Online detection method for GIS shell flange surface flatness based on structured light three-dimensional point cloud

Abstract

The present application relates to the technical field of data processing, more particularly, the present application relates to a GIS shell flange surface flatness online detection method based on structured light three-dimensional point cloud, the method comprises: using a scanner to scan the GIS shell flange surface all around, obtaining an initial three-dimensional point cloud set containing a plurality of sampling points, obtaining the adaptive search radius of each sampling point, and constructing a neighborhood sampling point set based on the adaptive search radius, performing eigenvalue decomposition on the neighborhood sampling point set to obtain a local topological dispersion index, calculating the fitting weight coefficient of each sampling point according to the local topological dispersion index, and constructing a weighted objective function to solve the optimal reference plane equation; according to the local topological dispersion index, the effective sampling points are screened out, the flatness detection value of the flange surface is obtained based on the distance from the effective sampling points to the optimal reference plane, and the sealing risk assessment of the flange surface is completed, and the accuracy of the flange surface flatness detection is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to an online method for detecting the flatness of GIS shell flange surfaces based on structured light 3D point clouds. Background Technology

[0002] As a core component of high-voltage power transmission systems, the processing quality of the flange faces at the casing connections of GIS (Gas Insulated Switchgear) is crucial. The flatness of the flange face directly determines the compression uniformity of the sealing ring. If the flatness exceeds the standard, it can lead to minor SF6 gas leakage, or even serious power safety accidents. Currently, in the online inspection stage after flange face finishing, the mainstream technology uses a line structured light scanner integrated on a robotic arm to acquire a three-dimensional point cloud of the casing surface. Then, the traditional least squares method is used to fit a theoretical reference plane to calculate the deviation of each sampling point from this plane, thereby determining the flatness.

[0003] GIS flanges are typically discontinuous and complex structures with densely packed bolt holes and sealing grooves of varying depths. These bolt holes and sealing grooves are formed through machining, resulting in chamfered or rounded edges. This causes a height shift in the sampling points at the bolt hole edges and sealing groove boundaries relative to the flange reference plane. Consequently, in traditional least-squares plane fitting, the sampling points at the bolt hole edges and sealing groove boundaries exhibit a significant residual pulling effect, causing a slight tilt or height shift in the fitted theoretical reference plane. This ultimately leads to misjudgments or missed detections in the flange flatness inspection results.

[0004] Meanwhile, the GIS flange face is a metal surface with significant specular reflection characteristics. When using structured light scanning to collect point cloud data of the flange face, the scanning results will generate a large number of local clusters of flying point noise due to factors such as the micro-texture and oil stains on the metal surface. The three-dimensional coordinates of this type of noise deviate significantly from the actual flange surface, which will also cause a slight tilt or height shift in the fitted theoretical reference plane, ultimately resulting in misjudgment or missed detection of the flange flatness detection results. Summary of the Invention

[0005] To address the issue that flying point noise generated by discontinuous complex topography and metallic mirror reflection on GIS flange surfaces can lead to residual drag effects in traditional least-squares plane fitting, causing tilting or translation of the theoretical reference plane and ultimately resulting in errors in flange flatness detection, this invention proposes an online flatness detection method for GIS shell flange surfaces based on structured light 3D point clouds. This method includes the following steps:

[0006] A scanner is used to perform a full-circumference scan of the flange surface of the GIS shell to obtain an initial three-dimensional point cloud set containing several sampling points;

[0007] Based on the distance between each sampling point and its nearest neighbors, an adaptive search radius is obtained for each sampling point; based on the adaptive search radius, a neighborhood sampling point set for each sampling point is obtained; the neighborhood covariance matrix of the neighborhood sampling point set is decomposed into eigenvalues ​​to obtain several eigenvalues; based on the eigenvalues, a local topological dispersion index for each sampling point is obtained.

[0008] Based on the local topological dispersion index, obtain the fitting weight coefficient for each sampling point; construct a weighted objective function based on the fitting weight coefficient to solve the optimal reference plane equation;

[0009] Valid sampling points are selected based on the local topological dispersion index; the flatness detection value of the flange face is obtained based on the distance from each valid sampling point to the optimal reference plane; and the flange face sealing risk assessment is completed based on the flatness detection value.

