A three-dimensional laser point cloud decluttering method for defect detection in moving steel pipes

Abstract

This invention provides a three-dimensional laser point cloud de-jitter method for defect detection of moving steel pipes, belonging to the field of machine vision inspection technology. The method includes: Step 1, fixing three fixed reference marker points at the beginning, end, and middle of the axial direction of the moving steel pipe surface; collecting original point cloud data of the steel pipe surface containing the marker points; separating the point cloud of the fixed reference marker point region; calculating the geometric center coordinates of the point cloud of each marker point region; and outputting a three-dimensional coordinate set of the marker points; Step 2, based on the original point cloud data, segmenting it along the axial direction of the steel pipe into an original 3D ring point cloud set; and fitting the axis of the ideal motion trajectory of the steel pipe using spatial linear regression based on the three-dimensional coordinate set. This invention can effectively eliminate point cloud distortion caused by the jitter of the moving steel pipe, improve the accuracy of defect detection and dimensional measurement while retaining the true defect characteristics, and provide a reliable three-dimensional point cloud model for steel pipe quality assessment.

Description

Technical Field

[0001] This invention relates to the field of machine vision inspection technology, and in particular to a three-dimensional laser point cloud anti-shake method for detecting defects in moving steel pipes. Background Technology

[0002] In practical applications, especially when performing online inspection of continuously moving steel pipes, there is a significant technical challenge: the mechanical vibration and unstable movement of the steel pipe during transmission (commonly known as "jitter") can cause distortion in the raw 3D point cloud data acquired by the laser scanner.

[0003] Axial movement: Unexpected, slight back-and-forth movement of the steel pipe along its axial direction;

[0004] Radial runout: The central axis of the steel pipe does not move in an ideal straight line in space, but is offset in a plane perpendicular to the transmission direction;

[0005] Uneven rotation: Ideally, the steel pipe rotates at a uniform speed, but in reality, there may be fluctuations in speed or momentary jamming. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a three-dimensional laser point cloud anti-jitter method for defect detection of moving steel pipes, which can compensate for the jitter effect during the movement of the steel pipe.

[0007] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0008] In a first aspect, a three-dimensional laser point cloud decluttering method for detecting defects in moving steel pipes is provided, the method comprising:

[0009] Step 1: Fix three fixed reference markers at the beginning, end and middle of the moving steel pipe surface along the axis. Collect the original point cloud data of the steel pipe surface containing the markers, separate the point cloud of the fixed reference marker area, calculate the geometric center coordinates of the point cloud of each marker area, and output the three-dimensional coordinate set of the markers.

[0010] Step 2: Based on the original point cloud data, segment the steel pipe along its axial direction into an original 3D ring point cloud set; fit the steel pipe's ideal motion trajectory axis using spatial linear regression based on the three-dimensional coordinate set.

[0011] Step 3: Perform axial position calibration on the original 3D loop point cloud, and perform statistical outlier filtering on the calibrated 3D loop to generate an optimized 3D loop point cloud;

[0012] Step 4: Based on the optimized 3D ring point cloud, extract the initial center coordinates of each ring, and project each initial center vertically onto the motion trajectory axis to obtain the corrected center position; translate all the point clouds of each 3D ring along the direction from the initial center to the corrected center position to generate a 3D ring point cloud with corrected center.

[0013] Step 5: Refit the standard circle with the corrected center position as the reference to obtain the radius values ​​of each ring, and calculate the average value of all ring radius values ​​as the global reference radius;

[0014] Step 6: Adjust each point in the corresponding ring along the radial direction from the center of the circle to the corresponding point to the global reference radius distance to generate a contour-standardized 3D ring point cloud. Integrate all contour-standardized 3D ring point clouds to generate a jitter-free 3D point cloud model of the steel pipe.

[0015] In a second aspect, a computing device includes:

[0016] One or more processors;

[0017] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0018] Thirdly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0019] The above-described solution of the present invention has at least the following beneficial effects:

[0020] By eliminating point cloud distortion caused by steel pipe vibration, false detection and missed detection of surface defects can be effectively avoided; precise correction of the center and contour can preserve the local features of real defects on the steel pipe surface (such as pits and protrusions), while removing the overall distortion interference caused by vibration, thus improving the reliability of defect identification.

[0021] After jitter reduction processing, the measured dimensional parameters of the steel pipe, such as straightness, outer diameter, ellipticity, and circumference, are closer to the actual values. The setting of a global reference radius and the contour standardization operation reduce dimensional measurement deviations caused by jitter, providing accurate data support for steel pipe quality assessment. Multi-step point cloud optimization (axial calibration, outlier filtering, center correction, etc.) improves the stability and consistency of the 3D point cloud. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating a three-dimensional laser point cloud anti-shake method for detecting defects in moving steel pipes, provided by an embodiment of the present invention. Detailed Implementation

[0023] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0024] like Figure 1 As shown, an embodiment of the present invention proposes a three-dimensional laser point cloud decluttering method for defect detection in moving steel pipes, the method comprising the following steps:

[0025] Step 1: Fix three fixed reference markers at the beginning, end and middle of the moving steel pipe surface along the axis. Collect the original point cloud data of the steel pipe surface containing the markers, separate the point cloud of the fixed reference marker area, calculate the geometric center coordinates of the point cloud of each marker area, and output the three-dimensional coordinate set of the markers.

[0026] Step 2: Based on the original point cloud data, segment the steel pipe along its axial direction into an original 3D ring point cloud set; fit the steel pipe's ideal motion trajectory axis using spatial linear regression based on the three-dimensional coordinate set.

[0027] Step 3: Perform axial position calibration on the original 3D loop point cloud, and perform statistical outlier filtering on the calibrated 3D loop to generate an optimized 3D loop point cloud;

[0028] Step 4: Based on the optimized 3D ring point cloud, extract the initial center coordinates of each ring, and project each initial center vertically onto the motion trajectory axis to obtain the corrected center position; translate all the point clouds of each 3D ring along the direction from the initial center to the corrected center position to generate a 3D ring point cloud with corrected center.

[0029] Step 5: Refit the standard circle with the corrected center position as the reference to obtain the radius values ​​of each ring, and calculate the average value of all ring radius values ​​as the global reference radius;

[0030] Step 6: Adjust each point in the corresponding ring along the radial direction from the center of the circle to the corresponding point to the global reference radius distance to generate a contour-standardized 3D ring point cloud. Integrate all contour-standardized 3D ring point clouds to generate a jitter-free 3D point cloud model of the steel pipe.

