Multi-layer and multi-weld-seam feature recognition method based on active vision

Abstract

The invention discloses an active vision-based multi-layer and multi-pass welding seam feature identification method, and relates to the technical field of XX. According to the method, laser images are obtained through active vision, a center line extraction and correction technology is adopted, geometric features of welding seams are extracted, a corner recognition and correction technology is combined, welding points are positioned, the welding process is divided into backing welding, filling welding and cosmetic welding, and center line correction and corner recognition and correction methods are selectively adopted for welding of each layer according to welding characteristics. Therefore, the geometric features of the laser center lines of the multiple welding seams are recognized and corrected, automatic welding of the welding robot in multi-layer and multi-pass welding operation can be achieved, manual participation is reduced, and the welding efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of welding robots, and in particular to a multi-layer and multi-pass weld feature recognition method based on active vision. Background Art

[0002] With the rapid development of my country's infrastructure construction, large-scale infrastructure such as cross-sea bridges and high-speed railways have put forward higher requirements for welding technology.

[0003] Multi-layer, multi-pass welding is widely used in this type of project because it ensures high-quality welding of thick and heavy plates. However, this method creates complex welds with irregular shapes, multiple geometric variations, and overlapping welds, making them difficult for welding robots to identify.

[0004] Traditional methods usually require manual assistance, which reduces automation efficiency.

[0005] Therefore, there is an urgent need for a technical solution that can help welding robots identify the characteristics of multi-layer and multi-pass welds. Summary of the Invention

[0006] The purpose of the present invention is to provide a multi-layer and multi-pass weld feature recognition method based on active vision to solve the technical problems raised in the background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a multi-layer and multi-pass weld feature recognition method based on active vision, comprising at least the following steps:

[0008] S1: Acquire laser images through active vision;

[0009] S2: Extract the weld geometry features based on the centerline extraction and correction technology;

[0010] S3: Combine corner point recognition and correction technology to locate welding points;

[0011] S4: Determine the welding process of multi-layer and multi-pass welding, which is divided into base welding, filling welding and cap welding. Apply steps S1-S3, and selectively adopt centerline correction and corner point recognition correction methods according to the welding characteristics of each layer.

[0012] Furthermore, the process of applying S1 to S3 to the bottom welding in S4 includes at least the following steps:

[0013] The centerline outlier correction method is used to improve the accuracy of centerline recognition by detecting the position of outliers and correcting their coordinates;

[0014] In the case of strong light interference or laser transmitter offset during welding, which causes discontinuities in the centerline recognition results, a centerline repair method is used to ensure that the repaired centerline retains the geometric characteristics of the weld through linear interpolation of the discontinuous area.

[0015] Finally, the corner points where the centerline shape changes significantly are detected by a derivation algorithm and used as welding points. Furthermore, the centerline outlier correction method used in the root welding includes at least the following steps:

[0016] First, a voting algorithm is used to detect the densest area of ​​laser stripes, and outliers are determined based on the set threshold. The algorithm formula is as follows:

[0017]

[0018] Where: R(x,y) represents the weight score of pixel (x,y), δ(x i ,yy i ) represents the point (x,y) and (x i ,y i ) distance function, ω i is the weight coefficient, N is the number of sample points, R median is the median of all point weight scores, and T is the threshold;

[0019] Then, the position of the outlier is corrected according to the center line coordinates around the outlier through the median fitting method. The outlier correction algorithm is as follows:

[0020]

[0021] Where: y left and y right It is the y coordinate of the nearest non-outlier point on the left and right sides of the outlier. If the difference between the left and right y coordinates is greater than the threshold K, only the left coordinate is used.

[0022] Furthermore, the bottom welding centerline repair method includes at least the following steps:

[0023] First calculate the distance between adjacent points on the laser center line. If d i,i+1 If the set threshold is exceeded, it will be identified as discontinuous.

[0024]

[0025] Then, linear interpolation is performed on the discontinuous areas.

[0026]

[0027] Among them: k=i+1,…,j-1 is the interpolation point of the discontinuous interval, i and j are the starting point and end point of the discontinuous area respectively.

