Welding seam point cloud region segmentation method and system based on multi-feature dynamic threshold analysis

Abstract

The invention belongs to the technical field of three-dimensional measurement, and particularly discloses a welding seam point cloud region segmentation method and system based on multi-feature dynamic threshold analysis. Comprising the following steps: acquiring three-dimensional point cloud data containing a welding seam region, and selecting any point in a welding seam as an initial seed point for region growth; calculating a normal vector divergence feature and a tangent plane deviation feature of the seed point, and performing three-dimensional point cloud feature fusion; the prior feature information is modeled into a Gaussian process, and dynamic threshold value updating of the region growing process is achieved; adopting a dynamic threshold value updating method to judge whether the seed points belong to a welding seam area, if yes, bringing the K neighborhood of the welding seam feature points into a seed set to serve as a seed point set for subsequently judging whether the seed points belong to the welding seam area, if not, judging whether the seed point set is empty, if not, updating the seed points, and if yes, updating the seed points; and judging that the complete welding seam region is segmented. The method can effectively extract the weld point cloud of the partial weak curvature feature area of the weld surface.

Description

Technical Field

[0001] The present invention belongs to the field of three-dimensional measurement technology, and more specifically, relates to a weld point cloud region segmentation method and system based on multi-feature dynamic threshold analysis. Background Art

[0002] Robotic welding technology has now achieved almost complete automation, and real-time evaluation of welding results has become one of the core issues in the welding field. Weld extraction is an important step in evaluating welding results, and it is also a prerequisite for subsequent surface quality processing. The principle of the weld extraction process is to achieve semantic segmentation of the weld and different areas of the workpiece by extracting the weld surface information and contour information. However, for welds with weak geometric features, or when there is no obvious distinction between the weld and the workpiece surface features, it is necessary to adopt a reasonable evaluation method and a more robust algorithm to realize the weld extraction process under different working conditions.

[0003] Weld extraction algorithms mainly include two-dimensional and three-dimensional weld extraction. Among them, two-dimensional extraction relies on visual images. Common methods such as structured light centerline algorithm and laser stripe grayscale distribution extraction algorithm are real-time, but there are some shortcomings, such as complex extraction process, low precision, and large limitations. Three-dimensional extraction is often processed on point clouds. Common methods include regional clustering and feature extraction algorithms, point cloud similarity region growing algorithms, etc. The clustering and extraction ideas have large errors in the extraction of target features in reality. The feature extraction process of the region growing algorithm is highly dependent on experience, and the same feature is not universal for different types of welds. At present, with the development of machine learning, some people have begun to try to use machine learning to realize weld extraction, and use the learned model to extract features of new welds, but such methods require a large amount of training data and high computing resources.

[0004] The characteristics of the weld area vary significantly, and the geometric features at different positions have certain differences, which will lead to the following problems: a single feature description of the weld area is difficult to fully cover the weld surface, the selection of feature thresholds is highly dependent on experience, and the noise information on the surface of the workpiece base material will have a great impact on the segmentation of the feature area. There is currently no suitable solution to the above problems in weld extraction. Summary of the invention

[0005] In response to the above defects or improvement needs of the prior art, the present invention provides a method and system for weld point cloud region segmentation based on multi-feature dynamic threshold analysis, which significantly improves the accuracy and robustness of three-dimensional point cloud weld region segmentation through multi-feature fusion and dynamic threshold analysis. This method utilizes the combination of point cloud normal vector divergence features and tangent plane deviation features to effectively identify weld regions, including those with weak curvature features, while avoiding mis-extraction caused by surface noise of the workpiece parent material. In addition, by modeling the weld region features as a Gaussian process and dynamically updating the threshold, the method can adapt to different working conditions, reduce dependence on manual experience, and thus improve the scope of application and flexibility of the weld extraction algorithm.

[0006] To achieve the above object, according to one aspect of the present invention, a weld point cloud region segmentation method based on multi-feature dynamic threshold analysis is proposed, comprising the following steps:

[0007] Step 1: Obtain 3D point cloud data containing the weld area, and select any point in the weld as the initial seed point for region growth;

[0008] Step 2: Calculate the normal vector divergence feature and the tangent plane deviation feature of the seed point, and perform three-dimensional point cloud feature fusion based on the feature to form a point cloud fusion feature;

[0009] Step 3: According to the workpiece point cloud fusion features, a dynamic threshold update method is used to determine whether the seed point belongs to the weld area. If so, the K neighborhood of the weld feature point is included in the seed set as the seed point set for subsequent determination of whether it belongs to the weld area. If not, determine whether the seed point set is empty. If not, update the seed point and return to step 2. If it is empty, determine that the complete weld area segmentation has been completed and enter step 4.