[0031] In this embodiment of the invention, the setting and coordinate extraction of fixed reference markers avoid the deviation of relying solely on dynamic point cloud fitting of the axis, ensuring a more accurate axis for the ideal motion trajectory of the steel pipe; axial position calibration eliminates the axial offset of the point cloud, and statistical outlier filtering removes noise interference. By projecting the initial center of each ring onto the reference axis and translating the point cloud, the problem of center deviation caused by jitter is effectively corrected, unifying the center of each ring onto the ideal axis, laying a spatial consistency foundation for dimensional measurement and defect detection.

[0032] The global reference radius is calculated and the contour is standardized. While preserving the real defects (such as local concavity and convexity), the overall contour distortion caused by jitter (such as ellipticization) is eliminated, making the point cloud closer to the real shape of the steel pipe. The final generated anti-jitter point cloud model reduces the false detection and false detection rate of defects, ensures the accuracy of dimensional measurement data such as straightness and outer diameter, and provides a reliable basis for steel pipe quality assessment.

[0033] In a preferred embodiment of the present invention, step 1 above, which involves fixing three fixed reference marker points at the first, last, and middle ends of the moving steel pipe surface along its axial direction, collecting original point cloud data of the steel pipe surface containing the marker points, separating the point cloud of the fixed reference marker point region, calculating the geometric center coordinates of the point cloud of each marker point region, and outputting a three-dimensional coordinate set of the marker points, may include:

[0034] Step 100: From the raw point cloud data collected by the 3D laser profilometer, identify the point cloud regions corresponding to the three fixed reference markers at the beginning, end and middle of the axis, and extract all the 3D coordinate data within each marker region.

[0035] Step 101: Calculate the average X-axis coordinate, average Y-axis coordinate, and average Z-axis coordinate of all points within each marked point area;

[0036] Step 102: Combine the average values ​​of the X-axis coordinate, Y-axis coordinate, and Z-axis coordinate to form the three-dimensional geometric center coordinates of the corresponding marker points, and output a three-dimensional coordinate set consisting of the geometric center coordinates of the first marker point, the last marker point, and the middle marker point.

[0037] In this embodiment of the invention, a three-dimensional laser profilometer is used to perform a comprehensive scan of the moving steel pipe, acquiring raw point cloud data containing all features of the steel pipe surface (including three fixed reference marker points). This raw data is presented in the form of a large number of discrete three-dimensional coordinate points (X, Y, Z), covering the entire outer surface of the steel pipe. Using preset marker point recognition rules (e.g., differences between the marker point's color, shape, size, etc., and the steel pipe surface), the point cloud regions corresponding to the three fixed reference marker points (head, tail, and middle) are distinguished from the raw point cloud data. For example, if the marker point is a circular patch of a specific color, the point cloud range of each marker point can be accurately located through color threshold filtering and shape recognition algorithms. The three-dimensional coordinate data of all points within each marker point region are extracted, i.e., the X-axis coordinates, Y-axis coordinates, and Z-axis coordinates of each point in the head marker point region, tail marker point region, and middle marker point region are recorded respectively, forming three independent coordinate datasets.

[0038] Step 101: For the three-dimensional coordinate dataset of the first-end marker point region extracted in Step 100, traverse the X-axis coordinates of all points in the region, add up these X-axis coordinate values ​​and divide by the total number of points to obtain the average X-axis coordinate of the first-end marker point region; use the same method to calculate the average Y-axis coordinate and average Z-axis coordinate of all points in the region respectively.

[0039] Perform the same operation as above on the 3D coordinate dataset of the tail marker region: traverse the X-axis coordinates of all points and calculate the average value to obtain the average X-axis coordinate of the tail marker region; similarly, calculate the average Y-axis coordinate and average Z-axis coordinate of the region.

[0040] For the three-dimensional coordinate dataset of the central marked point region, repeat the above calculation process to obtain the average X-axis coordinate, average Y-axis coordinate, and average Z-axis coordinate of the central marked point region.

[0041] Step 102: Combine the average X-axis, average Y-axis, and average Z-axis coordinates of the first marker region calculated in Step 101 to form the three-dimensional geometric center coordinates of the first marker, for example, (X_first_average, Y_first_average, Z_first_average). Similarly, combine the average X-axis, average Y-axis, and average Z-axis coordinates of the tail marker region to obtain the three-dimensional geometric center coordinates of the tail marker (X_tail_average, Y_tail_average, Z_tail_average). Combine the average X-axis, average Y-axis, and average Z-axis coordinates of the middle marker region to obtain the three-dimensional geometric center coordinates of the middle marker (X_middle_average, Y_middle_average, Z_middle_average). Finally, integrate the three-dimensional geometric center coordinates of the first, tail, and middle markers to form a complete three-dimensional coordinate set, i.e., {(X_first_average, Y_first_average, Z_first_average), (X_tail_average, Y_tail_average, Z_tail_average), (X_middle_average, Y_middle_average, Z_middle_average)}, and use this set as the basic data output for subsequent fitting of the ideal motion trajectory axis.

[0042] By setting fixed reference markers at key axial positions of the steel pipe and accurately calculating their geometric center coordinates, the interference of steel pipe vibration on the establishment of the benchmark is reduced. Using the average coordinates of all points within the area as the geometric center coordinates of the markers effectively offsets the measurement errors of individual point coordinates, making the three-dimensional coordinates of the markers closer to their true positions and improving the accuracy of the benchmark data.

[0043] In a preferred embodiment of the present invention, step 2 above, which involves dividing the original point cloud data into an original 3D ring point cloud set along the axial direction of the steel pipe, and fitting the ideal motion trajectory axis of the steel pipe through spatial linear regression based on the three-dimensional coordinate set, may include:

[0044] Step 200: Use the geometric center coordinates of the first marker point, the tail marker point, and the middle marker point as input data.

[0045] Step 201: Calculate the three-dimensional line parameters based on the spatial coordinates of the three marker points. The line parameters include:

[0046] Reference point coordinates: the three-dimensional coordinates of any point on the straight line except for the marked point;

[0047] Direction vector: the three-dimensional component of the direction of a straight line's extension;

[0048] Step 202: Calculate the vertical distance from the geometric center of each marker point to the current fitted line, and optimize and adjust the coordinates of the reference point and the direction vector components to minimize the sum of the squares of the vertical distances of the three marker points. Output the final determined three-dimensional coordinates of the reference point and the direction vector components as the complete parameters of the axis of the ideal motion trajectory of the steel frame.