[0010] The innovation of this invention lies in obtaining a neighborhood sampling point set for each sampling point by adaptively adjusting the search radius of each sampling point. Then, eigenvalue decomposition is performed on the neighborhood covariance matrix of the neighborhood sampling point set to obtain the local topological dispersion index for each sampling point. This effectively distinguishes between sampling points in flat areas of the flange surface and those at the bolt hole edges and sealing groove boundaries of the flange. Based on the local topological dispersion index, the fitting weight coefficients for each sampling point are obtained to construct a weighted objective function, which in turn solves the optimal reference plane equation. This improves the accuracy of the optimal reference plane and avoids the influence of sampling points at the bolt hole edges and sealing groove boundaries of the flange on the fitted plane.

[0011] Preferably, obtaining the adaptive search radius for each sampling point includes:

[0012] ;

[0013] In the formula, The adaptive search radius represents the i-th sampling point; This represents the number of nearest neighbors of the i-th sampling point; Represents the spatial coordinates of the i-th sampling point; This represents the spatial coordinates of the j-th nearest neighbor of the i-th sampling point.

[0014] This ensures that there are enough sampling points for feature extraction in sparse areas, while avoiding the introduction of feature points from other areas into dense areas.

[0015] Preferably, the acquisition of the nearest neighbors of the i-th sampling point includes:

[0016] Preset number of nearest neighbors ; Obtain the Euclidean distance between the i-th sampling point and each other sampling point, and record the K sampling points closest to the i-th sampling point as the nearest neighbors of the i-th sampling point.

[0017] Preferably, obtaining the neighborhood sampling point set of each sampling point includes:

[0018] Centered on the i-th sampling point, all sampling points within its adaptive search radius are taken as the neighborhood sampling point set of the i-th sampling point.

[0019] Preferably, the eigenvalue decomposition of the neighborhood covariance matrix of the neighborhood sampling point set yields several eigenvalues, including:

[0020] Based on the neighborhood sampling point set of the i-th sampling point, calculate its neighborhood covariance matrix, and perform eigenvalue decomposition on the neighborhood covariance matrix to obtain three eigenvalues. Then, sort the three eigenvalues ​​in descending order. ; The first eigenvalue represents the neighborhood covariance matrix of the i-th sampling point; The second eigenvalue represents the neighborhood covariance matrix of the i-th sampling point. The third eigenvalue represents the neighborhood covariance matrix of the i-th sampling point.

[0021] Preferably, obtaining the local topological dispersion index for each sampling point includes:

[0022] ;

[0023] In the formula, The local topological dispersion index represents the i-th sampling point; The first eigenvalue represents the neighborhood covariance matrix of the i-th sampling point; The second eigenvalue represents the neighborhood covariance matrix of the i-th sampling point. The third eigenvalue represents the neighborhood covariance matrix of the i-th sampling point; exp() represents an exponential function with the natural constant as the base.

[0024] It can accurately quantify the probability of a sampling point belonging to the flat area of ​​the flange surface from the three-dimensional spatial distribution dimension.

[0025] Preferably, obtaining the fitting weight coefficients for each sampling point includes:

[0026] ;

[0027] In the formula, The fitting weight coefficients represent the i-th sampling point; The local topological dispersion index represents the i-th sampling point; This represents the threshold for judgment; represents the steepness factor; exp() represents an exponential function with the natural constant as the base.

[0028] Preferably, the step of constructing a weighted objective function based on the fitted weight coefficients and solving the optimal reference plane equation includes:

[0029] Preset plane equation: ,in, Let be the plane parameters to be solved; In the formula, Represents the weighted objective function; The fitting weight coefficients represent the i-th sampling point; This represents the number of sampling points in the 3D point cloud set. Represents the spatial coordinates of the i-th sampling point; || represents the absolute value function;

[0030] exist At that time, for the plane parameters in the weighted objective function Taking the partial derivatives and setting them to zero, we obtain four partial derivatives. These four partial derivatives are then used as a system of linear equations, which we solve to obtain... Then, substitute the equations into the preset plane equations to obtain the optimal reference plane equations.

[0031] It fundamentally eliminates the fitting drift caused by holes and slots, and improves the accuracy of plane fitting.

