Weld seam scanning and tracking processing method

Abstract

A weld seam scanning and tracking processing method is provided. Based on the laser scanning of the weld seam, the scanned laser data is converted into three-dimensional space point cloud data using the robot coordinate transformation relationship, and the spatial point cloud structure is used to remove noise in the single measurement data in a small range, fill or interpolate the point cloud data of the missing position, and more accurately reflect the weld seam morphology. A unique weld seam feature library is established, and the weld seam is converted from 2D image data to more intuitive 3D point cloud data. This solves the problem of inaccurate measurement results caused by optical lens distortion of the laser or perspective camera during weld seam tracking. After multiple weld bead feature recognition and correction of the segmented scanned data, the accurate weld seam morphology characteristics and position are determined through weld bead information reconstruction, eliminating the problem of poor accuracy of single measurement results. The coherent measurement processing method can accurately reflect the true situation of the weld bead and has high weld seam measurement accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of laser weld tracking, and in particular relates to a weld scanning tracking processing method. Background Art

[0002] Weld seam scanning and tracking systems typically collect weld data, track welds, calculate weld position information, and calculate the welding process required to meet the needs of automated welding. Existing weld seam tracking systems primarily rely on lasers or binocular cameras installed at the end of industrial robots. As the robot moves, they directly calculate and process the weld seam to obtain the real-time weld seam position. Conventional weld seam calculation schemes rely solely on single measurement results or simply perform histogram statistical averaging of multiple measurement results. This presents several issues: ① The weld seam tracking system can suffer from inaccurate measurement results at different scanning angles due to optical lens distortion of the laser or visual camera; ② Directly calculating weld seam information from a single measurement result can lead to inaccurate measurement results due to single measurement inaccuracies or low signal-to-noise ratios; and ③ The histogram statistical averaging of multiple measurement results cannot accurately reflect the actual weld conditions due to the independence of the measured cross-sections of the weld bead. Therefore, improvements are necessary to address these issues. Summary of the Invention

[0003] The technical problem solved by the present invention is as follows: A weld scanning and tracking processing method is provided. Based on the weld scanning by the laser at the end of the robot, the scanned laser data is converted into three-dimensional space point cloud data using the robot coordinate transformation relationship, and the spatial point cloud structure is used to remove the noise in the single measurement data in a small range, fill or interpolate the point cloud data of the lost position, more accurately reflect the weld morphology, establish a unique weld feature library, and convert the weld from 2D image data to more intuitive 3D point cloud data, which solves the problem of inaccurate measurement results caused by optical lens distortion of the laser or perspective camera during weld tracking. After the weld features are identified and corrected for the segmented scanned data multiple times, the accurate weld morphology features and position are determined by weld information reconstruction, eliminating the problem of poor accuracy of single measurement results. The coherent measurement processing method can accurately reflect the true situation of the weld and has high weld measurement accuracy.

[0004] The technical solution adopted by the present invention is a weld seam scanning and tracking processing method, which includes the following steps:

[0005] 1) Weld seam pre-scanning: According to the preset trajectory, the laser installed at the end of the robot first scans the laser data of a small section of the weld seam. The laser data is 2D point data formed in the X and Z directions. According to the robot coordinate conversion relationship, all scanned laser data are converted to a spatial 3D coordinate system. On the 3D spatial data, the above process is repeated until the laser data of the first V-shaped line segment corresponding to the weld seam is found, which is the starting position of the weld seam. In the laser data of the first V-shaped line segment, the center point position (X center ,Y center ,Z center ), according to the preset starting position of the laser weld scanning (X start , Y start , Z start ), calculate the difference between the center point position and the preset starting position through the distance formula between two points in space, and obtain the corrected position of the laser on the robot (ΔX=X start -X center ,ΔY=Y start -Y center ,ΔZ=Z start -Z center ), then adjust the laser to the corrected position and start scanning the weld after correction;

[0006] 2) Weld seam scanning: Repeat step 1), and at each laser correction position, collect the weld seam raw data according to the preset robot scanning trajectory. Use the robot coordinate transformation relationship in step 1) to convert all weld seam raw data into 3D spatial data, establish a 3D point cloud dataset Ω0, and complete the scanning process;