[0010] Step 4: extract the weld point cloud and contour information after segmentation to complete the weld area segmentation of the three-dimensional point cloud.

[0011] As a further preferred embodiment, in step one, a line laser scanner is used to scan the weld area to obtain a narrow point cloud including the weld area, and the narrow point cloud is used as three-dimensional point cloud data.

[0012] As further preferred, in step 2, the calculating of the seed point normal vector divergence feature includes:

[0013] (211) Calculate the K-neighborhood of each point in the three-dimensional point cloud data and calculate its L1 median point;

[0014] (212) Based on the L1 median point, a PCA method is used to estimate the normal vector of each point in the three-dimensional point cloud data;

[0015] (213) Calculate a local normal vector divergence characteristic factor of the feature point based on the normal vector.

[0016] As a further preferred embodiment, in step (211), let the three-dimensional point cloud data point set P k ={p 1 ,p 2 ,…,p j ,…,p k}, where p k ∈R 3 Yes j Nearby k different points, β i For p j The weight of p j A point q∈R in the neighborhood 3 , C(q) is the sum of q and p 1 ,p 2 ,…,p k The weighted sum of the distances, the equation of C(q) includes:

[0017]

[0018] Among them, d i (q)=||qp i || represents q and p i The Euclidean distance of

[0019] Find a point q∈R 3 To minimize the equation for C(q):

[0020] M=M(p 1 ,…,p k β 1 ,…β k ) = argmin{C(q); q∈R 3}

[0021] Among them, the solution of M is the L1 median point in the k-neighborhood;

[0022] Preferably, step (212) comprises:

[0023] Construction C 3×3 The covariance matrix of is:

[0024]

[0025] Among them, K is the number of neighborhood points;

[0026] Calculate the eigenvalues ​​of the covariance matrix and use the eigenvector corresponding to the minimum eigenvalue as the normal vector;

[0027] Preferably, step (213) comprises:

[0028] Define the local normal vector mean:

[0029]

[0030] The local divergence characteristic factor calculation model is constructed according to the local normal vector mean:

[0031]

[0032] Among them, n ij is the unit normal vector of each point in the K neighborhood, K is the number of neighborhood points, α i is the local divergence characteristic factor.

[0033] As further preferred, in step 2, the calculating of the seed point tangent plane deviation feature includes:

[0034] (221) Slice the feature points along the scanning line direction and calculate the projection of each point on the tangent plane to form a two-dimensional coordinate;

[0035] (222) performing curve fitting on the two-dimensional coordinates using a sliding average filter;

[0036] (223) constructing a deviation characteristic factor and calculating the tangent plane deviation characteristic, wherein the deviation characteristic factor includes:

[0037]

[0038] Where, d i is the deviation between the original point of the two-dimensional coordinate and its corresponding point on the generating curve.

[0039] As further preferred, step (221) comprises:

[0040] Let feature point p A =[x A ,y A ,z A ] T , click p j The distance from the tangent plane C is:

[0041] d j =|[p A -p j ] T ·n C |

[0042] Among them, n C is the normal vector of the tangent plane, when the distance d j Less than the threshold d thre When p j Located on the tangent plane;

[0043] The tangent plane coordinate system is constructed with the tangent plane normal vector, the scan line direction vector and the direction vector perpendicular to the two. The homogeneous transformation matrix T between the tangent plane coordinate system and the spatial coordinate system is calculated to convert the three-dimensional point cloud coordinates into two-dimensional coordinates in the tangent plane coordinate system.

[0044] As further preferred, in step (222), the formula of the sliding average filter includes:

[0045]

[0046] Where N is the length of the filter, p M-i Represents the value within the filter range;

[0047] The deviation d between the original point of the two-dimensional coordinate and its corresponding point on the mother curve is calculated by the following formula: i :

[0048]

[0049] Among them, [x pi ,y pi ] represents the corresponding coordinate point after smoothing, [x i ,y i ] represents the initial coordinates before smoothing.

