An automatic profiling and attaching optimization method for the front and rear pillar foams in data modeling

Abstract

The present invention discloses an automatic profiling and attaching optimization method for the front and rear pillar foams in data modeling, which relates to the technical field of pillar foam attaching and includes: obtaining a three-dimensional shape data set of the target front and rear pillars of the skylight from multiple perspectives, performing denoising processing and coordinate transformation to obtain a shape space data set of the front and rear pillars of the skylight; performing three-dimensional reconstruction and scoring optimization to build a space model of the front and rear pillars of the skylight; performing profiling and attaching simulation on the space model of the front and rear pillars of the skylight based on the foam material information and thickness information according to the skylight foam attaching standard to generate a front and rear pillar foam attaching model; performing surface feature recognition and attaching path planning to determine the target foam attaching path; executing the target foam attaching path through an automatic attaching device and monitoring the foam attaching to obtain an attaching monitoring data stream, and introducing a closed-loop feedback mechanism to optimize the foam attaching. The present invention solves the technical problem of low foam attaching accuracy in the prior art and achieves the technical effect of improving the foam attaching accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of column foam attachment, and particularly relates to an optimized method for automatic profiling attachment of front and rear column foams for skylights in data modeling. Background Art

[0002] With the continuous development of automobile manufacturing technology, the requirements for vehicle sealing and sound insulation performance are increasing day by day. The precise attachment of front and rear column foams for skylights has become a key link in improving the overall vehicle quality. However, with the increasing complexity of automobile design and the improvement of production automation level, the precision control in the foam attachment process faces many challenges. The traditional foam attachment methods mainly rely on manual operation or fixed mechanical attachment, often having the problem of low attachment precision and being difficult to meet the requirements of high precision and high consistency in modern automobile manufacturing. Summary of the Invention

[0003] This application provides an optimized method for automatic profiling attachment of front and rear column foams for skylights in data modeling, which is used to solve the technical problem of low foam attachment precision in the prior art.

[0004] In view of the above problems, this application provides an optimized method for automatic profiling attachment of front and rear column foams for skylights in data modeling.

[0005] This application provides an optimized method for automatic profiling attachment of front and rear column foams for skylights in data modeling. The method includes:

[0006] Using a three-dimensional scanning device to collect and obtain a three-dimensional shape data set of the target front and rear columns of the skylight from multiple perspectives, performing denoising processing and coordinate transformation on the three-dimensional shape data set to obtain a shape space data set of the front and rear columns of the skylight; based on the shape space data set of the front and rear columns of the skylight, performing three-dimensional reconstruction and scoring optimization to build a space model of the front and rear columns of the skylight; obtaining foam material information and thickness information, and performing profiling attachment simulation on the space model of the front and rear columns of the skylight according to the skylight foam attachment standard based on the foam material information and thickness information to generate a front and rear column foam attachment model; based on the front and rear column foam attachment model, performing surface feature recognition and attachment path planning to determine the target foam attachment path; using an automatic attachment device to execute the target foam attachment path to monitor the foam attachment of the target front and rear columns of the skylight to obtain an attachment monitoring data stream, and introducing a closed-loop feedback mechanism to optimize the foam attachment based on the attachment monitoring data stream.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] This application uses a three-dimensional scanning device to collect and obtain a three-dimensional shape dataset of the front and rear pillars of the target sunroof from multiple perspectives. The three-dimensional shape dataset is denoised and coordinate-transformed to obtain a shape space dataset of the front and rear pillars of the sunroof. Based on the shape space dataset of the front and rear pillars of the sunroof, three-dimensional reconstruction and scoring optimization are performed to build a space model of the front and rear pillars of the sunroof. The foam material information and thickness information are obtained, and based on the foam material information and thickness information, a profiling attachment simulation is performed on the space model of the front and rear pillars of the sunroof according to the sunroof foam attachment standard to generate a foam attachment model for the front and rear pillars. Based on the foam attachment model for the front and rear pillars, surface feature recognition and attachment path planning are performed to determine the target foam attachment path. The target foam attachment path is executed by an automatic attachment device to monitor the foam attachment to the front and rear pillars of the target sunroof, obtaining an attachment monitoring data stream, and a closed-loop feedback mechanism is introduced to optimize the foam attachment based on the attachment monitoring data stream. The present invention solves the technical problem of low foam attachment accuracy in the prior art, and achieves the technical effect of improving the foam attachment accuracy through three-dimensional scanning modeling, profiling attachment simulation, automatic attachment execution, and closed-loop feedback optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for description in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings without creative efforts based on these drawings.

