A method for identifying the parting line of castings based on point cloud

Abstract

This invention provides a method for identifying the parting line of a casting based on a point cloud, which lies at the intersection of three-dimensional industrial inspection and intelligent manufacturing. The method comprises the following steps: preprocessing the original three-dimensional point cloud data of the casting; forming a primary candidate set of point clouds whose curvature change rate exceeds a curvature threshold and satisfies a normal mutation condition in the voxel three-dimensional point cloud data; using the three-dimensional DBSCAN clustering algorithm to screen connected domains from the primary candidate set, and setting a connected domain aspect ratio threshold and a length threshold to eliminate unqualified connected domains to obtain candidate connected domains; performing dilation and erosion operations on each candidate connected domain, and performing topological constraint verification on the obtained results to obtain a consistent connected domain; and sequentially performing three-level geometric verification on the consistent connected domains. This invention can effectively improve the accuracy of parting line identification.

Description

Technical Field

[0001] The present invention belongs to the intersection of three-dimensional industrial detection and intelligent manufacturing, and specifically relates to a casting parting line recognition method based on point cloud. Background Art

[0002] The foundry industry is one of the foundational pillars of the manufacturing industry, widely used in the automotive, aerospace, and energy equipment sectors. The parting line, an inevitable product of the casting process, has a direct impact on the sealing performance, assembly accuracy, and fatigue life of the casting. For example, in an automobile engine cylinder block, a parting line deviation exceeding 0.3mm can lead to seal failure between the block and cylinder head, resulting in oil leakage and potentially resulting in recall costs of millions of dollars per batch.

[0003] Existing technologies for detecting parting lines rely on either manual inspection or two-dimensional visual inspection systems. The former suffers from low efficiency and a high rate of missed detections, while the latter suffers from a high rate of false positives. Three-dimensional digital inspection is used for parting line detection, but existing point cloud processing algorithms lag significantly behind hardware development. When analyzing casting surfaces, these algorithms are susceptible to interference from stray noise. For example, when calculating surface normals, a small number of outliers can lead to errors in overall orientation, which can snowball into errors. While the improved moving least squares method can alleviate some of these issues, its fixed calculation range parameters make it difficult to account for variations in point cloud density across different regions. This can overly blur details in dense areas and fail to accurately capture features in sparse areas. For subtle edge features like parting lines, existing methods use a uniform curvature threshold (e.g., 0.05). However, actual casting materials vary significantly: cast iron has a rough surface, while aluminum alloys are relatively smooth. This one-size-fits-all threshold setting results in a high rate of missed detections for thin-walled castings. Even more challenging is that when the parting line intersects or bifurcates, conventional clustering algorithms can easily misidentify it as multiple discontinuous segments. Deep learning solutions, which rely on massive data training, are not only time-consuming and labor-intensive to label, but recognition performance can also significantly degrade after changing scanning equipment. Furthermore, traditional breakpoint repairing can destroy the original shape like an eraser, while uniform expansion and erosion operations can distort the width of the parting line. Summary of the Invention

[0004] The purpose of this invention is to propose a casting parting line recognition method based on point cloud, which can effectively improve the parting line recognition accuracy.

[0005] The present invention is achieved through the following technical solutions:

[0006] A method for identifying the parting line of a casting based on point cloud comprises the following steps:

[0007] Step S1, preprocessing the acquired original three-dimensional point cloud data of the casting, the preprocessing comprising obtaining a normal vector of each point cloud in the original three-dimensional point cloud data, performing noise suppression on the original three-dimensional point cloud data to obtain denoised three-dimensional point cloud data, and performing voxel filtering on the denoised three-dimensional point cloud data to obtain voxel three-dimensional point cloud data;

[0008] Step S2: forming a primary candidate set of point clouds whose curvature change rate is greater than a curvature threshold and that meet a normal mutation condition in the voxel 3D point cloud data. The normal mutation condition is related to the angle between the normals of adjacent points in the point cloud neighborhood.

[0009] Step S3: Using the 3D DBSCAN clustering algorithm to screen connected domains from the primary candidate set, and setting a connected domain aspect ratio threshold and a length threshold to eliminate unqualified connected domains, thereby obtaining candidate connected domains. The neighborhood radius in the 3D DBSCAN clustering algorithm is dynamically adjusted according to the curvature change rate.