[0007] 3) Weld space data processing, the specific steps are as follows:

[0008] a. First, perform denoising, cropping and filling preprocessing on the original data;

[0009] b. Weld bead feature recognition: Based on the input weld surrounding morphological features including the number of weld scanning planes, weld groove shape, weld angle, weld depth, and weld width, RANSAC point cloud recognition is performed on the weld raw data to obtain valid weld data. The specific steps are as follows:

[0010] b1 In the overall 3D spatial data after preprocessing in step a, first, according to the weld angle, find the area in the overall 3D spatial data that meets the weld angle in the X direction along the Y direction. According to the weld width, find the area in the overall 3D spatial data that meets the weld angle in the Z direction along the Y direction. The area intercepted in the X and Z directions is the spatial area where the weld is located. According to the weld groove shape and the weld edge data information obtained by scanning, the weld edge is located in the spatial area where the weld is located. Linear fitting is performed in 3D space to determine the linear equations of the two weld edges, and the direction vectors ν1 and ν2 of the two weld edge straight lines are calculated. According to the direction vectors ν1 and ν2, plane fitting is performed on the planes on both sides of the weld respectively to determine the two plane equations A11 and A21, and then the normal vectors ν3 and ν4 of the two plane equations A11 and A21 are calculated;

[0011] b2 sets the sampling area along the traveling direction Y of the weld scanning, with a distance of 10mm in the Y-axis direction as a sampling area segment, and determines the two spatial plane equations P11 and P21 in the sampling area through the plane normal vectors ν3 and ν4 on both sides of the weld. Then, the point cloud data located in the two spatial plane equations P11 and P21 are classified in the sampling area through the spatial distance calculation formula. The intersecting spatial line equation Q1 is obtained from the two spatial plane equations P11 and P21, and all point cloud data in the cylindrical space area H1 with the spatial line as the axis and a radius of 1.5mm are obtained. The average direction vectors of v1 and v2 are used to perform a quadratic fitting line in the above cylindrical space area H1, and all real point sets in the cylindrical space area H2 with the quadratic fitting line as the axis and a radius of 0.5mm are obtained. The real point set is the center point of the bottom of all welds in the sampling area;

[0012] b3 Repeat step b2 until the Y-axis traversal is completed, and the weld bottom center point dataset Ω1 is obtained. The robot will track the weld based on the dataset Ω1;

[0013] 4) Weld bead information reconstruction: poll the point p in the weld dataset Ω1 in sequence i ,i=1,…,end, take this point as the center point, take the set XOZ plane as the reference welding plane, and establish p i The spatial plane equation P31 in the reference welding plane is calculated, and the spatial distance between the original data and the spatial plane equation P31 is calculated to obtain all points on the spatial plane equation P31, that is, the points constitute the welding section plane Pn1;

[0014] 5) Repeat step 4) and after traversing Ω1, the welding cross-section plane Pn obtained each time is iA new weld dataset Ω2 is formed by repeating b2 in step 3) above to obtain the plane equations P1 on both sides of the welds. i and P2 i This constitutes the weld bead dataset Ω3;

[0015] 6) Weld measurement: Take the data set Ω2 as the calculation data set, and at each section plane Pn i The weld depth is calculated using the weld plane equation in Ω3 as the difference between the highest and lowest points on each side of the weld plane in the Z direction, and the weld width is calculated using the difference between the two end points in the X direction in the Ω2 data set, thus completing the weld measurement function.

[0016] In the above step 1), the robot coordinate transformation relationship is as follows:

[0017]

[0018] Among them, R is the rotation matrix from the manually calibrated laser to the robot, T is the translation matrix from the manually calibrated laser to the robot, X L and Z L The X direction and Z direction data are obtained from the laser data respectively. w ,Y w ,Z w The data corresponding to the X, Y, and Z directions in 3D space; all scanned laser data are converted to a spatial 3D coordinate system.

