welding wire 3d passes through the wire feeder 3a, passes through the arc motion driver 3b, and then extends obliquely from the conductive rod mechanism 3c, generates a welding arc 2 in the welding groove 6, and performs periodic arc-shaped shaking or unidirectional rotation around the center line of the welding torch. One end of the welding power supply 3e is connected to the arc motion driver 3b, and the other end is connected to the workpiece where the welding groove 6 is located. When the arc swings or rotates to the nearest position of the left wall or the right wall of the groove, the narrow gap welding control cabinet 14 sends an arc position signal P arc During pulse welding, when the arc swings to the nearest position of the left wall or the right wall of the groove, the current sensor 13 detects the arc base current signal i b , and the arc position signal P arc After the AND operation, it is sent to the signal trigger 9. The developed real-time image transmission module 10 controls the infrared camera 8 to collect the welding image in real time, and completes the output and storage of the image at the same time.

[0049] The infrared camera 8 is coaxially mounted with the filter system 7 and is placed in front of the welding groove 6 at a depression angle θ (10-80°) and at a welding speed V w Moving forward, the infrared camera 8 has a focal length range of 18 to 45 mm and an aperture range of f / 5.6 to 36; the filter system 7 includes a narrow-band filter with a central wavelength range of 800 to 1100 nm, a neutral density filter and a UV filter. Through the infrared camera 8, the filter system 7, the signal trigger 9 and the real-time image transmission module 10, the global image 15 of the welding area is collected in real time and sent to the computer image processing module 11 to calculate the weld deviation value; and the weld deviation adjustment amount is obtained through the weld tracking controller, and the obtained weld deviation adjustment amount is sent to the narrow gap welding control cabinet 14 through the data communication module 12 to correct the narrow gap welding torch.

[0050] The overall scheme of the narrow gap welding visual sensing weld tracking method of the present invention is: Figure 1 The narrow gap welding visual sensing weld tracking system obtains a global image 15 of the welding area, such as Figure 2 As shown; for the global image 15 of the welding area obtained, by Figure 3 The global image processing method shown in the figure includes median filtering, histogram analysis, global threshold and morphological operation to obtain the highest point position A of arc 2.i , so as to determine the positions of the wire ROI window 17 and the groove edge ROI window 16;

[0051] Through the wire ROI window 17, the wire ROI image is intercepted. Through Figure 3 The wire ROI window processing method shown, including median filtering, local adaptive thresholding, morphological operations, and Canny edge operator, extracts the right center (or left center) position point of the wire; along the height direction of the wire ROI window, the mean value or fitted value of the left center and the right center position points of the wire is calculated to form the original position point distribution of the torch center;

[0052] Through the groove edge ROI window 16, the groove edge ROI image on the side opposite to the arc 2 is intercepted. Through Figure 3 The groove edge ROI image processing method shown, including median filtering, contrast stretching, segmented thresholding, morphological operations, and Canny edge operator, extracts the left (or right) edge of the groove; along the height direction of the groove edge ROI window, by calculating the mean value of the left edge and the right edge position points of the groove, the original position point distribution of the groove center is formed; through the original position point distribution of the torch center and the original position point distribution of the groove center, the torch center position and the groove center position are calculated, thereby calculating the weld deviation value.

[0053] As Figure 4 shown, it is the flow chart of the anti-interference algorithm for pattern recognition based on the coefficient of variation of the present invention. The specific steps include: according to the extracted left and right edges of the groove, obtain the original position point distribution of the groove center located in the data window x - O g -y The data window x - O g -y has a height of h3 pixels. Divide the h3 original position points g g (j = 1, 2,..., h3) of the groove center contained in the data window x - O j -y into n (n < h3) equal parts on average, and calculate the coefficient of variation CV n (n = 1, 2,..., n) of the (h3 / n) original position points of the groove center contained in each part, and compare the sizes of each CV n , and retain the original position points of the groove center in the part corresponding to the minimum coefficient of variation as the true position points of the groove center;

[0054] The actual position point of the groove center is globally matched with h3 original position points of the groove center, and the original position points of the groove center with the same value as the actual position points of the groove center are retained in situ as the effective position points of the groove center; the sampling value of the groove center position is obtained by linear fitting or calculating the mean of the effective position points of the groove center; and the sampling value of the groove center position is filtered by a numerical filtering algorithm to obtain the groove center position value, so as to realize the accurate detection of the groove center position under welding interference conditions.

