A method for detecting welding quality of a grouting anchor pipe welding end

Abstract

The present application relates to the technical field of image processing, more particularly, the present application relates to a kind of grouting anchor pipe welding end head welding quality detection method, the method comprises, obtaining welding gray image, obtain several anchor points in welding gray image, obtain the texture structure difference of each anchor point;Texture normal range interval is obtained, according to whether the texture structure difference of each anchor point belongs to texture normal range interval, the anchor point is screened, obtains several edge points, each edge point is edge tracking, if there is welding tumor in the process of tracking, according to the convex situation of image welding tumor, the length of span is obtained, the welding tumor is crossed, obtains two continuous edge lines, according to two edge lines, whether the welding end head and the rod body part are coaxial after welding is judged, and then the welding quality is judged, the complete edge line of the present application is obtained, and the influence of welding tumor edge is avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing. More particularly, the present application relates to a method for detecting the welding quality of a grouting anchor pipe welding end. BACKGROUND

[0002] The grouting anchor pipe is a reinforcing element frequently used in geotechnical engineering, and the connection quality between the welding end and the rod body directly relates to the safety and reliability of the entire anchoring system. Welding is a key process for connecting the two, and one of the core quality indicators is to ensure that the welding end and the rod body part are strictly coaxial.

[0003] On the automatic production line, machine vision technology is usually used to detect the coaxiality after welding. The basic principle is to collect the image of the welding area of the grouting anchor pipe, extract the contour edge line of the anchor pipe through edge detection algorithm, and then analyze the geometry of the edge line to judge the coaxiality.

[0004] However, in the actual welding process, "welding bumps" will inevitably be produced at the weld. Welding bumps are irregular metal nodules formed after molten metal splashes and solidifies, and their surfaces are uneven, appearing as a strong interference area with complex texture and irregular edges on the image. Existing edge detection algorithms, such as the EDline algorithm (Exponential Decay Line), face great challenges when processing such images: when the algorithm tracks to the welding bump area, due to its irregular shape and dramatic change in gradient direction, it will cause the edge tracking to be interrupted, and a complete anchor pipe contour cannot be formed. In addition, the complex edges of the welding bumps are also mistakenly detected as a large number of small and chaotic "pseudo-edges", which seriously interferes with the subsequent analysis of the true edges, making it extremely difficult to accurately judge the coaxiality.

[0005] Traditional edge detection algorithms cannot effectively distinguish between the true anchor pipe edge and the welding bump edge, and are prone to misdetecting the complex contour of the welding bump as a large number of chaotic pseudo-edges, which seriously pollutes the detection results. When the true anchor pipe edge line extends to the welding bump area, due to the physical obstruction of the welding bump and the dramatic change in image features, the edge tracking process of the traditional algorithm will be interrupted at this point, and a continuous and complete anchor pipe contour cannot be extracted, thus making it impossible to perform effective coaxiality analysis. SUMMARY

[0006] In order to solve the problem that due to the physical obstruction of the welding bump and the dramatic change in image features, the edge tracking process of the traditional algorithm will be interrupted at this point, and a continuous and complete anchor pipe contour cannot be extracted, thus making it impossible to perform effective coaxiality analysis, the present application proposes a method for detecting the welding quality of a grouting anchor pipe welding end, which comprises the following steps:

[0007] Collecting a welding grayscale image and extracting a plurality of anchor points;

[0008] calculate the texture structure difference of each anchor point; screen the anchor points based on the texture structure difference to obtain real edge points;

[0009] edge tracking is performed on the edge points to form edge lines, when a welding bump is encountered in the tracking process, a toe starting point is obtained; based on the toe starting point, tracking is continued to locate a peak point of the texture structure difference of the welding bump area; based on the change of the texture structure difference between the toe starting point and the peak point, an average slope is obtained; based on the average slope, a crossing length of the toe starting point is obtained; the toe starting point is jumped by a distance of the crossing length along the tracking direction of the edge point, and edge tracking is continued to obtain two edge lines;

[0010] Based on the two edge lines, the coaxiality of the welded end head and the rod body part of the grouting anchor pipe is judged to evaluate the welding quality.