[0010] Figure 1 Schematic flow chart of an automatic profiling attachment optimization method for the front and rear pillar foams of a sunroof with data modeling provided by an embodiment of this application;

[0011] Figure 2 Schematic flow chart of building a space model of the front and rear pillars of a sunroof in an automatic profiling attachment optimization method for the front and rear pillar foams of a sunroof with data modeling provided by an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0012] This application provides an automatic profiling attachment optimization method for the front and rear pillar foams of a sunroof with data modeling, which is used to solve the technical problem of low foam attachment accuracy in the prior art, and achieves the technical effect of improving the foam attachment accuracy through three-dimensional scanning modeling, profiling attachment simulation, automatic attachment execution, and closed-loop feedback optimization.

[0013] Next, in combination with the accompanying drawings in the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts belong to the scope of protection of the present application.

[0014] It should be noted that any variations of the terms "including" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0015] Embodiments, such as Figure 1 As shown, the present application provides an automatic profiling and attaching optimization method for the front and rear pillar foams in data modeling. The method includes:

[0016] Step S100: Use a three-dimensional scanning device to collect and obtain a three-dimensional shape data set of the target front and rear pillars of the skylight from multiple perspectives, perform denoising processing and coordinate transformation on the three-dimensional shape data set to obtain a shape space data set of the front and rear pillars of the skylight.

[0017] In the embodiments of the present application, the three-dimensional shape data of the target front and rear pillars of the skylight is collected by using a three-dimensional scanning device (such as a laser scanner or a structured light scanner). These devices emit laser or structured light and receive the reflected signals to obtain the point cloud data of the target surface, thereby forming a three-dimensional shape data set. Since the data collected during the scanning process may be affected by ambient light, device errors, or the reflection characteristics of the target surface, the data will contain noise. Therefore, denoising processing is performed on the three-dimensional shape data set, that is, by identifying the type (such as random noise, systematic noise), intensity, distribution, and change trend of the noise, and selecting a suitable data filter (such as a Gaussian filter or a median filter) to filter the data, removing invalid or abnormal points to obtain an available three-dimensional shape data set.

[0018] Next, in order to unify the scanned data into a consistent coordinate system, a camera coordinate system is constructed based on the three-dimensional scanning device, and the available three-dimensional shape data set is mapped into this coordinate system to obtain a three-dimensional coordinate set of the front and rear pillars. However, since the requirements for foam attachment may involve specific process requirements, it is necessary to customize the foam attachment coordinate system according to the actual application scenario. By converting the three-dimensional coordinate set of the front and rear pillars into the foam attachment coordinate system, a shape space data set of the front and rear pillars of the skylight is finally obtained.

[0019] Furthermore, in the method provided by the embodiments of the application, the obtaining of the shape space data set of the front and rear pillars of the skylight further includes:

[0020] Identify the noise variation pattern of the three-dimensional shape dataset to obtain noise data characteristic information, where the noise data characteristic information includes noise type, noise intensity, noise distribution, and noise variation trend; select a data filter according to the noise data characteristic information, and use the data filter to perform denoising filtering on the three-dimensional shape dataset to obtain an available three-dimensional shape dataset; construct a camera coordinate system based on the three-dimensional scanning device, map the available three-dimensional shape dataset to the camera coordinate system, and obtain the three-dimensional coordinates of the front and rear columns; customize the foam attachment coordinate system according to the foam attachment requirements, and perform coordinate transformation on the three-dimensional coordinates of the front and rear columns based on the foam attachment coordinate system to obtain the shape space dataset of the front and rear columns of the sunroof.

