Three-dimensional model data processing method for automobile assembly tolerance analysis

Abstract

The invention relates to the technical field of computer vision and three-dimensional point cloud processing, in particular to a three-dimensional model data processing method for automobile assembly tolerance analysis, and the method comprises the steps: calculating the probability of point cloud data belonging to an initial seed point according to the curvature, reflection intensity and local density of the point cloud data, determining an initial seed point in the point cloud data; obtaining a growth area by using the initial seed point, and calculating the fitting probability of the edge point through the curvature variance and the normal vector angle variance of the edge point; fitting is carried out after fitting edge points are determined from the edge points of the growth area based on the fitting probability, and a spline curve of the growth area is obtained; the virtual points are inserted in the tangential direction of the spline curve of the growth area, the growth area and the virtual points are fitted, and a detection path is planned after the automobile surface contour is obtained, so that the assembly tolerance analysis of the automobile surface contour is realized, and the accuracy of the assembly tolerance analysis of the automobile surface contour is effectively improved.

Description

Technical Field

[0001] This invention relates to the fields of computer vision and 3D point cloud processing technology. More specifically, this invention relates to a 3D model data processing method for tolerance analysis in automotive assembly. Background Technology

[0002] In the automotive manufacturing industry, assembly tolerance analysis is a core step in ensuring the precision of component fit and the quality of the whole vehicle. The three-dimensional dimensional deviations of key components such as body panels and chassis structural components directly affect appearance quality, wind noise level and long-term reliability.

[0003] The above process typically utilizes 3D scanning equipment to acquire point cloud data of body panels or structural components. This data is then registered and compared with the original design model to calculate and evaluate 3D dimensional deviations. To achieve automated inspection, one core step is to use point cloud segmentation technology to identify key features such as mounting holes and mating boundaries. Existing technologies widely employ region growing algorithms. The basic logic of this algorithm is to select seed points based on attributes such as curvature, and then, based on growth criteria, check other points near the seed points for growth until no neighboring points that meet the growth criteria are found, at which point the iteration stops. Ultimately, assembly tolerances can be calculated and visualized.

[0004] However, automotive component surfaces often contain reflective surfaces or complex assembly obstructions, resulting in missing data areas in the 3D point cloud data of the automotive surface contour region acquired by vision sensors. When the region growing algorithm reaches such a missing area, it stops growing because there are no neighboring points that meet the growth criteria. This prevents the algorithm from completely and continuously segmenting the feature region contour, such as a complete mounting surface contour, and only outputs fragmented feature segments.

[0005] Therefore, how to accurately extract the complete feature region contours of the car surface to improve the accuracy of the car assembly tolerance analysis results is a problem that needs to be solved. Summary of the Invention

[0006] To address the technical problem of accurately extracting the complete feature region contours of a vehicle surface, thereby improving the accuracy of vehicle assembly tolerance analysis results, this invention proposes a three-dimensional model data processing method for vehicle assembly tolerance analysis. This method includes the following steps: The process involves acquiring the coordinates, normal vector, curvature, and reflection intensity of each point cloud data point in the automotive component to be analyzed; constructing a spherical neighborhood of the point cloud data to obtain its local density; calculating the probability that a point cloud data point belongs to an initial seed point to determine the initial seed point, as the probability is positively correlated with the curvature, reflection intensity, and local density of that point cloud data; using the initial seed point to perform region growth to obtain a growth region, and calculating the fitting probability of each edge point based on its curvature variance and normal vector angle variance; determining the average interval of the point cloud data based on its coordinates, and then fitting edge points from the edge points of the growth region based on the fitting probability to obtain the spline curve of the growth region; inserting virtual points with an average interval along the tangent direction of the spline curve of the growth region, fitting the growth region with the virtual points to obtain the automotive surface contour, and planning a detection path based on the automotive surface contour to achieve automotive surface contour assembly tolerance analysis.