[0011] The present application obtains the texture structure difference index of each anchor point, and effectively filters out most of the false edge points caused by the welding bump through adaptive threshold screening, to obtain a plurality of edge points; then the edge points are tracked, and the physical occlusion area of the welding bump area is adaptively skipped in the tracking process, ensuring that even in the presence of welding bumps, continuous and complete anchor pipe profiles can still be extracted, which completely solves the problem of edge interruption of traditional edge detection algorithms under the interference of welding bumps, making the subsequent evaluation of the coaxiality of the welded end head and the rod body part of the grouting anchor pipe based on the edge lines more accurate.

[0012] Preferably, the calculation of the texture structure difference of each anchor point comprises:

[0013] a circular neighborhood window is constructed around the anchor point, and the circular neighborhood window is divided into two symmetrical semicircular regions along the gradient direction of the anchor point;

[0014] the local window coding sequence of each semicircular region of each anchor point is obtained;

[0015] ；

[0016] In the formula, represents the texture structure difference of the i th anchor point; represents the number of all runs in the local window coding sequence of the first semicircular region of the i th anchor point; represents the number of all runs in the local window coding sequence of the second semicircular region of the i th anchor point; represents the number of all elements in the local window coding sequence of any semicircular region of the i th anchor point; || represents the absolute value symbol.

[0017] According to the symmetry of the texture on both sides of the real edge point, the texture structure difference of the anchor point is quantified, so that the real edge point and the weld bead edge point can be effectively distinguished.

[0018] Preferably, the local window coding sequence of each semicircular region of each anchor point is obtained by:

[0019] The contrast pixel points of each semicircular region of each anchor point are obtained.

[0020] If the gray value of any contrast pixel point in any semicircular region of any anchor point is greater than the gray value of the anchor point, 1 is taken as the target coding of the contrast pixel point; if the gray value of the contrast pixel point is less than or equal to the gray value of the anchor point, 0 is taken as the target coding of the contrast pixel point; the sequence composed of the target coding of all contrast pixel points in the semicircular region is recorded as the local window coding sequence of the semicircular region; similarly, the local window coding sequence of the other semicircular region of the anchor point is obtained.

[0021] The local window coding sequences of the two semicircular regions of the anchor point are compared to obtain the texture structure difference of the anchor point.

[0022] Preferably, the contrast pixel points of each semicircular region of each anchor point are obtained by:

[0023] A division parameter t is preset, and the neighborhood circular window of the anchor point is equally divided into 2t sectors by t straight lines passing through the anchor point; the intersection point of each straight line and the neighborhood circular window of the anchor point is recorded as a target intersection point; all pixel points corresponding to all target intersection points in each sector of each semicircular region are recorded as the contrast pixel points of each semicircular region.

[0024] Preferably, the anchor points are screened based on the texture structure difference to obtain the real edge points, comprising:

[0025] The gray value of the anchor point is obtained by:

[0026] For any anchor point in the welded gray image, if the texture structure difference of the anchor point is in the texture normal range interval, the anchor point is recorded as an edge point, and a plurality of edge points in the welded gray image are obtained.

[0027] Preferably, the edge tracking is performed on the edge points to form an edge line, and when a weld bead is encountered in the tracking process, a slope foot starting point is obtained.

[0028] The tracking direction of the edge point is obtained, edge tracking is performed in any tracking direction of any edge point, and a pixel point adjacent to the edge point in the tracking direction of the edge point is obtained and recorded as a first to-be-tracked pixel point; if the texture structure difference of the first to-be-tracked pixel point is in the texture normal range interval, a pixel point adjacent to the first to-be-tracked pixel point in the tracking direction is obtained and recorded as a second to-be-tracked pixel point, and the process is repeated until the texture structure difference of the latest to-be-tracked pixel point is not in the texture normal range interval, and then the latest to-be-tracked pixel point is recorded as a toe starting point and all other to-be-tracked pixel points are recorded as edge points.

[0029] The toe starting point is facilitated to be crossed subsequently.

