Method and system for identifying geometric features of defects of petrochemical equipment
By adjusting the attitude of point cloud data from petrochemical equipment and creating a reference point cloud, the problem of measuring density, location, and depth in pitting corrosion detection was solved, enabling automated and accurate defect feature identification and statistical analysis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA PETROLEUM & CHEMICAL CORP
- Filing Date
- 2024-10-15
- Publication Date
- 2026-04-17
AI Technical Summary
Existing technologies cannot simultaneously and accurately measure the density, location, opening area, and depth of pitting corrosion in petrochemical equipment. Furthermore, the selection of a reference plane is difficult during the post-processing of point cloud data in three-dimensional measurement, resulting in low detection efficiency.
By adjusting the pose of the point cloud data, a reference point cloud is created, defect point clouds are filtered, and defect feature parameters, including defect density, opening area, and depth, are calculated. The offset angle of the cylindrical surface axis relative to the coordinate axis is calculated by slicing twice. The reference point cloud is automatically created by combining circle fitting, and the defect volume and opening area are calculated using curvature and coordinates.
It enables automated and accurate defect detection and statistical analysis, improving detection efficiency and accuracy without the need for manual intervention.
Smart Images
Figure CN121883345A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of non-contact non-destructive testing technology for petrochemical equipment, and in particular to a method and system for identifying the geometric features of defects in petrochemical equipment. Background Technology
[0002] Petrochemical equipment operates under harsh conditions with highly corrosive internal media. Under the influence of sulfur, chlorine, and other factors, the equipment is prone to corrosion and cracking. Studies of pitting corrosion typically focus on observing the morphology of severe pitting in relevant scenarios, but this approach cannot quantitatively assess the severity and distribution of pitting corrosion.
[0003] Methods for quantitative assessment of pitting corrosion include visual inspection, quality loss measurement, depth measurement of pre-selected pits by vertical cross-section, and rating using standard icons.
[0004] ASME G46-94 provides a standard graphical method for describing the severity of pitting corrosion in metallic materials, characterizing the severity of pitting corrosion from three dimensions: distribution density, average pitting area, and average depth. The data obtained by this method is easy to store and can be conveniently and quickly compared with other test data. However, both slicing and measuring the depth of all pits and manual counting are not feasible and are time-consuming and labor-intensive. Furthermore, these methods cannot simultaneously and accurately measure the density of pitting corrosion and the location of pits.
[0005] However, three-dimensional measurement can acquire the surface morphology of an object, offering advantages such as high accuracy and repeatability. This method uses point cloud data to represent the shape of the measured object; that is, by establishing several points and specifying the position of each point in three-dimensional space, the geometry of the object is described. It can simultaneously and accurately measure the location and size information of pitting corrosion pits, making it suitable for data acquisition in pitting corrosion statistical analysis or rating work. However, in the post-processing of point cloud data, a reference plane needs to be manually selected to calculate the required geometric information, such as the opening area, volume, and depth of defects. For densely distributed pitting corrosion, this process is time-consuming and laborious, hindering the acquisition of statistical analysis results for surface defects, especially for non-planar point clouds or point cloud data with uncertain spatial positions, where reference plane selection is difficult. Furthermore, the relative error of the reference plane selection operation gradually increases as the pitting corrosion area decreases, making it difficult to achieve objective and accurate evaluation.
[0006] Furthermore, deep learning methods have received considerable attention for defect identification and counting. These methods are widely used in 3D object detection tasks, but they require a large amount of relevant data for model training, which also necessitates high-quality manual annotation. Data acquisition and annotation hinder the development of this approach. In addition, deep learning methods require data downsampling, making it difficult to capture the features of small targets and prone to missed detections on smaller pitting defects.
[0007] In summary, the present invention provides a point cloud data post-processing scheme based on the geometric features of petrochemical equipment to solve one or more of the problems in the existing technology, namely, that traditional pitting detection technology cannot simultaneously obtain geometric feature information such as the density, location, initiation area and depth of pitting, and that the selection of reference surface is difficult when post-processing point cloud data through three-dimensional measurement. Summary of the Invention
[0008] The purpose of this invention is to provide a post-processing scheme for point cloud data based on the geometric features of petrochemical equipment, in order to solve one or more of the problems in the existing technology, namely, that traditional pitting detection technology cannot simultaneously obtain geometric feature information such as the density, location, initiation area and depth of pitting, and that it is difficult to select a reference surface when post-processing point cloud data through three-dimensional measurement.
[0009] To address the aforementioned technical problems, embodiments of the present invention provide a method for identifying the geometric features of defects in petrochemical equipment, comprising: obtaining raw point cloud data characterizing the surface features of the petrochemical equipment to be inspected, and adjusting its orientation to obtain first point cloud data; creating a reference point cloud containing the position information of all point cloud points based on the first point cloud data; filtering point cloud points belonging to defects from the reference point cloud to obtain a defect point cloud; and directly calculating defect feature parameters based on the defect point cloud, wherein the defect feature parameters include defect density and the opening area, depth, and volume of each defect.
[0010] Preferably, the step of obtaining raw point cloud data characterizing the surface features of the petrochemical equipment to be inspected and adjusting its attitude to obtain first point cloud data includes: when the object to be inspected is a plane, solving for the normal vector of the current plane based on the raw point cloud data of the plane, and using the normal vector to adjust the attitude of the raw point cloud to obtain the first point cloud data; when the object to be inspected is a cylinder, calculating the offset feature of the raw point cloud based on the raw point cloud data of the cylinder, and using the offset feature to adjust the attitude of the raw point cloud to obtain the first point cloud data.
[0011] Preferably, the step of solving for the normal vector of the current plane based on the original point cloud data of the plane, and using the normal vector to adjust the attitude of the original point cloud to obtain the first point cloud data, includes: performing an initial plane fitting on the original point cloud data of the object being detected on the current plane and obtaining the general equation of the fitted plane, thereby obtaining the normal vector of the initial fitted plane; using the normal vector of the initial fitted plane, calculating a first offset angle of the current normal vector relative to the x-axis or y-axis of the original point cloud in the Cartesian coordinate system, and using the first offset angle to rotate the original point cloud data; performing a second plane fitting on the point cloud data after plane rotation and obtaining the general equation of the fitted plane, thereby obtaining the normal vector of the second fitted plane; using the normal vector of the second fitted plane, calculating a second offset angle of the current normal vector relative to the z-axis of the rotated point cloud in the Cartesian coordinate system, and using the second offset angle to rotate the point cloud data after plane rotation to obtain the first point cloud data.