[0019] In step a of step 3) above, the steps of denoising, cropping, and padding the original data are as follows:

[0020] a1 uses a median filter and sets a sliding window size of 5-10 points to filter out salt and pepper noise interference;

[0021] a2 uses a low-pass filter to convert the spatial 3D data into the frequency domain, sets the filter cutoff frequency to 1KHz, filters out the high-frequency noise in the frequency domain data, and converts it back to the spatial domain data;

[0022] a3 uses the K-means unsupervised clustering method, sets the number of categories to 3-8, performs cluster classification, retains the first three categories with the largest number of point clouds, and eliminates the categories with small number of point clouds;

[0023] a4 uses the nearest neighbor interpolation method and sets the sliding window size to 20-50 point clouds to fill the vacant areas.

[0024] In step 6) above, the specific calculation process for calculating the weld depth value using the difference in the Z direction between the highest point and the lowest point in the plane on each side of the weld is as follows: the average value of the difference in the Z direction between the highest point and the lowest point of the weld plane on one side of the weld and the difference in the Z direction between the highest point and the lowest point of the weld plane on the other side is the weld depth.

[0025] In the above step 4), the welding cross-section plane is a cross section of the welding plane in the original data space.

[0026] The advantages of the present invention compared with the prior art are:

[0027] 1. Based on the laser scanning of the weld by the end-user robot, this technical solution uses the robot coordinate transformation relationship to convert the scanned laser data into three-dimensional spatial point cloud data. The spatial point cloud structure is then used to remove noise from single measurement data within a small range, fill in or interpolate point cloud data at missing locations, and more accurately reflect the weld morphology. A unique weld feature library is established, and the weld is converted from 2D image data to more intuitive 3D point cloud data. This solves the problem of inaccurate measurement results caused by optical lens distortion of the laser or view camera during weld tracking.

[0028] 2. This technical solution identifies and corrects the weld features of the segmented scanned data multiple times, and then reconstructs the weld information to determine the accurate weld morphology and position, eliminating the problem of poor accuracy of single measurement results. The continuous measurement processing method can accurately reflect the actual situation of the weld, and the weld measurement accuracy is high.

[0029] 3. This technical solution solves the measurement error caused by the mismatch between the measurement plane of the line scanning laser and the actual welding plane, and has high measurement accuracy;

[0030] 4. This technical solution adds a segmented sampling method to the calculation of the weld direction of the total data sample. It refers to the direction vector of the total sample and calculates the actual sample situation of each segment, which is more in line with the actual weld situation, can better adapt to the non-linear morphology of the weld on the real workpiece, and has a more accurate tracking effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a simplified structural diagram of the weld bead within the spatial region where the weld seam is located as determined in step 3) of the present invention. DETAILED DESCRIPTION

[0032] The following is a combination of the embodiments of the present invention Figure 1The technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0033] It should be noted that, as used herein, the terms "comprise," "include," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article, or apparatus comprising the elements.

[0034] The weld seam scanning and tracking processing method includes the following steps:

[0035] 1) Weld seam pre-scanning: According to the preset trajectory, the laser installed at the end of the robot first scans the laser data of a small section of the weld seam. The laser data is 2D point data formed in the X and Z directions. According to the robot coordinate conversion relationship, all scanned laser data are converted to a spatial 3D coordinate system. On the 3D spatial data, the above process is repeated until the laser data of the first V-shaped line segment corresponding to the weld seam is found, which is the starting position of the weld seam. In the laser data of the first V-shaped line segment, the center point position (X center ,Y center ,Z center ), according to the preset starting position of the laser weld scanning (X start , Y start , Z start ), calculate the difference between the center point position and the preset starting position through the distance formula between two points in space, and obtain the corrected position of the laser on the robot (ΔX=X start -X center ,ΔY=Y start -Y center ,ΔZ=Z start -Z center ), then adjust the laser to the corrected position and start scanning the weld after correction;

[0036] Specifically, the robot coordinate transformation relationship is as follows:

[0037]

[0038] Among them, R is the rotation matrix from the manually calibrated laser to the robot, T is the translation matrix from the manually calibrated laser to the robot, X L and Z L The X direction and Z direction data are obtained from the laser data respectively. w ,Y w ,Z w The data corresponding to the X, Y, and Z directions in 3D space; all scanned laser data are converted to a spatial 3D coordinate system.