[0030] Preferably, the texture structure difference of the toe starting point is tracked based on the toe starting point to locate a peak point, and the method comprises the following steps:

[0031] A pixel point adjacent to the toe starting point in the tracking direction of the edge point is obtained and recorded as a first target pixel point; if the difference in texture structure difference between the first target pixel point and the toe starting point is greater than 0, a pixel point adjacent to the first target pixel point is obtained and recorded as a second target pixel point, and the process is repeated until the difference in texture structure difference between the latest target pixel point and the previous target pixel point is less than 0, and then the previous target pixel point is recorded as a peak point.

[0032] Preferably, the average slope is obtained according to the change in texture structure difference between the toe starting point and the peak point, and the crossing length of the toe starting point is obtained based on the average slope, and the method comprises the following steps:

[0033] A sequence of texture structure differences of all pixel points between the toe starting point and the peak point is recorded as a texture structure difference sequence; Z represents the average slope; represents the number of texture structure difference data in the texture structure difference sequence; represents the difference between the (i+1)th texture structure difference data and the ith texture structure difference data in the texture structure difference sequence;

[0034] , represents the crossing length of the toe starting point; represents the actual average width of the welding bump measured by manual measurement; norm() represents a normalization function.

[0035] The crossing length of the toe starting point is obtained, which facilitates subsequent crossing of the toe starting point based on the crossing length and solves the problem of edge breakage.

[0036] Preferably, the tracking direction of the edge point is jumped over the distance of the span length, and the edge tracking is continued to obtain two edge lines, including:

[0037] Along the tracking direction of the edge point, L pixel points adjacent to the toe starting point are obtained as edge points, and the edge tracking is continued after the span to stop until the image edge is tracked to obtain two edge lines. The span length represents the toe starting point.

[0038] Preferably, based on the two edge lines, the coaxial condition of the welded end head and the rod body part of the grouting anchor pipe is judged to evaluate the welding quality, including:

[0039] A preset fitting threshold T is set, for any pixel point on any edge line, the pixel point on the other edge line perpendicular to the pixel point is recorded as the matching pixel point of the pixel point, the matching pixel points of all pixel points on the edge line are obtained, the midpoint of each pixel point and its matching pixel point on the edge line is obtained, all midpoints are fitted as a straight line using the least square method, and the determination coefficient is obtained, if the determination coefficient is greater than the fitting threshold, the welding quality is qualified, otherwise, the welding quality is unqualified.

[0040] The present application has the following beneficial effects: the purpose of the present application is to first obtain a plurality of anchor points in the image, then analyze the gray scale around each anchor point, obtain the texture structure difference index of each anchor point, and effectively filter out most of the false edge points caused by the welding bump through adaptive threshold screening to obtain a plurality of edge points; then the edge points are tracked, and the physical occlusion area of the welding bump area is adaptively skipped in the tracking process, ensuring that even in the presence of welding bumps, continuous and complete anchor pipe contours can still be extracted, which completely solves the problem of edge interruption of traditional edge detection algorithms under the interference of welding bumps, making the subsequent evaluation of the coaxial condition of the welded end head and the rod body part of the grouting anchor pipe based on the edge line more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 It is a step flow chart of a grouting anchor pipe welded end head welding quality detection method according to an embodiment of the present application. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application.

[0043] Please refer to Figure 1 which shows a step flow chart of a grouting anchor pipe welded end head welding quality detection method according to an embodiment of the present application, and the method includes the following steps:

[0044] S001, a welded gray scale image is obtained.

[0045] It should be noted that the welding end of the grouting anchor pipe is generally used to connect the anchor to transmit the anchoring force, and the welding end of the grouting anchor pipe and the rod body are welded together to form a complete grouting anchor pipe, and the purpose of the present application is to measure whether the welding end and the rod body part of the welded grouting anchor pipe are coaxial to judge the welding quality.

[0046] In the embodiment of the present application, an industrial camera is used to photograph the welding area of the grouting anchor pipe, and it is ensured that the welding end and the rod body part of the grouting anchor pipe can be photographed to obtain a welding RGB image, and the welding RGB image is subjected to grayscale processing to obtain a welding grayscale image for subsequent analysis.