[0049] In this embodiment of the invention, the input data consists of three structured three-dimensional coordinates:

[0050] Geometric center coordinates of the first marker point: including specific values ​​in the three directions of X-axis, Y-axis, and Z-axis (for example, it can be understood as the specific position of the point in three-dimensional space, left and right, front and back, and up and down).

[0051] Geometric center coordinates of the tail end marker: also includes specific values ​​in the X, Y, and Z directions, corresponding to the center position of the marker at the axial end of the steel pipe;

[0052] The geometric center coordinates of the middle marker point: including the specific values ​​in the X, Y, and Z directions, corresponding to the center position of the marker point in the middle of the steel pipe's axial direction.

[0053] Before inputting the data, the accuracy of these three coordinates must be confirmed: for example, check whether the value of each coordinate is within a reasonable range (no abnormal values ​​that significantly exceed the length or diameter of the steel pipe), and whether the Z-axis coordinates of the three points increase (or decrease) roughly along the length of the steel pipe, conforming to the physical characteristics of the axial extension of the steel pipe. After confirming that there are no errors, organize these three coordinates into an ordered dataset in the order of "beginning end → middle → end".

[0054] Step 201: Initially determine the direction vector and the coordinates of the reference point, which together constitute the basic parameters of the three-dimensional straight line.

[0055] Preliminary calculation of direction vector:

[0056] The direction vector describes the extension direction of a line in three-dimensional space, and consists of components in three dimensions: X, Y, and Z (each component reflects the extension trend of the line along the corresponding axis). The specific calculation process is as follows:

[0057] The first step is to calculate the initial vector from the beginning to the end: taking the coordinates of the beginning marker point as the starting point and the coordinates of the end marker point as the ending point, calculate the differences between the two on the X-axis, Y-axis, and Z-axis respectively (i.e., the X value of the end coordinate minus the X value of the beginning coordinate; the same applies to the Y-axis and Z-axis). These three differences together constitute an initial vector, which roughly reflects the overall extension direction of the steel pipe from the beginning to the end.

[0058] The second step involves fine-tuning the vector using the midpoint marker: Calculate the vector from the midpoint marker to the first marker (midpoint X minus first marker X; Y and Z are calculated similarly). Observe whether this vector is "in the same direction" as the initial vector obtained in the first step (i.e., whether the positive and negative values ​​of each dimension component are consistent and whether the proportions are close). If there is a deviation (e.g., the proportion of the Y-axis component from the midpoint to the first marker differs significantly from that of the initial vector), then fine-tune each component of the initial vector. For example, if the Y-axis component of the initial vector is 5, while the Y-axis component from the midpoint to the first marker is 4.5, the Y-axis component of the initial vector can be adjusted to (5+4.5) / 2=4.75. The same applies to the X-axis and Z-axis components, making the vector more closely match the overall trend of the three points. The final direction vector needs to retain specific values ​​(e.g., X component is 10.2, Y component is 0.5, Z component is 80.3) to fully describe the extension direction of the line.

[0059] (2) Preliminary calculation of the coordinates of the reference point:

[0060] A reference point is a fixed point (unmarked point) on a straight line, used to anchor the spatial position of the line. The calculation process is as follows:

[0061] The first step is to project the three marker points onto the line containing the direction vector: For each marker point (start, middle, and end), determine its projection point on the line by "finding the nearest point along the direction vector". For example, the projection point of the start marker point is the point on the line closest to the start endpoint, and its X, Y, and Z coordinates can be obtained by "approaching" the start endpoint onto the line (the specific operation can be understood as: keeping the direction vector unchanged, adjusting the coordinate values ​​so that the point is on the line).

[0062] The second step is to calculate the average coordinates of the projection points: take the average X-axis coordinate, average Y-axis coordinate, and average Z-axis coordinate of the three projection points, and combine these three average values ​​into a new coordinate (for example, X=20.5, Y=15.3, Z=30.1). This coordinate is the initial coordinate of the reference point.

[0063] If the projection calculation is complex, a simplified method can be used: move the starting point along the direction vector to the ending point by 1 / 3 of the distance (proportionally equal to the length of the direction vector) to obtain a new point as the reference point. For example, if the total length of the direction vector is 100 units, the point 33 units from the starting point to the ending point is the reference point.

[0064] Step 202: Through iterative optimization, the straight line is made as close as possible to the three marked points, and the ideal axis parameters are finally determined.

[0065] Calculate the perpendicular distance from the marked point to the line:

[0066] For the preliminary straight line (including the reference point and direction vector) obtained in step 201, the perpendicular distance from each marker point to the straight line needs to be calculated. The specific process is as follows:

[0067] For the initial marker point: Draw a perpendicular line from this point to the line (forming a 90° angle with the line), and find the intersection of the perpendicular line and the line (i.e., the foot of the perpendicular); by measuring the straight-line distance between the initial marker point and the foot of the perpendicular (e.g., measuring the length of the straight line between the two points in space with a ruler), obtain the perpendicular distance from the initial point to the line (e.g., 2.3 mm). Repeat the above operation for the middle and tail marker points to obtain the perpendicular distance from the middle point to the line (e.g., 1.8 mm) and the perpendicular distance from the tail point to the line (e.g., 2.1 mm), respectively.

[0068] Optimize parameters to minimize the sum of squared distances:

[0069] By fine-tuning the line parameters to minimize the sum of the squares of the three perpendicular distances, the specific process is as follows:

[0070] The first step is to calculate the initial sum of squares: square the three perpendicular distances respectively and add them together to get the initial deviation value.

[0071] The second step is to fine-tune the direction vector: First, adjust the X component of the direction vector (for example, from 10.2 to 10.25), redefine the line, and then calculate the vertical distance and sum of squares from the three marked points to the new line (for example, 12.80). If the sum of squares decreases, retain the adjustment; if it increases, restore the original value. Then, fine-tune the Y and Z components in sequence, repeating the above process.

[0072] The third step is to fine-tune the coordinates of the reference point: After the direction vector is determined, fine-tune the X coordinate of the reference point (for example, from 20.5 to 20.48), redetermine the straight line, calculate the sum of squares (for example, 12.75), retain the adjustments that reduce the sum of squares, and then fine-tune the Y and Z coordinates.

[0073] The fourth step is to repeat the iteration: alternately fine-tune the direction vector and the coordinates of the reference point, gradually reducing the adjustment range each time (for example, from 0.05 units to 0.01 units) until after 5 consecutive adjustments, the change in the sum of squares is less than 0.001 (for example, from 12.75 to 12.749, and then to 12.748, the change is minimal). At this point, it can be considered that the minimum value has been reached.