[0007] This invention, by acquiring the contours of automotive surface components, can maintain stable operation and improve analysis efficiency during assembly tolerance analysis. During the acquisition of the automotive surface contours, this invention considers that some areas may experience breakage or missing point cloud data due to specular reflection. Therefore, this invention analyzes the features of the edges of the grown regions obtained through region growing and performs virtual point interpolation on the missing areas to ensure the physical continuity of the automotive surface contours, thereby improving the accuracy of detection path planning. During region growing, this invention considers that if the initial seed point falls in a non-automotive surface contour area, it may lead to incorrect identification of the final automotive surface contour area. Therefore, this invention determines the initial seed point by assessing the probability that each point cloud data conforms to the characteristics of the automotive surface contour area. This effectively avoids the possibility of significant deviations between the automotive surface contour fragments and the actual automotive surface contour shape caused by growing non-automotive surface contour areas, effectively improving the accuracy of automotive surface contour assembly tolerance analysis.

[0008] According to the present invention, a three-dimensional model data processing method for automobile assembly tolerance analysis is provided. The step of obtaining the coordinates, normal vector, curvature, and reflection intensity of each point cloud data in the automobile component to be analyzed includes: collecting the three-dimensional coordinates of the point cloud data from different angles and performing preprocessing to obtain the coordinates of each point cloud data; determining the normal vector of the point cloud data based on the coordinates of the point cloud data; and collecting the curvature data and reflection data of the point cloud data and performing preprocessing to obtain the curvature and reflection intensity of each point cloud data.

[0009] This invention takes into account that light will be reflected on the surface of the car. Traditional use of single-angle sensors will result in incomplete signal reception and inconsistencies in the dimensions of curvature and reflection data. Therefore, this invention can perform preprocessing after data acquisition to improve the overall quality of point cloud data.

[0010] According to the present invention, a three-dimensional model data processing method for automobile assembly tolerance analysis is provided, wherein constructing a spherical neighborhood of point cloud data to obtain the local density of point cloud data includes: presetting the radius of the spherical neighborhood; obtaining the number of point cloud data in the radius of the spherical neighborhood of point cloud data; and recording the ratio of the number of point cloud data in the spherical neighborhood to the total number of point cloud data as the local density of the point cloud data.

[0011] According to the present invention, a three-dimensional model data processing method for automobile assembly tolerance analysis is provided, wherein calculating the probability that the point cloud data belongs to an initial seed point includes: ; , , , Let be the probability, curvature, reflection intensity, and local density of the i-th point cloud data belonging to the initial seed point, respectively. , , , respectively, are the weights of curvature, reflection intensity, and local density, and e is the natural constant.

[0012] This invention obtains the morphological characteristics of point cloud data through curvature and reflection intensity, and reflects the data integrity of point cloud data through local density. By combining the two, the confidence level of the point cloud data on the surface contour of the car can be comprehensively evaluated. Using point cloud data with high confidence as the initial seed point can effectively avoid the possibility of using point cloud data with high curvature and low density generated by specular reflection as the initial seed point.

[0013] According to the present invention, a three-dimensional model data processing method for automobile assembly tolerance analysis is provided. The step of calculating the probability that point cloud data belongs to an initial seed point to determine the initial seed point in the point cloud data includes: sorting all point cloud data in ascending order according to the probability of all point cloud data belonging to the initial seed point, and using the third quartile of the sort as the seed threshold; if the probability of point cloud data belonging to the initial seed point is greater than the seed threshold, then the point cloud data is the initial seed point.

[0014] According to the present invention, a three-dimensional model data processing method for automobile assembly tolerance analysis is provided. The method of obtaining a growth region by growing a region using an initial seed point includes: setting a growth criterion; if the curvature and normal vector between the seed point and the neighboring point cloud data meet the growth criterion, the neighboring point cloud data is included in the initial growth region, and other neighboring point cloud data is continuously acquired until no other neighboring point cloud data meets the growth criterion, at which point the iteration stops, the growth process of the initial growth region stops, and the growth region is obtained.