[0047] The rod body part (without welding) of the grouting anchor pipe is photographed using an industrial camera and then subjected to grayscale processing to obtain a rod body grayscale image.

[0048] S002, a plurality of anchor points in the welding grayscale image are obtained, the texture structure difference of each anchor point is obtained, a texture normal range interval is obtained, and the anchor points are screened according to whether the texture structure difference of each anchor point belongs to the texture normal range interval to obtain a plurality of edge points.

[0049] It should be noted that the grouting anchor pipe is obtained after the welding end and the rod body part of the grouting anchor pipe are welded, and the two edges of the grouting anchor pipe in the welding grayscale image are detected, and whether the welding end and the rod body part are coaxial can be judged according to the boundary of the grouting anchor pipe, so it is necessary to detect the edges in the welding grayscale image.

[0050] It is known that in the process of edge tracking, the traditional EDline algorithm checks whether the gradient direction of a pixel point is consistent with the gradient direction of the previous pixel point to determine whether the pixel point is an edge point, but since the welding end and the rod body part are welded together, welding beads will appear in the welding area, and when the EDline algorithm tracks the welding bead, the consistency of the gradient direction will be broken due to the irregular shape of the welding bead. At this time, the Edline algorithm will interrupt the tracking, resulting in the inability to obtain a complete edge, and the irregular shape of the welding bead itself will also be incorrectly detected by EDline as many small and chaotic short line segments. These "pseudo edges" will be mixed in the real edges, making it impossible to accurately determine whether the welding end and the rod body part are coaxial.

[0051] It needs to be further explained that the welding bump is an irregular metal bump or bead formed by the solidification of the molten metal after splashing, and the surface is uneven and rough, so the gray value of the pixel points in the welding bump will change sharply, and the gray value of the pixel points in the smooth area without welding bump usually changes gently, so the EDline line segment extraction algorithm is used first to obtain a plurality of anchor points in the welding gray image, the anchor point is a pixel point with large gradient change in the image, that is, a pixel point on the edge or welding bump edge;

[0052] The two sides of the edge point are the background area or the grouting anchor pipe area, and the gray distribution of the two areas is uniform, so a neighborhood circular window is constructed for the anchor point, and the neighborhood circular window is divided into two symmetrical half circular areas according to the gradient direction of the anchor point, if the gray values of the pixel points on the two symmetrical half circular areas are more similar, it means that the anchor point is an edge point, otherwise, there are welding bump area pixel points in the neighborhood circular window of the anchor point, which means that the anchor point may be a welding bump edge point or the anchor point is an edge point but there are welding bump area pixel points in the neighborhood circular window, at this time, the anchor point needs to be excluded and not edge tracking (the anchor point which is an edge point but has welding bump area pixel points in the neighborhood circular window is excluded, and the subsequent edge point will be tracked), so the texture structure difference of each anchor point is obtained by constructing a neighborhood circular window for the anchor point, and then the edge point is obtained.

[0053] In the embodiment of the application, the EDline algorithm is used to obtain a plurality of anchor points in the welding gray image; it needs to be noted that the EDline algorithm for obtaining anchor points is a known technology, and in the embodiment of the application, it will not be described in more detail;

[0054] A radius parameter r=5 is preset, for any anchor point, a circular window with a radius of r is constructed with the anchor point as the center, which is recorded as the neighborhood circular window of the anchor point;

[0055] The neighborhood circular window of the anchor point is divided into two symmetrical half circular areas by a straight line passing through the anchor point and along the gradient direction of the anchor point;

[0056] A division parameter t=18 is preset, and the neighborhood circular window of the anchor point is equally divided into 2t sector areas by t straight lines passing through the anchor point; the intersection point of each straight line and the neighborhood circular window of the anchor point is recorded as a target intersection point;

[0057] All target intersection points in all sector areas in each half circular area are recorded as contrast pixel points of each half circular area;