[0074] When the sum of squared distances reaches its minimum value, record the current coordinates of the reference point (e.g., X=20.47, Y=15.29, Z=30.05) and the direction vector (e.g., X=10.23, Y=0.48, Z=80.25). Combine these two parameters to form the complete parameters of the steel pipe's ideal motion trajectory axis. This axis can best approximate the spatial distribution of the three marker points and represent the true centerline of the steel pipe when it is not shaking.

[0075] By using three fixed marker points as a reference and combining "preliminary calculation + iterative optimization", the fitted axis is made to fit the true center line of the steel pipe as closely as possible, reducing the axis offset error caused by jitter. Using the fixed marker points as anchor points avoids the deviation when directly using the cross-sectional center of the jittering steel pipe to fit the axis, ensuring the reliability of the axis as the reference for subsequent point cloud correction. The accurate ideal motion trajectory axis provides a unified spatial reference for subsequent axial calibration, center correction and other steps, ensuring the continuity and accuracy of the entire jitter reduction process.

[0076] In a preferred embodiment of the present invention, step 3 above, which involves axial position calibration of the original 3D loop point cloud and statistical outlier filtering of the calibrated 3D loop to generate an optimized 3D loop point cloud, may include:

[0077] Step 300: Based on the axis direction vector components, calculate the projected displacement of the center of each original 3D ring point cloud along the axis direction;

[0078] Step 301: Translate the entire point cloud of the current 3D ring along the axis direction. The translation vector is equal to the projected displacement.

[0079] Step 302, perform the following for each individual 3D loop point cloud that has completed axial position calibration:

[0080] Calculate the average spatial distance between each point within the ring and a preset number of neighboring points;

[0081] Based on the average distance data of all points, determine the overall statistical characteristics of the distance distribution of the entire ring;

[0082] The effective distance interval is determined based on overall statistical characteristics;

[0083] Delete points whose distance from the mean exceeds the valid distance interval;

[0084] Step 303: Output the optimized 3D loop cloud set after removing outliers.

[0085] In this embodiment of the invention, the axial offset distance of each original 3D ring is determined based on the direction vector of the ideal axis. The specific process is as follows:

[0086] The first step is to determine the center of the original 3D ring: For each original 3D ring (i.e., a cross-sectional point cloud of the steel pipe), calculate its geometric center - traverse the X, Y, and Z coordinates of all points in the ring, calculate the average value of each, and obtain the center coordinates of the ring (which can be understood as the "center point" of the cross-section).

[0087] The second step is to clarify the "axial trend" of the axis direction vector: The direction vector of an ideal axis contains three components: X, Y, and Z. The Z component usually reflects the length direction (axial direction) of the steel pipe. Therefore, the Z component of the direction vector is used as the main reference to determine the "axial extension direction" of the axis (for example, if the Z component is positive, the axis is along the positive Z-axis direction, and if it is negative, the opposite is true).

[0088] The third step is to calculate the projection of the center along the axis: project the center coordinates of the original 3D ring onto the ideal axis, that is, find the point on the axis that is closest to the center (projection point). The Z coordinate (axial coordinate) of this projection point represents the axial position of the ring in the ideal state.

[0089] The fourth step is to calculate the projected displacement: Subtract the Z-coordinate of the projected point from the Z-coordinate of the original 3D ring center. The difference is the "projected displacement." If the difference is positive, it means the ring is ahead of the ideal position in the axial direction; if it is negative, it is behind the ideal position. This displacement directly reflects the degree of axial offset of the ring due to vibration.

[0090] Step 301: Through translation, eliminate the axial offset of the original 3D rings, aligning the positions of each ring with the ideal axis. The specific process is as follows:

[0091] The first step is to determine the translation direction: the translation direction is determined based on the sign of the projected displacement. If the displacement is positive (the ring is ahead), the translation is performed in the opposite direction of the axis (negative direction of the Z-axis); if the displacement is negative (the ring is behind), the translation is performed in the positive direction of the axis (positive direction of the Z-axis).

[0092] The second step is to perform a translation operation: adjust the coordinates of all point cloud points within the 3D ring according to the projected displacement. Subtract the displacement from the Z coordinate (axial coordinate) of each point (if the displacement is positive, it is equivalent to moving in the negative direction of the Z axis; if it is negative, it is equivalent to moving in the positive direction of the Z axis), while keeping the X and Y coordinates unchanged (because only the axial position needs to be calibrated).

[0093] The third step is to verify the translation effect: After translation, recalculate the center coordinates of the ring and check whether its Z coordinate is consistent with the Z coordinate of the projection point (or the deviation is within a very small range, such as less than 0.01mm) to ensure that the ring is aligned to the ideal position in the axial direction.

[0094] Step 302 involves removing noise points (outliers) from the calibrated 3D loop through statistical analysis to ensure the purity of the point cloud data. The specific process is as follows:

[0095] The first step is to calculate the average distance between each point and its neighbors: For each point within the calibrated 3D ring, a "neighborhood range" is set (for example, the 50 nearest points around the point are selected as neighbors, and the number can be adjusted according to the point cloud density). The spatial distance (i.e., the straight-line distance in 3D coordinates) from the point to each neighbor is calculated, and then the average of these distances is calculated to obtain the "average neighborhood distance" of the point. This value reflects the "clustering degree" of the point and its surrounding points (the smaller the value, the closer the point is to its surrounding points, and the more likely it is to be a valid point).

[0096] The second step is to determine the statistical characteristics of the distance distribution across the ring: collect the "mean neighborhood distance" of all points within the ring, and calculate the overall statistical characteristics of these means—including the "overall mean" (the average of all means) and the "standard deviation" (reflecting the dispersion of the means). For example, an overall mean of 2 mm and a standard deviation of 0.5 mm indicate that the mean neighborhood distance of most points is between 1.5 mm and 2.5 mm.

[0097] The third step is to set the effective distance interval: Based on statistical characteristics, the effective range is defined. Usually, the effective interval is "total mean ± 2 standard deviations" (in the example above, the interval is 2 ± 1 mm, i.e. 1 mm to 3 mm). This interval includes the mean neighborhood distance of most normal points. Points outside this interval are likely to be noise (outliers).

[0098] The fourth step is to remove outliers: Traverse all points within the ring. If the mean neighborhood distance of a point exceeds the valid range (e.g., greater than 3mm or less than 1mm in the example above), it is identified as an outlier and deleted. For example, if the mean neighborhood distance of a point is 4mm, which is much greater than 3mm, it indicates that it is too far away from its surrounding points, which may be noise caused by scanning errors and needs to be removed.