[0012] Preferably, the step of calculating the offset feature of the original point cloud based on the original point cloud data of the cylinder, and using the offset feature to adjust the attitude of the original point cloud to obtain the first point cloud data, includes: slicing the original point cloud data of the current cylinder detection object at least twice along the z-axis, and fitting a two-dimensional ellipse to the point set of the at least two slices to obtain the center of the fitted ellipse of the at least two slices; and calculating the third offset of the line connecting the centers of the at least two slices relative to the x-axis or y-axis of the original point cloud in the Cartesian coordinate system based on the obtained centers of the at least two slices. Angle, thereby using the third offset angle to rotate the original point cloud data once; for the point cloud data after one rotation, slice it at least twice along the z-axis, and fit the point set of the at least two slices into a two-dimensional ellipse to obtain the center of the fitted ellipse of the at least two slices; based on the center of the at least two slices obtained now, calculate the fourth offset angle of the line connecting the current center relative to the z-axis of the point cloud after one rotation in the Cartesian coordinate system, thereby using the fourth offset angle to rotate the point cloud data after one rotation a second time to obtain the first point cloud data.
[0013] Preferably, when the object being detected is a plane, the step of creating a reference point cloud containing the position information of all point cloud points based on the first point cloud data includes: using the first point cloud data as the reference point cloud.
[0014] Preferably, when the object being detected is a cylinder, the step of creating a reference point cloud containing the position information of all point cloud points based on the first point cloud data includes: extracting a target cross-sectional plane from the first point cloud data, and performing two-dimensional planar circle fitting on the target cross-sectional plane to obtain the center and radius of the currently fitted circle; and using the center and radius of the currently fitted circle as a reference, performing planar unfolding on the first point cloud data to obtain the reference point cloud.
[0015] Preferably, the step of expanding the first point cloud data to obtain the reference point cloud, using the center and radius of the currently fitted circle as boundaries, includes: aligning the axis of the first point cloud data of the current cylinder with the first coordinate axis representing the z-axis in the Cartesian coordinate system based on the center coordinates of the currently fitted circle; labeling the other two coordinate axes in the Cartesian coordinate system as the second and third coordinate axes, respectively; denoteing the plane formed by the second and third coordinate axes as the basic plane; dividing the first point cloud data into four parts in the basic plane according to quadrants based on the center coordinates of the currently fitted circle under the basic plane; and converting the first point cloud data of the cylinder into a reference point cloud by performing coordinate transformation on each point cloud point. This includes: during the coordinate transformation of each point cloud point in a clockwise order, keeping the coordinate value of each point cloud point on the first coordinate axis unchanged, using the arc length of the cylinder on the cross section where each point cloud point is located as the coordinate value of that point on the second coordinate axis representing the y-axis, and using the distance of each point cloud point to the cylinder axis as the coordinate value of that point on the third coordinate axis representing the x-axis; or, during the coordinate transformation of each point cloud point in a counterclockwise order, keeping the coordinate value of each point cloud point on the first coordinate axis unchanged, using the arc length of the cylinder on the cross section where each point cloud point is located as the coordinate value of that point on the third coordinate axis representing the y-axis, and using the distance of each point cloud point to the cylinder axis as the coordinate value of that point on the second coordinate axis representing the x-axis.
[0016] Preferably, the step of extracting and filtering point cloud points belonging to defects from the reference point cloud to obtain a defect point cloud containing information on the distribution location of each defect includes: filtering point cloud points belonging to defects by calculating the curvature of each point cloud point in the reference point cloud to obtain an initial defect point cloud; performing cluster analysis on the defect point cloud points in the initial defect point cloud, determining the number of defects based on the clusters, thereby obtaining the defect point cloud used to distinguish the distribution location of each defect region.
[0017] Preferably, the curvature of each point in the reference point cloud and the average curvature of all points are calculated; based on the average curvature, a curvature threshold for distinguishing the curvature of defective points from that of normal points is determined; and point cloud points in the reference point cloud whose curvature exceeds the curvature threshold are taken as defective point cloud points to obtain the initial defective point cloud.
[0018] Preferably, the step of directly calculating the feature parameters of each defect based on the defect point cloud includes: taking the difference between the maximum and minimum coordinates of the point cloud points in each defect region on the z-axis as the depth of the corresponding defect region based on the coordinates of each defect point cloud point in the defect point cloud; determining the convex hull of each defect region point set based on the coordinates of each defect point cloud point in the defect point cloud, and taking the volume of each convex hull as the volume of the corresponding defect region; projecting each defect region point set onto a plane perpendicular to the z-axis, determining the corresponding convex hull for the projected point set, and taking the projected convex hull as the opening area of the corresponding defect region; projecting all defect region point sets onto a plane perpendicular to the z-axis, determining the area of the minimum bounding rectangle of all projected points, and obtaining the defect distribution density of the current detected region by calculating the ratio of the number of current defect regions to the area of the minimum bounding rectangle.
[0019] On the other hand, the present invention provides a computer-readable storage medium comprising a series of instructions for performing the steps of the method described above.
[0020] In addition, this invention also provides a system for identifying the geometric features of defects in petrochemical equipment, comprising: an attitude optimization module configured to obtain raw point cloud data characterizing the surface features of the petrochemical equipment to be inspected, and to adjust the attitude of the raw point cloud data to obtain first point cloud data; a reference point cloud generation module configured to create a reference point cloud containing the position information of all point cloud points based on the first point cloud data; a defect identification module configured to filter point cloud points belonging to defects from the reference point cloud to obtain a defect point cloud; and a defect statistical analysis module configured to directly calculate defect feature parameters based on the defect point cloud, wherein the defect feature parameters include defect density and the opening area, depth, and volume of each defect.
[0021] Compared with the prior art, one or more embodiments of the above solutions may have the following advantages or beneficial effects:
[0022] This invention proposes a method and system for identifying the geometric features of defects in petrochemical equipment. Addressing the problem that visual inspection and two-dimensional image recognition of pitting corrosion cannot accurately and quickly obtain parameters such as depth and volume, this method and system proposes a point cloud data post-processing method based on the equipment's geometric features. It calculates the offset angle of the cylindrical surface's axis relative to the coordinate axes using a two-slice method, and then obtains the cylindrical surface radius by combining circle fitting, developing a cylindrical surface fitting method insensitive to initial values. It automatically creates a reference point cloud by unfolding a curved surface into a plane, eliminating the need for complex manual selection. It distinguishes defective and normal parts by calculating the curvature of each point, and obtains the location of each defect through density clustering. It calculates the volume and opening area of each defect using coordinates and solving for the convex hull, and calculates the depth of the defect using coordinates, ultimately achieving statistical analysis of defect feature parameters. The entire process of the defect feature identification method provided by this invention requires no manual intervention, automatically creates a reference point cloud, and then performs defect detection and statistical analysis, effectively improving the accuracy of defect detection and the efficiency of automated detection.