[0055] Four specific embodiments of the narrow gap welding visual sensing weld tracking method of the present invention.

[0056] Example 1

[0057] Figure 5 The figure shows an example of obtaining the edge and center line of the welding wire in the global image of the narrow gap welding area. The test conditions include: using a CMOS infrared camera as the infrared camera 8, and the signal trigger 9 receiving the arc position signal P arc and arc base current signal i b After that, the infrared camera 8 is controlled by an external trigger to start collecting the global image 15 of the welding area, and the real-time collected global image 15 of the narrow gap welding area is sent to the computer image processing module 11 through the real-time image transmission module 10 to extract the welding groove edge and the welding wire edge for subsequent weld deviation detection, thereby performing weld tracking control. The aperture of the infrared camera 8 is F16, the exposure time is 0.3ms, and the shooting angle θ is less than 30°; the filter system 7 includes a UV mirror, a neutral dimming filter with a transmittance of 30%, and a narrow band filter with a central wavelength of 970nm and a bandwidth of ±20nm. The base material of the welding specimen is Q370qE steel plate, and the width of the welding groove 6 is 13.82mm; the average current of the pulse arc 2 is 302.5A, the average arc voltage is 28.7V, and the welding speed V w The arc oscillation frequency of arc 2 is 2.5 Hz, the oscillation angle is 82°, and the dwell time on both sides of the groove is 100 ms.

[0058] Figure 5 As shown in FIG. 1 , the global image of the narrow gap welding area when the arc is located on the left and right walls of the groove respectively according to the two consecutive frames collected is shown in FIG. Figure 5 As shown in (a) and (b), the arc passes through the highest point position A i The vertical coordinate of the welding wire ROI window is used to obtain the vertical position of the welding wire ROI window. Together with the horizontal coordinate of the welding wire position detected before welding, the welding wire ROI window position with a height of h4 and a width of 80 pixels is adaptively determined, such as Figure 5 As shown in the white square boxes in (a) and (b). The left and right ROI images of the welding wire are captured through the welding wire ROI window, as shown in Figure 5 (c) and (d). The grayscale distribution of the left and right ROI images of the welding wire is analyzed by grayscale histogram, as shown in Figure 5 As shown in (e) and (f), it is determined that the local adaptive threshold method is used to perform threshold segmentation on the welding wire, and the morphological operation of the 3×3 rectangular morphological kernel and the Canny operator are integrated to extract the left and right edges of the welding wire, and the left center line and the right center line of the welding wire are calculated, as shown in FIG. Figure 5 Indicated by the center line in (g) and (h). Figure 5 In (g), because the arc is located on the left side wall at this time, the left edge of the welding wire is called the inner edge of the welding wire, and the right edge of the welding wire is the outer edge of the welding wire; Figure 5 (h), at this time, the arc is located on the right side wall, so the left edge of the welding wire is called the outer edge of the welding wire, and the right edge of the welding wire is called the inner edge of the welding wire; according to the left and right edges of the welding wire, the inner edge of the welding wire, the outer edge of the welding wire and the left and right center lines of the welding wire in two adjacent frames of welding images, the current center position of the welding torch can be calculated by the single-side edge detection method and the double-side edge detection method.