[0032] Preferably, obtaining the flatness detection value of the flange surface includes:

[0033] ;

[0034] In the formula, The flatness test value represents the flange face; This represents the maximum distance among all valid sampling points to the optimal reference plane; This represents the minimum distance among all valid sampling points to the optimal reference plane.

[0035] This improves the accuracy of flatness test values.

[0036] Preferably, the step of completing the flange face sealing risk assessment based on the flatness detection value includes:

[0037] The system has a preset flatness threshold T2. If the flatness detection value of the flange face is greater than the flatness threshold, the system will automatically determine that the flange has a sealing risk and trigger an alarm mechanism to prompt the staff to handle it.

[0038] The present invention has the following beneficial effects: First, the present invention performs eigenvalue decomposition on the neighborhood covariance matrix of the neighborhood sampling point set of each sampling point to obtain the local topological dispersion index of each sampling point, which can effectively distinguish the sampling points of the flat area of ​​the flange surface and the sampling points of the bolt hole edge and the sealing groove boundary of the flange; based on the local topological dispersion index, the fitting weight coefficient of each sampling point is obtained to construct a weighted objective function, and then the optimal reference plane equation is solved, which improves the accuracy of the optimal reference plane, avoids the influence of the sampling points of the bolt hole edge and the sealing groove boundary of the flange on the fitting plane, and greatly improves the sealing reliability of the flange surface. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the steps of the online detection method for the flatness of a GIS shell flange surface based on structured light 3D point cloud according to an embodiment of the present invention.

[0040] Figure 2 This is a schematic diagram of the reference surface fitted by the traditional algorithm;

[0041] Figure 3 This is a schematic diagram of the optimal reference plane obtained by the solution of the present invention. Detailed Implementation

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

[0043] Please see Figure 1 The diagram illustrates a flowchart of an online detection method for the flatness of a GIS shell flange surface based on structured light 3D point clouds, according to an embodiment of the present invention. The method includes the following steps:

[0044] S001. Use a scanner to perform a full-circumference scan of the GIS shell flange surface to obtain an initial three-dimensional point cloud set containing several sampling points.

[0045] In this embodiment of the invention, an industrial-grade line structured light scanner is used to perform a full-circumference rotational scan on the flange surface of the GIS shell to obtain an initial three-dimensional point cloud set, wherein the initial three-dimensional point cloud set contains several sampling points.

[0046] S002. Determine the adaptive search radius based on the distribution of nearest neighbors of each sampling point, and construct a neighborhood sampling point set based on the adaptive search radius. Perform eigenvalue decomposition on the neighborhood sampling point set to obtain the local topological dispersion index.

[0047] It should be noted that the purpose of this invention is to distinguish between sampling points at the bolt hole edge and sealing groove boundary, flying noise points, and sampling points in the flat area of ​​the flange surface in the initial three-dimensional point cloud set. It is known that principal component analysis of the neighborhood point set of the sampling point is a quantitative expression of the local geometric feature distribution of the sampling point in three-dimensional space. The distribution law of the three eigenvalues ​​obtained by eigenvalue decomposition of the covariance matrix of the neighborhood point set can directly reflect the geometric characteristics of the region where the corresponding sampling point is located. This characteristic enables it to accurately distinguish between the flat area of ​​the flange, the hole edge, the sealing groove boundary, and flying noise points.

[0048] In online flange inspection scenarios, the sampling density of point clouds in the flat areas and hole edges of flanges exhibits significant non-uniformity due to fluctuations in the movement speed of the scanner's robotic arm, changes in the sensor's scanning angle, and differences in the reflectivity of the metal surface. If a fixed search radius is directly used to select the neighborhood point set of the sampling point for analysis, in sparsely sampled areas, the fixed radius will result in insufficient effective points in the neighborhood point set, and the covariance matrix constructed based on this point set will be easily distorted by interference from a single noise point. In densely sampled areas, the fixed radius will result in too many neighborhood point sets, making it difficult to ensure that all neighborhood points are within the same geometric plane, and easily mixing in sampling points with heterogeneous features, causing misjudgment of local geometric features. Therefore, this invention calculates the local average point spacing of the sampling points and dynamically adjusts the search radius for each sampling point, enabling precise physical matching between local geometric feature analysis and local sampling density. This avoids the introduction of irrelevant features and effectively suppresses the interference of microscopic noise, ensuring the accuracy of geometric feature analysis in different density areas.