[0050] As a further preferred embodiment, in step 2, the three-dimensional point cloud feature fusion includes: weighted combination of normal vector divergence feature and tangent plane deviation feature information to form a new weld feature factor:

[0051] θ i =ω 1 α i +ω 2 β i

[0052] Among them, ω 1 and ω 2 All are weight values.

[0053] As further preferred, step three comprises the following steps:

[0054] (311) Collect and prepare data samples, and combine some of the acquired prior weld feature information to form a weld area feature set.

[0055] (312) Train the Gaussian process model, input the feature information into the Gaussian process model, and set a suitable kernel function k(x, x') to model the correlation of the input data. By training the Gaussian process model, the output mean function μ(x) and covariance function σ(x) are obtained;

[0056] (313) Predict the probability distribution of new points. For a new feature point xnew , using the Gaussian process model to predict its mean μ(x new ) and variance σ(x new ), calculate P(X>Threshold) using the cumulative probability function F(z) of the normal distribution;

[0057] (314) Calculate the cumulative probability to reach a threshold of 99%;

[0058] (315) Determine whether the seed point belongs to a weld feature point based on the threshold. If so, include the point in the prior data sample for subsequent threshold update. If not, the threshold remains unchanged. Determine whether the seed point set is empty. If not, update the seed point and return to step 2. If it is empty, determine that the complete weld area segmentation has been completed and enter step 4.

[0059] According to another aspect of the present invention, a weld point cloud region segmentation system based on multi-feature dynamic threshold analysis is also provided, comprising:

[0060] The first main control module is used to obtain the three-dimensional point cloud data containing the weld area and select any point in the weld as the initial seed point for region growth;

[0061] The second main control module is used to calculate the normal vector divergence feature and the tangent plane deviation feature of the seed point, and perform three-dimensional point cloud feature fusion based on the calculated feature to form a point cloud fusion feature;

[0062] The third main control module is used to determine whether the seed point belongs to the weld area according to the workpiece point cloud fusion feature and adopts a dynamic threshold update method. If so, the K neighborhood of the weld feature point is included in the seed set as the seed point set for subsequent judgment of whether it belongs to the weld area. If not, it is determined whether the seed point set is empty. If not, the seed point is updated and the updated seed point is input into the second main control module. If it is empty, it is determined that the complete weld area segmentation has been completed.

[0063] The fourth main control module is used to extract the weld point cloud and contour information after the segmentation is completed after the third main control module determines that the complete weld area segmentation has been completed, so as to complete the weld area segmentation of the three-dimensional point cloud.

[0064] In general, the above technical solution conceived by the present invention has the following technical advantages compared with the prior art:

[0065] 1. The present invention constructs a multi-feature dynamic threshold analysis algorithm framework to perform multi-feature extraction on the preprocessed complete point cloud data to output the feature value of the feature point cloud to be judged, and then substitutes the acquired prior feature information into the dynamic threshold update algorithm framework to accurately judge the region to which the feature point belongs, and complete the weld point cloud region extraction. Accordingly, by weighting and fusing different feature information, the feature extraction process under different working conditions can be adapted. At the same time, by introducing Gaussian process modeling and dynamically adjusting the threshold of the region growth process through different confidence ranges, the adaptability of the region segmentation process is improved. Therefore, the present invention has the characteristics of high accuracy and good robustness.

[0066] 2. The present invention can effectively extract weld point clouds in some areas with weak curvature characteristics on the weld surface by combining the point cloud normal vector divergence feature and the tangent plane deviation feature, thereby improving the situation where weld point clouds may be omitted in the single feature extraction process. This method can better resist the influence of neighborhood feature changes, thereby improving the accuracy and robustness of weld extraction.

[0067] 3. The present invention incorporates the cutting plane deviation item feature into the feature item. This method can effectively avoid the situation where the local high curvature point cloud in some areas of the workpiece base material surface is incorrectly extracted due to the noise point cloud, thereby improving the accuracy of weld point cloud extraction.

[0068] 4. The present invention realizes dynamic threshold update of regional growth process by modeling weld regional features as Gaussian process, avoiding the inapplicability of manually set thresholds for different working conditions. This dynamic threshold update mechanism improves the scope of application of weld extraction algorithm, making it adaptable to different welding working conditions and conditions.