[0059] Example 2

[0060] Figure 6 The figure shows an example of global image processing of a narrow gap welding area when the welding torch is tilted to the left. The test conditions include: the base material of the welding specimen is Q370qE steel plate, the width of the welding groove 6 is 13.8mm; the average current of the pulse arc 2 is 322.5A, the average arc voltage is 299.3V, and the welding speed is V w The arc 2 has a circular arc shaking frequency of 2.5 Hz, a shaking angle of 82°, and a dwell time of 100 ms on both sides of the groove. The aperture of the infrared camera 8 is F16, the exposure time is 0.3 ms, and the shooting angle θ is less than 30°. The parameters of the filter system 7 are the same as those of the infrared camera 8. Figure 5 same.

[0061] Figure 6 (a) and (b) are two adjacent frames of the global image of the welding area 15, respectively. The arc is located at the right wall and the left wall of the groove, respectively, through the groove edge ROI window and the welding wire ROI window with a height and width of 80 pixels, as shown in FIG. Figure 6 As shown in the four white boxes in (a) and (b), the ROI images of the left and right edges of the groove and the right and left ROI images of the welding wire are captured.

[0062] For the intercepted left and right edge ROI images of the groove, the groove edge ROI image processing method is used, including median filtering, contrast stretching, image grayscale histogram analysis, four-equal partition threshold method, and fusion of 3×3 rectangular morphological kernel morphological closing operation and Canny edge detection operator to obtain the left edge and right edge of the groove, respectively. Figure 6 Indicated by white lines in (c) and (d). Figure 6 In (c), due to the presence of welding spatter, the extracted left edge of the groove convexes to the right, which makes it difficult to detect the center of the groove. Therefore, it is necessary to develop a new algorithm to remove the influence of interference such as welding spatter on the accuracy of groove center detection, so as to improve the accuracy of weld tracking control.

[0063] For the intercepted right and left ROI images of the welding wire, the right edge of the welding wire, the left edge of the welding wire, the right center line of the welding wire, and the left center line of the welding wire are extracted through the welding wire ROI image processing method, including median filtering denoising, local adaptive threshold processing, morphological operation, and Canny edge operator, respectively. Figure 6 As shown by the white lines in (e) and (f), the steps of obtaining the left and right center lines of the welding wire include: calculating the mean of the left and right edges of the left (or right) contour of the welding wire and rounding it off to an integer to form the left (or right) center line of the welding wire; calculating the mean of the left and right center lines of the welding wire to obtain the original position point distribution of the center of the welding torch.

[0064] Example 3

[0065] Figure 7 Shown is an example of groove edge and centerline distribution, weld deviation detection based on coefficient of variation, wire center and torch centerline distribution. Figure 6 Taking the two adjacent frames of the global image of the welding area captured when the welding torch is tilted to the left as shown in (a) and (b), the process of obtaining the weld deviation value by the pattern recognition anti-interference algorithm based on the coefficient of variation is described. For the left and right ROI images of the groove edge with a width and height of 80 pixels, the groove edge ROI image processing method is used to extract Figure 6 (c) and (d) the left and right edges of the groove; along the groove edge ROI window height h3 direction, calculate the mean of the corresponding position points of the left and right edges of the groove and take the integer value (when performing coefficient of variation analysis, the data is not rounded), and obtain the distribution line of the original position point of the groove center. The horizontal coordinates of the left edge of the groove, the right edge of the groove and the center line of the groove are converted to the global image of the welding area, respectively as follows Figure 7 (a) Indicated by the white line in the center.

[0066] According to the distribution of the obtained groove center original position points, the weld deviation detection steps based on the coefficient of variation include: first, the groove center position is calculated according to the obtained groove center original position points. The h3 = 80 groove center original position points are evenly divided into n = 4 equal parts, and the h3 in each equal part is calculated from bottom to top by using expressions (1) and (2). m (h m =80÷4=20) The coefficients of variation of the original position points of the groove center are CV1=0.000867, CV2=0.002432, CV3=0.001202, CV4=0.002263, respectively. Figure 7 (b) as shown.

[0067]

[0068]

[0069] In the formula, s, and CV correspond to h in each equal partition m The standard deviation, mean and coefficient of variation of the data; h m is the number of data in the equally divided interval, expressed by h m =(h3 / n), where h3 is the height of the ROI window at the edge of the groove, and n is the number of equal partitions; g jd is the original position point of any groove center in each equal partition, j d (jd≤h m ) is g jd The position number in the equal partition, jd = 1, 2, 3, ..., h m .