[0049] In this embodiment of the invention, the number of nearest neighbors is preset. In other embodiments, the implementer may preset the number of nearest neighbor points according to the specific implementation method; obtain the Euclidean distance between the i-th sampling point and each other sampling point, and record the K sampling points closest to the i-th sampling point as the nearest neighbors of the i-th sampling point;

[0050] Obtain the adaptive search radius for each sampling point:

[0051] ;

[0052] In the formula, The adaptive search radius represents the i-th sampling point; This represents the number of nearest neighbors of the i-th sampling point; Represents the spatial coordinates of the i-th sampling point; Represents the spatial coordinates of the j-th nearest neighbor of the i-th sampling point;

[0053] The value reflects the average distance between a sampling point and its nearest neighbors. The larger the value, the more sparse the sampling point is in, thus automatically expanding the search radius to ensure sufficient geometric support information is obtained. Conversely, the smaller the value, the more dense the sampling point is in, thus reducing the radius to preserve the fine structure of the edge.

[0054] It should be noted that the known flat region of the flange surface has two-dimensional planar topological characteristics. Its surface is flat and continuous without abrupt changes. The sampling points in the region are concentrated only in the two-dimensional plane, and there is no obvious positional fluctuation in the normal direction perpendicular to the reference plane. After performing eigenvalue decomposition on the covariance matrix of the sampling points in this region based on principal component analysis, the principal component energy of the sampling points will be concentrated in the plane dominant direction corresponding to the first two eigenvalues. The third eigenvalue corresponds to the normal distribution dimension, which is close to zero due to its extremely small dispersion.

[0055] The bolt holes and sealing grooves of the flange are formed by machining, and their edges have machined chamfers or rounded corners. This causes the normal vector of the sampling points in the edge region to change drastically with the hole contour. The height value of the sampling point is offset relative to the flange reference plane. At the same time, the sampling points at the edge of the hole and groove are linearly discrete along the circumference of the hole, rather than being uniformly distributed in a two-dimensional plane. After performing eigenvalue decomposition on the covariance matrix of the sampling points in the edge region based on principal component analysis, the principal component energy of the sampling points will be concentrated in the first eigenvalue, which is significantly increased, and the second eigenvalue is small, which represents the high dispersion of the linear extension direction of the edge. The third eigenvalue corresponds to a significant increase in the normal distribution dimension.

[0056] Therefore, this invention performs eigenvalue decomposition on the neighborhood covariance matrix of each sampling point to obtain three eigenvalues, and constructs a topological dispersion index based on the three eigenvalues ​​to characterize the probability that the sampling point belongs to the flat area of ​​the flange surface.

[0057] In this embodiment of the invention, taking the i-th sampling point as the center, all sampling points within its adaptive search radius are considered as the neighborhood sampling point set of the i-th sampling point; based on the neighborhood sampling point set of the i-th sampling point, its neighborhood covariance matrix is ​​calculated, and the neighborhood covariance matrix is ​​decomposed into eigenvalues ​​to obtain three eigenvalues, which are then sorted in descending order. ; The first eigenvalue represents the neighborhood covariance matrix of the i-th sampling point; The second eigenvalue represents the neighborhood covariance matrix of the i-th sampling point. The third eigenvalue represents the neighborhood covariance matrix of the i-th sampling point;

[0058] Obtain the local topological dispersion index of the i-th sampling point:

[0059] ;

[0060] In the formula, The local topological dispersion index represents the i-th sampling point; The first eigenvalue represents the neighborhood covariance matrix of the i-th sampling point; The second eigenvalue represents the neighborhood covariance matrix of the i-th sampling point. The third eigenvalue represents the neighborhood covariance matrix of the i-th sampling point; exp() represents an exponential function with the natural constant as the base.