[0023] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description, claims, and drawings. Attached Figure Description
[0024] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:
[0025] Figure 1 This is a flowchart illustrating the steps of a method for identifying defective geometric features of petrochemical equipment, as described in an embodiment of this application.
[0026] Figure 2 This is a schematic diagram of the original point cloud data distribution of a cylindrical inspection object in the method for identifying defect geometric features of petrochemical equipment according to an embodiment of this application.
[0027] Figure 3 This is a schematic diagram of the point cloud data distribution of the slice point set in the method for identifying defect geometric features of petrochemical equipment according to an embodiment of this application.
[0028] Figure 4 (a) is a schematic diagram of the fitting result of the fitted ellipse 1 during the initial attitude adjustment in the method for identifying defect geometric features of petrochemical equipment according to an embodiment of this application.
[0029] Figure 4(b) is a schematic diagram of the fitting result of the fitted ellipse 2 during the initial attitude adjustment in the method for identifying defect geometric features of petrochemical equipment according to an embodiment of this application.
[0030] Figure 5 This is a schematic diagram of the point cloud data distribution after attitude adjustment of a cylindrical detection object in the method for identifying defect geometric features of petrochemical equipment according to an embodiment of this application.
[0031] Figure 6 This is a schematic diagram of the point cloud data distribution of a cylindrical detection object after unfolding, in the method for identifying defect geometric features of petrochemical equipment according to an embodiment of this application.
[0032] Figure 7 This is a schematic diagram of the point cloud data distribution of the defect point cloud set in the method for identifying defect geometric features of petrochemical equipment according to an embodiment of this application.
[0033] Figure 8 This is a schematic diagram of a defect point cloud in a method for identifying defect geometric features of petrochemical equipment according to an embodiment of this application.
[0034] Figure 9 This is a schematic diagram of the convex hull of a point set of a defect region in a method for identifying the defect geometric features of petrochemical equipment according to an embodiment of this application.
[0035] Figure 10 (a) is a schematic diagram of the frequency distribution of the defect depth of the object to be detected in the method for identifying the defect geometric features of petrochemical equipment according to an embodiment of this application.
[0036] Figure 10 (b) is a schematic diagram of the defect volume frequency distribution of the object to be detected in the method for identifying defect geometric features of petrochemical equipment according to an embodiment of this application.
[0037] Figure 11 This is a schematic diagram of the distribution of the original point cloud data of the planar inspection object in the method for identifying the defect geometric features of petrochemical equipment according to an embodiment of this application.
[0038] Figure 12 This is a schematic diagram of the structure of a system for identifying defective geometric features of petrochemical equipment according to an embodiment of this application. Detailed Implementation
[0039] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings and examples, so that the process of how the present invention uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly. It should be noted that, as long as there is no conflict, the various embodiments and features in the various embodiments of the present invention can be combined with each other, and the resulting technical solutions are all within the protection scope of the present invention.
[0040] Furthermore, the steps illustrated in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Also, although a logical order is shown in the flowcharts, in some cases the steps shown or described may be performed in a different order than that shown here.
[0041] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments. Unless the context clearly indicates otherwise, the singular forms “a” and “an” as used herein are also intended to include the plural. It should also be understood that the terms “comprising” and / or “including” as used herein specify the presence of the stated features, integers, steps, operations, units, and / or components, without excluding the presence or addition of one or more other features, integers, steps, operations, units, components, and / or combinations thereof.
[0042] Petrochemical equipment operates under harsh conditions with highly corrosive internal media. Under the influence of sulfur, chlorine, and other factors, the equipment is prone to corrosion and cracking. Studies of pitting corrosion typically focus on observing the morphology of severe pitting in relevant scenarios, but this approach cannot quantitatively assess the severity and distribution of pitting corrosion.
[0043] Methods for quantitative assessment of pitting corrosion include visual inspection, quality loss measurement, depth measurement of pre-selected pits by vertical cross-section, and rating using standard icons.
[0044] ASME G46-94 provides a standard graphical method for describing the severity of pitting corrosion in metallic materials, characterizing the severity of pitting corrosion from three dimensions: distribution density, average pitting area, and average depth. The data obtained by this method is easy to store and can be conveniently and quickly compared with other test data. However, both slicing and measuring the depth of all pits and manual counting are not feasible and are time-consuming and labor-intensive. Furthermore, these methods cannot simultaneously and accurately measure the density of pitting corrosion and the location of pits.
[0045] However, three-dimensional measurement can acquire the surface morphology of an object, offering advantages such as high accuracy and repeatability. This method uses point cloud data to represent the shape of the measured object; that is, by establishing several points and specifying the position of each point in three-dimensional space, the geometry of the object is described. It can simultaneously and accurately measure the location and size information of pitting corrosion pits, making it suitable for data acquisition in pitting corrosion statistical analysis or rating work. However, in the post-processing of point cloud data, a reference plane needs to be manually selected to calculate the required geometric information, such as the opening area, volume, and depth of defects. For densely distributed pitting corrosion, this process is time-consuming and laborious, hindering the acquisition of statistical analysis results for surface defects, especially for non-planar point clouds or point cloud data with uncertain spatial positions, where reference plane selection is difficult. Furthermore, the relative error of the reference plane selection operation gradually increases as the pitting corrosion area decreases, making it difficult to achieve objective and accurate evaluation.
[0046] Furthermore, deep learning methods have received considerable attention for defect identification and counting. These methods are widely used in 3D object detection tasks, but they require a large amount of relevant data for model training, which also necessitates high-quality manual annotation. Data acquisition and annotation hinder the development of this approach. In addition, deep learning methods require data downsampling, making it difficult to capture the features of small targets and prone to missed detections on smaller pitting defects.
[0047] In summary, the present invention provides a point cloud data post-processing scheme based on the geometric features of petrochemical equipment to solve one or more of the problems in the existing technology, namely, that traditional pitting detection technology cannot simultaneously obtain geometric feature information such as the density, location, initiation area and depth of pitting, and that the selection of reference surface is difficult when post-processing point cloud data through three-dimensional measurement.
[0048] To address one or more of the technical problems in the prior art, this invention provides a method and system for identifying the geometric features of defects in petrochemical equipment. This method and system achieve statistical analysis of surface defects in equipment by performing attitude adjustment, geometric parameter determination, surface curvature calculation, defect location determination, and defect volume calculation on the obtained point cloud data.