[0099] Step 303: Integrate all 3D loop point clouds that have undergone axial calibration and outlier filtering into the final optimized dataset:

[0100] For each 3D ring, repeat steps 300-302—first calibrate the axial position, then remove outliers to obtain a single optimized 3D ring point cloud. Arrange all optimized 3D ring point clouds in axial order of the steel pipe (from the first end to the last end) to form a complete "optimized 3D ring point cloud set". Each ring in this set is axially aligned and has high point cloud purity.

[0101] By calculating the projected displacement and calibrating the translation, all 3D rings are aligned to the ideal position in the axial direction, eliminating the axial offset caused by the vibration of the steel pipe and ensuring the consistency of the point cloud of each section in the spatial position.

[0102] Statistical outlier filtering eliminates noise points caused by scanning errors or interference through quantitative analysis, avoiding interference from outliers in subsequent circle center fitting and contour correction, and improving the reliability of point cloud data.

[0103] In a preferred embodiment of the present invention, step 4 above, based on the optimized 3D ring point cloud, extracts the initial center coordinates of each ring, projects each initial center vertically onto the motion trajectory axis to obtain the corrected center position; and translates all point clouds of each 3D ring along the direction from the initial center to the corrected center position to generate a 3D ring point cloud with corrected center, may include:

[0104] Step 400: Use the random sampling consensus algorithm to perform circle fitting on each optimized 3D ring point cloud, and output the initial 3D coordinates of the center of each ring, specifically including:

[0105] Step 4010: Randomly select a preset number of point cloud points from the current optimized 3D ring point cloud to generate candidate circle models;

[0106] Step 4011: Calculate the radial distance from all point cloud points of the corresponding ring to the candidate circle model, and mark the points whose distance is less than the set threshold as inner points;

[0107] Step 4012: Repeatedly execute random sampling modeling operation and interior point screening operation, update the maximum number of interior points record in each loop until the preset maximum number of iterations is reached, and select the candidate circle model with the most interior points and uniform spatial distribution in the entire iteration process as the final fitted circle.

[0108] Step 4013: Use the final three-dimensional center coordinates of the fitted circle as the initial center coordinates of the current loop;

[0109] Step 401: Project each initial circle center coordinate onto the axis along a direction perpendicular to the axis of the ideal motion trajectory of the steel pipe to obtain the corresponding corrected circle center coordinates;

[0110] Step 402: For each 3D ring, calculate the three-dimensional spatial displacement vector between the initial center coordinates and the corrected center coordinates;

[0111] Step 403: Based on the spatial displacement vector, translate all the point cloud data corresponding to the 3D ring, and output the 3D ring point cloud set with the center position corrected as the 3D ring point cloud set with the center corrected.

[0112] In this embodiment of the invention, the initial center of each ring is fitted from the optimized 3D ring point cloud using the Random Sample Consensus (RANSAC) algorithm, which has strong noise resistance. The specific process is as follows:

[0113] Step 4010: For a single optimized 3D ring point cloud (i.e., a cross-sectional point cloud after axial calibration and denoising), randomly select 3 points from all points of the ring (the preset number is 3, since 3 non-collinear points can uniquely determine a circle); calculate a "candidate circle model" based on the spatial coordinates of these 3 points: that is, assuming that these 3 points are all on the same circle, determine the temporary center coordinates (3D) and temporary radius of the circle through geometric relationships (the distance from the center to the 3 points is approximately equal); for example, if the coordinates of the 3 points are A, B, and C, the calculated center of the candidate circle is O1, and the radius is the distance from O1 to A.

[0114] Step 4011: For all point cloud points within the current 3D ring, calculate the "radial distance" from each point to the candidate circle model: that is, the difference between the distance from the point to the center of the candidate circle and the radius of the candidate circle (if the difference is positive, the point is outside the circle; if it is negative, it is inside the circle). Set an "inner point threshold" (determined based on the steel pipe size and scanning accuracy, for example, 0.3mm). If the absolute value of the radial distance of a point is less than the threshold, it means that the point is likely to belong to this circle and is marked as an "inner point"; otherwise, it is marked as an "outer point" (it may be noise or a defect, but this is not distinguished at this time). For example, if the radius of the candidate circle is 50mm, the distance from a point to the center is 50.2mm, and the radial distance is 0.2mm (less than 0.3mm), then it is marked as an inner point.

[0115] Step 4012: Set the "maximum number of iterations" (based on the point cloud density, e.g., 200 times). Repeat steps 4010-4011 for each iteration: randomly select 3 points to generate a new candidate circle model, calculate and count the number of interior points in the model. After each iteration, if the number of interior points in the current candidate circle is greater than the previously recorded "maximum number of interior points," update the maximum number of interior points and record the parameters (center, radius) of the candidate circle. After the iteration is complete, select the model with the "most interior points" and "uniform spatial distribution of interior points" from all candidate circles as the final fitted circle. For example, if two candidate circles have the same number of interior points (both 1000), select the candidate circle where the interior points are evenly distributed on the ring (rather than concentrated in a local area) to avoid fitting bias caused by dense local points.

[0116] Step 4013: Use the three-dimensional center coordinates of the final fitted circle (including the specific values ​​of the X, Y, and Z directions) as the "initial center coordinates" of the current optimized 3D ring; for example, if the center of the final fitted circle is (X=100.5mm, Y=20.3mm, Z=500.2mm), then this coordinate is the initial center of the ring.

[0117] Step 401: Through projection, the initial center of the circle is "attached" to the axis of the ideal motion trajectory to obtain the corrected center position. The specific process is as follows:

[0118] Define the parameters of the ideal trajectory axis of the steel pipe (reference point coordinates and direction vector, already determined in step 202); draw a "perpendicular line" (at a 90-degree angle with the axis) from the initial center coordinates to the ideal axis, and find the intersection point of the perpendicular line and the axis—this intersection point is the "vertical projection point" of the initial center on the axis; calculate the three-dimensional coordinates of this projection point: determine the point on the axis closest to the initial center through geometric relationships (because the perpendicular distance is the shortest distance), and the X, Y, and Z coordinates of this point are the "corrected center coordinates"; for example, if the initial center is (100.5, 20.3, 500.2), after drawing a perpendicular line to the axis, the intersection point coordinates are (100.0, 20.0, 500.2), then these coordinates are the corrected center.