[0061] The significance of the neighborhood sampling point set of the i-th sampling point in the most dominant extension direction;

[0062] Given that the sampling points within the flat region of the flange surface are concentrated only in a two-dimensional plane, with no significant positional fluctuations in the normal direction perpendicular to the reference plane, the first and second eigenvalues ​​are relatively large, while the third eigenvalue is extremely small. The larger the value, and When the value approaches 1, The larger the value, the greater the probability that the i-th sampling point belongs to the flat area of ​​the flange surface;

[0063] The normals of the sampling points at the bolt hole edges and sealing groove boundaries of the flange undergo drastic changes with the hole contour. The height values ​​of the sampling points are offset relative to the flange reference plane. Furthermore, the sampling points at the hole and groove edges are linearly discrete along the hole circumference, not uniformly distributed points on a two-dimensional plane. Therefore, according to the principal component analysis results, the first and third eigenvalues ​​are significantly increased, while the second eigenvalue is relatively small. When the third eigenvalue increases, the denominator increases. The value decreases, and The larger the value, the more likely it is to cause... The smaller the value, the better. The smaller the value, the greater the probability that the i-th sampling point belongs to the edge of the flange hole and the boundary of the sealing groove;

[0064] The high reflectivity of metal surfaces generates flying point noise in spatial regions far from the flange reference plane. This noise appears as randomly distributed clusters, exhibiting significant three-dimensional isotropic characteristics at the microscopic level, meaning it is discretely distributed in the x, y, and z spatial dimensions. Reflected in the principal component analysis results, this manifests as the disappearance of the dominance of the first eigenvalue, a significant increase in the third eigenvalue, and its amplitude approaching that of the second eigenvalue, leading to... The value decreases and The smaller the value, the better. The smaller the value, the more likely the i-th sampling point is to be a noise point.

[0065] S003. Calculate the fitting weight coefficient of each sampling point according to the local topological dispersion index, construct a weighted objective function to solve the optimal reference plane equation, select effective sampling points according to the local topological dispersion index, calculate the distance range from the effective sampling points to the optimal reference plane, and obtain the flatness detection value of the flange surface.

[0066] It should be noted that if the flange hole edge, sealing groove boundary and noise are directly removed based on the local topological dispersion index, the fitting plane will produce a mathematical residual jump at the physical boundary of the flange, causing the fitting result of the reference plane to deviate from the true physical neutral surface of the flange. Therefore, this invention obtains the fitting weight coefficient of each sampling point based on the local topological dispersion index, which is convenient for subsequent fitting of the plane, and assigns small weight coefficients to the sampling points of flange hole edge, sealing groove boundary and noise.

[0067] In this embodiment of the invention, the fitting weight coefficients for each sampling point are obtained:

[0068] ;

[0069] In the formula, The fitting weight coefficients represent the i-th sampling point; The local topological dispersion index represents the i-th sampling point; This represents the threshold for judgment; Represents the steepness factor, used for rapid amplification. The value of ; exp() represents an exponential function with the natural constant as the base; in this embodiment of the invention, a preset judgment threshold is used. Preset steepness factor In other embodiments, implementers may preset the judgment threshold and the steepness factor value according to the specific implementation situation;

[0070] When the local topological dispersion index is less than the judgment threshold, it indicates that the sampling point is more likely to be the edge of the flange bolt hole, the boundary of the sealing groove, or a noise point. The value is positive because of the steepness factor. A large value results in a large exponential term in the formula. The rapid increase leads to a decrease in weight. It rapidly approaches 0;

[0071] When the local topological dispersion index of the i-th sampling point is greater than the judgment threshold, it indicates that the probability of the i-th sampling point belonging to the flat region of the flange surface is higher. The value is negative due to the steepness factor. The value is relatively large. The value rapidly approaches 0, causing the weight to... It rapidly approaches 1.

[0072] It should be noted that the height values ​​of the sampling points at the bolt hole edges and sealing groove boundaries are offset relative to the flange reference plane. If the traditional least squares method is used directly for plane fitting, the global average error convergence characteristic of the algorithm will cause the fitting plane to tend towards the spatial center of all sampling points. This will cause the reference plane to be pulled downward by the concave areas of the bolt holes and sealing grooves, resulting in unavoidable fitting deviations. Therefore, this invention utilizes the fitting weight coefficient of each sampling point to assign greater weight to sampling points in the flat areas of the flange surface during the fitting process, and to assign less weight to the bolt hole edges, sealing groove boundaries, or noise points of the flange, thereby improving the accuracy of the fitting plane.