[0069] 5. The present invention adopts the point with the smallest weighted distance as the neighborhood center point to estimate the point cloud normal vector, and uses the L1 median as the data center to highlight the differences between feature points in adjacent points. It is more resistant to the influence of feature points in the neighborhood than the average value and is closer to the center of the data body. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is a flow chart of a method for segmenting weld point cloud regions based on multi-feature dynamic threshold analysis according to an embodiment of the present invention;

[0071] Figure 2 is a flow chart of point cloud feature value calculation involved in an embodiment of the present invention;

[0072] Figure 3 This is a cloud feature judgment threshold acquisition process involved in an embodiment of the present invention;

[0073] Figure 4 It is a schematic diagram of the weld point cloud extraction effect of an embodiment of the present invention. DETAILED DESCRIPTION

[0074] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0075] like Figure 1 As shown, an embodiment of the present invention provides a weld point cloud region segmentation method based on multi-feature dynamic threshold analysis, and the weld region segmentation process of the three-dimensional point cloud includes:

[0076] (1) Since the volume of the weld itself will interfere with the line laser, the transition area between the weld and the workpiece base material surface will be blocked, resulting in the loss of some point cloud details. Therefore, it is necessary to control the relative posture of the scanner and the weld workpiece to collect complete weld point cloud data. The weld has a certain length feature in its main extension direction. After a single scan of the weld area by the line laser scanner, a narrow point cloud including the weld area will be generated.

[0077] (2) After the acquisition is completed, the normal vector of the acquired point cloud needs to be estimated to facilitate the subsequent feature extraction process. Due to the exposure problem in the process of line structured light scanning of metal surfaces and the accuracy of camera equipment, the collected point cloud data will introduce noise and outliers. Therefore, it is necessary to perform pre-processing such as point cloud filtering and uniform downsampling on the collected point cloud, and remove and clean up the unnecessary surface point cloud data in the thickness direction.

[0078] Specifically, in the present invention, the feature extraction module includes two parts: point cloud normal vector divergence feature and tangent plane deviation feature. A multi-feature fusion method is used to form a seed point feature value, which is used to determine whether the point cloud belongs to the weld area in the subsequent process. The complete feature extraction process is as follows: Figure 2 shown.

[0079] (3) The divergence feature used in this patent estimates the normal vector of the feature point and the points in its K neighborhood, and constructs the divergence feature value according to the degree of dispersion of the normal vector.

[0080] ①Step 1: Calculate the L1 median of the point cloud and its K neighborhood. Suppose the point set P k ={p 1 ,p 2 ,…,p j ,…,p k}, where p k ∈R 3 Yes j Nearby k different points, β i For p jThe weight of p j A point q∈R in the neighborhood 3 , C(q) is the sum of q and p 1 ,p 2 ,…,p k The equation of C(q) is as follows:

[0081]

[0082] Among them, d i (q)=||qp i || represents q and p i The Euclidean distance of .

[0083] The L1 median point in the k-neighborhood is the point set p i The centroid of the region formed. Find a point q∈R 3 To minimize equation (1). This process can be expressed as follows:

[0084] M=M(p 1 ,…,p k β 1 ,…β k ) = arg min{C(q); q∈R 3} (2)

[0085] The solution of M is the L1 median point in its k-neighborhood.

[0086] ②Step 2: PCA method estimates the point cloud normal vector. The L1 median in equation (2) is used to optimize the principal component analysis of the point cloud. j The neighboring point minus the L1 median, C 3×3 The covariance matrix of is as follows:

[0087]

[0088] Covariance matrix eigenvalue λ 1 <λ 2 <λ 3 , then the eigenvector corresponding to the minimum eigenvalue is the normal vector.

[0089] ③Step 3: Calculate the normal vector divergence feature. Define the local normal vector mean as follows:

[0090]

[0091] The local divergence characteristic factor can be defined by the local normal vector mean as follows:

[0092]

[0093] Among them, n ijis the unit normal vector of each point in the K neighborhood, K is the number of neighborhood points, α i is the local divergence characteristic factor.

[0094] (4) The deviation feature used in this patent is processed by slicing the feature points, extracting the projection of the adjacent point cloud on the cutting plane, smoothing the raised contour of the local weld area using the moving average filtering method, and constructing the deviation feature value according to the degree of deviation between the original projection point and the smoothed point.