[0058] The preset two encoding parameters are b1=1 and b2=0, if the gray value of any contrast pixel point in any semicircular region is greater than the gray value of the anchor point, b1 is taken as the target code of the contrast pixel point; if the gray value of the contrast pixel point is less than or equal to the gray value of the anchor point, b2 is taken as the target code of the contrast pixel point; a sequence formed by the target codes of all the contrast pixel points in the semicircular region is recorded as the local window encoding sequence of the semicircular region; similarly, the local window encoding sequence of another semicircular region is obtained;

[0059] The specific method for obtaining the texture structure difference of each anchor point according to the local window encoding sequences of the two semicircular regions of the anchor point is as follows:

[0060] ；

[0061] In the formula, represents the texture structure difference of the i-th anchor point; represents the number of all runs in the local window encoding sequence of the first semicircular region of the i-th anchor point; represents the number of all runs in the local window encoding sequence of the second semicircular region of the i-th anchor point; represents the number of all elements in the local window encoding sequence of any semicircular region of the i-th anchor point; The smaller the value of is, the more consistent the number of all runs in the local window encoding sequences of the two semicircular regions of the i-th anchor point is, at this time, the jumps of the gray values on the two semicircular regions are consistent, at this time, the i-th anchor point is more likely to be an edge point, on the contrary, the more inconsistent the gray values on the two semicircular regions of the i-th anchor point are, at this time, the neighborhood of the i-th anchor point is more likely to be a weld nodule edge point or there is a weld nodule region pixel point in the circular window of the neighborhood of the anchor point.

[0062] It should be noted that the rod body part of the grouting anchor pipe is photographed by using an industrial camera, and there is no weld nodule in the obtained rod body gray image, therefore, according to the method for obtaining the texture structure difference of the anchor point of the welding gray image, the texture structure differences of all the anchor points in the rod body gray image are obtained, all the anchor points in the rod body gray image are real edge points, therefore, according to the texture structure differences of all the anchor points in the rod body gray image, the texture normal range interval is obtained, which is convenient for screening the anchor points in the welding gray image.

[0063] According to the above operation, the mean value and the standard deviation of the texture structure differences of all the anchor points in the rod body gray image are obtained; the difference value between the mean value and the standard deviation is recorded as the left boundary; the sum value between the mean value and the standard deviation is recorded as the right boundary; the interval formed by the left boundary and the right boundary is recorded as the texture normal range interval;

[0064] For any anchor point in the welding grayscale image, if the texture structure difference of the anchor point is within the normal range of texture, then the anchor point is recorded as an edge point, and several edge points in the welding grayscale image are obtained.

[0065] S003. Perform edge tracking for each edge point. If there is a weld bead during the tracking process, obtain the crossing length based on the protrusion of the weld bead in the image, cross the weld bead, and obtain a continuous edge line.

[0066] It should be noted that during the process of tracking edge points, the texture structure difference of each pixel is calculated along the tracking direction. If the texture structure difference is within the normal range of texture, the pixel is an edge point. Conversely, if the texture structure difference is not within the normal range of texture, the pixel is the starting point of the slope. If there is a weld bead in front of the tracking direction, the span length of the starting point of the slope needs to be obtained to cross the weld bead and obtain a continuous edge line.

[0067] It should be further explained that the common weld bead morphology is a bun-shaped protrusion, where the portion of the weld surface that protrudes excessively above the base material and has an uneven shape, resembling small hills piled up on the weld. Therefore, the weld bead area in the welding grayscale image appears as a small peak. The texture structure difference at the starting point of the slope is not within the normal range because one side of the neighboring circular window of the starting point of the slope contains a small number of weld bead area pixels. If the starting point of the slope moves along the tracking direction, the number of weld bead area pixels in the neighboring circular window will first increase and then gradually decrease. Therefore, when the starting point of the slope moves along the tracking direction, the texture structure difference will show a process of first increasing and then decreasing. Therefore, this invention obtains the peak point corresponding to the maximum texture structure difference during the movement. Based on the difference in texture structure difference of pixels between the starting point of the slope and the peak point, the average slope is obtained. The average slope can reflect the degree of protrusion of the weld bead.