[0070] By comparing the size of the coefficient of variation, we can see that the coefficient of variation CV1 of the data in the first equal partition is the smallest, indicating that the dispersion of the original position points of the groove center contained in it is the smallest. Therefore, the data in the first equal partition is used as the true position point of the groove center. Then, the position point with the same value as the true position point of the groove center in the original position point of the groove center is searched globally and retained in situ to form the effective position point of the groove center, such as Figure 7 (c) As shown in the figure, according to the retained effective position points of the groove center, the groove center position x is obtained by calculating the mean value. g The center position x of the groove is obtained by directly calculating the mean of the original position point of the groove center instead of using the pattern recognition anti-interference algorithm based on the coefficient of variation to remove the data formed by splashing. g The value is 291.96 pixels. Due to the presence of welding spatter, the obtained groove center position is significantly smaller than the groove center position value obtained by the proposed algorithm. This shows that the proposed algorithm reduces the influence of welding spatter on the detection accuracy of the groove center position.

[0071] Secondly, according to the obtained original position point of the welding torch center, the welding torch center position is calculated by the welding wire double-side edge detection method. Figure 6 The right center position point and the left center position point of the welding wire obtained in (e) and (f) are calculated along the height direction of the ROI window. The mean of the left and right center position points of the welding wire at the corresponding height position is calculated and the integer value is taken to form the original position point distribution of the welding torch center. The horizontal coordinates of the left and right centers of the welding wire and the center of the welding torch are converted to the global image of the welding area, as shown in Figure 7 (d) is shown by the white line. By calculating the mean of the original position points of the welding torch center, the welding torch center position x is obtained. e is 274.13 pixels.

[0072] Finally, according to the obtained groove center position x g and torch center position x e , calculate the weld deviation value. The specific steps are: through the expression △x i =(x g -x e ) Calculate the difference between the center position of the groove and the center position of the welding torch to obtain the weld deviation value. The weld deviation value obtained by the pattern recognition anti-interference algorithm based on the coefficient of variation is +24.62 pixels, while the weld deviation value obtained by the conventional mean method is +17.83 pixels. It shows that the pattern recognition anti-interference algorithm based on the coefficient of variation reduces the influence of welding spatter interference on weld deviation detection when calculating the groove center, indicating that the pattern recognition anti-interference algorithm based on the coefficient of variation has good adaptability to the welding environment and high detection accuracy.

[0073] In the above embodiment, due to the existence of the weld deviation value, the welding torch is in a left-biased state relative to the center of the groove, and the arc is biased toward the left wall of the groove; at this time, the welding torch needs to be adjusted to the right by a corresponding distance to ensure that the two sides of the groove form a uniform side wall penetration. Therefore, it is necessary to correct the welding torch to achieve weld tracking control. In order to achieve weld tracking, according to the obtained weld deviation value, the weld deviation adjustment amount is calculated through the designed weld tracking controller, and the weld deviation adjustment amount is sent to the weld tracking control cabinet 14 through the data communication module 12, and the motor is controlled to drive the welding torch to perform the correction operation, so as to achieve real-time tracking control of the welding process, thereby improving the welding quality.

[0074] Example 4

[0075] Figure 8The figure shows an example of guiding the initial position of the welding torch before welding. Through the developed real-time image transmission module, the pre-weld groove image when the welding wire stays at the groove center and shakes to the groove side wall is collected in real time. By calculating the welding torch center position and groove center position of the pre-weld groove image, the initial welding torch position guidance before welding of narrow gap welding is achieved.