[0049] Example 1
[0050] Figure 1 This is a flowchart illustrating the steps of a method for identifying defective geometric features in petrochemical equipment, according to an embodiment of this application. The following refers to... Figure 1 The specific steps of the method for identifying the defect geometric features of petrochemical equipment (also known as the "defect feature identification method") described in the embodiments of the present invention will be explained.
[0051] Step S110 obtains raw point cloud data characterizing the surface features of the petrochemical equipment to be tested, and adjusts the attitude of the raw point cloud data to obtain the first point cloud data.
[0052] In this embodiment of the invention, due to the different types of petrochemical equipment to be tested, the area to be tested in the petrochemical equipment is either a planar object or a cylindrical object.
[0053] When the object in the detected region is a plane, the normal vector of the current plane is calculated based on the original point cloud data of the plane. The attitude of the original point cloud data is then adjusted using the obtained plane normal vector to obtain the first point cloud data.
[0054] For an example of raw point cloud data for planar object detection, see [link to example]. Figure 11 .
[0055] When adjusting the pose of the original point cloud data of the plane detection object, the normal vector of the current plane is calculated based on the original point cloud data of the plane. The original point cloud data is then translated and rotated using the current normal vector to obtain the first point cloud data.
[0056] In this embodiment of the invention, when adjusting the attitude of the original point cloud data of the planar detection object, the object can be rotated first and then translated, or translated first and then rotated, to obtain the first point cloud data after attitude optimization and adjustment.
[0057] In one embodiment, when adjusting the pose of the raw point cloud data of a planar detection object, the pose adjustment is performed according to the following method for the process of rotation followed by translation:
[0058] The initial plane fitting is performed on the original point cloud data of the current plane detection object, and the general equation of the fitted plane is obtained, thereby obtaining the normal vector of the initially fitted plane.
[0059] Using the normal vector of the initially fitted plane, calculate the offset angle of the current normal vector relative to the x-axis or y-axis of the original point cloud in the Cartesian coordinate system, and denot it as the first offset angle. Then, use this first offset angle to rotate the original point cloud data so that the original point cloud is transformed into a state where the plane normal is parallel to the xoz-axis or yoz-axis plane.
[0060] A second-order plane fitting is performed on the point cloud data after planar rotation, and the general equation of the fitted plane is obtained, thereby obtaining the normal vector of the second-order fitted plane.
[0061] Using the normal vector of the quadratic fitting plane, calculate the offset angle of the current normal vector relative to the z-axis of the rotated point cloud in the Cartesian coordinate system, denoted as the second offset angle. Then, use this second offset angle to rotate the point cloud data after plane rotation to obtain the first point cloud data, so that the rotated point cloud is transformed into a state where the plane normal is parallel to the z-axis.
[0062] It should be noted that the attitude adjustment process of the original point cloud data can first be performed with the plane formed by any two coordinate axes in the Cartesian coordinate system as the reference for the initial rotation, and then the original point cloud data after the initial rotation is rotated a second time with the last coordinate axis as the reference, thereby obtaining the first point cloud data after attitude optimization and adjustment. This embodiment of the invention does not impose a specific limitation on the order of the coordinate axes involved in the attitude adjustment process; those skilled in the art can design it according to the actual situation.
[0063] When the object in the detected area is a cylinder, the offset features of the original point cloud are calculated based on the original point cloud data of the cylinder. The attitude of the original point cloud is then adjusted using the currently obtained offset features to obtain the first point cloud data.
[0064] For an example of raw point cloud data for a cylindrical object being inspected (e.g., the shell of a pipe or heat exchanger), see [link to example]. Figure 2 The left and right images in the middle.
[0065] When adjusting the pose of the original point cloud data of the cylindrical object being detected, this embodiment of the invention can perform several (e.g., more than or equal to 2) slices along the z-axis (see example of slice point set). Figure 3 The process involves sequentially fitting a two-dimensional ellipse to the point set obtained from each slice, thus obtaining the center of each fitted ellipse. (See [link to documentation]). Figure 4 (a) and Figure 4 (b) Then, based on at least two fitted elliptical centers, obtain the offset angle of the line connecting the centers relative to one of the coordinate axes (x-axis or y-axis); adjust the position of the original point cloud of the cylinder based on the offset angle. Repeat this process at least once more, rotating the axis of the cylinder to be detected until it is parallel to the z-axis. For the point cloud data after attitude optimization, see [link to relevant documentation]. Figure 5 .
[0066] In this embodiment of the invention, the offset angle is obtained by: based on the center coordinates of at least two fitted ellipses, and according to the trigonometric function relationship, calculating the offset angle of the original point cloud of the cylinder in one of the coordinate axes in the Cartesian coordinate system.
[0067] When adjusting the pose of the raw point cloud data of the cylindrical object, the following method should be used:
[0068] For the original point cloud data of the current cylindrical object, slice it at least twice along the z-axis, and fit a two-dimensional ellipse to the point set of the at least two slices to obtain the center of the fitted ellipse of the at least two slices.
[0069] Based on the center of the current slice obtained at least twice, calculate the offset angle of the current center line relative to the x-axis or y-axis of the original point cloud in the Cartesian coordinate system, and record it as the third offset angle. Then, use the third offset angle to rotate the original point cloud data once, so that the original point cloud is transformed into a state where the axis of the cylinder is parallel to the xoz-axis or yoz-axis plane.
[0070] For the point cloud data that has been rotated once, continue to slice it at least twice along the z-axis, and continue to fit a two-dimensional ellipse to the point set of the current at least two slices to obtain the center of the fitted ellipse of the at least two slices.
[0071] Based on the center of the at least two slices obtained so far, calculate the offset angle of the line connecting the current center relative to the z-axis of the point cloud after one rotation in the Cartesian coordinate system, and denot it as the fourth offset angle. Then, use this fourth offset angle to perform a second rotation on the point cloud data after one rotation to obtain the first point cloud data, so that the point cloud after one rotation is transformed into a state where the axis of the cylinder is parallel to the z-axis.
[0072] After completing the pose optimization and adjustment of the original point cloud data, proceed to step S120.
[0073] Step S120: Based on the first point cloud data obtained in step S110, create a reference point cloud containing the location information of all point cloud points.
[0074] When the object being detected is a plane, step S120 will directly use the first point cloud data of the plane detection object as the reference point cloud, and then proceed to step S130.