[0068] In this embodiment of the invention, a straight line perpendicular to the gradient direction of the edge point is denoted as the target straight line; both directions of the target straight line are denoted as the tracking directions of the edge point.

[0069] Taking any tracking direction of any edge point as an example, the specific process of edge tracking is as follows:

[0070] In the tracking direction of the edge point, the pixel adjacent to the edge point is obtained and recorded as the first pixel to be tracked; if the texture structure difference of the first pixel to be tracked is within the normal range of texture, the pixel adjacent to the first pixel to be tracked is obtained in the tracking direction and recorded as the second pixel to be tracked, and so on, until the texture structure difference of the latest pixel to be tracked is not within the normal range of texture, then the latest pixel to be tracked is recorded as the starting point of the slope, and all other pixels to be tracked are recorded as edge points;

[0071] In the tracking direction of the edge point, a pixel point adjacent to the slope foot starting point is obtained and recorded as a first target pixel point; if the difference between the texture structure difference of the first target pixel point and the slope foot starting point is greater than 0, a pixel point adjacent to the first target pixel point is obtained and recorded as a second target pixel point, and so on, until the difference between the texture structure difference of the latest target pixel point and the texture structure difference of the previous target pixel point is less than 0, and the previous target pixel point is recorded as a peak point;

[0072] A sequence formed by the texture structure differences of all the pixel points between the slope foot starting point and the peak point is recorded as a texture structure difference sequence;

[0073] An average slope is obtained:

[0074] ;

[0075] In the formula, Z represents the average slope; represents the number of texture structure difference data in the texture structure difference sequence; represents the difference between the (i+1)th texture structure difference data and the ith texture structure difference data in the texture structure difference sequence;

[0076] It should be noted that in a stable and repeated welding process, the amount of melted metal per unit length is relatively fixed. If the shape of the welding bead needs to become steeper (i.e., the protrusion degree is greater), the bottom width of the welding bead will be reduced, and conversely, if the welding bead is pressed flatter (the protrusion degree is reduced), its bottom width will increase and expand to both sides. Because if the average width of the welding bead is larger and the average slope is smaller, the spanning length of the slope foot starting point is greater at this time; on the contrary, if the average width of the welding bead is smaller and the average slope is greater, the spanning length of the slope foot starting point is smaller at this time. Therefore, the spanning length of the slope foot starting point is obtained by measuring the average width of the welding bead and combining the above average slope.

[0077] In the embodiment of the present application, after the welding is completed, the width of the welding bead is measured at multiple different positions by manual selection, and the average of the measured multiple welding bead widths is taken as the actual welding bead average width measured by manual measurement. The specific way of obtaining the spanning length of the slope foot starting point according to the average width of the welding bead and the average slope is:

[0078] ;

[0079] In the formula, represents the spanning length of the slope foot starting point; represents the actual welding bead average width measured by manual measurement; Z represents the average slope; and norm() represents a normalization function.

[0080] Along the tracking direction of the edge points, obtain L consecutive adjacent pixels to the starting point of the slope as edge points. After crossing, continue edge tracking until the image edge is tracked and then stop to obtain the edge line.

[0081] The edge lines are obtained by simultaneously tracing all edge points.

[0082] It should be noted that when tracking edge points simultaneously, if any edge point encounters other edge points during the tracking process, tracking that edge point should be stopped to obtain a complete edge line.

[0083] S004. Based on the two edge lines, determine whether the welded end and the rod body are coaxial after welding, and then judge the welding quality.

[0084] It should be noted that after extracting the two edge lines of the grouting anchor pipe obtained after welding, the midpoints are extracted point by point from the pixels of the two edge lines. The least squares method is used to fit a straight line to all midpoints and obtain the coefficient of determination. The coefficient of determination reflects the fitting quality. The larger the value, the better the fitted straight line can reflect the change of the midpoint. After welding, the weld end and the rod body are coaxial, indicating good welding quality. Conversely, the smaller the coefficient of determination, the worse the fitting effect of the midpoint. After welding, the weld end and the rod body are not coaxial, indicating poor welding quality.