[0076] For the collected pre-weld groove image when the welding wire stays at the center of the groove, the pre-weld groove image processing method is used to extract the contour of the welding wire, and the mean of the corresponding position points on the left and right edges of the welding wire is calculated along the height direction of the welding wire ROI window, so as to obtain the center position value of the welding torch. Accordingly, in the pre-weld groove image, by setting the groove edge ROI window, the pre-weld groove edge ROI image processing method is used to calculate the mean of the position points on the left and right edges of the groove at the corresponding positions along the height direction of the groove edge ROI window, so as to obtain the center position value of the groove. Through the expression △x i =(x g -x e ) Calculate the difference between the center position of the groove and the center position of the welding torch, adjust the position of the welding torch before welding, and realize the tracking and guidance of the welding torch position before welding.

[0077] Figure 8 (a) is the groove image before welding, ROI image and its edge extraction result when the welding wire stays at the center of the groove. Figure 8 (a1) is a pre-weld groove image when the welding wire is stopped, which includes two groove edge ROI windows with a height of h5 pixels and one welding wire ROI window with a height of h6 pixels, where ROI-g1, ROI-g2 and ROI-w1 are the groove left and right edges and welding wire ROI windows respectively, and the above windows are determined by the midpoint method or the midline method, as well as the image size, groove width, welding torch height and the distance from the end of the welding wire to the bottom of the groove. The height and width of ROI-g1 and ROI-g2 are 80×80 pixels; the height and width of ROI-w1 are 100×100 pixels; the size of the window can be automatically set according to the actual image size collected.

[0078] Through the ROI-g1 and ROI-g2 windows, the ROI images of the left and right edges of the groove before welding are intercepted, and the left and right edges of the groove are extracted by using median filtering, morphological opening operation, adaptive threshold and Canny edge operator, as shown in Figure 8 (a2) and 8(a4); the ROI image of the welding wire before welding is intercepted through the ROI-w1 window, and the left and right edges of the welding wire are extracted by using median filtering, local adaptive threshold, morphological operation and Canny edge extraction. The intercepted welding wire ROI image and its edge extraction results are shown in Figure 8 As shown in (a3).

[0079] Along the height direction of the groove edge ROI window, the mean of the left and right edges of the groove is calculated, and the center position of the groove before welding is obtained as 269.48 pixels; along the height direction of the welding wire ROI window, by calculating the mean of the corresponding positions of the left and right edges of the welding wire, the center position of the welding torch is obtained as 268.837 pixels. By calculating the difference between the center position of the groove and the center position of the welding torch, it can be known that the welding torch is 0.023mm left-biased at this time. In the experiment, when the deviation is less than 0.05mm, it is approximately considered that the welding torch is in the centering state before welding. The above method realizes the monitoring, guidance and tracking functions of the centering state of the welding torch position before welding.

[0080] Figure 8 (b) is the extraction result of the welding wire contour and the center position of the welding torch when the welding wire is shaken to the left and right sides of the groove and stops. Figure 8 (b1) shows the image of the groove before welding when the welding wire including the welding wire ROI window is shaken to the left and right sides of the groove. At this time, the welding wire shakes at a certain frequency. The welding wire ROI window when the welding wire is located at the left and right sides of the groove is obtained by the distance from the end of the welding wire to the side wall of the groove, as shown in FIG. Figure 8 As shown in the two white square boxes in (b1), the ROI window height h6 and width are both 80 pixels. Figure 8 The ROI image of the welding wire before welding is processed by the ROI image processing method before welding to extract the left and right contours of the welding wire (including the left and right edges of the welding wire, respectively) when the welding wire is shaken to the left and right sides of the groove, as shown in (b2) and (b3). Figure 8 As shown in (b4) and (b5). Along the height direction of the welding wire ROI window, the mean of the left and right edges of the welding wire are calculated respectively to obtain the left and right center lines of the welding wire. By calculating the mean of the left and right center lines of the welding wire, the center position of the welding torch is obtained. Together with the obtained groove center position, the pre-welding position of the welding torch can be guided.

[0081] according to Figure 8 The collected pre-weld groove image is processed by the pre-weld wire ROI image processing method to extract the wire contour and wire center line, so as to obtain the center position of the welding torch in real time and realize the calibration of the initial welding torch position before welding. At the same time, the obtained initial position of the welding torch before welding is combined with the center position of the pre-weld groove to realize the monitoring, tracking and guidance of the welding torch position before welding.