[0010] Step S4: Perform dilation and erosion operations on each candidate connected domain, and perform topological constraint verification on the results to obtain a consistent connected domain. In the dilation operation, the expansion direction is dominated by the normal direction of the point cloud in the candidate connected domain. The topological constraint verification is used to verify whether the shape fidelity condition and density condition are met.

[0011] Step S5: Perform three-level geometric verification on the consistent connected domain in sequence. The first-level geometric verification is used to retain the consistent connected domain with continuous angles. The second-level geometric verification is used to retain the consistent connected domain with width and height differences that meet the set thresholds, and obtain the mold line trajectory based on the consistent connected domain. The third-level geometric verification is used to determine whether the length of the mold line trajectory and the standard mold line length are within the error range, and eliminate the mold line trajectory that is not within the error range.

[0012] Furthermore, in step S1, the point cloud in the original three-dimensional point cloud data is p i , use the moving least squares method to fit the local surface, and get the normal vector of the point cloud according to the gradient direction of the local surface n i .

[0013] Furthermore, in step S1, performing noise suppression on the original three-dimensional point cloud data includes:

[0014] Step S11: the point cloud in the original 3D point cloud data p i , when the total neighborhood weight of the point cloud W i satisfy When it is removed, the point cloud data is obtained. P Gauss ,in, , , q j Point Cloud p i In the neighborhood No. j point clouds, n i and n j Represent point clouds p i and point cloud q j The normal vector of Point Cloud p i The standard deviation of the distance between point clouds in the neighborhood, To set the normal vector deviation tolerance threshold, α is the proportionality coefficient, k is the number of neighborhood points;

[0015] Step S12: Point cloud data P Gauss Point cloud in p j , if it meets , then the point cloud is far-field noise, remove it, if it meets If the normal deviation angle is greater than the set threshold, it is near-field noise, which is removed to obtain denoised 3D point cloud data. P filter ,in, μ Point cloud data P Gauss The mean position of Point cloud data P Gauss The global standard deviation of μ j Point Cloud p j The local mean of the neighborhood of Point Cloud p j The standard deviation of the distance between points in the neighborhood.

[0016] Furthermore, in step S1, the voxel side length is used Perform voxel filtering on the denoised 3D point cloud data to obtain voxel 3D point cloud data P voxel ,in, r base is the average distance between point clouds in the denoised 3D point cloud data, is the curvature change rate of the point cloud in the denoised 3D point cloud data, The first curvature threshold is set.

[0017] Furthermore, in step S2, for the point cloud in the voxel three-dimensional point cloud data p i' , if it meets R i' > R t , then it satisfies the normal mutation condition, where , k is the number of neighborhood point clouds, Point Cloud p i' Point cloud in its neighborhood p j' The normal angle of It is an indicator function that returns 1 if the condition is met, otherwise it returns 0. is the normal angle threshold, R t =0.3, Point Cloud p i' The normal vector of Point Cloud p j' The normal vector of .

[0018] Furthermore, in step S3, the number of neighborhood point clouds of the three-dimensional DBSCAN clustering algorithm is expressed as ,in, To set the base radius, is the second curvature threshold set; when the aspect ratio of the connected domain is greater than the aspect ratio threshold or the length is less than the length threshold, the connected domain is determined to be an unqualified connected domain.

[0019] Furthermore, in step S4, according to the formula For candidate connected regions R i Point cloud in p i'' After the expansion operation, the new point cloud is obtained p i''_new , the shape fidelity condition is expressed as , the density condition is expressed as ,in, α' is the expansion step length, Point Cloud p i'' The normal vector of Represents novel point cloud p i''_new With point cloud p i'' The distance between l c is the distance threshold set, ρfinal Candidate connected domain R i The density of the new connected domain point cloud obtained after expansion and corrosion operations, ρ original Candidate connected domain R i The density of the point cloud, m ∈(0,1).

[0020] Furthermore, in step S5, the first-level geometric verification is performed by formula Conduct, among which , s j is the first j edge points, is the set angle threshold.