[0085] In this embodiment of the invention, for any pixel on any edge line, a pixel on another edge line perpendicular to that pixel is recorded as its matching pixel. The matching pixels of all pixels on the edge line are obtained, and the midpoint between each pixel on the edge line and its matching pixel is obtained. The least squares method is used to fit all midpoints to a straight line, and the determination coefficient is obtained. If the determination coefficient is greater than the fitting threshold, the welded end and the rod body of the grouting anchor pipe are coaxial, and the welding quality is qualified; otherwise, the welded end and the rod body of the grouting anchor pipe are not coaxial, and the welding quality is unqualified.

[0086] The preset fitting threshold T=0.9. In other embodiments, the implementer can set the value of the fitting threshold according to the specific implementation situation.

[0087] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting the welding quality of a welding end head of a grouting anchor pipe, characterized in that, The method comprises the following steps: Collecting a welding grayscale image, and extracting a plurality of anchor points; Calculating the texture structure difference of each anchor point; Screening the anchor points based on the texture structure difference to obtain real edge points; Edge tracking is performed on the edge points to form edge lines, and when a welding bump is encountered during the tracking process, a toe starting point is obtained; Based on the toe starting point, the peak point of the texture structure difference of the welding bump area is located; According to the change of the texture structure difference between the toe starting point and the peak point, the average slope is obtained; Based on the average slope, the crossing length of the toe starting point is obtained; The toe starting point is jumped by a distance of the crossing length along the tracking direction of the edge point, and edge tracking is continued to obtain two edge lines; Based on the two edge lines, the coaxiality of the welding end of the grouting anchor pipe and the rod body part is judged to evaluate the welding quality.

2. The method for detecting the welding quality of the welding end of the grouting anchor pipe according to claim 1, characterized in that, The calculation of the texture structure difference of each anchor point comprises: A circular neighborhood window is constructed around the anchor point, and the circular neighborhood window is divided into two symmetrical semicircular regions along the gradient direction of the anchor point; The local window coding sequence of each semicircular region of each anchor point is obtained; ； wherein the texture structure difference of the i-th anchor point; the number of all runs in the local window coding sequence of the first semicircular region of the i-th anchor point; the number of all runs in the local window coding sequence of the second semicircular region of the i-th anchor point; the number of all elements in the local window coding sequence of the arbitrary semicircular region of the i-th anchor point; and || represents the absolute value symbol.

3. The method for detecting the welding quality of the welding end of the grouting anchor pipe according to claim 2, characterized in that, The local window coding sequence of each semicircular region of each anchor point is obtained, which comprises: Obtaining the contrast pixel points of each semicircular region of each anchor point; If the gray value of any contrast pixel point in any semicircular region of any anchor point is greater than the gray value of the anchor point, 1 is taken as the target coding of the contrast pixel point; if the gray value of the contrast pixel point is less than or equal to the gray value of the anchor point, 0 is taken as the target coding of the contrast pixel point; the sequence composed of the target coding of all contrast pixel points in the semicircular region is recorded as the local window coding sequence of the semicircular region; similarly, the local window coding sequence of the other semicircular region of the anchor point is obtained.

4. The method for detecting the welding quality of the welding end of the grouting anchor pipe according to claim 3, characterized in that, The contrast pixel points of each semicircular region of each anchor point are obtained, which comprises: A division parameter t is preset, and the neighborhood circular window of the anchor point is equally divided into 2t sector regions by t straight lines passing through the anchor point; the intersection point of each straight line and the neighborhood circular window of the anchor point is recorded as a target intersection point; all target intersection points in each semicircular region are recorded as the contrast pixel points of each semicircular region.