[0021] In the embodiment of the present application, first, identify the noise variation pattern of the three-dimensional shape dataset to extract the noise data characteristic information. This step uses statistical analysis methods to perform outlier detection and distribution analysis on the point cloud in the three-dimensional shape dataset. By calculating the density distribution and outlier characteristics of the point cloud, identify the type of noise (such as random noise, systematic noise, or environmental noise), noise intensity (the degree of influence of the noise on the data), noise distribution (the spatial distribution of the noise in the dataset), and noise variation trend (the variation pattern of the noise with the scanning conditions or time). These characteristic information provide a scientific basis for subsequent denoising processing.

[0022] Next, based on the noise data characteristic information, select a suitable data filter to perform denoising filtering on the three-dimensional shape dataset. For example, for random noise, use a Gaussian filter to calculate the weighted average of each point through a convolution kernel to smooth the noise; for impulse noise, use a median filter to replace the current point with the median of the neighboring points to remove isolated noise points; for scenarios where edge details need to be retained, use a bilateral filter to retain the target edge information while denoising. Through the filtering process, remove the invalid or abnormal points in the dataset to obtain an available three-dimensional shape dataset, which has higher accuracy and reliability.

[0023] Next, construct a camera coordinate system based on the three-dimensional scanning device, which is the reference coordinate system used by the scanning device when collecting data. This step is achieved through camera calibration technology, using a calibration board or known geometric features to determine the internal parameters (such as focal length, principal point coordinates) and external parameters (such as camera position and attitude) of the camera. Then, map the point cloud coordinates in the available three-dimensional shape dataset to the camera coordinate system through matrix transformation (such as rotation matrix and translation vector) to obtain the three-dimensional coordinates of the front and rear columns. This coordinate set describes the precise geometric position and shape information of the target front and rear columns of the sunroof in the camera coordinate system.

[0024] Finally, according to the process requirements of the foam attachment, customize the foam attachment coordinate system. This step is completed using a coordinate system definition tool (such as CAD software) to determine the origin, axis directions, and scale of the foam attachment coordinate system, ensuring its alignment with the attachment device or process requirements. Then, through a coordinate transformation matrix (such as rotation, translation, and scaling matrices), the three-dimensional coordinate sets of the front and rear pillars are transformed from the camera coordinate system to the foam attachment coordinate system, and finally, the shape space data set of the front and rear pillars of the skylight is obtained. This data set not only contains the accurate geometric information of the target front and rear pillars of the skylight but also ensures the coordinate consistency of the data with the subsequent attachment process, providing a basis for subsequent three-dimensional reconstruction and profiling attachment.

[0025] Step S200: Based on the shape space data set of the front and rear pillars of the skylight, perform three-dimensional reconstruction and scoring optimization to build a three-dimensional space model of the front and rear pillars of the skylight.

[0026] In the embodiment of the present application, based on the shape space data set of the front and rear pillars of the skylight, first, a set of feature points and a set of multi-dimensional descriptors of the feature points of the front and rear pillars of the skylight are obtained through feature point extraction and multi-dimensional description, providing key geometric information for three-dimensional reconstruction; then, using these feature information, a three-dimensional model reconstruction is performed on the data set to generate an initial three-dimensional space model of the front and rear pillars; subsequently, a model space data set is collected and alternately trained based on it to construct a model generation adversarial network including a model generator and a model scorer; finally, the network is used to iteratively score and optimize the initial three-dimensional space model of the front and rear pillars, gradually improving the model accuracy, and finally building a high-precision three-dimensional space model of the front and rear pillars of the skylight.