[0073] In this embodiment of the invention, a predefined plane equation is: ,in, Let be the plane parameters to be solved;

[0074] Construct a weighted objective function:

[0075] ;

[0076] In the formula, Represents the weighted objective function; The fitting weight coefficients represent the i-th sampling point; This represents the number of sampling points in the 3D point cloud set. Represents the spatial coordinates of the i-th sampling point; || represents the absolute value function; Let represent the distance from the i-th sampling point to the plane, where the constraint condition is... This constraint is useful when solving the system of equations later, as it is necessary to combine it with the solution to ensure the uniqueness of the solution.

[0077] If the fitting weight coefficient of the i-th sampling point is larger, it indicates that the i-th sampling point is located in a flat area of ​​the flange surface, thus increasing the distance from the i-th sampling point to the plane; if the fitting weight coefficient of the i-th sampling point is smaller, it indicates that the i-th sampling point is more likely to be the edge of the flange bolt hole, the boundary of the sealing groove, or a noise point, thus reducing the distance from the i-th sampling point to the plane.

[0078] It should be noted that the objective is to minimize the sum of the squared distances from all weighted sampling points to the plane (minimizing the weighted objective function). The value of is used to obtain the optimal reference plane equation.

[0079] In this embodiment of the invention, the plane parameters in the weighted objective function are... Taking the partial derivatives of each partial derivative and setting them to zero yields four partial derivatives. These four partial derivatives are then treated as a system of linear equations and solved to obtain the following result. ;Will Substituting into the preset plane equation, the optimal reference plane equation is obtained.

[0080] It should be noted that by judging the threshold, only the sampling points in the flat area of ​​the flange surface are retained as valid points. The difference between the maximum and minimum distances from the valid points to the optimal reference plane is used to quantify the actual flatness of the flange surface. This calculation method can accurately avoid the interference of invalid points on the flatness detection results.

[0081] In this embodiment of the invention, if the local topological dispersion index of any sampling point is greater than or equal to the determination threshold... When this happens, the sampling point is considered a valid sampling point;

[0082] Obtain the flatness test value of the flange face:

[0083] ;

[0084] In the formula, The flatness test value represents the flange face; This represents the maximum distance among all valid sampling points to the optimal reference plane; This represents the minimum distance among all valid sampling points to the optimal reference plane.

[0085] S004. Complete the flange face sealing risk assessment based on the flatness test value.

[0086] It should be noted that the flatness test value of the flange face is compared with the flatness threshold to complete the quantitative assessment of whether there is a sealing risk to the flange.

[0087] In this embodiment of the invention, the preset flatness threshold T2 = 0.15 mm. If the flatness detection value of the flange surface is greater than the flatness threshold, it indicates that the unevenness of the flange surface has exceeded the allowable range for safe sealing. The system will automatically determine that the flange has a sealing risk and trigger an alarm mechanism to prompt the staff to handle it.

[0088] Figure 2 This is a schematic diagram of the reference plane fitted by the traditional algorithm. Due to the inability to identify non-planar features, and the combined influence of hole depth and high-position flying points, the fitted plane deviates significantly from the real physical plane of the flange, presenting an incorrect tilt angle. Figure 3 This is a schematic diagram of the optimal reference plane obtained by the solution of this invention. The algorithm accurately identifies the hole edges and flying point clusters and assigns them extremely low weights. The red fitting plane is steadily locked on the machining surface of the flange. Even in areas with dense holes, it maintains an extremely high flatness reference, verifying the advanced nature of the solution.