[0119] Step 402: Calculate the coordinate differences between the initial center and the corrected center in the X, Y, and Z directions respectively: X component of displacement vector = X coordinate of initial center - X coordinate of corrected center; Y component = Y coordinate of initial center - Y coordinate of corrected center; Z component = Z coordinate of initial center - Z coordinate of corrected center; For example, if the initial center is (100.5, 20.3, 500.2) and the corrected center is (100.0, 20.0, 500.2), then the displacement vector is (0.5mm, 0.3mm, 0mm), indicating that the initial center deviates by 0.5mm in the X direction, 0.3mm in the Y direction, and has no deviation in the Z direction.

[0120] Step 403: For all point cloud points within the current 3D ring, translate them in the opposite direction of the displacement vector: subtract the X component of the displacement vector from the X coordinate of each point, subtract the Y component of the displacement vector from the Y coordinate, and subtract the Z component of the displacement vector from the Z coordinate (i.e., "point coordinate = original coordinate - displacement vector").

[0121] Taking the above example, the original coordinates of a point are (101.5, 21.3, 500.2). After subtracting the displacement vector (0.5, 0.3, 0), the new coordinates are (101.0, 21.0, 500.2). After translation, the initial center of the circle will coincide with the corrected center of the circle (because the initial center coordinates - displacement vector = corrected center coordinates). The point cloud of the entire 3D ring moves with the center, completing the generation of the "center-corrected 3D ring point cloud".

[0122] Overall process description:

[0123] For each ring within the optimized 3D ring point cloud, repeat steps 400-403: first extract the initial center using the RANSAC algorithm, then project it onto the ideal axis to obtain the corrected center, calculate the displacement vector and translate the point cloud, and finally obtain the 3D ring point cloud with the center corrected for all rings, forming a complete set of center corrected point clouds.

[0124] The Random Sample Consensus (RANSAC) algorithm effectively resists the interference of point cloud noise and local defects. By selecting candidate circles with the most internal points and uniform distribution, it ensures that the initial circle center coordinates are closer to the true center of the ring. Projecting the initial circle center vertically onto the ideal axis and translating the point cloud completely eliminates the problem of the ring centers deviating from the axis due to steel pipe vibration, unifying the centers of all rings onto the ideal axis. The overall translation of the point cloud preserves the relative positional relationships of the points within the ring (i.e., preserves the local features of the true defect), correcting only the overall offset caused by vibration, avoiding the destruction of defect information, and providing reliable data for defect detection and dimensional measurement.

[0125] In a preferred embodiment of the present invention, step 5 above, which involves refitting the standard circle based on the corrected center position to obtain the radius values ​​of each ring, and calculating the average value of all ring radius values ​​as the global reference radius, may include:

[0126] Step 500: For each independent 3D ring point cloud in the 3D ring point cloud set with the center-corrected center coordinates of the corresponding ring as the fixed center position, calculate the average distance value as the fitting radius value of the corresponding ring based on the set of distance values ​​from all point cloud data of the corresponding ring to the fixed center position, and output the fitting radius value dataset of each ring.

[0127] Step 501: Perform an average calculation on the radius values ​​of all rings in the fitted radius value dataset for each ring, and output the calculation result as the global baseline radius.

[0128] In this embodiment of the invention, using the corrected center of the circle as a reference, the fitting radius specific to 3D ring point cloud computing is corrected for each center of the circle. The specific process is as follows:

[0129] Determine the fixed center: For a specific 3D ring point cloud in the 3D ring point cloud set for center correction, determine its corresponding "corrected center coordinates" (the coordinates have been determined in step 401 and are located on the axis of the steel pipe's ideal motion trajectory). This coordinate is the only fixed reference point for fitting the standard circle in this step and will not be adjusted again.

[0130] Calculate the distance from all points within the ring to the fixed center: Traverse every point in the 3D ring's point cloud and calculate the spatial straight-line distance from each point to the fixed center. For example, if the fixed center's coordinates are (X0, Y0, Z0) and a point within the ring has coordinates (X1, Y1, Z1), then the straight-line length (i.e., distance value) between these two points needs to be calculated; summarize the distance values ​​of all points to form a "distance value set".

[0131] Calculate the average distance as the fitting radius: Sum all distance values ​​in the above "distance value set", then divide by the total number of point cloud points within the ring. The average value obtained is the "fitting radius value" of the ring. For example, if there are 1000 points within a ring, and the total distance from all points to the fixed center is 50000mm, then the fitting radius value of the ring is 50000÷1000=50mm.

[0132] Output the fitted radius dataset for each ring: For each independent 3D ring point cloud in the 3D ring point cloud set for center correction, repeat the above operation—determine a fixed center, calculate the distance from all points to the center, and calculate the average distance to obtain the fitted radius. Arrange the fitted radius values ​​of all rings in axial order of the steel pipe (from the beginning to the end) to form the "fitted radius value dataset for each ring".

[0133] Step 501: Based on the fitted radius values ​​of all loops, calculate a unified global baseline radius. The specific process is as follows:

[0134] Collect the fitting radius of all rings: Extract the fitting radius values ​​of all 3D rings from the "fitting radius value dataset of each ring" output in step 500 (for example, a steel pipe has 500 rings, corresponding to 500 fitting radius values, ranging from 49.8mm to 50.2mm).

[0135] Perform an averaging operation: Sum all the fitted radius values, then divide the sum by the total number of loops (i.e., the total number of fitted radius values). The result is the "global reference radius". For example, if the sum of the fitted radius values ​​of 500 loops is 25000 mm, then the global reference radius is 25000 ÷ 500 = 50 mm.

[0136] Output global reference radius: The result of the above average calculation (e.g., 50mm) is used as the final global reference radius output. This value represents the average radius of the entire steel pipe under ideal conditions.

[0137] The average distance is calculated using the corrected center as a fixed reference, avoiding the influence of center offset on radius calculation. This makes the fitted radius of each ring closer to the true radius of the cross-section, accurately reflecting the local dimensional characteristics of various parts of the steel pipe. The global reference radius is obtained by calculating the average of the fitted radii of all rings, integrating the dimensional information of the entire steel pipe and ensuring consistent contour correction scales for different rings. The fitted radius includes the true dimensional information of the steel pipe but does not filter out surface defects (such as protrusions and pits, which can cause local distance values ​​to be too large or too small, thus affecting the fitted radius of that ring), while the global reference radius balances the dimensional differences among the rings.

[0138] In a preferred embodiment of the present invention, step 6 above, which involves adjusting each point within the corresponding ring along the radial direction from the center of the circle to the corresponding point to a global reference radius distance, generating a contour-standardized 3D ring point cloud, and integrating all contour-standardized 3D ring point clouds to generate a jitter-free 3D point cloud model of the steel pipe, may include:

[0139] Step 600: Using the corrected center position as the reference point, calculate the original radial distance from each point cloud point in the ring to the reference point;

[0140] Step 601: Move each point cloud point along the line connecting itself to the reference point, so that the distance between the moved point and the reference point is equal to the global reference radius. After completing the standardized movement of all point cloud points of the corresponding ring, generate a contour-standardized 3D ring point cloud.