[0028] Furthermore, the process of applying S1 to S3 in the filling welding in S4 includes at least the following steps:

[0029] First, the centerline outlier correction and discontinuous centerline correction are performed on the weld laser image in the same way as in the root welding process.

[0030] Afterwards, considering that the slight change in the centerline geometry caused by the irregular shape of the previous layer of welds will affect the recognition accuracy of subsequent corner points, a centerline smoothing method is used to ensure that the centerline only retains the most critical weld geometric features;

[0031] Subsequently, in the corner point identification of the filling weld, considering that the overly obvious geometric feature mutations caused by factors such as weld overlap will be retained, resulting in recognition errors in corner point identification, a corner point correction method is adopted, which includes initial corner point positioning and corner point positioning during welding.

[0032] Furthermore, the centerline smoothing method of the fill weld comprises at least the following steps:

[0033] First, a small window of fixed size is generated with each point in the centerline image as the center;

[0034] Then, the median of the vertical coordinates and the absolute median difference of the points in the window are calculated as reference standards for determining whether the point is offset;

[0035] If the difference between the vertical coordinate of any point in the window and the vertical median coordinate is less than the set threshold, it is considered a valid point;

[0036] If it is greater than the set threshold, its vertical coordinate is replaced by the vertical median. The algorithm equation is as follows:

[0037]

[0038] M i =Median(y j ),j∈W i (7)

[0039] MAD i =Median(|y j -M i |),j∈W i (8)

[0040] Where: W i (x i ,y i ) is the center line point set within the window generated by the center, M i and MAD iare the median and absolute median difference of the y coordinates of the centerline point set in the window, respectively, and the threshold T = 3·C·MAD i , scaling factor C = 1.4826.

[0041] Furthermore, the initial corner point in the initial corner point positioning refers to the starting position of the welding gun during each layer of welding, corresponding to the first geometric feature change point of the welding in the groove. The initial corner point positioning is based on the known groove information and combined with the corner point position of the previous weld to construct a numerical model for welding prediction, which is used to infer the corner point position of each subsequent filler weld;

[0042] The thickness of each weld layer is measured by infrared ranging to detect the height difference before and after welding, and the groove angle is determined according to the welding conditions;

[0043] The initial corner point positioning comprises at least the following steps:

[0044] First, the horizontal position coordinates of the corner points are calculated based on the numerical model.

[0045] x pred =x i ±Δx (9)

[0046] Δx=h·tanα (10)

[0047] Where: x pred is the predicted initial corner point horizontal coordinate, x i is the horizontal coordinate of the initial corner point of the previous layer of weld, h is the weld height, and α is the groove inclination angle;

[0048] Secondly, the corner point positions are preliminarily calculated using the corner point recognition algorithm:

[0049]

[0050] Where: k i For point (x i ,y i ) at the curvature, if k i If it is greater than the set threshold, it is determined to be a corner point;

[0051] Then, the predicted corner point position is compared with the predicted corner point position calculated by the numerical model. When the difference between the recognized position and the calculated position is less than the set threshold, the corner point recognition position is valid. If it is greater than the set threshold, it is considered that the corner point recognition is offset. The initial corner point positioning is performed according to the predicted position. The algorithm formula is as follows:

[0052]

[0053] Where: x pred is the horizontal coordinate of the predicted initial corner point, x initis the horizontal coordinate of the identified initial corner point, Δx is the set offset threshold;

[0054] Finally, the initial corner point of the filler weld of this layer is used as a reference to predict the initial corner point of the next layer of welding.

[0055] Furthermore, the corner point positioning during welding includes at least the following steps:

[0056] First, on the basis of determining the initial welding corner point of each weld, the coordinate positions of the two corner points are recorded simultaneously;

[0057] Then, after obtaining the data of the next pair of corner points, compare them with the coordinates of the initial corner point. When the offset of the two corner points on the horizontal coordinate relative to the initial corner point is less than the threshold, the corner point is determined to be valid;

[0058] For cases where the value is greater than the threshold or no corner point can be detected, the horizontal coordinate of the initial corner point will be used as the corrected position of the corner point. The algorithm formula is as follows:

[0059]

[0060] Among them, x i is the horizontal coordinate of the corner point detected during welding, x i-1 is the horizontal coordinate of the corner point during the previous welding, and Δx is the set offset threshold.