[0089] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for online detection of the flatness of GIS shell flange surface based on structured light 3D point cloud, characterized in that, include: A scanner is used to perform a full-circumference scan of the flange surface of the GIS shell to obtain an initial three-dimensional point cloud set containing several sampling points; Based on the distance between each sampling point and its nearest neighbors, obtain the adaptive search radius of each sampling point; based on the adaptive search radius, obtain the neighborhood sampling point set of each sampling point; The neighborhood covariance matrix of the neighborhood sampling point set is decomposed into eigenvalues ​​to obtain several eigenvalues. Based on the feature values, the local topological dispersion index of each sampling point is obtained; Based on the local topological dispersion index, the fitting weight coefficients for each sampling point are obtained, including: , The fitting weight coefficients represent the values ​​for the i-th sampling point. The local topological dispersion index represents the i-th sampling point. This represents the threshold for judgment. The steepness factor is represented by exp(), which represents an exponential function with the natural constant as the base. Based on the fitted weight coefficients, a weighted objective function is constructed to solve for the optimal reference plane equation, including: a preset plane equation: ,in, Let be the plane parameters to be solved; In the formula, Represents the weighted objective function. This represents the number of sampling points in the 3D point cloud set. Represents the spatial coordinates of the i-th sampling point, and || represents the absolute value function; in At that time, for the plane parameters in the weighted objective function Taking the partial derivatives and setting them to zero, we obtain four partial derivatives. These four partial derivatives are then used as a system of linear equations, which we solve to obtain... Substitute these equations into the preset plane equations to obtain the optimal reference plane equations; Valid sampling points are selected based on the local topological dispersion index; the flatness detection value of the flange face is obtained based on the distance from each valid sampling point to the optimal reference plane; and the flange face sealing risk assessment is completed based on the flatness detection value.

2. The online detection method for the flatness of GIS shell flange surface based on structured light three-dimensional point cloud according to claim 1, characterized in that, The process of obtaining the adaptive search radius for each sampling point includes: ； In the formula, The adaptive search radius represents the i-th sampling point; This represents the number of nearest neighbors of the i-th sampling point; Represents the spatial coordinates of the i-th sampling point; This represents the spatial coordinates of the j-th nearest neighbor of the i-th sampling point.

3. The online detection method for the flatness of GIS shell flange surface based on structured light three-dimensional point cloud according to claim 2, characterized in that, The acquisition of the nearest neighbors of the i-th sampling point includes: Preset number of nearest neighbors ; Obtain the Euclidean distance between the i-th sampling point and each other sampling point, and record the K sampling points closest to the i-th sampling point as the nearest neighbors of the i-th sampling point.

4. The online detection method for the flatness of GIS shell flange surface based on structured light three-dimensional point cloud according to claim 1, characterized in that, The step of obtaining the neighborhood sampling point set for each sampling point includes: Centered on the i-th sampling point, all sampling points within its adaptive search radius are taken as the neighborhood sampling point set of the i-th sampling point.

5. The online detection method for the flatness of GIS shell flange surface based on structured light three-dimensional point cloud according to claim 1, characterized in that, The eigenvalue decomposition of the neighborhood covariance matrix of the neighborhood sampling point set yields several eigenvalues, including: Based on the neighborhood sampling point set of the i-th sampling point, calculate its neighborhood covariance matrix, and perform eigenvalue decomposition on the neighborhood covariance matrix to obtain three eigenvalues. Then, sort the three eigenvalues ​​in descending order. ; The first eigenvalue represents the neighborhood covariance matrix of the i-th sampling point; The second eigenvalue represents the neighborhood covariance matrix of the i-th sampling point; The third eigenvalue represents the neighborhood covariance matrix of the i-th sampling point.

6. The online detection method for the flatness of GIS shell flange surface based on structured light three-dimensional point cloud according to claim 1, characterized in that, The process of obtaining the local topological dispersion index for each sampling point includes: ； In the formula, The local topological dispersion index represents the i-th sampling point; The first eigenvalue represents the neighborhood covariance matrix of the i-th sampling point; The second eigenvalue represents the neighborhood covariance matrix of the i-th sampling point. The third eigenvalue represents the neighborhood covariance matrix of the i-th sampling point; exp() represents an exponential function with the natural constant as the base.

7. The online detection method for the flatness of GIS shell flange surface based on structured light three-dimensional point cloud according to claim 1, characterized in that, The process of obtaining the flatness test value of the flange surface includes: ； In the formula, The flatness test value represents the flange face; This represents the maximum distance among all valid sampling points to the optimal reference plane; This represents the minimum distance among all valid sampling points to the optimal reference plane.

8. The online detection method for the flatness of GIS shell flange surface based on structured light three-dimensional point cloud according to claim 1, characterized in that, The step of completing the flange face sealing risk assessment based on the flatness test value includes: The system has a preset flatness threshold T2. If the flatness detection value of the flange face is greater than the flatness threshold, the system will automatically determine that the flange has a sealing risk and trigger an alarm mechanism to prompt the staff to handle it.

Images