[0141] Step 602: Aggregate all contour-standardized 3D ring point clouds in axial order of the steel pipe to generate a complete 3D point cloud model of the steel pipe with jitter eliminated.

[0142] In this embodiment of the invention, using the corrected center of the circle as a reference, the radial distance from all points within the ring to the center is measured. The specific process is as follows:

[0143] Define the reference point: For a 3D ring point cloud with a certain contour to be standardized, determine its corresponding "corrected center position" (this coordinate has been determined in step 401 and is located on the axis of the ideal motion trajectory of the steel pipe), and use this position as the reference point for calculating the radial distance.

[0144] Traverse all points within the ring: Extract the three-dimensional coordinates (including specific values ​​in the X, Y, and Z directions) of each point within the 3D ring point cloud.

[0145] Calculate the original radial distance: For each point, calculate its straight-line distance to the reference point (the center of the corrected circle)—this distance is the "original radial distance," reflecting the point's radial position relative to the center in the current ring. For example, if the reference point coordinates are (100, 200, 300) and a point's coordinates are (105, 205, 300), then the straight-line distance from that point to the reference point is its original radial distance.

[0146] Record distance data: Associate the original radial distance of each point with the coordinates of that point to form the "original radial distance set" of the ring.

[0147] Step 601: By adjusting the positions of the points, the distance from all points within the ring to the center is made uniform to the global reference radius. The specific process is as follows:

[0148] Determine the adjustment direction: For each point within the ring, determine the direction of the line connecting it to the reference point (the center of the circle after correction)—that is, the straight line direction (radial direction) from the reference point to that point. For example, if the reference point is at (100, 200, 300) and a point is at (105, 205, 300), then the direction of the line is diagonally outward from (100, 200, 300) to (105, 205, 300).

[0149] The distance to move is determined based on the difference between the original radial distance at that point and the global reference radius.

[0150] If the original radial distance is greater than the global reference radius, it means that the point needs to be moved towards the reference point (inward), and the movement range is "original radial distance - global reference radius";

[0151] If the original radial distance is less than the global reference radius, it means that the point needs to be moved away from the reference point (outward), and the movement range is "global reference radius - original radial distance".

[0152] Perform standardized movement: Move each point along the direction of the line mentioned above by the corresponding amount, ensuring that the distance between the point and the reference point after the movement is exactly equal to the global reference radius. For example, if the global reference radius is 50mm and the original radial distance of a point is 52mm, then it needs to be moved inward by 2mm, and the final distance from the reference point is 50mm; if the original radial distance of another point is 48mm, then it needs to be moved outward by 2mm, and the final distance is also 50mm.

[0153] Once all points within the ring have completed the aforementioned movement, the distance from all points to the reference point is unified to the global reference radius. At this point, the ring's outline is standardized, forming a "contour-standardized 3D ring point cloud".

[0154] Step 602: Integrate all standardized ring point clouds in axial order to form the final dejitter model. The specific process is as follows:

[0155] Determine the axial order: Based on the Z-axis coordinates (axial position, already calibrated in step 301) corresponding to the standardized 3D ring point cloud of each contour, sort the ring point clouds from the beginning to the end of the steel pipe (e.g., arrange the Z-axis coordinates from smallest to largest). Merge all the sorted standardized 3D ring point clouds into a complete point cloud dataset, ensuring that the point clouds of adjacent rings are continuously connected in the axial direction (without obvious gaps or overlaps). The merged point cloud dataset completely covers the outer surface of the steel pipe, and the contour of each ring is based on the global reference radius, eliminating the overall contour distortion caused by jitter (such as ellipticization and radial offset), while retaining the local features of surface defects (such as pits and protrusions, which will appear as small local deviations in the point cloud), ultimately forming a "jitter-free 3D point cloud model of the steel pipe".

[0156] By adjusting each point to the global reference radius, the contour scale of each ring was unified, completely eliminating the overall radial deviation caused by steel pipe vibration (such as a ring's radius being too large or too small due to vibration), making the point cloud model closer to the ideal shape of the steel pipe when it is vibration-free. The adjustment process only scales radially to the global reference radius, so the local features of surface defects (such as local protrusions that cause the original radial distance to be slightly larger than the reference radius, but still retain a slight overshoot after adjustment; pits that retain a slight undershoot) are completely preserved, avoiding the miscorrection or filtering of defects. The aggregated 3D point cloud model maintains a high degree of consistency in both the axial and radial directions, providing a precise data foundation for subsequent defect detection (reducing false positives / false negatives) and dimensional measurements (such as diameter, ellipticity, and straightness), improving the reliability of the detection results. From single-ring contour standardization to full-pipe model aggregation, the entire process systematically eliminates the influence of vibration on the 3D point cloud. The final model can realistically reflect the morphological characteristics of the steel pipe, providing an intuitive and accurate visualization basis for production quality assessment.

[0157] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0158] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0159] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for removing wobble of a three-dimensional laser point cloud for defect detection of a moving steel pipe, characterized in that, The method comprises: Step 1, three fixed reference marker points are fixedly arranged on the axial head end, tail end and middle part of the surface of the moving steel pipe, original point cloud data of the steel pipe surface containing the marker points is collected, the fixed reference marker point area point cloud is separated, the geometric center coordinates of each marker point area point cloud are calculated, and a three-dimensional coordinate set of the marker points is output; Step 2, based on the original point cloud data, the steel pipe is divided into an original 3D ring point cloud set along the axial direction; and based on the three-dimensional coordinate set, a steel management ideal motion trajectory axis is fitted by spatial linear regression; Step 3, axial position calibration is performed on the original 3D ring point cloud set, and statistical outlier filtering is performed on the calibrated 3D ring to generate an optimized 3D ring point cloud set; Step 4, based on the optimized 3D ring point cloud set, initial circle center coordinates of each ring are extracted, each initial circle center is vertically projected onto the motion trajectory axis to obtain a corrected circle center position; and all point clouds of each 3D ring are translated in the direction from the initial circle center to the corrected circle center position to generate a circle center corrected 3D ring point cloud; Step 5, a standard circle is refitted with the corrected circle center position as the reference to obtain a radius value of each ring, and an average value of all ring radius values is calculated as a global reference radius; Step 6, each point in the corresponding ring is adjusted to the global reference radius distance in the radial direction from the circle center to the corresponding point to generate a contour standardized 3D ring point cloud, and all contour standardized 3D ring point clouds are integrated to generate a steel pipe three-dimensional point cloud model eliminating jitter.