[0039] 2) Weld seam scanning: Repeat step 1), and at each laser correction position, collect the weld seam raw data according to the preset robot scanning trajectory. Use the robot coordinate transformation relationship in step 1) to convert all weld seam raw data into 3D spatial data, establish a 3D point cloud dataset Ω0, and complete the scanning process;

[0040] 3) Weld space data processing, the specific steps are as follows:

[0041] a. First, perform denoising, cropping, and padding preprocessing on the original data. The specific steps for denoising, cropping, and padding preprocessing on the original data are as follows:

[0042] a1 uses a median filter and sets a sliding window size of 5-10 points to filter out salt and pepper noise interference;

[0043] a2 uses a low-pass filter to convert the spatial 3D data into the frequency domain, sets the filter cutoff frequency to 1KHz, filters out the high-frequency noise in the frequency domain data, and converts it back to the spatial domain data;

[0044] a3 uses the K-means unsupervised clustering method, sets the number of categories to 3-8, performs cluster classification, retains the first three categories with the largest number of point clouds, and eliminates the categories with small number of point clouds;

[0045] a4 uses the nearest neighbor interpolation method and sets the sliding window size to 20-50 point clouds to fill the vacant areas.

[0046] b. Weld bead feature recognition: Based on the input weld surrounding morphological features including the number of weld scanning planes, weld groove shape, weld angle, weld depth, and weld width, RANSAC point cloud recognition is performed on the weld raw data to obtain valid weld data. The specific steps are as follows:

[0047] b1 In the overall 3D spatial data after preprocessing in step a, first, according to the weld angle, find the area in the overall 3D spatial data that meets the weld angle in the X direction along the Y direction. According to the weld width, find the area in the overall 3D spatial data that meets the weld angle in the Z direction along the Y direction. The area intercepted in the X and Z directions is the spatial region 1 where the weld is located. According to the weld groove shape and the weld edge data information obtained by scanning, the edge of weld bead 2 is located in the spatial region 1 where the weld is located. Linear fitting is performed in 3D space to determine the linear equations of the two edges of weld bead 2, and the direction vectors ν1 and ν2 of the two weld edge straight lines are calculated. According to the direction vectors ν1 and ν2, plane fitting is performed on the planes on both sides of the weld respectively to determine the two plane equations A11 and A21, and then the normal vectors ν3 and ν4 of the two plane equations A11 and A21 are calculated;

[0048] b2 sets the sampling area along the traveling direction Y of the weld scanning, with a distance of 10mm in the Y-axis direction as a sampling area segment, and determines the two spatial plane equations P11 and P21 in the sampling area through the plane normal vectors ν3 and ν4 on both sides of the weld. Then, the point cloud data located in the two spatial plane equations P11 and P21 are classified in the sampling area through the spatial distance calculation formula. The intersecting spatial line equation Q1 is obtained from the two spatial plane equations P11 and P21, and all point cloud data in the cylindrical space area H1 with the spatial line as the axis and a radius of 1.5mm are obtained. The average direction vectors of v1 and v2 are used to perform a quadratic fitting line in the above cylindrical space area H1, and all real point sets in the cylindrical space area H2 with the quadratic fitting line as the axis and a radius of 0.5mm are obtained. The real point set is the center point of the bottom of all welds in the sampling area;

[0049] b3 Repeat step b2 until the Y-axis traversal is completed, and the weld bottom center point dataset Ω1 is obtained. The robot will track the weld based on the dataset Ω1;

[0050] 4) Weld bead information reconstruction: poll the point p in the weld dataset Ω1 in sequence i ,i=1,…,end, take this point as the center point, take the set XOZ plane as the reference welding plane, and establish p i The spatial plane equation P31 in the reference welding plane is calculated, and the spatial distance between the original data and the spatial plane equation P31 is calculated to obtain all points on the spatial plane equation P31, that is, these points constitute the welding cross-sectional plane Pn1; specifically, the welding cross-sectional plane is the cross section of the welding plane in the original data space;

[0051] 5) Repeat step 4) and after traversing Ω1, the welding cross-section plane Pn obtained each time is iA new weld dataset Ω2 is formed by repeating b2 in step 3) above to obtain the plane equations P1 on both sides of the welds. i and P2 i This constitutes the weld bead dataset Ω3;

[0052] 6) Weld measurement: Take the data set Ω2 as the calculation data set, and at each section plane Pn i In the weld plane equation in Ω3, the weld depth value is calculated by the difference between the highest point and the lowest point in the Z direction on each side of the weld plane, and the weld width is calculated by the difference between the two end points in the X direction in the Ω2 data set, thus completing the weld measurement function. Specifically, the specific calculation process for calculating the weld depth value by the difference between the highest point and the lowest point in the Z direction on each side of the weld plane is as follows: the average value of the difference between the highest point and the lowest point in the Z direction on one side of the weld plane and the difference between the highest point and the lowest point in the Z direction on the other side of the weld plane is the weld depth.