[0075] When the object to be detected is a cylinder, step S120 first extracts the target cross-sectional plane from the first point cloud data of the cylinder detection object obtained in step S110, and performs two-dimensional planar circle fitting on the target cross-sectional plane to obtain the center and radius of the current fitted circle; then, using the center and radius of the current fitted circle as the reference (limit), the first point cloud data obtained in step S110 is expanded in a plane to obtain the reference point cloud.
[0076] In this embodiment of the invention, the extraction process of the target cross-sectional plane can be performed by slicing the cylindrical point cloud data (i.e., the first point cloud data) after attitude transformation in S110 along a coordinate axis parallel to the axis (e.g., along the z-axis) to use the current sliced point set as the target interception plane. Alternatively, the cylindrical point cloud data (i.e., the first point cloud data) after attitude transformation in S110 can be projected onto a plane perpendicular to the z-axis to use the projected point set as the target cross-sectional plane.
[0077] Then, a two-dimensional planar circle fitting is performed on the current target cross-section plane using, for example, the least squares method, and the (two-dimensional) center coordinates and radius of the fitted circle are obtained.
[0078] In the process of unfolding the first point cloud data in a plane using the two-dimensional center coordinates and radius of the currently fitted circle as a reference: First, based on the center coordinates, the axis of the first point cloud data of the cylinder is moved until it coincides with the z-axis. This axis is called the first coordinate axis. The other two coordinate axes (x-axis and y-axis) are called the second and third coordinate axes, and the plane formed by the second and third coordinate axes is denoted as the fundamental plane. Then, using the second and third coordinate axes, the first point cloud data is divided into four parts according to the quadrants. Coordinate transformation is performed on each point cloud point in a clockwise or counterclockwise order, thereby converting the first point cloud data of the cylinder into a planar point cloud, which is the reference point cloud. The unfolded point cloud data is shown below. Figure 6 .
[0079] Specifically, based on the two-dimensional center coordinates of the currently fitted circle, the axis of the first point cloud data of the current cylinder is aligned with the first coordinate axis representing the z-axis in the Cartesian coordinate system, and the other two coordinate axes in the Cartesian coordinate system are marked as the second coordinate axis and the third coordinate axis, respectively.
[0080] Then, the plane formed by the second and third coordinate axes is called the basic plane. Based on the center coordinates of the current fitted circle under the basic plane (i.e., two-dimensional center coordinates), the first point cloud data is divided into four parts in the basic plane according to the quadrant. The first point cloud data of the cylinder is converted into the reference point cloud by performing coordinate transformation on each point cloud point.
[0081] In the first embodiment, during the coordinate transformation of each point cloud point in a sequential manner, the coordinate value of each point cloud point on the first coordinate axis (z-axis) remains unchanged. The arc length of the cylinder on the cross section where each point cloud point is located is used as the coordinate value of that point on the second coordinate axis representing the y-axis, and the distance of each point cloud point to the cylinder axis is used as the coordinate value of that point on the third coordinate axis representing the x-axis. Alternatively,
[0082] In the second embodiment, during the process of performing coordinate transformation on each point cloud point in a counterclockwise order, the coordinate value of each point cloud point on the first coordinate axis remains unchanged, the arc length of the cylinder on the cross section where each point cloud point is located is taken as the coordinate value of the point on the third coordinate axis representing the y-axis, and the distance of each point cloud point to the cylinder axis is taken as the coordinate value of the point on the second coordinate axis representing the x-axis.
[0083] Thus, after completing the reference point cloud of the cylindrical object, the process proceeds to step S130.
[0084] Step S130: From the reference point cloud obtained in step S120, select the point cloud points that belong to defects to obtain the defect point cloud.
[0085] In this embodiment of the invention, the defect point cloud contains information about the distribution location characteristics of each defect region.
[0086] In step S130, firstly, by calculating the curvature of each point cloud point in the reference point cloud obtained in step S110, point cloud points belonging to defects are selected, and thus the initial defect point cloud is obtained by retaining the defect point cloud points.
[0087] Then, step S130 further performs cluster analysis on the defect point cloud points in the initial defect point cloud, determines the number of defect regions based on the clusters, and thus obtains a defect point cloud used to distinguish the distribution location of each defect region. Noise points identified during the cluster analysis are treated as non-defect points.
[0088] In a specific example of filtering point cloud points, the first step is to calculate the curvature of each point in the baseline point cloud, as well as the average curvature of all points.
[0089] Then, based on the currently calculated average curvature, a curvature threshold is determined to distinguish the curvature of defective points from that of normal points (which are non-defective points). The curvature threshold can be calculated using the following expression:
[0090]
[0091] Where threshold represents the curvature threshold, δ i denoted by , where represents the curvature of the i-th point in the reference point cloud, m represents the number of all point points in the reference point cloud, and α represents a preset coefficient.
[0092] Finally, point cloud points in the baseline point cloud with curvature exceeding the above curvature threshold are taken as defect point cloud points. All defect point cloud points are retained and all point cloud points with curvature not exceeding the curvature threshold are deleted to obtain the initial defect point cloud.
[0093] In this embodiment of the invention, an example of an initial defect point cloud composed of all the selected defect point cloud points can be found in [reference needed]. Figure 7 .
[0094] In addition, during the process of screening defective point cloud points, this embodiment of the invention can also set a standard straight line in the reference point cloud, calculate the distance of each point cloud point to the standard straight line, retain point cloud points whose distance exceeds the distance threshold, and delete all point cloud points whose distance does not exceed the distance threshold, thereby obtaining the initial defective point cloud.
[0095] After obtaining the initial defect point cloud, step S130 further employs, for example, a density-based DBSCAN algorithm to perform cluster analysis on the retained point set in the defect point cloud plane. Simultaneously, noise points identified by the DBSCAN algorithm are excluded from subsequent analysis (i.e., noise points are treated as normal points). This allows for the determination of clustering results obtained after the cluster analysis, such as... Figure 8 As shown, a defect point cloud is obtained to determine the number of defects and distinguish the specific location of each defect.
[0096] Continue to refer to Figure 1 After obtaining the defect point cloud, proceed to step S140.
[0097] Step S140: Based on the defect point cloud obtained in step S130, directly calculate the defect feature parameters. These defect feature parameters include: the defect density of the current area to be detected, and the opening area, depth, and volume of each defect area.
[0098] In the first embodiment of step S140, based on the coordinates of each defect point in the defect point cloud, the difference between the maximum and minimum coordinates of the point cloud points in each defect region of the clustered defect point cloud on the z-axis is used as the depth of the corresponding defect region.
[0099] In other words, for each defect region, it is necessary to determine the maximum and minimum z-axis coordinate values among all defect point cloud points within the corresponding defect region. The difference between the maximum and minimum z-axis coordinate values of each defect region is used as the defect depth of the corresponding defect region. Figure 10 (a) shows the frequency distribution of defect depth features for all defect regions in a region to be detected.