[0061] Furthermore, the cap weld is the last layer of a multi-layer multi-pass weld, and the welding conditions are the same as those of the fill weld, so the same weld feature recognition process as that of the fill weld is adopted.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] The present invention acquires laser images through active vision, adopts centerline extraction and correction technology to extract weld geometric features, and combines corner point recognition and correction technology to locate welding points. The welding process is divided into base welding, filling welding and cover welding. Each layer of welding selectively adopts centerline correction and corner point recognition and correction methods according to welding characteristics, thereby realizing laser centerline geometric feature recognition and correction of multi-pass welds, and can realize automated welding of welding robots in multi-layer and multi-pass welding operations, reduce manual participation and improve welding efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0065] Figure 1This is a schematic diagram of the process of the multi-layer and multi-pass weld feature recognition technology of the present invention;

[0066] Figure 2 This is a schematic diagram of the multi-layer multi-pass welding process of the present invention;

[0067] Figure 3 This is a schematic diagram of the centerline correction technology of the present invention;

[0068] Figure 4 Schematic diagram of the corner point recognition and correction technology of the present invention;

[0069] Figure 5 This is a schematic diagram of centerline outlier correction in the present invention;

[0070] Figure 6 This is a schematic diagram of the discontinuous centerline correction of the present invention;

[0071] Figure 7 This is a schematic diagram of the centerline smoothing process of the present invention;

[0072] Figure 8 This is a schematic diagram of the initial corner point identification and correction of the present invention;

[0073] Figure 9 This is a schematic diagram of corner point identification and correction during welding according to the present invention;

[0074] Figure 10 Schematic diagram of corner point identification and correction according to the present invention. DETAILED DESCRIPTION

[0075] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0076] See also Figure 1-Figure 4 A multi-layer and multi-pass weld feature recognition method based on active vision includes at least the following steps:

[0077] S1: Acquire laser images through active vision;

[0078] S2: Extract the weld geometry features based on the centerline extraction and correction technology;

[0079] S3: Combine corner point recognition and correction technology to locate welding points;

[0080] S4: Determine the welding process of multi-layer and multi-pass welding. The welding process is divided into base welding, filling welding and cover welding. Apply steps S1-S3. According to the welding characteristics of each layer, selectively adopt the center line correction and corner point recognition correction methods.

[0081] The process of applying S1-S3 to the root welding in S4 includes at least the following steps:

[0082] The centerline outlier correction method is used to improve the accuracy of centerline recognition by detecting the position of outliers and correcting their coordinates;

[0083] In the case of strong light interference or laser transmitter offset during welding, which causes discontinuities in the centerline recognition results, a centerline repair method is used to ensure that the repaired centerline retains the geometric characteristics of the weld through linear interpolation of the discontinuous area.

[0084] Finally, the corners where the centerline shape changes significantly are detected by the derivation algorithm and used as welding points

[0085] The centerline outlier correction method used in the root pass welding includes at least the following steps:

[0086] First, a voting algorithm is used to detect the densest area of ​​laser stripes, and outliers are determined based on the set threshold. The algorithm formula is as follows:

[0087]

[0088] Where: R(x,y) represents the weight score of pixel (x,y), δ(x i ,yy i ) represents the point (x,y) and (x i ,y i ) distance function, ω i is the weight coefficient, N is the number of sample points, R median is the median of all point weight scores, and T is the threshold;

[0089] The coefficient ω is graded based on the distance calculated by the distance function. A greater distance gives a larger weight coefficient, indicating a more significant offset relative to the rest of the points. R(x,y) is then compared to the median weight. If the weight is significantly greater than the median (i.e., outside the threshold T), it is considered an offset.

[0090] Then, the position of the outlier is corrected according to the center line coordinates around the outlier through the median fitting method. The outlier correction algorithm is as follows:

[0091]

[0092] Where: y left and y right Is the y coordinate of the nearest non-outlier point on the left and right sides of the outlier. If the difference between the left and right y coordinates is greater than the threshold K, only the left coordinate is used. The correction effect is as follows Figure 5 shown.