[0027] Further, as Figure 2 shown, in the method provided by the embodiment of the application, the building of the three-dimensional space model of the front and rear pillars of the skylight further includes:

[0028] Perform feature point extraction and multi-dimensional description on the shape space data set of the front and rear pillars of the skylight to obtain a set of feature points and a set of multi-dimensional descriptors of the feature points of the front and rear pillars of the skylight; based on the set of feature points and the set of multi-dimensional descriptors of the feature points of the front and rear pillars of the skylight, perform three-dimensional model reconstruction on the shape space data set of the front and rear pillars of the skylight to generate an initial three-dimensional space model of the front and rear pillars; collect and obtain a model space data set, and perform alternating training based on the model space data set to obtain a model generation adversarial network, where the model generation adversarial network includes a model generator and a model scorer; based on the model generator and the model scorer, perform iterative scoring and optimization on the initial three-dimensional space model of the front and rear pillars to build a three-dimensional space model of the front and rear pillars of the skylight.

[0029] In the embodiments of the present application, based on the skylight front and rear pillar shape space dataset, first, a feature point extraction algorithm (such as Harris corner detection or SIFT algorithm) is used to identify the key geometric feature points in the dataset. Specifically, Harris corner detection identifies the regions with significant curvature changes as feature points by calculating the gradient changes of pixels or points in the image or point cloud; the SIFT algorithm detects local extreme points as feature points by constructing a scale space. Then, each feature point is described using a multi-dimensional descriptor (such as FPFH or SHOT). For example, FPFH (Fast Point Feature Histograms) generates a histogram describing the local geometric information of the feature point by calculating the normal vector differences in the neighborhood around the feature point; SHOT (Signature of Histograms of OrienTations) describes the geometric distribution of the feature point by combining the local coordinate system and the histogram. Finally, a skylight front and rear pillar feature point set and a feature point multi-dimensional descriptor set are generated.

[0030] Next, based on the extracted skylight front and rear pillar feature point set and the feature point multi-dimensional descriptor set, first, the similarity of the feature points is calculated to generate a feature point similarity set. Specifically, the matching degree between the feature points is quantified by calculating the Euclidean distance or cosine similarity between the feature point descriptors. Then, the feature points are registered according to the feature point similarity set to generate a front and rear pillar matching feature point set. The nearest neighbor search algorithm is used in the registration process to quickly find the matching point pairs, and the RANSAC algorithm is used to eliminate the mismatched points to ensure the registration accuracy. Then, plane projection and grid conversion are performed on the front and rear pillar matching feature point set to generate a front and rear pillar triangular grid. Specifically, the matching feature points are projected onto a two-dimensional plane, and the Delaunay triangulation algorithm is used to generate a triangular grid, and then the grid is mapped back to the three-dimensional space. Finally, the skylight front and rear pillar shape space dataset mapping is applied to the front and rear pillar triangular grid, and the grid vertex information is filled by interpolation or fitting algorithms to complete the three-dimensional model reconstruction and generate an initial front and rear pillar space model.

[0031] To optimize the initial model, a model space dataset is first collected. Specifically, the true geometric data of the front and rear pillars of the skylight is obtained through a 3D scanner or high-precision measurement device to generate a high-precision point cloud or mesh model. Then, the initial model is compared with the true data, and information such as point cloud deviation and geometric error is calculated to generate a model space dataset. For example, the deviation of the model is quantified by calculating the Hausdorff distance or point cloud registration error between the initial model and the true data. Next, based on this dataset, an alternating training is carried out to construct a model generative adversarial network (GAN). The GAN consists of a model generator and a model scorer. The generator is responsible for generating an optimized model, and the scorer evaluates the similarity between the generated model and the true data. Specifically, the generator adopts a convolutional neural network (CNN) or a graph neural network (GNN) structure, takes the initial model as input, and outputs an optimized model; the scorer evaluates the accuracy of the model by calculating the geometric error (such as Chamfer distance) between the generated model and the true data. Through alternating training, the generator continuously optimizes the model generation ability, and the scorer improves the model evaluation accuracy, and finally a generative adversarial network that can efficiently generate high-quality models is obtained.