[0082] The above description is only a part of the specific implementation of the present invention. Of course, the present invention may also have other various embodiments. Without departing from the spirit and essence of the present invention, any technician familiar with the technical field can make various corresponding equivalent changes and deformations according to the present invention, which should all fall within the protection scope of the claims attached to the present invention.

Claims

1. A narrow gap welding visual sensing weld seam tracking method, using a narrow gap welding visual sensing weld seam tracking system, the narrow gap welding visual sensing weld seam tracking system comprising a narrow gap welding torch, a narrow gap welding control cabinet (14), an infrared camera (8), a filter system (7), a current sensor (13), a signal trigger (9), an image real-time transmission module (10), a computer image processing module (11) and a data communication module (12); the infrared camera (8) is aimed at the welding groove (6) and is arranged at an angle θ with the horizontal plane (θ=10-80°), and is connected to the output end of the signal trigger (9) and the input end of the image real-time transmission module (10) through a video cable; the input end of the signal trigger (9) is connected to the narrow gap welding control cabinet (14) and the current sensor (13) through a wire, and the narrow gap welding control cabinet (14) and the current sensor (13) are connected to each other through a wire. The control cabinet (14) is connected to the arc motion driver (3b) of the narrow gap welding torch; one end of the welding power source (3e) in the narrow gap welding torch is connected to the arc motion driver (3b), and the other end of the connecting cable is connected to the workpiece where the welding groove (6) is located after passing through the detection ring of the current sensor (13) or connecting the current sensor (13) in series; the real-time image transmission module (10) includes an image acquisition card and image acquisition software, and is sequentially connected to a computer image processing module (11) including an edge extraction module, a weld deviation value obtaining module, and a weld deviation adjustment amount obtaining module, and a data communication module (12) including a communication protocol, communication software and a communication interface, and a narrow gap welding control cabinet (14) through a video cable; the filter system (7) is coaxially arranged at the front end of the infrared camera (8), and is composed of a filter, a dimming filter and a UV filter, and is characterized in that: The steps include: 1) Real-time image acquisition: The arc current base value signal i detected by the current sensor (13) is b and arc position signal P arc , a single channel or a dual channel is sent to a signal trigger (9), controlling an infrared camera (8) to start collecting a pre-weld groove image with a width and height of h1 (h1≤1024) and h2 (h2≤1024) pixels respectively or a global image of the welding area synchronized with the arc position (15), and developing an image acquisition program through C or C++ language and image processing software to form an image real-time transmission module (10) with real-time image acquisition, storage and output functions, thereby realizing real-time image transmission and providing real-time images for pre-weld torch position guidance and weld seam tracking control of narrow gap welding; 2) Extract the groove edge and obtain the groove center position x g :Including obtaining the groove center position x of the groove image before welding or the global image of the welding area (15) g ; The groove center position x of the groove image before welding g The method comprises the following steps: according to a single frame of a pre-weld groove image acquired in real time, adaptively positioning a groove left and right edge ROI window with a height of h5 (5≤h5≤h2) pixels by a midpoint method or a midline method, and adaptively intercepting the groove left and right edge ROI images; extracting the pre-weld groove left edge (4) and the groove right edge (5) by a pre-weld groove edge ROI image processing method in a computer image processing module (11); obtaining the original position point distribution of the groove center by calculating the mean value or the fitting value along the groove edge ROI window height direction; and calculating the groove center position x in the pre-weld groove image by the mean value method. g ; The groove center position x of the global image (15) of the welding area g The method is to extract the position of the arc (2) in the global image (15) of the welding area through a global image processing method in a computer image processing module (11), and adaptively set a groove edge ROI window (16) on the opposite side of the arc (2) through the position of the arc (2), intercept