5. The method for detecting the welding quality of the welding end of the grouting anchor pipe according to claim 1 or 2, characterized in that, The screening of the anchor points based on the texture structure difference to obtain real edge points comprises: Collecting a rod body grayscale image without welding, obtaining the mean value and standard deviation of the texture structure difference of all anchor points in the rod body grayscale image; the difference between the mean value and the standard deviation is recorded as the left boundary; the sum of the mean value and the standard deviation is recorded as the right boundary; the interval formed by the left boundary and the right boundary is recorded as the texture normal range interval; For any anchor point in the welding grayscale image, if the texture structure difference of the anchor point is in the texture normal range interval, the anchor point is recorded as an edge point, and a plurality of edge points in the welding grayscale image are obtained.

6. The method for detecting the welding quality of the welding end of the grouting anchor pipe according to claim 1, characterized in that, The edge tracking of the edge points to form edge lines, and the acquisition of the toe starting point when a welding bump is encountered during the tracking process comprises: The tracking direction of the edge point is acquired, edge tracking is performed in any tracking direction of any edge point, and a pixel point adjacent to the edge point in the tracking direction of the edge point is acquired and recorded as a first to-be-tracked pixel point. If the texture structure difference of the first to-be-tracked pixel point is in a texture normal range interval, a pixel point adjacent to the first to-be-tracked pixel point in the tracking direction is acquired and recorded as a second to-be-tracked pixel point, and the process is repeated until the texture structure difference of the latest to-be-tracked pixel point is not in the texture normal range interval. Then, the latest to-be-tracked pixel point is recorded as a toe starting point, and all other to-be-tracked pixel points are recorded as edge points.

7. The method for detecting the welding quality of the welding end of the grouting anchor pipe according to claim 1 or 6, characterized in that, The toe starting point is used to continue tracking and locate a peak point of the texture structure difference of the welding bump region, and the method comprises the following steps: In the tracking direction of the edge point, a pixel point adjacent to the toe starting point is acquired and recorded as a first target pixel point. If the difference between the texture structure difference of the first target pixel point and the texture structure difference of the toe starting point is greater than 0, a pixel point adjacent to the first target pixel point is acquired and recorded as a second target pixel point, and the process is repeated until the difference between the texture structure difference of the latest target pixel point and the texture structure difference of the previous target pixel point is less than 0. Then, the previous target pixel point is recorded as the peak point.

8. The method for detecting the welding quality of the welding end of the grouting anchor pipe according to claim 1, characterized in that, The average slope is acquired according to the change of the texture structure difference between the toe starting point and the peak point. The span length of the toe starting point is acquired based on the average slope, and the method comprises the following steps: A sequence of texture structure difference of all pixel points between the starting point of the slope foot and the peak point is denoted as a texture structure difference sequence; Z represents the average slope; N represents the number of texture structure difference data in the texture structure difference sequence; Δi represents the difference between the i+1th texture structure difference data and the ith texture structure difference data in the texture structure difference sequence. , the length of the span representing the toe of the slope; represents the average actual weld flash width from manual measurements; norm() represents the normalization function.

9. The method for detecting the welding quality of the welding end of the grouting anchor pipe according to claim 1, characterized in that, The toe starting point is skipped by a distance of the span length along the tracking direction of the edge point, and edge tracking is continued to obtain two edge lines, and the method comprises the following steps: Along the tracking direction of the edge point, L pixel points adjacent to the start point of the slope foot are obtained as the edge point, and the edge tracking is continued after crossing to stop until the image edge is tracked to obtain two edge lines; The crossing length representing the start point of the slope foot.

10. The method for detecting the welding quality of the welding end of the grouting anchor pipe according to claim 1, characterized in that, The coaxiality of the welding end head and the rod body part of the grouting anchor pipe is judged based on the two edge lines to evaluate the welding quality, and the method comprises the following steps: A preset fitting threshold T is set. For any pixel point on any edge line, a pixel point on another edge line perpendicular to the pixel point is recorded as a matching pixel point of the pixel point. The matching pixel points of all pixel points on the edge line are acquired, the midpoint of each pixel point on the edge line and its matching pixel point is acquired, all midpoints are fitted into a straight line by using the least square method, and a determination coefficient is acquired. If the determination coefficient is greater than the fitting threshold, the welding quality is qualified. Otherwise, the welding quality is unqualified.

Images