[0021] Furthermore, in step S5, the secondary geometry verification specifically includes the following steps:

[0022] Step S51: Receive the continuous edge path obtained after the first-level geometric verification , for any point cloud in the continuous edge path , construct a local coordinate system ,in, Point Cloud The tangent vector of Point Cloud The normal vector of Point Cloud The binormal vector of N is the number of point clouds in the continuous edge path;

[0023] Step S52: Point cloud Each neighboring point cloud in the neighborhood of q k Projection to normal vector Go up and get , then the point cloud The neighborhood width is If the width meets the set width threshold range, the width of the point cloud is determined to be qualified, otherwise it is marked as an abnormal point;

[0024] Step S53: Extract the height value sequence of all point clouds on the continuous edge path , according to the formula Calculate the height change between adjacent point clouds and use the formula Calculate the overall height fluctuation range. If the height changes If the height difference of the continuous edge path is less than the set height threshold and the overall height fluctuation range is less than the set height threshold, the height difference of the continuous edge path is determined to be qualified, otherwise it is marked as a defective area and removed. is the first in the height value sequence j'' values.

[0025] Furthermore, in step S5, a bidirectional dynamic Dijkstra algorithm is used to obtain the parting line trajectory.

[0026] The present invention has the following beneficial effects:

[0027] 1. The present invention first pre-processes the original three-dimensional point cloud data of the casting, and screens the primary candidate set from the pre-processed data according to the curvature change rate and normal mutation conditions. Then, the three-dimensional DBSCAN clustering algorithm is used to screen the connected domain from the primary candidate set, and the aspect ratio threshold and length threshold of the connected domain are set to eliminate unqualified connected domains to obtain candidate connected domains. Then, expansion and erosion operations are performed on each candidate connected domain, and the obtained results are subjected to topological constraint verification to obtain consistent connected domains. Finally, the consistent connected domains are subjected to three-level geometric verification in sequence to obtain the parting line trajectory that meets the design requirements, thereby effectively improving the parting line recognition accuracy. degree; in the preprocessing stage, it can effectively distinguish between noise and valid points. When screening connected domains, the neighborhood radius is dynamically adjusted according to the curvature change rate, and the connected domain aspect ratio threshold and length threshold are set, which can solve the problem of over-segmentation of sharp edges and enhance the ability to screen connected domains of complex structures. In the expansion and corrosion operations, the shape fidelity condition and density condition are set, which can effectively fill the holes in the point cloud and smooth the surface, and ensure the stability of the geometric structure. The three-level geometric verification is used to optimize the continuity of the mold line. In this way, it is superior to traditional technologies in noise suppression, feature retention, clustering robustness and repair accuracy, and is suitable for high-precision industrial detection scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The present invention will be further described in detail below with reference to the accompanying drawings.

[0029] Figure 1 Flowchart of the present invention.

[0030] Figure 2 Schematic diagram of normal vectors of original three-dimensional point cloud data of the present invention.

[0031] Figure 3 This is a schematic diagram of the present invention removing point clouds that do not meet the curvature threshold and normal mutation conditions.

[0032] Figure 4 This is a schematic diagram of eliminating unqualified connected domains according to the present invention.

[0033] Figure 5 This is a schematic diagram of the consistent connected domain after expansion and corrosion of the present invention.

[0034] Figure 6 This is a schematic diagram of a consistent connected domain whose width difference and height difference both meet the threshold value according to the present invention.

[0035] Figure 7 Schematic diagram of the standard parting line of the present invention.

[0036] Figure 8 Schematic diagram of the parting line trajectory of the present invention. DETAILED DESCRIPTION

[0037] like Figure 1 As shown in FIG, the method for identifying the parting line of a casting based on point cloud includes the following steps:

[0038] Step S1, preprocessing the acquired original three-dimensional point cloud data of the casting, the preprocessing comprising obtaining a normal vector of each point cloud in the original three-dimensional point cloud data, performing noise suppression on the original three-dimensional point cloud data to obtain denoised three-dimensional point cloud data, and performing voxel filtering on the denoised three-dimensional point cloud data to obtain voxel three-dimensional point cloud data;

[0039] The specific steps include:

[0040] Step S10: original 3D point cloud data The i Point Cloud , use the moving least squares method to fit the local surface, and get the normal vector of the point cloud according to the gradient direction of the local surface n i ;