[0015] According to the present invention, a three-dimensional model data processing method for automotive assembly tolerance analysis includes calculating the fitting probability of an edge point based on the curvature variance and normal vector angle variance of the edge point in the growth region, comprising: ; Let be the fitted probability of the j-th edge point. , These are the curvature weights and normal vector weights of the edge points, respectively. , Let be the curvature variance and the normal vector angle variance in the spherical neighborhood of the j-th edge point, respectively, and e be the natural constant.

[0016] This invention takes into account that the edge points of the fracture area can be used as anchor points for virtual point interpolation, but these points contain noise points. Therefore, this invention analyzes the curvature fluctuation and normal vector fluctuation of the edge points and identifies points with low curvature fluctuation and low normal vector fluctuation as the most likely true edge points of the car surface contour area, effectively improving the accuracy of virtual point interpolation.

[0017] According to the present invention, a three-dimensional model data processing method for automotive assembly tolerance analysis is provided. The step of determining the average interval of point cloud data based on the coordinates of each point cloud data, and determining the fitting edge points from the edge points of the growth region based on the fitting probability, and then fitting the spline curve of the growth region, includes: using the Euclidean distance between adjacent point cloud data as the interval between point cloud data to obtain the average interval of all point cloud data; recording edge points with a fitting probability greater than a preset fitting probability threshold as fitting edge points, and using the least squares method to fit all fitting edge points to obtain the spline curve of the growth region.

[0018] According to the present invention, a three-dimensional model data processing method for automobile assembly tolerance analysis is provided. The step of planning a detection path based on the automobile surface contour to realize automobile surface contour assembly tolerance analysis includes: generating an assembly tolerance analysis path by equidistant offsetting the automobile surface contour, converting the assembly tolerance analysis path into robot TCP coordinates, and realizing smooth robot motion through S-shaped velocity curve planning.

[0019] This invention achieves smooth robot movement through S-shaped velocity curve planning, which can effectively reduce the possibility of impact when the robot starts, stops, or changes its path, ensuring operational safety and accuracy.

[0020] According to the present invention, a three-dimensional model data processing method for automobile assembly tolerance analysis is provided, wherein generating an assembly tolerance analysis path by equidistant offsetting the automobile surface contour includes: recording the growth region and virtual points as a complete point set, using a B-spline curve to globally fit the complete point set to obtain a smooth B-spline, and generating an assembly tolerance analysis path by equidistant normal offsetting.

[0021] The present invention has the following beneficial effects: Based on the above technical solutions, this invention provides a 3D model data processing method for automotive assembly tolerance analysis. By acquiring the contours of automotive surface components, it can maintain stable operation during the assembly tolerance analysis process, improving analysis efficiency. During the acquisition of the automotive surface contours, this invention considers that some areas may experience point cloud data breaks or gaps due to specular reflection. Therefore, this invention analyzes the features of the edges of the grown regions obtained through region growing and performs virtual point interpolation on the missing areas to ensure the physical continuity of the automotive surface contours, thereby improving the accuracy of detection path planning. During region growing, this invention considers that if the initial seed point falls in a non-automotive surface contour area, it may lead to incorrect identification of the final automotive surface contour area. Therefore, this invention determines the initial seed point by acquiring the probability that each point cloud data conforms to the characteristics of the automotive surface contour area, effectively avoiding the possibility of significant deviations between the automotive surface contour fragments and the actual automotive surface contour shape caused by the growth of non-automotive surface contour areas, effectively improving the accuracy of automotive surface contour assembly tolerance analysis. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the steps of a three-dimensional model data processing method for automobile assembly tolerance analysis according to an embodiment of the present invention. Detailed Implementation

[0023] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0024] Please see Figure 1 , Figure 1 This is a flowchart illustrating the steps of a three-dimensional model data processing method for automobile assembly tolerance analysis according to an embodiment of the present invention.