[0032] Finally, the trained model generator and model scorer are used to iteratively score and optimize the initial front and rear pillar space model. In each iteration, the generator optimizes the model according to the feedback of the scorer to correct the defects in the model (such as uneven surface or feature loss), and the scorer re-evaluates the optimized model. Specifically, the generator reduces the geometric error with the true data by adjusting the vertex positions or mesh structure of the model; the scorer generates a new scoring result by calculating the deviation between the optimized model and the true data. Through multiple iterations, the geometric accuracy and detail restoration degree of the model are gradually improved, and finally a high-precision skylight front and rear pillar space model is built. This process realizes the complete construction from data to high-quality 3D models through the combination of feature extraction, 3D reconstruction, and generative adversarial networks, providing reliable technical support for the precise design and manufacturing of the skylight front and rear pillars.

[0033] Furthermore, in the method provided by the application embodiment, the generation of the initial front and rear pillar space model further includes:

[0034] Calculating the similarity of the skylight front and rear pillar feature point sets based on the feature point multi-dimensional descriptor set to obtain a feature point similarity set; registering the skylight front and rear pillar feature point sets according to the feature point similarity set to obtain a front and rear pillar matching feature point set; performing planar projection and mesh conversion on the front and rear pillar matching feature point set to obtain a front and rear pillar triangular mesh; mapping and applying the skylight front and rear pillar shape space dataset to the front and rear pillar triangular mesh for 3D model reconstruction to generate the initial front and rear pillar space model.

[0035] In the embodiment of the present application, based on the set of feature points of the front and rear pillars of the sunroof and the set of multi-dimensional descriptors of the feature points, a set of feature point similarities is first generated through similarity calculation. Specifically, a measurement method such as Euclidean distance or cosine similarity is used to calculate the similarity between the feature point descriptors. For example, for the FPFH descriptor, a histogram is generated by calculating the difference in normal vectors in the neighborhood around the feature point, and the similarity calculation is achieved by comparing the distribution differences of the histograms. The Euclidean distance quantifies the similarity of feature points by calculating the geometric distance between descriptor vectors; the cosine similarity measures the direction consistency of feature points by calculating the cosine value of the angle between vectors. Through this process, a set of feature point similarities is finally obtained, providing a matching basis for subsequent feature point registration.

[0036] Next, according to the set of feature point similarities, feature point registration is performed on the set of feature points of the front and rear pillars of the sunroof to generate a set of matching feature points for the front and rear pillars. Specifically, a nearest neighbor search algorithm (such as KD-Tree) is used to quickly find matching point pairs. KD-Tree constructs a multi-dimensional space index to efficiently search for the nearest neighbor points, significantly improving the matching efficiency. Then, the RANSAC algorithm is used to eliminate mis-matched points to ensure the registration accuracy. The RANSAC algorithm screens out the optimal matching point pairs through random sampling and consistency verification, effectively improving the robustness of the registration. Finally, a set of matching feature points for the front and rear pillars is obtained, providing aligned point cloud data for subsequent mesh generation.

[0037] Subsequently, plane projection and mesh conversion are performed on the set of matching feature points for the front and rear pillars. Specifically, the matching feature points are projected onto a two-dimensional plane, and the Delaunay triangulation algorithm is used to generate a triangular mesh. Delaunay triangulation generates a uniform and non-overlapping triangular mesh by maximizing the minimum angle principle, ensuring the mesh quality. The projection process reduces the three-dimensional point cloud data to a two-dimensional plane, facilitating subsequent mesh generation. Finally, a triangular mesh for the front and rear pillars is obtained, providing a basic geometric framework for three-dimensional model reconstruction.

[0038] Finally, the shape space data set mapping of the front and rear pillars of the sunroof is applied to the triangular mesh of the front and rear pillars, and the mesh vertex information is filled by bilinear interpolation to complete the three-dimensional model reconstruction and generate an initial spatial model of the front and rear pillars. The interpolation algorithm combines the point cloud data with the mesh structure to ensure the integrity and continuity of the geometric information of the model. The finally obtained initial spatial model of the front and rear pillars provides a basis for subsequent optimization and refinement processing.