[0095] ① Step 1: Slice the feature points along the scan line direction and calculate the projection of each point on the tangent plane to form a two-dimensional coordinate. Let the feature point p A =[x A ,y A ,z A ] T , click p j The distance from the tangent plane C is:

[0096] d j =|[p A -p j ] T ·n C | (6)

[0097] Among them, n C is the normal vector of the tangent plane. j Less than the threshold d thre When j Located on the cutting plane.

[0098] Then, the tangent plane coordinate system is constructed with the tangent plane normal vector, the scan line direction vector and the direction vector perpendicular to the two, and the homogeneous transformation matrix T between the tangent plane coordinate system and the space coordinate system is calculated to convert the three-dimensional point cloud coordinates into two-dimensional coordinates in the tangent plane coordinate system.

[0099] ②Step 2: Use a moving average filter to fit the curve. The sampled data set is adjusted by creating a series of average values ​​of different subsets. As the filter length increases, the smoothness of the output increases, and sharp disturbances in the data can be significantly eliminated. The formula for the moving average filter can be expressed as follows.

[0100]

[0101] Where N is the length of the filter, p M-i Represents the value within the filter range.

[0102] The present invention adopts a moving average filter to obtain the fitting points of the mother curve, and at the same time calculates the deviation between the original point and its corresponding point on the mother curve by the following formula.

[0103]

[0104] Among them, [x pi ,y pi ] represents the corresponding coordinate point after smoothing, [x i ,y i ] represents the initial coordinates before smoothing.

[0105] ③ Step 3: Calculate the tangent plane deviation feature. This patent uses the following formula to construct the deviation feature factor:

[0106]

[0107] (5) The two feature information are weighted and combined to form a new weld feature factor for subsequent feature judgment.

[0108] θ i =ω 1 α i +ω 2 β i (10)

[0109] Among them, ω 1 and ω 2 are the weights of the two features respectively.

[0110] The point cloud feature information obtained in steps (3), (4), and (5) will be used in the feature point threshold determination process of the subsequent region growing process. The specific process of obtaining the subsequent feature threshold will be introduced in detail below. Figure 3 shown.

[0111] (5) The embodiment of the present invention uses part of the weld feature information acquired a priori for the dynamic threshold update process, and models the continuous weld area as a Gaussian process, and uses the interval with a cumulative probability of 99% as the feature threshold for subsequent determination of whether a feature point belongs to the weld point cloud. By continuously updating the a priori information, weld features belonging to the same distribution can be extracted, and the feature threshold is continuously updated to achieve adaptive threshold update.

[0112] ① Step 1: Collect and prepare data samples. Combine some of the acquired prior weld feature information to form a weld area feature set.

[0113] ②Step 2: Train the Gaussian process model. Input the feature information into the Gaussian process model and set a suitable kernel function k(x,x') to model the correlation of the input data. By training the Gaussian process model, the output mean function μ(x) and covariance function σ(x) are obtained.

[0114] ③Step 3: predict the probability distribution of new points. For a new feature point x new, using the Gaussian process model to predict its mean μ(x new ) and variance σ(x new ). Since the output of the Gaussian process follows a normal distribution, the cumulative probability function F(z) of the normal distribution can be used to calculate P(X>Threshold).

[0115] ④ Step 4: Calculate the cumulative probability to reach the 99% threshold. For the normal distribution N(μ,σ 2 ), the threshold with a cumulative probability of 0.99 can be expressed as:

[0116] θ threshold =μ(x new )-z 0.99 ×σ(x new ) (11)

[0117] Among them, z 0.99 is the quantile of the standard normal distribution.

[0118] ⑤ Step 5: Use the threshold obtained in step 4 to determine whether the seed point belongs to the weld feature point. If so, include the point in the prior data sample for subsequent threshold update. If not, the threshold remains unchanged and is still required in step 4.

[0119] (6) The weld feature judgment threshold in the region growing process is dynamically updated by the above method. In the region growing process, when the seed point set is empty, it is determined that the weld region segmentation has been completed. By extracting the weld point cloud and contour information after the segmentation is completed, the weld region segmentation of the three-dimensional point cloud can be accurately completed, such as Figure 4 shown.