[0025] It should be noted that during the point cloud segmentation of automotive surface contours using region growing algorithms, the growth process is forced to interrupt when it extends to areas with missing point cloud data because it cannot find continuous points in the neighborhood that meet the preset growth criteria. This results in a series of isolated and discontinuous fragments of the automotive surface contour. Since the automotive surface contour is convex and geometrically continuous, points on the contour typically have a higher curvature than other flat shell substrate surfaces in the contour region, and the difference in normal vectors between adjacent points is smaller. Furthermore, even if the point cloud data is broken due to reflection, the geometric orientation of the automotive surface contour itself remains smooth and continuous. Therefore, the reasonable shape of the missing portion can be inferred from the features of adjacent regions.

[0026] Based on this, embodiments of the present invention can screen initial seed points through local density and morphological features, prioritizing point cloud data with high curvature and high local density to improve the integrity of the point cloud data and the likelihood that the initial seed points are located within the vehicle surface contour region, thus focusing the initial growth range on the actual vehicle surface contour region. When the region-growing algorithm reaches the fracture region caused by reflection, virtual point interpolation can be performed on the missing fracture region based on the point cloud data of the fracture region edge to fit a continuous vehicle surface contour, thereby improving the accuracy of assembly tolerance analysis. The method specifically includes the following steps: S1: Obtain the coordinates, curvature, and reflection intensity of each point cloud data in the automotive component to be analyzed.

[0027] It should be noted that, because light reflects off the contours of a car's surface, traditional point cloud data acquisition using a single-angle sensor results in incomplete signal reception, easily leading to missing point cloud data. Therefore, in this embodiment of the invention, point cloud data can be acquired from different incident angles to cover as much of the blind spot point cloud data as possible caused by specular reflection.

[0028] For example, in an embodiment of the present invention, obtaining the coordinates, curvature, and reflection intensity of each point cloud data in the automotive component to be analyzed includes: preprocessing the three-dimensional coordinates of the point cloud data collected from different angles to obtain the coordinates of each point cloud data; and preprocessing the curvature data and reflection data of the point cloud data to obtain the curvature and reflection intensity of each point cloud data.

[0029] Specifically, when acquiring the coordinates of point cloud data, the contour of the car surface can be fixed, and a robot-driven 405nm blue laser profilometer can be used to perform data acquisition, avoiding infrared reflection interference from the car surface. During the scanning process, 12 scans can be performed from 3 different pitch angles and 4 circumference angles to generate a set of point cloud data, which is then processed using a data fusion algorithm to obtain all the point cloud data. The reflected light signal is captured by the photodetector built into the laser sensor, and the reflection intensity of the point cloud data is obtained by ADC analog-to-digital conversion.

[0030] The pitch angle can be set to 15 degrees, 30 degrees, and 45 degrees, and the surround angle can be set to 0 degrees, 90 degrees, 180 degrees, and 270 degrees, depending on actual needs. The data fusion algorithm can be set to the ICP registration algorithm.

[0031] It should be noted that the collected point cloud data set may contain noise and outliers, and the dimensions of the curvature and reflection intensity of the point cloud data may be different. Therefore, different preprocessing can be performed on the three-dimensional coordinates, curvature data and reflection intensity data of the collected point cloud data to improve the overall quality of the point cloud data.

[0032] For example, preprocessing the three-dimensional coordinates of point cloud data can include using bilateral filtering algorithms for data denoising, etc.

[0033] For example, the preprocessing of point cloud data curvature data and reflection intensity data can include using linear normalization to eliminate the influence of dimensions, etc., and the specific settings can be configured according to actual needs.

[0034] It is understood that the embodiments of the present invention process point cloud data with emission light interference. Therefore, after obtaining the reflection intensity of the point cloud using a laser scanner, the same car surface contour can be scanned from three different angles. If the number of angles with missing point cloud data for the car component to be analyzed is greater than or equal to two, it can be said that the car component to be analyzed has reflective interference and needs to be processed by the method provided in the embodiments of the present invention.

[0035] Based on the above steps, the coordinates, curvature, and reflection intensity of each point cloud data in the automotive component to be analyzed can be obtained. By analyzing the morphological characteristics and local density of the point cloud data to determine the initial seed point, the probability that the initial seed point is located in the automotive surface contour region can be increased, thereby improving the accuracy of obtaining the automotive surface contour by region growing.