[0100] In the second embodiment of step S140, the convex hull of each defect region point set in the defect point cloud is determined according to the coordinates of each defect point cloud point, and the volume of the convex hull of each defect region is used as the volume of the corresponding defect region.
[0101] Figure 9 An example diagram showing the convex hull of a point set in a defective region is presented.
[0102] Figure 10 (b) shows the frequency distribution of defect volume characteristics of all defect regions in a certain area to be detected.
[0103] In the third embodiment of step S140, each defect region point set in the defect point cloud is projected onto a plane perpendicular to the z-axis, and the corresponding convex hull is determined for each region's projected point set, thereby using the convex hull of the projected point set as the opening area of the corresponding defect region.
[0104] In the fourth embodiment of step S140, all defect region point sets in the defect point cloud are projected onto a plane perpendicular to the z-axis, and the area of the minimum bounding rectangle of all points after projection is determined. Then, the defect distribution density of the current detected region is obtained by calculating the ratio of the number of current defect regions obtained in step S130 to the area of the minimum bounding rectangle.
[0105] Example 2
[0106] Based on the defect feature recognition method of Embodiment 1 above, this embodiment of the invention provides a specific example of applying the defect feature recognition method to a cylindrical object for inspection. The defect feature recognition method includes the following steps:
[0107] S1: Point cloud data attitude adjustment:
[0108] ① Slice the point cloud of the cylindrical surface to be fitted twice along the z-axis, see... Figure 3 The resulting point set lies on an ellipse within the corresponding slice plane. Ellipse fitting is then performed on the point sets from both slices, as shown in the diagram. Figure 4 (a) Figure 4 (b). The centers (x1, y1, z1) and (x2, y2, z2) of the two ellipses can be found.
[0109] ② The center of the obtained ellipse lies on the axis of the point cloud of the cylindrical surface to be fitted. Calculate the angle of its axis relative to the y-axis based on the coordinates of the center. Rotate the point cloud according to the obtained offset angle so that the axis is parallel to the xoz plane;
[0110] ③ Repeat ① to obtain the coordinates of the two ellipse centers (x3, y3, z3) and (x4, y4, z4). Repeat ② to calculate the angle of the axis relative to the z-axis. Rotate the point cloud according to the obtained offset angle until its axis is parallel to the z-axis. The adjusted point cloud is shown below. Figure 5 .
[0111] The method for fitting the ellipse is to use the least squares method.
[0112] The two ellipse parameters obtained are:
[0113] Ellipse 1: The center coordinates are (3.2267, -3.2402, 5.0017);
[0114] Ellipse 2: The center coordinates are (-3.2243, 3.2355, -4.9978);
[0115] Based on the coordinates of the centers of at least two ellipses, the offset angle of the cylindrical point cloud relative to one of the coordinate axes is calculated using trigonometric functions.
[0116] The method for obtaining the offset angle is as follows:
[0117] The angle between the cylindrical axis and the x-axis
[0118] Based on the calculated offset angle, the point cloud is transformed until its axis coincides with the xoz plane.
[0119] Repeat the process of slicing, ellipse fitting, and calculating the offset angle to rotate the point cloud until its axis is parallel to the z-axis.
[0120] The ellipse obtained in this process is:
[0121] Ellipse 1, center coordinates (75.2154, 33.1361, 75.0022);
[0122] Ellipse 2, center coordinates (88.9256, 33.1526, 90.1229);
[0123] The angle between the cylindrical axis and the z-axis
[0124] Based on the calculated offset angle, the point cloud is transformed so that its axis is parallel to the z-axis.
[0125] S2: Surface unfolding:
[0126] For the cylindrical point cloud, after the S1 attitude transformation, the cylindrical point cloud is sliced along a coordinate axis parallel to its axis. The point set obtained from the slice is then fitted with a two-dimensional planar circle to obtain the center and radius of the fitted circle. Based on the center and radius of the fitted circle, the cylindrical point cloud is unfolded.
[0127] The method for fitting the circle is to use the least squares method. The radius of the fitted circle is 49.9836, and the coordinates of the center are (4.9236, 33.0740).
[0128] The unfolding method is as follows: the point cloud is divided into four quadrants using the x and y axes. Following the order of the first quadrant, fourth quadrant, third quadrant, and second quadrant, the cylindrical point cloud is unfolded into a planar point cloud through coordinate transformation. During the coordinate transformation (after transformation, the point cloud is perpendicular to the x-axis), the coordinate values on the z-axis remain unchanged; the arc length of the cylinder replaces the coordinate values on the y-axis; and the distance from the point to the axis line replaces the coordinate values on the x-axis.
[0129] Flattened dot clouds, such as Figure 6 As shown.
[0130] S3: Defect Detection
[0131] Compute the curvature of the obtained planar / cylindrical point cloud. Based on the distribution of surface curvature, remove the point cloud on the fitted plane, retaining the point cloud at the defect location. (See...) Figure 7 The retained points are clustered to determine the number of defects and their corresponding locations.
[0132] The method for removing point clouds on the plane is as follows: calculate the average surface curvature of all points, set a coefficient to calculate a threshold for curvature, and retain points whose surface curvature exceeds the threshold.
[0133] The clustering method employed was a density-based DBSCAN algorithm to cluster the defect point set. Noise points identified by the DBSCAN algorithm were excluded from subsequent analysis. The results of the density clustering are shown below. Figure 8 .
[0134] S4: Calculation of characteristic parameters:
[0135] After determining the location and number of defects through density clustering, the depth and volume are calculated sequentially based on the point cloud data for each defect location. The depth is calculated using the difference between the maximum and minimum z-axis coordinates of that point cloud portion; the convex hull of this point cloud portion is then calculated, and its volume is used as the volume of the defect. The convex hull is the minimal polyhedron that contains all points in the point cloud. (See...) Figure 9 After calculating the volume and depth, a statistical analysis of the frequency distribution of the defect depth and volume on the plane was performed, as shown in the figure. Figure 10 (a) and Figure 10 (b)
[0136] Example 3
[0137] Based on the defect feature recognition method of Embodiment 1 above, this embodiment of the invention provides a specific example of applying the defect feature recognition method to a planar inspection object. The defect feature recognition method includes the following steps:
[0138] S1: Point cloud data attitude adjustment:
[0139] By performing plane fitting on the plane, the general equation of the plane is obtained as follows:
[0140] 0.4771x-0.4770y+0.7381z-10.0006=0
[0141] (0.4771, -0.4770, -10.0006) is the normal vector of the plane. Calculate the offset angle of this normal vector relative to the y-axis: angel_y = arctan(|0.4771 / -0.4770|) = 0.7855.