a groove edge ROI image with a height of h3 (5≤h3≤h2) pixels, and extract the groove left edge (4) or the groove right edge (5) on the opposite side of the arc (2) through the groove edge ROI image processing method in the computer image processing module (11); according to the The groove left edge (4) and the groove right edge (5) extracted from the two adjacent frames of the global image (15) of the welding area are used to obtain the distribution of the original position points of the groove center by calculating the mean value or the fitting value along the height direction of the groove edge ROI window; the dispersion of the distribution of the original position points of the groove center is analyzed by a pattern recognition anti-interference algorithm based on the coefficient of variation, and after the abnormal original position points of the groove center are eliminated, the groove center position x of the global image (15) of the welding area is obtained by calculating the mean value or the fitting value from the retained effective position points of the groove center. g , providing a basis for correcting the position of the narrow gap welding torch; 3) Extract the welding wire contour and obtain the welding torch center position x e : The center position x of the welding torch including the image of the groove before welding and the global image of the welding area (15) e ; The center position of the welding torch before welding groove image x e The method further comprises obtaining a pre-weld groove image when the welding wire is stopped and a pre-weld groove image when the welding wire is shaken; for the pre-weld groove image when the welding wire is stopped or shaken acquired by the image real-time transmission module (10), a pre-weld wire ROI image with a height of h6 (5≤h6≤h2) pixels is intercepted through a pre-weld wire ROI window, and a pre-weld wire ROI image processing method is used to extract a welding wire contour, and obtain a welding torch center position x of the pre-weld groove image. e , initial torch position guidance for narrow gap welding; The center position x of the welding torch in the global image (15) of the welding area e The method comprises: capturing a welding wire ROI image with a height of h4 (5≤h4≤h2) pixels through a welding wire ROI window (17) on the same side of the arc (2); extracting the welding wire edge and the welding wire center through a welding wire ROI image processing method; obtaining a welding torch center position sampling value through a welding wire single-side edge detection method or a welding wire double-side edge detection method according to the welding wire left contour (including the welding wire left edge and the welding wire right edge), the welding wire left center or the welding wire right contour (including the welding wire left edge and the welding wire right edge), and the welding wire right center extracted from two adjacent frames of the welding area global image (15); and obtaining the welding torch center position x of the welding area global image (15) through a numerical filtering algorithm for the obtained welding torch center position sampling value. e ; 4) Obtain the weld deviation value △x i : Based on the groove center position and the welding torch center position respectively obtained from the groove image before welding or the global image of the welding area (15), the weld deviation detection module in the computer image processing module (11) is used to calculate the weld deviation according to the formula △x i =(x g -x e ) respectively calculates the deviation value of the welding torch relative to the groove center before welding or during welding, obtains the weld deviation of the groove image before welding or the global image of the welding area (15), and provides a basis for the welding torch position guidance and weld tracking control before welding; 5) Obtain the weld deviation adjustment amount: for the obtained weld deviation value △x i , through proportional, integral, differential controllers, or separable combinations, intelligent controllers, and composite controllers integrated with proportional, integral, and differential controllers, develop narrow gap weld tracking controllers to form weld tracking controllers that automatically match control parameters according to weld deviations, thereby obtaining weld deviation adjustment amounts; 6) Correction operation: for the obtained weld deviation adjustment amount, data communication software and communication interface are developed through computer language to form a data communication module (12) with data storage and output functions; the weld deviation adjustment amount is sent to the narrow gap welding control cabinet (14) through the data communication module (12) to guide the welding torch before welding and correct the deviation during welding; 7) Repeat the above steps 1) to 6) until the welding process is completed.