[0036] S2: Construct a spherical neighborhood of the point cloud data to obtain the local density of the point cloud data; calculate the probability that the point cloud data belongs to the initial seed point to determine the initial seed point in the point cloud data. The probability that the point cloud data belongs to the initial seed point is positively correlated with the curvature, reflection intensity and local density of the point cloud data.

[0037] It should be noted that point cloud data in the contour area of ​​a car surface usually corresponds to high curvature and high reflection intensity. Therefore, point cloud data with high curvature is more likely to belong to the contour area of ​​the car surface, while high local density points can ensure that the data in this area is complete and not severely affected by reflection.

[0038] Therefore, embodiments of the present invention can screen initial seed points based on the above features, improve the accuracy of seed point selection from the source, reduce the possibility of misselecting substrate points or noise points as seed points, and lay the foundation for the reliability of subsequent regional growth.

[0039] For example, in an embodiment of the present invention, constructing a spherical neighborhood of point cloud data to obtain the local density of the point cloud data includes: presetting the radius of the spherical neighborhood; obtaining the number of point cloud data in the radius of the spherical neighborhood of the point cloud data; and recording the ratio of the number of point cloud data in the spherical neighborhood to the total number of point cloud data as the local density of the point cloud data.

[0040] The radius of the spherical neighborhood can be set to 5, and the specific value can be set according to actual needs.

[0041] For example, in an embodiment of the present invention, the probability that a point cloud data belongs to an initial seed point is calculated based on the curvature, reflection intensity, and local density of the point cloud data. The calculation formula is as follows: ; Let be the probability that the i-th point cloud data belongs to the initial seed point. As curvature weight, Let be the curvature of the i-th point cloud data. As the reflection intensity weight, Let be the reflection intensity of the i-th point cloud data. Let be the local density of the i-th point cloud data. denoted by , where is the weight of the local density, and e is the natural constant.

[0042] in, It can be set to 0.5. It can be set to 0.5. It can be set to 0.1; the curvature and reflection intensity of point cloud data represent the geometric and material characteristics of point cloud data, respectively. They are independent but complementary physical quantities. Therefore, when setting the weights, the same weight can be set to avoid misjudgment caused by a single feature. For example, the reflection of a specular reflection point is high but the curvature is low. The specific value can be set according to actual needs.

[0043] In this calculation method, when the curvature, reflection intensity, and local density of the i-th point cloud data are all high, it indicates that the point cloud data is more likely to be located in the protruding area of ​​the car surface contour, and is also more likely to be used as an initial seed point.

[0044] For example, in an embodiment of the present invention, calculating the probability that point cloud data belongs to an initial seed point to determine the initial seed point in the point cloud data includes: sorting all point cloud data in ascending order according to the probability of all point cloud data belonging to the initial seed point, and using the third quartile of the sort as the seed threshold; if the probability of point cloud data belonging to the initial seed point is greater than the seed threshold, then the point cloud data is the initial seed point.

[0045] It is understandable that the point cloud data of the region with missing reflective properties is discontinuous. If the initial seed point is located in this region, it will be interrupted prematurely or expand in the wrong direction during the region growth process, resulting in a large deviation between the final segmented car surface contour and the actual car surface contour. The present invention can effectively avoid the possibility that the selected initial seed point is located in the region with missing reflective properties by combining the local density and morphological features of the point cloud data to select the seed point.

[0046] S3: Obtain the normal vector of each point cloud data in the automotive component to be analyzed, use the initial seed point to perform region growth to obtain the growth region, calculate the fitting probability of the edge points, and determine the fitting edge points from the edge points of the growth region based on the fitting probability, and then perform fitting to obtain the spline curve of the growth region.

[0047] Among them, the normal vector of the point cloud data can be determined based on the coordinates of the point cloud data.