[0053] The present invention is based on the laser scanning of the weld at the end of the robot, and uses the robot coordinate transformation relationship to convert the scanned laser data into three-dimensional space point cloud data. The spatial point cloud structure is used to remove the noise in the single measurement data in a small range, fill or interpolate the point cloud data of the missing position, and more accurately reflect the weld morphology. A unique weld feature library is established, and the weld is converted from 2D image data to more intuitive 3D point cloud data. The problem of inaccurate measurement results caused by optical lens distortion of the laser or perspective camera in the weld tracking process is solved. The weld feature recognition is performed on the segmented scanned data multiple times and the weld is detected and analyzed. After correction, the accurate weld morphology characteristics and position are determined through weld bead information reconstruction, eliminating the problem of poor accuracy of single measurement results. The continuous measurement processing method can accurately reflect the actual situation of the weld bead, and the weld measurement accuracy is high. It solves the measurement error caused by the mismatch between the measurement plane of the line scanning laser and the actual welding plane, and the measurement accuracy is high. In the calculation of the weld direction of the total data sample, a segmented sampling method is added, and the direction vector of the total sample is referred to to calculate the actual sample situation of each segment, which is more in line with the actual weld situation, can better adapt to the non-linear morphology of the weld on the real workpiece, and the tracking effect is more accurate.

[0054] 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.