[0120] According to another aspect of the present invention, a weld point cloud region segmentation system based on multi-feature dynamic threshold analysis is also provided, which is used to implement the method of any embodiment or a combination of multiple embodiments described above, including:

[0121] The first main control module is used to obtain the three-dimensional point cloud data containing the weld area and select any point in the weld as the initial seed point for region growth;

[0122] The second main control module is used to calculate the normal vector divergence feature and the tangent plane deviation feature of the seed point, and perform three-dimensional point cloud feature fusion based on the calculated feature to form a point cloud fusion feature;

[0123] The third main control module is used to determine whether the seed point belongs to the weld area according to the workpiece point cloud fusion feature and adopts a dynamic threshold update method. If so, the K neighborhood of the weld feature point is included in the seed set as the seed point set for subsequent judgment of whether it belongs to the weld area. If not, it is determined whether the seed point set is empty. If not, the seed point is updated and the updated seed point is input into the second main control module. If it is empty, it is determined that the complete weld area segmentation has been completed.

[0124] The fourth main control module is used to extract the weld point cloud and contour information after the segmentation is completed after the third main control module determines that the complete weld area segmentation has been completed, so as to complete the weld area segmentation of the three-dimensional point cloud.

[0125] It will be easily understood by those skilled in the art that 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 and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A weld point cloud region segmentation method based on multi-feature dynamic threshold analysis, characterized in that: The following steps are involved: Step 1: Obtain 3D point cloud data containing the weld area, and select any point in the weld as the initial seed point for region growth; Step 2: Calculate the normal vector divergence feature and the tangent plane deviation feature of the seed point, and perform three-dimensional point cloud feature fusion based on the feature to form a point cloud fusion feature; Step 3: According to the workpiece point cloud fusion features, a dynamic threshold update method is used to determine whether the seed point belongs to the weld area. If so, the K neighborhood of the weld feature point is included in the seed set as the seed point set for subsequent determination of whether it belongs to the weld area. If not, determine whether the seed point set is empty. If not, update the seed point and return to step 2. If it is empty, determine that the complete weld area segmentation has been completed and enter step 4. Step 4: extract the weld point cloud and contour information after segmentation to complete the weld area segmentation of the three-dimensional point cloud.

2. The weld point cloud region segmentation method based on multi-feature dynamic threshold analysis according to claim 1 is characterized in that: In step 1, a line laser scanner is used to scan the weld area to obtain a narrow point cloud including the weld area, and the narrow point cloud is used as three-dimensional point cloud data.

3. The weld point cloud region segmentation method based on multi-feature dynamic threshold analysis according to claim 1 is characterized in that: In step 2, the calculation of the seed point normal vector divergence feature includes: (211) Calculate the K-neighborhood of each point in the three-dimensional point cloud data and calculate its L1 median point; (212) Based on the L1 median point, a PCA method is used to estimate the normal vector of each point in the three-dimensional point cloud data; (213) Calculate a local normal vector divergence characteristic factor of the feature point based on the normal vector.

4. The weld point cloud region segmentation method based on multi-feature dynamic threshold analysis according to claim 3 is characterized in that: In step (211), let the three-dimensional point cloud data set P k ={p1,p2,…,p j ,…,p k }, where p k ∈R 3 Yes j Nearby k different points, β i For p j The weight of p j A point q∈R in the neighborhood 3 , C(q) is the sum of q and p1,p2,…,p k The weighted sum of the distances, the equation of C(q) includes: Among them, d i (q)=||qp i || represents q and p i The Euclidean distance of Find a point q∈R 3 To minimize the equation for C(q): M=M(p1,…,p k ;β1,…β k )=argmin{C(q);q∈R 3 } Among them, the solution of M is the L1 median point in the k-neighborhood; Preferably, step (212) comprises: Construction C 3×3 The covariance matrix of is: Among them, K is the number of neighborhood points; Calculate the eigenvalues ​​of the covariance matrix and use the eigenvector corresponding to the minimum eigenvalue as the normal vector; Preferably, step (213) comprises: Define the local normal vector mean: The local divergence characteristic factor calculation model is constructed according to the local normal vector mean: Among them, n ij is the unit normal vector of each point in the K neighborhood, K is the number of neighborhood points, α i is the local divergence characteristic factor.