[0093] The root weld centerline repair method includes at least the following steps:

[0094] Firstly, the distance between adjacent points on the laser center line is calculated, and if d i,i+1 If it exceeds the set threshold, it is identified as an interruption,

[0095]

[0096] Then, linear interpolation is performed on the interruption area,

[0097]

[0098] Where: k = i + 1, …, j - 1 is the interpolation point of the interruption interval, i and j are the start and end points of the interruption area respectively. The correction effect is shown in Figure 6 .

[0099] The filling welding in S4 applies the processes of S1-S3, including at least the following steps:

[0100] Firstly, the same as the process of primer welding, the laser image of the weld is sequentially processed by center line outlier correction and interruption center line correction;

[0101] Then, considering that the slight change in the center line geometry caused by the irregular shape of the previous layer of welds will affect the identification accuracy of the subsequent corner points, a center line smoothing method is used to ensure that the center line only retains the most critical weld geometry features;

[0102] Then, in the corner point identification of the filling welding, considering that the too obvious geometric feature mutation caused by factors such as weld overlap will be retained, resulting in identification error in corner point identification, a corner point correction method is used for processing, which includes initial corner point positioning and welding corner point positioning.

[0103] The center line smoothing method of the filling welding includes at least the following steps:

[0104] First, take each point in the center line image as the center to generate a small window of a fixed size;

[0105] Then, calculate the median and absolute median deviation of the vertical coordinates of the points in the window as the reference standard for determining whether the point is offset;

[0106] If the difference between the vertical coordinate of any point in the window and the median vertical coordinate is less than the set threshold, it is identified as a valid point;

[0107] If it is greater than the set threshold, replace its vertical coordinate with the median vertical coordinate, and the algorithm equation is as follows,

[0108]

[0109] M i = Median(yj ),j∈W i (7)

[0110] MAD i =Median(|y j -M i |),j∈W i (8)

[0111] Where: W i (x i ,y i ) is the center line point set within the window generated by the center, M i and MAD i are the median and absolute median difference of the y coordinates of the centerline point set in the window, respectively, and the threshold T = 3·C·MAD i , scaling factor C = 1.4826. The correction effect is as follows Figure 7 shown.

[0112] The initial corner point in the initial corner point positioning refers to the starting position of the welding gun during each layer of welding, corresponding to the first geometric feature change point of the welding in the groove. The initial corner point positioning is based on the known groove information and the corner point position of the previous weld to construct a numerical model for welding prediction, which is used to infer the corner point position of each subsequent filler weld;

[0113] The thickness of each weld layer is measured by infrared ranging to detect the height difference before and after welding, and the groove angle is determined according to the welding conditions;

[0114] See Figure 8 , the initial corner point positioning includes at least the following steps:

[0115] First, the horizontal position coordinates of the corner points are calculated based on the numerical model.

[0116] x pred =x i ±Δx(9)

[0117] Δx=h·tanα(10)

[0118] Where: x pred is the horizontal coordinate of the predicted initial corner point, x i is the horizontal coordinate of the initial corner point of the previous layer of weld, h is the weld height, and α is the groove inclination angle;

[0119] Secondly, the corner point positions are preliminarily calculated using the corner point recognition algorithm:

[0120]

[0121] Where: k i For point (x i ,yi ) at the curvature, if k i If it is greater than the set threshold, it is determined to be a corner point;

[0122] Then, the predicted corner point position is compared with the predicted corner point position calculated by the numerical model. When the difference between the recognized position and the calculated position is less than the set threshold, the corner point recognition position is valid. If it is greater than the set threshold, it is considered that the corner point recognition is offset. The initial corner point positioning is performed according to the predicted position. The algorithm formula is as follows:

[0123]

[0124] Where: x pred is the predicted initial corner point horizontal coordinate, x init is the horizontal coordinate of the identified initial corner point, Δx is the set offset threshold;

[0125] Finally, the initial corner point of the filler weld of this layer is used as a reference to predict the initial corner point of the next layer of welding.