[0142] Based on the calculated offset angle, the point cloud is transformed until its normal is parallel to the xoz plane.
[0143] By performing plane fitting on the rotated point cloud, the general equation of the plane is obtained as follows:
[0144] 0.6747x+0.0001y+0.7381z-13.6907=0
[0145] (0.6747, 0.0001, 0.7381) is the normal vector of the rotated plane. Calculate the offset angle of this normal vector relative to the z-axis: angel_z = arctan(|0.6747 / 0.7381|) = 0.7405.
[0146] Based on the calculated offset angle, the point cloud is transformed until its normal is parallel to the z-axis.
[0147] S2: Surface unfolding:
[0148] This step is skipped for planar point clouds.
[0149] S3: Defect Detection
[0150] Defect detection is performed on the point cloud after attitude adjustment, and points in the defective areas are retained for subsequent calculations. The defect detection method is as follows: in plane fitting, points belonging to the plane are considered normal parts, and points not belonging to the plane are considered defective parts.
[0151] The remaining implementation steps, such as defect detection and feature parameter calculation, are the same as in Examples 1 and 2.
[0152] Example 4
[0153] Based on the defect feature identification method described in Embodiments 1 to 3 above, this invention provides a computer-readable storage medium. The storage medium stores a computer program, which is executed to run a method for identifying defect geometric features in petrochemical equipment. The computer program is capable of executing computer instructions, which include computer program code. The computer program code can be in the form of source code, object code, executable file, or some intermediate form.
[0154] Computer-readable storage media can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.
[0155] It should be noted that the contents of computer-readable storage media may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, computer-readable storage media may not include electrical carrier signals and telecommunication signals.
[0156] Example 5
[0157] Based on the above-mentioned defect feature identification method, this invention also provides a system for identifying defect geometric features of petrochemical equipment (also referred to as a "defect feature identification system").
[0158] Figure 12 This is a schematic diagram of a system for identifying defective geometric features in petrochemical equipment, according to an embodiment of this application. Figure 12 As shown, the defect feature recognition system includes: a posture optimization module 1201, a reference point cloud generation module 1202, a defect recognition module 1203, and a defect statistical analysis module 1204.
[0159] In this embodiment of the invention, the attitude optimization module 1201 is implemented according to the method described in step S110 above, configured to obtain raw point cloud data characterizing the surface features of the petrochemical equipment to be inspected, and adjust its attitude to obtain first point cloud data; the reference point cloud generation module 1202 is implemented according to the method described in step S120 above, configured to create a reference point cloud containing the position information of all point cloud points based on the first point cloud data; the defect identification module 1203 is implemented according to the method described in step S130 above, configured to filter point cloud points belonging to defects from the reference point cloud, thereby obtaining a defect point cloud; the defect statistical analysis module 1204 is implemented according to the method described in step S140 above, configured to directly calculate defect feature parameters based on the defect point cloud.
[0160] In this embodiment of the invention, the defect characteristic parameters include defect density and the opening area, depth and volume of each defect.
[0161] This invention discloses a method and system for identifying the geometric features of defects in petrochemical equipment. Addressing the problem that visual inspection and two-dimensional image recognition of pitting corrosion cannot accurately and quickly obtain parameters such as depth and volume, this method and system proposes a point cloud data post-processing method based on equipment geometric features. It calculates the offset angle of the cylindrical surface's axis relative to the coordinate axes using a two-slice method, and then obtains the cylindrical surface radius by combining circle fitting, developing a cylindrical surface fitting method insensitive to initial values. It automatically creates a reference point cloud by unfolding a curved surface into a plane, eliminating the need for complex manual selection. It distinguishes defective and normal parts by calculating the curvature of each point, and obtains the location of each defect through density clustering. It calculates the volume and opening area of each defect using coordinates and solving for the convex hull, and calculates the depth of the defect using coordinates, ultimately achieving statistical analysis of defect feature parameters. The entire process of the defect feature identification method provided by this invention requires no manual intervention, automatically creates a reference point cloud, and then performs defect detection and statistical analysis, effectively improving the accuracy of defect detection and the efficiency of automated detection.
[0162] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
[0163] In the description of this invention, unless otherwise stated, "a plurality of" means two or more; the terms "upper," "lower," "left," "right," "inner," "outer," "front end," "rear end," "head," "tail," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance.
[0164] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "connected" and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0165] It should be understood that the embodiments disclosed herein are not limited to the specific structures, processing steps, or materials disclosed herein, but should be extended to equivalent substitutions of these features as understood by those skilled in the art. It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting.
[0166] The phrase "an embodiment" or "an embodiment" used in this specification means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the invention. Therefore, the phrase "an embodiment" or "an embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment.
[0167] While the embodiments disclosed in this invention are as described above, the content is merely for the purpose of facilitating understanding of the invention and is not intended to limit the invention. Any person skilled in the art to which this invention pertains may make any modifications and changes in form and detail of the implementation without departing from the spirit and scope disclosed herein; however, the scope of patent protection of this invention shall still be determined by the scope defined in the appended claims.
Claims
1. A method for identifying a defect geometry of a petrochemical plant, characterized in that, include: The original point cloud data characterizing the surface features of the petrochemical equipment to be inspected is obtained, and its attitude is adjusted to obtain the first point cloud data. Based on the first point cloud data, a reference point cloud containing the location information of all point cloud points is created; The defect point cloud is obtained by filtering point cloud points belonging to defects from the reference point cloud; Based on the defect point cloud, defect feature parameters are directly calculated, including defect density and the opening area, depth, and volume of each defect.
2. The method according to claim 1, characterized in that, The steps of obtaining raw point cloud data characterizing the surface features of the petrochemical equipment to be inspected, and adjusting its attitude to obtain the first point cloud data, include: When the object being detected is a plane, the normal vector of the current plane is calculated based on the original point cloud data of the plane, and the pose of the original point cloud is adjusted using the normal vector to obtain the first point cloud data. When the object being detected is a cylinder, the offset features of the original point cloud are calculated based on the original point cloud data of the cylinder, and the attitude of the original point cloud is adjusted using the offset features to obtain the first point cloud data.