[0041] The moving least squares fitting method is specifically to use a second-order polynomial Fit the neighborhood points and solve the parameters of the second-order polynomial by weighted least squares method. i The local surface of a point cloud is represented as , set the polynomial order to 2, the search radius , is the average distance between point clouds, is the Gaussian kernel function, h is the Gaussian kernel width parameter, Point Cloud p i of k points in the neighborhood, k is the neighborhood radius, which is dynamically adjusted according to the point cloud density and is usually k =10~30, Represents the midpoint of the point cloud p i With neighboring points The square of the distance is used to measure spatial proximity;

[0042] The quadratic surface fitting as described above can improve the accuracy of normal estimation of complex surfaces. The normal vector is obtained by the gradient direction of the fitted local surface, which can ensure the adaptability to complex surfaces. The point cloud normal vector is as follows: Figure 2 As shown;

[0043] Step S11: the point cloud in the original 3D point cloud data p i , when the total neighborhood weight of the point cloud W i satisfy When it is removed, the point cloud data is obtained. P Gauss ,in, , is the dynamic weight allocation formula, q j Point Cloud p i In the neighborhood No. j point clouds, n i and n j Represent point clouds p i and point cloud q j The normal vector of Point Cloud p i The standard deviation of the distance between points in the neighborhood (i.e., local density), The normal vector deviation tolerance threshold is set, usually set to 15°. α is the proportionality coefficient ( α =0.3∼0.5), k is the number of neighborhood points;

[0044] is the spatial weight form of the Gaussian kernel function, is the normal weight form of the Gaussian kernel function, w ij By imposing dual constraints on spatial distance and normal deviation, noise points and valid points can be effectively distinguished to improve filtering accuracy.

[0045] Step S12: Point cloud data P Gauss Point cloud in p j , if it meets , then the point cloud is far-field noise, remove it, if it meets If the normal deviation angle is greater than the set threshold (25°), it is near-field noise and is removed to obtain denoised 3D point cloud data. P filter ,in, μ Point cloud data P Gauss The mean position of Point cloud data P Gauss The global standard deviation of μ j Point Cloud p j The local mean of the neighborhood of Point Cloud p j The standard deviation of the distance between point clouds in the neighborhood, the normal deviation angle is the point cloud p j The angle between the normal and the neighboring points in the k-neighborhood set.

[0046] Mean location Calculated by the following formula: , Point Cloud p j The coordinates of n Point cloud data P Gauss The number of point clouds in Point Cloud p j The distance from the mean position;

[0047] This can effectively distinguish different types of noise points and adopt different processing methods according to their characteristics, thereby enhancing the filtering effect;

[0048] Step S13: Using voxel side length Perform voxel filtering on the denoised 3D point cloud data to obtain voxel 3D point cloud data P voxel ,in, r base is the average distance between point clouds in the denoised 3D point cloud data, is the curvature change rate of the point cloud in the denoised 3D point cloud data, and are the curvatures in the minimum and maximum principal directions, is the first curvature threshold set;

[0049] This allows for downsampling of flat areas using larger voxel sizes to improve efficiency, retaining the original resolution in high-curvature areas to maintain details, and avoiding the transition smoothing of details caused by traditional voxel filtering.

[0050] Step S2: forming a primary candidate set of point clouds whose curvature change rate is greater than a curvature threshold and that meet a normal mutation condition in the voxel 3D point cloud data. The normal mutation condition is related to the angle between the normals of adjacent points in the point cloud neighborhood.

[0051] Specifically, for the point cloud in voxel 3D point cloud data p i' , if it meets R i' > R t , then it satisfies the normal mutation condition, where , k is the number of neighborhood point clouds, Point Cloud p i' Point cloud in its neighborhood p j' The normal angle of It is an indicator function that returns 1 when the condition is met, otherwise it returns 0. is the normal angle threshold, represents the dot product, Point Cloud p i' The normal vector of Point Cloud p j' The normal vector of

[0052] In this embodiment, the curvature threshold is set to 0.8. R t = 0.3, remove the point clouds that do not meet the curvature threshold and normal mutation conditions, such as Figure 3 shown.

[0053] Step S3: Using the 3D DBSCAN clustering algorithm to screen connected domains from the primary candidate set, and setting a connected domain aspect ratio threshold and a length threshold to eliminate unqualified connected domains, thereby obtaining candidate connected domains. The neighborhood radius in the 3D DBSCAN clustering algorithm is dynamically adjusted according to the curvature change rate.