[0126] See Figure 9 , the corner point positioning during welding includes at least the following steps:

[0127] First, on the basis of determining the initial welding corner point of each weld, the coordinate positions of the two corner points are recorded simultaneously;

[0128] Then, after obtaining the data of the next pair of corner points, compare them with the coordinates of the initial corner point. When the offset of the two corner points on the horizontal coordinate relative to the initial corner point is less than the threshold, the corner point is determined to be valid;

[0129] For cases where the value is greater than the threshold or no corner point can be detected, the horizontal coordinate of the initial corner point will be used as the corrected position of the corner point. The algorithm formula is as follows:

[0130]

[0131] Among them, x i is the horizontal coordinate of the corner point detected during welding, x i-1 is the horizontal coordinate of the corner point during the previous welding, and Δx is the set offset threshold;

[0132] The above steps are then repeated to complete the entire welding task. The correction effect is as follows: Figure 10 shown.

[0133] Cap welding is the last layer of multi-layer multi-pass welding. The welding conditions are the same as those of filler welding, so the same weld feature recognition process as that of filler welding is adopted.

[0134] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A multi-layer and multi-pass weld feature recognition method based on active vision, characterized by: At least the following steps are included: S1: Acquire laser images through active vision; S2: Extract the weld geometry features based on the centerline extraction and correction technology; S3: Combine corner point recognition and correction technology to locate welding points; S4: Determine the welding process of multi-layer and multi-pass welding, which is divided into base welding, filling welding and cap welding. Apply steps S1-S3, and selectively adopt centerline correction and corner point recognition correction methods according to the welding characteristics of each layer.

2. The method for multi-layer and multi-pass weld feature recognition based on active vision according to claim 1, characterized in that: The process of applying S1-S3 to the bottom welding in S4 includes at least the following steps: The centerline outlier correction method is used to improve the accuracy of centerline recognition by detecting the position of outliers and correcting their coordinates; In the case of strong light interference or laser transmitter offset during welding, which causes discontinuities in the centerline recognition results, a centerline repair method is used to ensure that the repaired centerline retains the geometric characteristics of the weld through linear interpolation of the discontinuous area. Finally, the corner points where the centerline shape changes significantly are detected through the derivation algorithm and are used as welding points.

3. The method for multi-layer and multi-pass weld feature recognition based on active vision according to claim 2, characterized in that: The centerline outlier correction method used in the bottom welding process comprises at least the following steps: First, a voting algorithm is used to detect the densest area of ​​laser stripes, and outliers are determined based on the set threshold. The algorithm formula is as follows: Where: R(x,y) represents the weight score of pixel (x,y), δ(x i ,yy i ) represents the point (x,y) and (x i ,y i ) distance function, ω i is the weight coefficient, N is the number of sample points, R median is the median of all point weight scores, and T is the threshold; Then, the position of the outlier is corrected according to the center line coordinates around the outlier through the median fitting method. The outlier correction algorithm is as follows: Where: y left and y right It is the y coordinate of the nearest non-outlier point on the left and right sides of the outlier. If the difference between the left and right y coordinates is greater than the threshold K, only the left coordinate is used.

4. The method for multi-layer and multi-pass weld feature recognition based on active vision according to claim 3, characterized in that: The bottom welding center line repair method comprises at least the following steps: First calculate the distance between adjacent points on the laser center line. If d i,i+1 If the set threshold is exceeded, it will be identified as discontinuous. Then, linear interpolation is performed on the discontinuous areas. Among them: k=i+1,…,j-1 is the interpolation point of the discontinuous interval, i and j are the starting point and end point of the discontinuous area respectively.

5. The method for multi-layer and multi-pass weld feature recognition based on active vision according to claim 4, characterized in that: The process of applying S1-S3 to the filling welding in S4 includes at least the following steps: First, the centerline outlier correction and discontinuous centerline correction are performed on the weld laser image in the same way as in the root welding process. Afterwards, considering that the slight change in the centerline geometry caused by the irregular shape of the previous layer of welds will affect the recognition accuracy of subsequent corner points, a centerline smoothing method is used to ensure that the centerline only retains the most critical weld geometric features; Subsequently, in the corner point identification of the filling weld, considering that the overly obvious geometric feature mutations caused by factors such as weld overlap will be retained, resulting in recognition errors in corner point identification, a corner point correction method is adopted, which includes initial corner point positioning and corner point positioning during welding.