[0048] For example, in an embodiment of the present invention, a region is grown using an initial seed point to obtain a growing region, including: a preset growth criterion; if the curvature and normal vector between the seed point and neighboring point cloud data satisfy the growth criterion, the neighboring point cloud data is included in the initial growing region, and other neighboring point cloud data is continuously acquired until no other neighboring point cloud data satisfies the growth criterion, at which point the iteration stops, the growth process of the initial growing region stops, and the growing region is obtained.

[0049] Specifically, the growth criteria can be set to two conditions. The first condition is that the curvature value of the neighboring point cloud data of the initial seed point must be greater than the curvature threshold to ensure that the growth process of the seed point always stays in the convex car surface contour area. The second condition is that the angle between the average normal vector of the initial seed point and its neighboring point cloud data is less than the angle threshold to ensure that the growth process can be carried out along a continuous and smooth surface and will not jump to an irrelevant surface.

[0050] The curvature threshold and the included angle threshold can be set according to actual needs, and the embodiments of the present invention do not impose too many restrictions here.

[0051] It should be noted that using the initial seed point for region growing can yield a car surface contour growth region free from reflective interference. This process results in a break at the edge between the reflective area and the car surface contour growth region. However, the car surface contour is continuous, and this characteristic will be retained within the reflective area. Therefore, embodiments of the present invention can infer the morphology of the car surface contour in the reflective missing area based on the characteristics of the growth region connected to the reflective missing area, thus filling in the reflective gaps at the data level.

[0052] It is understandable that reflection-induced fractures will interrupt the growth of the region at the missing points. In this case, some edge points of the growth region may be noise points left by reflection. Therefore, before constructing virtual points based on the edge points of the growth region, edge points with high fitting probability can be selected according to the curvature fluctuation and normal vector fluctuation of the edge points of the growth region. These edge points with high fitting probability are the real fracture boundary points and can be used as anchor points of the car surface contour at the fracture point to achieve accurate interpolation of virtual points.

[0053] For example, the covariance matrix of the point cloud data can be constructed based on the coordinates of the point cloud data. The eigenvector corresponding to the smallest eigenvalue obtained by performing eigenvalue decomposition on the covariance matrix is ​​the normal vector of the point cloud data.

[0054] For example, in an embodiment of the present invention, calculating the fitting probability of an edge point based on the curvature variance and normal vector angle variance of the edge point in the growth region includes: ; Let be the fitted probability of the j-th edge point. The curvature weights of the edge points The weights of the normal vectors of the edge points. Let V be the curvature variance in the spherical neighborhood of the j-th edge point. Let be the variance of the normal vector angle in the spherical neighborhood of the j-th edge point, and e be the natural constant.

[0055] in, It can be set to 0.5. It can be set to 0.5, but the specific value can be adjusted according to actual needs.

[0056] In this calculation method, curvature variance is used to characterize curvature fluctuation, and normal vector angle variance is used to characterize normal vector fluctuation. When both curvature and normal vector fluctuations are small in the spherical neighborhood of the j-th edge point, it indicates that the point cloud data distribution in the spherical neighborhood of the edge point is more stable, the internal data similarity is higher, the probability of belonging to the car surface is higher, and the corresponding fitting probability is also higher.

[0057] For example, in an embodiment of the present invention, after determining the fitting edge points from the edge points of the growth region based on the fitting probability, fitting is performed to obtain the spline curve of the growth region, including: using the Euclidean distance between adjacent point cloud data as the interval between point cloud data to obtain the average interval of all point cloud data; recording edge points with fitting probabilities greater than a preset fitting probability threshold as fitting edge points, and using the least squares method to fit all fitting edge points to obtain the spline curve of the growth region.

[0058] For example, the average interval of the point cloud data is determined based on the coordinates of each point cloud data point.

[0059] When setting the probability threshold, the mean and standard deviation of the fitting probability of all edge points can be obtained. The probability threshold is obtained by subtracting twice the standard deviation from the mean. The specific probability threshold can be set according to actual needs.

[0060] In this way, the embodiments of the present invention insert virtual points in the data missing area according to the normal point cloud density, simulating the point cloud distribution that should exist in the area, which conforms to the actual physical shape of the car surface contour. By merging the real points in the growth area and the fitted virtual points, a continuous and complete car surface contour can be obtained, thereby eliminating the segmentation and breakage caused by reflection and preparing for the car assembly tolerance analysis.