2. The method of claim 1, wherein the method further comprises: Three fixed reference marker points are fixedly arranged on the axial head end, tail end and middle part of the surface of the moving steel pipe, original point cloud data of the steel pipe surface containing the marker points is collected, the fixed reference marker point area point cloud is separated, the geometric center coordinates of each marker point area point cloud are calculated, and a three-dimensional coordinate set of the marker points is output, comprising: From the original point cloud data collected by the three-dimensional laser profiler, the point cloud regions corresponding to the three fixed reference marker points on the axial head end, tail end and middle part are identified, and all three-dimensional coordinate data in each marker point region is extracted; The average value of the X-axis coordinate, the average value of the Y-axis coordinate and the average value of the Z-axis coordinate of all points in each marker point region are calculated respectively; The average value of the X-axis coordinate, the average value of the Y-axis coordinate and the average value of the Z-axis coordinate are combined into the three-dimensional geometric center coordinates of the corresponding marker point, and a three-dimensional coordinate set composed of the geometric center coordinates of the head end marker point, the geometric center coordinates of the tail end marker point and the geometric center coordinates of the middle marker point is output.

3. The method of claim 2, wherein the method further comprises: Based on the original point cloud data, the steel pipe is divided into an original 3D ring point cloud set along the axial direction; and based on the three-dimensional coordinate set, a steel management ideal motion trajectory axis is fitted by spatial linear regression, comprising: The geometric center coordinates of the head end marker point, the geometric center coordinates of the tail end marker point and the geometric center coordinates of the middle marker point are taken as input data; Three-dimensional straight line parameters are calculated according to the spatial coordinates of the three marker points, the straight line parameters comprising: Reference point coordinates: three-dimensional coordinates of any point on the straight line except the marker points; Direction vector: three-dimensional components of the extension direction of the straight line; Calculate the perpendicular distance from the geometric center of each marker point to the current fitting straight line, and optimize the adjustment of the reference point coordinates and the direction vector components to minimize the sum of the squares of the perpendicular distances of the three marker points, and output the final determined reference point three-dimensional coordinates and direction vector components as the complete parameters of the steel management ideal motion trajectory axis.

4. The method of claim 3, wherein the method further comprises: Perform axial position calibration on the original 3D ring point cloud set, and perform statistical outlier filtering on the calibrated 3D ring to generate an optimized 3D ring point cloud set, including: Based on the axial direction vector component, calculate the projection displacement of each original 3D ring point cloud center along the axial direction; Translate the entire point cloud of the current 3D ring along the axial direction, and the translation vector is equal to the projection displacement; For each independent 3D ring point cloud that has completed axial position calibration, perform the following: Calculate the average spatial distance of each point in the ring and a predetermined number of neighboring points; Determine the overall statistical characteristics of the entire ring distance distribution based on the distance average value data of all points; Set the effective distance interval based on the overall statistical characteristics; Delete points with distance averages outside the effective distance interval; Output the optimized 3D ring point cloud set after removing outliers.

5. The method of claim 4, wherein the method further comprises: Based on the optimized 3D ring point cloud set, extract the initial circle center coordinates of each ring, and vertically project each initial circle center to the motion trajectory axis to obtain the corrected circle center position; translate the entire point cloud of each 3D ring along the direction from the initial circle center to the corrected circle center position to generate a circle center corrected 3D ring point cloud, including: Use the random sample consensus algorithm to fit a circle to each optimized 3D ring point cloud, and output the initial circle center three-dimensional coordinates of each ring. Project each initial circle center coordinate onto the axis in a direction perpendicular to the steel management ideal motion trajectory axis to obtain the corresponding corrected circle center coordinate. For each 3D ring, calculate the three-dimensional spatial displacement vector between the initial circle center coordinate and the corrected circle center coordinate. Based on the spatial displacement vector, translate the entire point cloud data of the 3D ring, and output the 3D ring point cloud set that has completed circle center position correction as the circle center corrected 3D ring point cloud set.

6. The method of claim 5, wherein the method further comprises: Use the random sample consensus algorithm to fit a circle to each optimized 3D ring point cloud, and output the initial circle center three-dimensional coordinates of each ring, including: Randomly select a predetermined number of point cloud points from the current optimized 3D ring point cloud to generate a candidate circle model. Calculate the radial distance of the entire point cloud of the corresponding ring to the candidate circle model, and mark the points with a distance less than the set threshold as inliers. Loop the random sampling modeling operation and inlier selection operation, update the maximum inlier number record in each loop until the maximum iteration number is reached, select the candidate circle model with the most inliers and uniform spatial distribution in the entire iteration process as the final fitted circle, and output the initial circle center three-dimensional coordinates of the final fitted circle as the initial circle center coordinates of the current ring. Re-fit the standard circle based on the corrected circle center position to obtain the radius values of each ring, and calculate the average of all ring radius values as the global reference radius, including:

7. The method of claim 6, wherein the method further comprises: For each independent 3D ring point cloud in the circle center corrected 3D ring point cloud set, take the corrected circle center coordinate of the corresponding ring as the fixed circle center position, and based on the distance value set of all point cloud data of the corresponding ring to the fixed circle center position, calculate the distance average as the fitted radius value of the corresponding ring, and output the fitted radius value dataset of each ring. ​ Perform an average operation on the radius values of all rings in the radius value data set of each ring, and output the calculation result as a global reference radius.

8. The method of claim 7, wherein the method further comprises: Adjust each point in the corresponding ring to the global reference radius distance in the radial direction from the center to the corresponding point, generate a contour standardized 3D ring point cloud, integrate all contour standardized 3D ring point clouds, and generate a steel pipe three-dimensional point cloud model eliminating jitter, including: Take the corrected center position as a reference point, calculate the original radial distance from each point cloud point in the ring to the reference point; Move each point cloud point along the direction of the line connecting itself and the reference point, so that the distance between the moved point and the reference point is equal to the global reference radius. After the standardized movement of all point cloud points in the corresponding ring is completed, a contour standardized 3D ring point cloud is generated; Aggregate all contour standardized 3D ring point clouds in the axial sequence of the steel pipe to generate a complete steel pipe three-dimensional point cloud model eliminating jitter.

9. A computing device, comprising: Comprise: One or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, so that the one or more processors implement the method as claimed in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a program which, when executed by a processor, implements the method as claimed in any one of claims 1 to 8.

Images