[0039] Furthermore, in the method provided by the application embodiment, the building of the spatial model of the front and rear pillars of the sunroof further includes:

[0040] Use the model scorer to score the initial front and rear pillar space model to obtain front and rear pillar model effect scoring information; feedback the front and rear pillar model effect scoring information to the input end of the model generator for optimized generation to obtain a front and rear pillar optimized space model; based on the model scorer and the model scorer, perform iterative scoring and tuning on the front and rear pillar optimized space model to build the skylight front and rear pillar space model.

[0041] In the embodiment of the present application, first use the model scorer to score the initial front and rear pillar space model to generate front and rear pillar model effect scoring information. Specifically, the Hausdorff distance method is used to calculate the maximum deviation between the model surface and the original point cloud data to quantify the geometric error of the model. The Hausdorff distance evaluates the overall geometric accuracy of the model by calculating the farthest distance from each point on the model surface to the point cloud data. Through this process, the front and rear pillar model effect scoring information is obtained, providing a quantitative basis for model optimization.

[0042] Next, feedback the front and rear pillar model effect scoring information to the input end of the model generator for optimized generation to obtain a front and rear pillar optimized space model. Specifically, the Laplacian smoothing algorithm is used to optimize the model surface. Laplacian smoothing reduces surface noise and local distortion by adjusting the positions of the mesh vertices to move them towards the average value of the surrounding vertices, obtaining a front and rear pillar optimized space model.

[0043] Finally, based on the model scorer, perform iterative scoring and tuning on the front and rear pillar optimized space model to build the skylight front and rear pillar space model. Specifically, the Iterative Closest Point algorithm (ICP) is used to further optimize the feature point alignment accuracy. ICP iteratively calculates the closest point pairs between the model and the point cloud data and minimizes the distance between them to gradually improve the matching degree of the feature points. After each iteration, the model scorer re-evaluates the optimization result to generate the latest scoring information to guide the next round of optimization. When the model score reaches the preset quality threshold (such as the geometric error is less than 0.1 mm), the iterative process terminates, and the skylight front and rear pillar space model is obtained.

[0044] Step S300: Obtain the foam material information and thickness information, and perform profiling and attachment simulation on the skylight front and rear pillar space model based on the foam material information and thickness information according to the skylight foam attachment standard to generate a front and rear pillar foam attachment model.

[0045] In the embodiment of the present application, first obtain the foam material information and thickness information from a preset database. Among them, the foam material information includes elastic modulus, Poisson's ratio, and density.

[0046] After obtaining the foam material information and thickness information, determine the evaluation indicators (such as position deviation, overlap limit, porosity, and smoothness) according to the skylight foam attachment standard, and set the profiling attachment parameters. Map the foam material and thickness information to the skylight front and rear pillar space model, and perform profiling attachment simulation optimization through finite element analysis to generate a front and rear pillar foam attachment model that meets the standard.

[0047] Further, in the method provided by the application embodiment, the generating of the front and rear pillar foam attachment model further includes:

[0048] According to the skylight foam attachment standard, determine the foam attachment evaluation indicators, where the foam attachment evaluation indicators include a position deviation threshold, an overlap limit, a porosity, and a smoothness; according to the foam attachment evaluation indicators, set the profiling attachment parameters; map the foam material information and thickness information to the skylight front and rear pillar space model according to the profiling attachment parameters for profiling attachment simulation optimization, and generate the front and rear pillar foam attachment model.

[0049] In the embodiment of the present application, according to the skylight foam attachment standard, first clarify the evaluation indicators, including a position deviation threshold (the allowable deviation range between the foam attachment position and the target position), an overlap limit (the maximum allowable value of the foam edge overlap), a porosity (the void ratio between the foam and the model surface), and a smoothness (the smoothness of the foam attachment surface). These indicators are determined by technical experts.