2. The narrow gap welding visual sensing weld tracking method according to claim 1 is characterized in that: In step 2), the midpoint method or the centerline method includes: by searching the midpoint coordinates (h1 / 2, h2 / 2) of the pre-weld groove image, or finding the centerline segment position x=h1 / 2 or y=h2 / 2 of the pre-weld groove image, based on the initial groove width value, the midpoint or centerline segment position of the pre-weld groove image, after translating up and down and left and right for a certain distance, automatically finding the ROI window position of the left and right edges of the groove of the pre-weld groove image, and adaptively intercepting the ROI image of the pre-weld groove edge.

3. The narrow gap welding visual sensing weld tracking method according to claim 1 is characterized in that: In step 2), the pre-weld groove edge ROI image processing method includes: median filtering, morphological operation, adaptive threshold and edge extraction algorithm.

4. The narrow gap welding visual sensing weld tracking method according to claim 1 is characterized in that: In step 2), the steps of the global image processing method are: performing median filtering, histogram analysis, global threshold and morphological operations on the global image (15) of the welding area, extracting the morphology of the arc (2), obtaining the position of the highest point of the arc through a grayscale recognition method, and numerically filtering the ordinate values ​​of multiple consecutive highest point positions of the arc to reduce the influence of welding interference on the highest point position of the arc.

5. The narrow gap welding visual sensing weld tracking method according to claim 1 is characterized in that: In step 2), the steps of the groove edge ROI image processing method are: median filtering to remove noise, contrast stretching to improve image contrast, four-partition threshold processing of the ROI image, and then morphological operation to eliminate isolated points, and then using the Canny edge operator to extract the groove edge.

6. The narrow gap welding visual sensing weld tracking method according to claim 1 is characterized in that: In step 2), the steps of the pattern recognition anti-interference algorithm based on the coefficient of variation are: partitioning the obtained h3 groove center original position points into n equal parts, calculating the coefficient of variation of the groove center original position points in each equal partition by the coefficient of variation (CV) formula (CV = standard deviation / mean), retaining the groove center original position point in the equal part corresponding to the minimum coefficient of variation as the groove center true position point; The actual position point of the groove center is globally matched with h3 original position points of the groove center, and the original position points of the groove center with the same value as the actual position points of the groove center are retained in situ as the effective position points of the groove center; the sampling value of the groove center position is obtained by linear fitting or calculating the mean of the effective position points of the groove center; for the obtained continuous multiple sampling values ​​of the groove center position, the sampling values ​​of the groove center position are filtered by the numerical filtering algorithm to obtain the groove center position value x g .

7. The narrow gap welding visual sensing weld tracking method according to claim 1 is characterized by: In step 3), the steps of the welding wire ROI image processing method before welding are: performing median filtering, local adaptive threshold processing, morphological operation, and Canny edge operator on the intercepted welding wire ROI image before welding to extract the welding wire contour; the steps of the welding wire ROI image processing method are: performing median filtering denoising, local adaptive threshold processing, morphological operation, and Canny edge operator on the intercepted welding wire ROI image to extract the welding wire contour.

8. The narrow gap welding visual sensing weld tracking method according to claim 1 is characterized in that: In step 3), the single-side edge detection method of the welding wire is to use the mean or fitting value of the single-side edge position point distribution of the welding wire as the sampling value of the center position of the welding torch; including the inner edge detection method of the welding wire and the outer edge detection method of the welding wire; the inner edge detection method of the welding wire refers to obtaining the sampling value of the center position of the welding torch by calculating the mean or fitting value according to the distribution of the edge points of the welding wire that are closer to the groove edge on the same side where the welding wire stops; the outer edge detection method of the welding wire refers to obtaining the sampling value of the center position of the welding torch by calculating the mean or fitting value according to the distribution of the edge points of the welding wire that are farther from the groove edge on the same side where the welding wire stops; the double-side edge detection method of the welding wire is to first obtain the distribution of the left or right center line position points of the single-side welding wire according to the extracted left and right edges of the welding wire, and then obtain the sampling value of the center position of the welding torch of the current frame by calculating the mean or fitting value of the left and right center line position points of the welding wire in two adjacent frames of welding images; the numerical filtering algorithm includes median filtering, mean filtering, anti-pulse mean or median filtering, and developed numerical filtering algorithm.

9. The narrow gap welding visual sensing weld tracking method according to claim 1 is characterized in that In step 4), the expression △x i =(x g -x e ), also expressed as △x i =(x e -x g ).

Images