[0054] Specifically, the number of neighborhood point clouds of the 3D DBSCAN clustering algorithm is expressed as ,in, To set the base radius, is the second curvature threshold set; when the aspect ratio of a connected domain is greater than the aspect ratio threshold or the length is less than the length threshold, the connected domain is determined to be an unqualified connected domain. In this embodiment, the aspect ratio threshold is set to 5:1. Connected domains with too large an aspect ratio are mostly noise or scanning errors (such as flying spots). The length threshold is set to 10 mm. Connected domains with too small a length are usually caused by point cloud noise or insufficient local sampling, and do not conform to the actual object size.

[0055] Eliminate unqualified connected domains such as Figure 4 As shown in the figure, the clustering accuracy of sharp edges is significantly improved. Combined with the topological constraints of aspect ratio and length, non-physically connected domains are effectively removed, taking into account both algorithm robustness and computational efficiency.

[0056] Step S4: Perform dilation and erosion operations on each candidate connected domain, and perform topological constraint verification on the results to obtain a consistent connected domain. In the dilation operation, the expansion direction is dominated by the normal direction of the point cloud in the candidate connected domain. The topological constraint verification is used to verify whether the shape fidelity condition and density condition are met.

[0057] Specifically, according to the formula For candidate connected regions R i Point cloud in p i'' After the expansion operation, the new point cloud is obtained p i''_new , the shape fidelity condition (i.e., anisotropic diffusion strategy) is expressed as , the density condition is expressed as ,in, α' =0.1 mm is the expansion step length, Point Cloud p i'' The normal vector of Represents novel point cloud p i''_new With point cloud p i'' The distance between l c =0.3 mm is the distance threshold set, ρ final Candidate connected domain R i The density of the new connected domain point cloud obtained after expansion and corrosion operations, ρ original Candidate connected domain R i The density of the point cloud, m ∈(0,1).

[0058] As mentioned above, the morphological closing operation of dilation followed by erosion is performed on each candidate connected domain, which can fill holes in point clouds or voxel models, connect broken areas, and smooth surface geometric features. The dilation operation expands along the normal direction and introduces an anisotropic diffusion strategy to control the diffusion range; the erosion operation is based on the density preservation criterion and dynamically adjusts the threshold to avoid excessive damage to the original structure. The consistent connected domain obtained after dilation and erosion is as follows Figure 5 shown.

[0059] Step S5: Perform three-level geometric verification on the consistent connected domains in sequence. The first-level geometric verification is used to retain the consistent connected domains with continuous angles. The second-level geometric verification is used to retain the consistent connected domains whose width and height differences meet the set thresholds. Based on the consistent connected domains, the parting line trajectory is obtained. The third-level geometric verification is used to determine whether the length of the obtained parting line trajectory is within the error range with the standard parting line length, and the parting line trajectories that are not within the error range are eliminated.

[0060] The specific steps include:

[0061] Step S50: First-level geometry verification by formula If it meets the requirements, the current consistent connected domain is considered continuous and enters the second-level geometric verification, where , s j is the first of the consistent connected domains obtained in step S4 j edge points, is the set angle threshold.

[0062] Step S51: Receive the continuous edge path of the consistent connected domain obtained after the first-level geometric verification , for any point cloud in the continuous edge path , construct a local coordinate system ,in, Point Cloud p j' If the tangent vector of j'' =1 or j'' = N ,but , N is the number of point clouds of continuous edge paths, Point Cloud The normal vector of , which can be estimated in the neighborhood points by principal component analysis (PCA), Point Cloud The binormal vector of

[0063] Step S52: Point cloud Each neighboring point cloud in the neighborhood of q k Projection to normal vector Go up and get , then the point cloud The neighborhood width is If the width meets the set width threshold range, the width of the point cloud is determined to be qualified, otherwise it is marked as an abnormal point. The neighborhood radius is 2~3mm, and the width threshold range is [1.5mm, 4mm].