[0050] Next, based on the determined foam attachment evaluation indicators, set the profiling attachment parameters, such as attachment pressure, temperature, and time, through experimental tests and simulation verification. The setting of these parameters depends on material mechanics performance tests (such as tensile tests and compression tests) and process parameter optimization techniques to ensure that the foam can meet the requirements of position deviation, overlap limit, porosity, and smoothness during the attachment process.

[0051] Subsequently, map the foam material information (such as elastic modulus, Poisson's ratio, and density) and thickness information to the skylight front and rear pillar space model through the material property definition module in computer-aided engineering (CAE) software. This step combines three-dimensional scanning technology to obtain accurate geometric data of the skylight front and rear pillars, and uses the mesh generation tool in finite element analysis (FEA) to discretize the model into multiple small elements, providing a basis for subsequent profiling attachment simulation.

[0052] Finite element analysis technology is used to accurately simulate the contact and deformation between the foam and the front and rear pillar models of the skylight. During the simulation process, contact algorithms (such as penalty function method) are used to deal with the contact problem between the foam and the model surface to ensure that the attachment effect conforms to the actual physical properties. Through the iterative optimization algorithm, the contour attachment parameters are adjusted to gradually optimize the position deviation, overlap limit, void ratio and smoothness of the foam until the evaluation indicators are met. After the simulation optimization is completed, the final front and rear pillar foam attachment model is generated using computer-aided design (CAD) software. This model not only accurately reflects the thickness and material properties of the foam, but also ensures that the attached geometry matches the height of the front and rear pillar space model of the skylight.

[0053] Step S400: performing surface feature recognition and attachment path planning based on the front and rear pillar foam attachment model to determine a target foam attachment path.

[0054] In the embodiment of the present application, based on the front and rear pillar foam attachment model, the key features such as shape contour, curvature change, depression and protrusion are first extracted through surface feature recognition to form a front and rear pillar foam surface feature set. Subsequently, the model is segmented into N attachment areas based on these features. Combined with the attachment path planning objectives (such as efficiency, accuracy and interference avoidance), the path optimization is performed for each area, and the target foam attachment path is finally determined.

[0055] Furthermore, in the method provided in the embodiment of the application, the step of determining the target foam attachment path further includes:

[0056] Based on the front and rear pillar foam attachment model, surface feature recognition is performed to obtain a front and rear pillar foam surface feature set, wherein the front and rear pillar foam surface feature set includes shape contour, curvature change, depression and protrusion; the front and rear pillar foam attachment model is segmented into attachment areas according to the front and rear pillar foam surface feature set to obtain N front and rear pillar foam attachment areas; an attachment path planning target is obtained, and based on the attachment path planning target, an attachment path optimization is performed on the N front and rear pillar foam attachment areas to determine a target foam attachment path.

[0057] In the embodiment of the present application, based on the front and rear pillar foam attachment model, three-dimensional scanning technology (such as laser scanning or structured light scanning) is first used to collect high-precision data on the model surface to obtain point cloud data. Then, the point cloud data is processed using traditional geometric processing algorithms to extract key geometric features. The shape contour is identified by an edge detection algorithm (such as the Canny algorithm) to determine the outer shape boundary of the model; the curvature change is analyzed by a curvature calculation algorithm (such as a method based on local surface fitting) to reflect the degree of curvature of the surface; the depressions and protrusions are identified by a local surface fitting algorithm (such as least squares fitting) to mark the high and low undulating areas on the model surface. These features together constitute the surface feature set of the front and rear pillar foam.

[0058] Next, according to the surface feature sets of the front and rear column foams, the model is segmented using a region segmentation algorithm. For example, the curvature-based region growing algorithm can be used, starting from a seed point, and gradually expanding the region according to curvature similarity; or the watershed algorithm based on edge detection can be used to divide the region through gradient information. By analyzing features such as shape contours, curvature changes, and depressions and protrusions, the model is divided into N front and rear column foam attachment regions.