6. The method for multi-layer and multi-pass weld feature recognition based on active vision according to claim 5, characterized in that: The centerline smoothing method of the fill weld comprises at least the following steps: First, a small window of fixed size is generated with each point in the centerline image as the center; Then, the median of the vertical coordinates and the absolute median difference of the points in the window are calculated as reference standards for determining whether the point is offset; If the difference between the vertical coordinate of any point in the window and the vertical median coordinate is less than the set threshold, it is considered a valid point; If it is greater than the set threshold, its vertical coordinate is replaced by the vertical median. The algorithm equation is as follows: M i =Median(y j ),j∈W i (7) MAD i =Median(|y j -M i |),j∈W i (8) Where: W i (x i ,y i ) is the center line point set within the window generated by the center, M i and MAD i are the median and absolute median difference of the y coordinates of the centerline point set in the window, respectively, and the threshold T = 3·C·MAD i , scaling factor C = 1.4826.

7. The method for multi-layer and multi-pass weld feature recognition based on active vision according to claim 5, characterized in that: The initial corner point in the initial corner point positioning refers to the starting position of the welding gun during each layer of welding, corresponding to the first geometric feature change point of the welding in the groove. The initial corner point positioning is based on the known groove information and combined with the corner point position of the previous weld to construct a numerical model for welding prediction, which is used to infer the corner point position of each subsequent filler weld; The thickness of each weld layer is measured by infrared ranging to detect the height difference before and after welding, and the groove angle is determined according to the welding conditions; The initial corner point positioning comprises at least the following steps: First, the horizontal position coordinates of the corner points are calculated based on the numerical model. x pred =x i ±Δx(9) Δx=h·tanα (10) Where: x pred is the horizontal coordinate of the predicted initial corner point, x i is the horizontal coordinate of the initial corner point of the previous layer of weld, h is the weld height, and α is the groove inclination angle; Secondly, the corner point positions are preliminarily calculated using the corner point recognition algorithm: Where: k i For point (x i ,y i ) at the curvature, if k i If it is greater than the set threshold, it is determined to be a corner point; Then, the predicted corner point position is compared with the predicted corner point position calculated by the numerical model. When the difference between the recognized position and the calculated position is less than the set threshold, the corner point recognition position is valid. If it is greater than the set threshold, it is considered that the corner point recognition is offset. The initial corner point positioning is performed according to the predicted position. The algorithm formula is as follows: Where: x pred is the horizontal coordinate of the predicted initial corner point, x init is the horizontal coordinate of the identified initial corner point, Δx is the set offset threshold; Finally, the initial corner point of the filler weld of this layer is used as a reference to predict the initial corner point of the next layer of welding.

8. The method for multi-layer and multi-pass weld feature recognition based on active vision according to claim 5, characterized in that: The corner point positioning during welding comprises at least the following steps: First, on the basis of determining the initial welding corner point of each weld, the coordinate positions of the two corner points are recorded simultaneously; Then, after obtaining the data of the next pair of corner points, compare them with the coordinates of the initial corner point. When the offset of the two corner points on the horizontal coordinate relative to the initial corner point is less than the threshold, the corner point is determined to be valid; For cases where the value is greater than the threshold or no corner point can be detected, the horizontal coordinate of the initial corner point will be used as the corrected position of the corner point. The algorithm formula is as follows: Among them, x i is the horizontal coordinate of the corner point detected during welding, x i-1 is the horizontal coordinate of the corner point during the previous welding, and Δx is the set offset threshold.

9. The method for multi-layer and multi-pass weld feature recognition based on active vision according to claim 5, characterized in that: The cap weld is the last layer of a multi-layer, multi-pass weld, and the welding conditions are the same as those of the fill weld, so the same weld feature recognition process as that of the fill weld is adopted.

Images