[0061] S4: Insert virtual points with an average interval along the spline curve tangent direction of the growth region, fit the growth region with the virtual points to obtain the vehicle surface profile, and plan the detection path according to the vehicle surface profile to realize the assembly tolerance analysis of the vehicle surface profile.

[0062] For example, in an embodiment of the present invention, a detection path is planned based on the vehicle surface contour to realize the assembly tolerance analysis of the vehicle surface contour, including: generating an assembly tolerance analysis path by equidistant offsetting the vehicle surface contour, converting the assembly tolerance analysis path into robot TCP coordinates, and realizing smooth robot motion through S-shaped velocity curve planning.

[0063] For example, in an embodiment of the present invention, generating an assembly tolerance analysis path by equidistant offsetting the surface contour of a car includes: recording the growth region and virtual points as a complete point set, using a B-spline curve to globally fit the complete point set, obtaining a smooth B-spline, and generating an assembly tolerance analysis path by equidistant normal offsetting.

[0064] Specifically, before predicting the path, the inserted virtual points can be filtered and smoothed. By fitting the complete point set with a B-spline curve, the curvature of the entire car surface contour can be ensured to be continuous and the change is gradual.

[0065] For example, when generating a detection path through equidistant offset, the offset distance can be determined first based on the radius of the detection tool and the single-sided margin of the car surface contour, and the offset direction can be determined based on the normal vector of the car surface contour. After calculating the spatial equidistant offset, it can be checked whether the offset path has self-intersection. If an intersection occurs, adjustments are made to ensure that the starting point and ending point of the offset path are consistent with the normal vector direction of the corresponding position, so as to avoid abrupt turns at the beginning and end of the path.

[0066] Following the steps outlined above, assembly tolerance analysis of the vehicle's surface contours can ultimately be achieved.

[0067] As can be seen, in this embodiment of the invention, when predicting the inspection route for automotive assembly tolerances, the coordinates, normal vector, curvature, and reflection intensity of each point cloud data in the automotive component to be analyzed can be obtained; a spherical neighborhood of the point cloud data can be constructed to obtain the local density of the point cloud data; the probability of the point cloud data belonging to the initial seed point can be calculated to determine the initial seed point in the point cloud data. The probability of the point cloud data belonging to the initial seed point is positively correlated with the curvature, reflection intensity, and local density of the point cloud data; the initial seed point is used to perform region growth to obtain the growth region, and the fitting probability of the edge point is calculated based on the curvature variance and normal vector angle variance of the edge point in the growth region; the average interval of the point cloud data is determined based on the coordinates of each point cloud data, and the fitting edge point is determined from the edge point of the growth region based on the fitting probability and then fitted to obtain the spline curve of the growth region; virtual points with an interval of the average interval are inserted along the tangent direction of the spline curve of the growth region, and the growth region is fitted with the virtual points to obtain the automotive surface contour, and the inspection path is planned based on the automotive surface contour to realize the automotive surface contour assembly tolerance analysis, which effectively improves the accuracy of the automotive surface contour assembly tolerance analysis.