5. The weld point cloud region segmentation method based on multi-feature dynamic threshold analysis according to claim 1, characterized in that: In step 2, the calculation of the seed point tangent plane deviation feature includes: (221) Slice the feature points along the scanning line direction and calculate the projection of each point on the tangent plane to form a two-dimensional coordinate; (222) performing curve fitting on the two-dimensional coordinates using a sliding average filter; (223) constructing a deviation characteristic factor and calculating the tangent plane deviation characteristic, wherein the deviation characteristic factor includes: Where, d i is the deviation between the original point of the two-dimensional coordinate and its corresponding point on the generating curve.

6. The method for weld point cloud region segmentation based on multi-feature dynamic threshold analysis according to claim 5, characterized in that: Step (221) comprises: Let feature point p A =[x A ,y A ,z A ] T , click p j The distance from the tangent plane C is: d j =|[p A -p j ] T ·n C | Among them, n C is the normal vector of the tangent plane, when the distance d j Less than the threshold d thre When p j Located on the tangent plane; The tangent plane coordinate system is constructed with the tangent plane normal vector, the scan line direction vector and the direction vector perpendicular to the two. The homogeneous transformation matrix T between the tangent plane coordinate system and the spatial coordinate system is calculated to convert the three-dimensional point cloud coordinates into two-dimensional coordinates in the tangent plane coordinate system.

7. The method for weld point cloud region segmentation based on multi-feature dynamic threshold analysis according to claim 6, characterized in that: In step (222), the formula of the sliding average filter includes: Where N is the length of the filter, p M-i Represents the value within the filter range; The deviation d between the original point of the two-dimensional coordinate and its corresponding point on the mother curve is calculated by the following formula: i : Among them, [x pi ,y pi ] represents the corresponding coordinate point after smoothing, [x i ,y i ] represents the initial coordinates before smoothing.

8. The weld point cloud region segmentation method based on multi-feature dynamic threshold analysis according to claim 1, characterized in that: In step 2, the three-dimensional point cloud feature fusion includes: weighted combination of normal vector divergence feature and tangent plane deviation feature information to form a new weld feature factor: i i =ω1a i +ω2β i Among them, ω1 and ω2 are weight values.

9. A weld point cloud region segmentation method based on multi-feature dynamic threshold analysis according to any one of claims 1 to 8, characterized in that: Step three includes the following steps: (311) Collect and prepare data samples, and combine some of the acquired prior weld feature information to form a weld area feature set. (312) Train the Gaussian process model, input the feature information into the Gaussian process model, and set a suitable kernel function k(x, x') to model the correlation of the input data. By training the Gaussian process model, the output mean function μ(x) and covariance function σ(x) are obtained; (313) Predict the probability distribution of new points. For a new feature point x new , using the Gaussian process model to predict its mean μ(x new ) and variance σ(x new ), calculate P(X>Threshold) using the cumulative probability function F(z) of the normal distribution; (314) Calculate the cumulative probability to reach a threshold of 99%; (315) Determine whether the seed point belongs to a weld feature point based on the threshold. If so, include the point in the prior data sample for subsequent threshold update. If not, the threshold remains unchanged. Determine whether the seed point set is empty. If not, update the seed point and return to step 2. If it is empty, determine that the complete weld area segmentation has been completed and enter step 4.

10. A weld point cloud region segmentation system based on multi-feature dynamic threshold analysis, characterized in that: include: The first main control module is used to obtain the three-dimensional point cloud data containing the weld area and select any point in the weld as the initial seed point for region growth; The second main control module is used to calculate the normal vector divergence feature and the tangent plane deviation feature of the seed point, and perform three-dimensional point cloud feature fusion based on the calculated feature to form a point cloud fusion feature; The third main control module is used to determine whether the seed point belongs to the weld area according to the workpiece point cloud fusion feature and adopts a dynamic threshold update method. If so, the K neighborhood of the weld feature point is included in the seed set as the seed point set for subsequent judgment of whether it belongs to the weld area. If not, it is determined whether the seed point set is empty. If not, the seed point is updated and the updated seed point is input into the second main control module. If it is empty, it is determined that the complete weld area segmentation has been completed. The fourth main control module is used to extract the weld point cloud and contour information after the segmentation is completed after the third main control module determines that the complete weld area segmentation has been completed, so as to complete the weld area segmentation of the three-dimensional point cloud.

Images