[0064] Step S53: Extract the height value sequence of all point clouds on the continuous edge path , according to the formula Calculate the height change between adjacent point clouds and use the formula Calculate the overall height fluctuation range. If the height changes If the height difference of the continuous edge path is less than the set height threshold and the overall height fluctuation range is less than the set height threshold, the height difference of the continuous edge path is determined to be qualified, otherwise it is marked as a defective area and removed. , is the first in the height value sequence j'' values, with a height threshold of 0.2 mm ;

[0065] The width difference and height difference both meet the threshold of the consistent connected domain such as Figure 6 shown.

[0066] In the parting line trajectory obtained based on the consistent connected domain, the bidirectional dynamic Dijkstra algorithm is used to obtain the parting line trajectory. This algorithm not only considers the geometric distance, but also introduces the curvature change and signal-to-noise ratio. The formula is: , ;

[0067] Geometric distance: Represents the geometric distance of each edge in the path, α =0.6 is the weight coefficient;

[0068] Curvature change: Indicates the curvature change of each edge in the path, β =0.3 is the weight coefficient. The curvature change reflects the smoothness of the path. A large curvature change may make the path unnatural or difficult to achieve.

[0069] Signal-to-noise ratio: represents the inverse of the signal-to-noise ratio, γ =0.1 is the weight coefficient, which reflects the reliability of the path. A lower signal-to-noise ratio may mean that the path is unreliable under certain conditions.

[0070] In the three-level geometric verification, such as Figure 7 The standard parting line shown is obtained from the CAD model of the casting. Get the length of the parting line trajectory according to the formula Judge whether it is qualified to ensure that the manufacturing meets the design requirements, The first i The three-dimensional coordinates of the point, N is the total number of points on the parting line trajectory, is the standard parting line length, .

[0071] The above description is merely a preferred embodiment of the present invention and therefore cannot be used to limit the scope of the present invention. In other words, equivalent changes and modifications made according to the scope of the patent application and the contents of the specification should still fall within the scope of the patent of the present invention.

Claims

1. A method for identifying the parting line of a casting based on point cloud, characterized by: The steps include: Step S1, preprocessing the acquired original three-dimensional point cloud data of the casting, the preprocessing comprising obtaining a normal vector of each point cloud in the original three-dimensional point cloud data, performing noise suppression on the original three-dimensional point cloud data to obtain denoised three-dimensional point cloud data, and performing voxel filtering on the denoised three-dimensional point cloud data to obtain voxel three-dimensional point cloud data; Step S2: forming a primary candidate set of point clouds whose curvature change rate is greater than a curvature threshold and that meet a normal mutation condition in the voxel 3D point cloud data. The normal mutation condition is related to the angle between the normals of adjacent points in the point cloud neighborhood. Step S3: Using the 3D DBSCAN clustering algorithm to screen connected domains from the primary candidate set, and setting a connected domain aspect ratio threshold and a length threshold to eliminate unqualified connected domains, thereby obtaining candidate connected domains. The neighborhood radius in the 3D DBSCAN clustering algorithm is dynamically adjusted according to the curvature change rate. Step S4: Perform dilation and erosion operations on each candidate connected domain, and perform topological constraint verification on the results to obtain a consistent connected domain. In the dilation operation, the expansion direction is dominated by the normal direction of the point cloud in the candidate connected domain. The topological constraint verification is used to verify whether the shape fidelity condition and density condition are met. Step S5: Perform three-level geometric verification on the consistent connected domains in sequence. The first-level geometric verification is used to retain the consistent connected domains with continuous angles. The second-level geometric verification is used to retain the consistent connected domains whose width and height differences meet the set thresholds. Based on the consistent connected domains, the parting line trajectory is obtained. The third-level geometric verification is used to determine whether the length of the parting line trajectory is within the error range with the standard parting line length, and the parting line trajectories that are not within the error range are eliminated. In step S2, for the point cloud in the voxel three-dimensional point cloud data p i' , if it meets R i' > R t , then it satisfies the normal mutation condition, where , k is the number of neighborhood point clouds, Point Cloud p i' Point cloud in its neighborhood p j' The normal angle of It is an indicator function that returns 1 if the condition is met, otherwise it returns 0. is the normal angle threshold, R t =0.3, Point Cloud p i' The normal vector of Point Cloud p j' The normal vector of .

2. The method for identifying the parting line of a casting based on point cloud according to claim 1, wherein: In step S1, the point cloud in the original three-dimensional point cloud data is p i , use the moving least squares method to fit the local surface, and get the normal vector of the point cloud according to the gradient direction of the local surface n i .