[0068] 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 three-dimensional model data processing method for automobile assembly tolerance analysis, characterized by, The method comprises the following steps: acquiring the coordinates, normal vector, curvature and reflection intensity of each point cloud data in the automobile part to be analyzed; constructing the spherical neighborhood of the point cloud data to acquire the local density of the point cloud data; calculating the probability of the point cloud data belonging to the initial seed point to determine the initial seed point in the point cloud data, wherein the probability of the point cloud data belonging to the initial seed point is positively correlated with the curvature, reflection intensity and local density of the point cloud data; using the initial seed point to perform region growing to obtain a growing region, calculating the fitting probability of an edge point in the growing region according to the curvature variance and normal vector angle variance of the edge point, determining the average interval of the point cloud data according to the coordinates of each point cloud data, and performing fitting on the fitting edge point determined from the edge points in the growing region based on the fitting probability to obtain a spline curve of the growing region; 2. The three-dimensional model data processing method for automobile assembly tolerance analysis according to claim 1, characterized in that, inserting a virtual point with an interval equal to the average interval in the tangential direction of the spline curve of the growing region, fitting the growing region and the virtual point to obtain an automobile surface contour, and planning a detection path according to the automobile surface contour to realize automobile surface contour assembly tolerance analysis. The method comprises the following steps: acquiring the coordinates, normal vector, curvature and reflection intensity of each point cloud data in the automobile part to be analyzed; 3. The three-dimensional model data processing method for automobile assembly tolerance analysis according to claim 1, characterized in that, acquiring the three-dimensional coordinates of the point cloud data at different angles, and then performing preprocessing to obtain the coordinates of each point cloud data; and determining the normal vector of the point cloud data according to the coordinates of the point cloud data; acquiring the curvature data and reflection data of the point cloud data, and then performing preprocessing to obtain the curvature and reflection intensity of each point cloud data. The method comprises the following steps:

4. The three-dimensional model data processing method for automobile assembly tolerance analysis according to claim 1, characterized in that, presetting the radius of the spherical neighborhood; ； 、 、 、 Pi, C, R, D are the probability, curvature, reflectance intensity, local density of the i-th point cloud data belonging to the initial seed point, respectively, 、 、 are the weights of curvature, reflectance intensity, local density, respectively, and e is the natural constant.

5. The three-dimensional model data processing method for automobile assembly tolerance analysis according to claim 1, characterized in that, acquiring the number of point cloud data in the spherical neighborhood radius of the point cloud data, and taking the ratio of the number of point cloud data in the spherical neighborhood to the total number of point cloud data as the local density of the point cloud data. The method comprises the following steps:

6. The three-dimensional model data processing method for automobile assembly tolerance analysis according to claim 1, characterized in that, The method comprises the following steps: arranging all the point cloud data in ascending order according to the probability of the point cloud data belonging to the initial seed point, taking the third quartile of the arranged point cloud data as a seed threshold, and taking the point cloud data whose probability of belonging to the initial seed point is greater than the seed threshold as the initial seed point. The method comprises the following steps:

7. The three-dimensional model data processing method for automobile assembly tolerance analysis according to claim 1, characterized in that, presetting a growth criterion; ； the fitting probability of the jth edge point, , the curvature weight, the normal vector weight of the edge point, respectively, , the curvature variance, the normal vector angle variance in the spherical neighborhood of the jth edge point, respectively, e is the natural constant.

8. The three-dimensional model data processing method for automobile assembly tolerance analysis according to claim 1, characterized in that, if the curvature and normal vector between the seed point and the adjacent point cloud data satisfy the growth criterion, then the adjacent point cloud data is included in the initial growing region, and other adjacent point cloud data is continuously acquired until there is no other adjacent point cloud data satisfying the growth criterion, and the growth process of the initial growing region is stopped to obtain the growing region. The method comprises the following steps: The method comprises the following steps: taking the Euclidean distance between adjacent point cloud data as the interval between the point cloud data to obtain the average interval of all the point cloud data. The edge point with the fitting probability greater than the preset fitting probability threshold is recorded as a fitted edge point, and all the fitted edge points are fitted using a least square method to obtain a spline curve of the growth region.

9. The three-dimensional model data processing method for automobile assembly tolerance analysis according to claim 1, characterized in that, The method comprises the following steps: The automobile surface profile is equidistantly offset to generate an assembly tolerance analysis path, and the assembly tolerance analysis path is converted into a robot TCP coordinate to realize smooth movement of the robot through S-shaped velocity curve planning.

10. The three-dimensional model data processing method for automobile assembly tolerance analysis according to claim 9, wherein, The method comprises the following steps: The growth region and the virtual point are recorded as a complete point set, the complete point set is globally fitted using a B-spline curve to obtain a smooth B-spline, and an assembly tolerance analysis path is generated through equidistant normal offset.

Images