[0055] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. Weld seam scanning and tracking processing method, characterized in that The following steps are involved: 1) Weld seam pre-scanning: According to the preset trajectory, the laser installed at the end of the robot first scans the laser data of a small section of the weld seam. The laser data is 2D point data formed in the X and Z directions. According to the robot coordinate conversion relationship, all scanned laser data are converted to a spatial 3D coordinate system. On the 3D spatial data, the above process is repeated until the laser data of the first V-shaped line segment corresponding to the weld seam is found, which is the starting position of the weld seam. In the laser data of the first V-shaped line segment, the center point position (X center ,Y center ,Z center ), according to the preset starting position of the laser weld scanning (X start , Y start , Z start ), calculate the difference between the center point position and the preset starting position through the distance formula between two points in space, and obtain the corrected position of the laser on the robot (ΔX=X start -X center ,ΔY=Y start -Y center ,ΔZ=Z start -Z center ), then adjust the laser to the corrected position and start scanning the weld after correction; 2) Weld seam scanning: Repeat step 1), and at each laser correction position, collect the weld seam raw data according to the preset robot scanning trajectory. Use the robot coordinate transformation relationship in step 1) to convert all weld seam raw data into 3D spatial data, establish a 3D point cloud dataset Ω0, and complete the scanning process; 3) Weld space data processing, the specific steps are as follows: a. First, perform denoising, cropping and filling preprocessing on the original data; b. Weld bead feature recognition: Based on the input weld surrounding morphological features including the number of weld scanning planes, weld groove shape, weld angle, weld depth, and weld width, RANSAC point cloud recognition is performed on the weld raw data to obtain valid weld data. The specific steps are as follows: b1 In the overall 3D spatial data after preprocessing in step a, first, according to the weld angle, find the area in the overall 3D spatial data that meets the weld angle in the X direction along the Y direction, and according to the weld width, find the area in the overall 3D spatial data that meets the weld angle in the Z direction along the Y direction. The area intercepted in the X and Z directions is the spatial area (1) where the weld is located. According to the weld groove shape and combined with the weld edge data information obtained by scanning, the weld (2) edge is located in the spatial area (1) where the weld is located, and a straight line fitting is performed in the 3D space to determine the straight line equations of the two weld (2) edges, and the direction vectors ν1 and ν2 of the two weld edge straight lines are calculated. According to the direction vectors ν1 and ν2, plane fitting is performed on the planes on both sides of the weld respectively to determine two plane equations A11 and A21, and then the normal vectors ν3 and ν4 of the two plane equations A11 and A21 are calculated; b2 sets the sampling area along the traveling direction Y of the weld scanning, with a distance of 10mm in the Y-axis direction as a sampling area segment, and determines the two spatial plane equations P11 and P21 in the sampling area through the plane normal vectors ν3 and ν4 on both sides of the weld. Then, the point cloud data located in the two spatial plane equations P11 and P21 are classified in the sampling area through the spatial distance calculation formula. The intersecting spatial line equation Q1 is obtained from the two spatial plane equations P11 and P21, and all point cloud data in the cylindrical space area H1 with the spatial line as the axis and a radius of 1.5mm are obtained. The average direction vectors of v1 and v2 are used to perform a quadratic fitting line in the above cylindrical space area H1, and all real point sets in the cylindrical space area H2 with the quadratic fitting line as the axis and a radius of 0.5mm are obtained. The real point set is the center point of the bottom of all welds in the sampling area; b3 Repeat step b2 until the Y-axis traversal is completed, and the weld bottom center point dataset Ω1 is obtained. The robot will track the weld based on the dataset Ω1; 4) Weld bead information reconstruction: poll the point p in the weld dataset Ω1 in sequence i ,i=1,…,end, take this point as the center point, take the set XOZ plane as the reference welding plane, and establish p i The spatial plane equation P31 in the reference welding plane is calculated, and the spatial distance between the original data and the spatial plane equation P31 is calculated to obtain all points on the spatial plane equation P31, that is, the points constitute the welding section plane Pn1; 5) Repeat step 4) and after traversing Ω1, the welding cross-section plane Pn obtained each time is i A new weld dataset Ω2 is formed by repeating b2 in step 3) above to obtain the plane equations P1 on both sides of the welds. i and P2 i This constitutes the weld bead dataset Ω3; 6) Weld measurement: Take the data set Ω2 as the calculation data set, and at each section plane Pn i The weld depth is calculated using the weld plane equation in Ω3 as the difference between the highest and lowest points on each side of the weld plane in the Z direction, and the weld width is calculated using the difference between the two end points in the X direction in the Ω2 data set, thus completing the weld measurement function.

2. The weld seam scanning and tracking method according to claim 1, characterized in that: In the above step 1), the robot coordinate transformation relationship is as follows: Among them, R is the rotation matrix from the manually calibrated laser to the robot, T is the translation matrix from the manually calibrated laser to the robot, X L and Z L The X direction and Z direction data are obtained from the laser data respectively. w ,Y w ,Z w The data corresponding to the X, Y, and Z directions in 3D space; all scanned laser data are converted to a spatial 3D coordinate system.

3. The weld seam scanning and tracking method according to claim 1, characterized in that: In step a of step 3) above, the steps of denoising, cropping, and padding the raw data are as follows: a1 uses a median filter and sets a sliding window size of 5-10 points to filter out salt and pepper noise interference; a2 uses a low-pass filter to convert the spatial 3D data into the frequency domain, sets the filter cutoff frequency to 1KHz, filters out the high-frequency noise in the frequency domain data, and converts it back to the spatial domain data; a3 uses the K-means unsupervised clustering method, sets the number of categories to 3-8, performs cluster classification, retains the first three categories with the largest number of point clouds, and eliminates the categories with small number of point clouds; a4 uses the nearest neighbor interpolation method and sets the sliding window size to 20-50 point clouds to fill the vacant areas.

4. The weld seam scanning and tracking method according to claim 1, characterized in that: In step 6) above, the specific calculation process for calculating the weld depth value using the difference in the Z direction between the highest point and the lowest point in the plane on each side of the weld is as follows: the average value of the difference in the Z direction between the highest point and the lowest point of the weld plane on one side of the weld and the difference in the Z direction between the highest point and the lowest point of the weld plane on the other side is the weld depth.

5. The weld seam scanning and tracking method according to claim 1, characterized in that: In the above step 4), the welding cross-section plane is a cross section of the welding plane in the original data space.

Images