3. The method for identifying the parting line of a casting based on point cloud according to claim 2, wherein: In step S1, performing noise suppression on the original three-dimensional point cloud data includes: Step S11: the point cloud in the original 3D point cloud data p i , when the total neighborhood weight of the point cloud W i satisfy When it is removed, the point cloud data is obtained. P Gauss ,in, , , q j Point Cloud p i In the neighborhood No. j point clouds, n i and n j Represent point clouds p i and point cloud q j The normal vector of Point Cloud p i The standard deviation of the distance between point clouds in the neighborhood, To set the normal vector deviation tolerance threshold, α is the proportionality coefficient, k is the number of neighborhood points; Step S12: Point cloud data P Gauss Point cloud in p j , if it meets , then the point cloud is far-field noise, remove it, if it meets If the normal deviation angle is greater than the set threshold, it is near-field noise, which is removed to obtain denoised 3D point cloud data. P filter ,in, μ Point cloud data P Gauss The mean position of Point cloud data P Gauss The global standard deviation of μ j Point Cloud p j The local mean of the neighborhood of Point Cloud p j The standard deviation of the distance between points in the neighborhood.

4. The method for identifying the parting line of a casting based on point cloud according to claim 3, wherein: In step S1, the voxel side length is used Perform voxel filtering on the denoised 3D point cloud data to obtain voxel 3D point cloud data P voxel ,in, r base is the average distance between point clouds in the denoised 3D point cloud data, is the curvature change rate of the point cloud in the denoised 3D point cloud data, The first curvature threshold is set.

5. The method for identifying the parting line of a casting based on point cloud according to claim 4, characterized in that: In step S3, the number of neighborhood point clouds of the three-dimensional DBSCAN clustering algorithm is expressed as ,in, To set the base radius, is the second curvature threshold set; when the aspect ratio of the connected domain is greater than the aspect ratio threshold or the length is less than the length threshold, the connected domain is determined to be an unqualified connected domain.

6. The method for identifying the parting line of a casting based on point cloud according to claim 5, characterized in that: In step S4, according to the formula For candidate connected regions R i Point cloud in p i'' After the expansion operation, the new point cloud is obtained p i''_new , the shape fidelity condition is expressed as , the density condition is expressed as ,in, α' is the expansion step length, Point Cloud p i'' The normal vector of Represents novel point cloud p i''_new With point cloud p i'' The distance between l c is the distance threshold set, ρ final Candidate connected domain R i The density of the new connected domain point cloud obtained after expansion and corrosion operations, ρ original Candidate connected domain R i The density of the point cloud, m ∈(0,1).

7. The method for identifying the parting line of a casting based on point cloud according to claim 6, characterized in that: In step S5, the first-level geometry verification is performed by formula Conduct, among which , s j is the first j edge points, is the set angle threshold.

8. The method for identifying the parting line of a casting based on point cloud according to claim 7, characterized in that: In step S5, the secondary geometry verification specifically includes the following steps: Step S51: Receive the continuous edge path obtained after the first-level geometric verification , for any point cloud in the continuous edge path , construct a local coordinate system ,in, Point Cloud The tangent vector of Point Cloud The normal vector of Point Cloud The binormal vector of N is the number of point clouds in the continuous edge path; Step S52: Point cloud Each neighboring point cloud in the neighborhood of q k Projection to normal vector Go up and get , then the point cloud The neighborhood width is If the width meets the set width threshold range, the width of the point cloud is determined to be qualified, otherwise it is marked as an abnormal point; Step S53: Extract the height value sequence of all point clouds on the continuous edge path , according to the formula Calculate the height change between adjacent point clouds and use the formula Calculate the overall height fluctuation range. If the height changes If the height difference of the continuous edge path is less than the set height threshold and the overall height fluctuation range is less than the set height threshold, the height difference of the continuous edge path is determined to be qualified, otherwise it is marked as a defective area and removed. is the first in the height value sequence j'' values.

9. The method for identifying the parting line of a casting based on point cloud according to claim 8, characterized in that: In step S5, a bidirectional dynamic Dijkstra algorithm is used to obtain the parting line trajectory.

Images