3. The method according to claim 2, characterized in that, The step of calculating the normal vector of the current plane based on the original point cloud data of the plane, and using the normal vector to adjust the attitude of the original point cloud to obtain the first point cloud data, includes: The initial plane fitting is performed on the original point cloud data of the current plane detection object, and the general equation of the fitted plane is obtained, thereby obtaining the normal vector of the initially fitted plane. Using the normal vector of the initially fitted plane, calculate the first offset angle of the current normal vector relative to the x-axis or y-axis of the original point cloud in the Cartesian coordinate system, and then use the first offset angle to rotate the original point cloud data; A quadratic plane fitting is performed on the point cloud data after planar rotation to obtain the general equation of the fitted plane, thereby obtaining the normal vector of the quadratic fitted plane. Using the normal vector of the quadratic fitting plane, calculate the second offset angle of the current normal vector relative to the z-axis of the rotated point cloud in the Cartesian coordinate system. Then, use this second offset angle to rotate the point cloud data after plane rotation to obtain the first point cloud data.
4. The method according to claim 2 or 3, characterized in that, The step of calculating the offset features of the original point cloud based on the original point cloud data of the cylinder, and using the offset features to adjust the attitude of the original point cloud to obtain the first point cloud data, includes: For the original point cloud data of the current cylindrical object, slice it at least twice along the z-axis, and fit a two-dimensional ellipse to the point set of the at least two slices to obtain the center of the fitted ellipse of the at least two slices. Based on the center of the at least two currently obtained slices, calculate the third offset angle of the line connecting the current center relative to the x-axis or y-axis of the original point cloud in the Cartesian coordinate system, and then use this third offset angle to rotate the original point cloud data once. For the point cloud data that has been rotated once, slice it at least twice along the z-axis, and fit a two-dimensional ellipse to the point set of the at least two slices to obtain the center of the fitted ellipse of the at least two slices. Based on the centers of at least two currently obtained slices, calculate the fourth offset angle of the line connecting the current centers relative to the z-axis of the point cloud in the Cartesian coordinate system after one rotation. Then, use this fourth offset angle to perform a second rotation on the point cloud data after one rotation to obtain the first point cloud data.
5. The method according to any one of claims 2 to 4, characterized in that, When the object being detected is a plane, the step of creating a reference point cloud containing the position information of all point cloud points based on the first point cloud data includes: The first point cloud data is used as the reference point cloud.
6. The method according to any one of claims 2 to 5, characterized in that, When the object being detected is a cylinder, the step of creating a reference point cloud containing the position information of all point cloud points based on the first point cloud data includes: Extract the target cross-sectional plane from the first point cloud data, and perform two-dimensional planar circle fitting on the target cross-sectional plane to obtain the center and radius of the currently fitted circle; Using the center and radius of the currently fitted circle as a reference, the first point cloud data is unfolded in a plane to obtain the reference point cloud.
7. The method according to claim 6, characterized in that, The step of expanding the first point cloud data to obtain the reference point cloud, using the center and radius of the currently fitted circle as boundaries, includes: Based on the center coordinates of the currently fitted circle, the axis of the first point cloud data of the current cylinder is aligned with the first coordinate axis representing the z-axis in the Cartesian coordinate system, and the other two coordinate axes in the Cartesian coordinate system are labeled as the second coordinate axis and the third coordinate axis, respectively. The plane formed by the second and third coordinate axes is denoted as the basic plane. Based on the center coordinates of the currently fitted circle under the basic plane, the first point cloud data is divided into four parts according to quadrants within the basic plane. The first point cloud data of the cylinder is converted into a reference point cloud by performing coordinate transformation on each point cloud point, including: During the coordinate transformation of each point cloud point in a sequential manner, the coordinate values of each point cloud point on the first coordinate axis are kept unchanged. The arc length of the cylinder on the cross section containing each point cloud point is used as the coordinate value of that point on the second coordinate axis, which represents the y-axis. The distance of each point cloud point to the axis of the cylinder is used as the coordinate value of that point on the third coordinate axis, which represents the x-axis. During the process of performing coordinate transformation on each point cloud point in a counterclockwise order, the coordinate value of each point cloud point on the first coordinate axis remains unchanged. The arc length of the cylinder on the cross section where each point cloud point is located is taken as the coordinate value of that point on the third coordinate axis representing the y-axis, and the distance of each point cloud point to the cylinder axis is taken as the coordinate value of that point on the second coordinate axis representing the x-axis.
8. The method according to claim 6 or 7, characterized in that, The step of extracting and filtering point cloud points belonging to defects from the reference point cloud to obtain a defect point cloud containing information on the distribution location of each defect includes: The initial defect point cloud is obtained by calculating the curvature of each point in the reference point cloud and filtering out point cloud points that belong to defects. Cluster analysis is performed on the defect point cloud points in the initial defect point cloud, and the number of defects is determined based on the clusters, thereby obtaining the defect point cloud used to distinguish the distribution location of each defect region.
9. The method according to claim 8, characterized in that, Calculate the curvature of each point in the reference point cloud and the average curvature of all points; Based on the average curvature, a curvature threshold is determined to distinguish between the curvature of defective points and the curvature of normal points; Points in the reference point cloud whose curvature exceeds the curvature threshold are taken as defect point cloud points to obtain the initial defect point cloud.
10. The method according to claim 8 or 9, characterized in that, The step of directly calculating the feature parameters of each defect based on the defect point cloud includes: Based on the coordinates of each defect point in the defect point cloud, the difference between the maximum and minimum coordinates of the point in each defect region on the z-axis is taken as the depth of the corresponding defect region. Based on the coordinates of each defect point in the defect point cloud, the convex hull of each defect region point set is determined, and the volume of each convex hull is used as the volume of the corresponding defect region. Project each defect region point set onto a plane perpendicular to the z-axis, determine the corresponding convex hull for the projected point set, and use the projected convex hull as the opening area of the corresponding defect region. Project all defect region points onto a plane perpendicular to the z-axis, determine the area of the minimum bounding rectangle of all projected points, and then obtain the defect distribution density of the current detected region by calculating the ratio of the number of current defect regions to the area of the minimum bounding rectangle.
11. A computer-readable storage medium, characterized in that, It includes a series of instructions for performing the method steps as described in any one of claims 1 to 10.
12. A system for identifying defective geometric features of petrochemical equipment, characterized in that, include: The attitude optimization module is configured to obtain raw point cloud data characterizing the surface features of the petrochemical equipment to be inspected, and adjust its attitude to obtain the first point cloud data. The reference point cloud generation module is configured to create a reference point cloud containing the location information of all point cloud points based on the first point cloud data. The defect identification module is configured to filter point cloud points belonging to defects from the reference point cloud to obtain the defect point cloud; The defect statistical analysis module is configured to directly calculate defect feature parameters based on the defect point cloud. The defect feature parameters include defect density and the opening area, depth and volume of each defect.