[0059] Subsequently, the pre-set attachment path planning objectives are obtained, including attachment efficiency, etc. Based on the attachment path planning objectives, a path optimization algorithm is used to optimize the paths of the N attachment regions. For example, the Dijkstra algorithm can be used to search for the shortest path in the region grid graph; or a heuristic search algorithm (such as the simulated annealing algorithm) can be used to find the optimal path in complex regions. During the optimization process, traditional collision detection techniques (such as detection methods based on bounding boxes or distance fields) are combined to ensure that the path does not interfere with other parts of the model. At the same time, kinematic simulation (such as a simulation tool based on the robot kinematic model) is used to verify the feasibility and efficiency of the path, ensuring that the attachment tool (such as a robotic arm) can execute smoothly along the planned path.

[0060] Through the above steps, a target foam attachment path that meets the attachment path planning objectives is generated.

[0061] Furthermore, in the method provided by the application embodiment, the determination of the target foam attachment path further includes:

[0062] Based on the attachment path planning objectives, the attachment paths of the N front and rear column foam attachment regions are respectively optimized to obtain N front and rear column foam attachment paths; using the distribution position information of the N front and rear column foam attachment regions as a constraint condition, the region connection path of the N front and rear column foam attachment paths is optimized to determine the target foam attachment path.

[0063] In the embodiment of the present application, for the N front and rear column foam attachment regions, respectively taking the attachment path planning objectives as the optimization direction, a traditional path optimization algorithm (such as the Dijkstra algorithm) is used for path planning. For example, in a flat area, a straight path is preferentially planned to improve efficiency; in a high-curvature area, the curve path is segmented and fitted to ensure accuracy; in a complex concave-convex area, collision detection techniques (such as the detection method based on the bounding box) are combined to avoid interference. Through the above method, N front and rear column foam attachment paths are generated.

[0064] After completing the path optimization of N regions, extract the distribution position information of the front and rear column foam attachment regions of N regions, including the boundary coordinates, central position of each region, and the spatial relationship between adjacent regions. First, use an edge detection algorithm (such as the Canny algorithm) to extract the boundary contour of each region from the point cloud data. Then, through a geometric center calculation algorithm, calculate the geometric center point of each region, for example, use the average value of the point cloud data as the center point. Finally, use a spatial index structure (such as a KD tree or quadtree) to analyze the spatial relationship between adjacent regions, and determine the connection order and transition path between regions. Through this process, obtain the distribution position information of the front and rear column foam attachment regions of N regions.

[0065] Subsequently, taking the distribution position information of N attachment regions as a constraint condition, use the A* algorithm or heuristic search algorithm to optimize the region connection path. First, generate the start and end points of the connection path according to the central points and boundary information of adjacent regions. Then, use the A* algorithm or heuristic search algorithm to search for the connection path in the global grid map. For example, by calculating the shortest path or optimal transition path between adjacent regions, ensure that the connection path avoids interference while meeting the attachment efficiency. During the path search process, combine the collision detection method based on the distance field to ensure that the path does not interfere with other parts of the model. For example, by calculating the minimum distance between the path and the model surface, exclude the paths that may cause collisions. Finally, use spline curve fitting to smooth the connection path to ensure that the attachment tool (such as a robotic arm) can execute smoothly, and finally generate a region connection path that meets the constraint conditions.

[0066] Through the above steps, integrate the N front and rear column foam attachment paths and the region connection paths to generate a complete target foam attachment path.

[0067] Step S500: Use an automatic attachment device to execute the target foam attachment path to monitor the foam attachment of the front and rear columns of the target skylight, obtain an attachment monitoring data stream, and introduce a closed-loop feedback mechanism to optimize the foam attachment based on the attachment monitoring data stream.

[0068] In the embodiment of the present application, use an automatic attachment device to execute the target foam attachment path to monitor the foam attachment of the front and rear columns of the target skylight, obtain an attachment monitoring data stream, and introduce a closed-loop feedback mechanism to optimize the foam attachment based on the attachment monitoring data stream.

[0069] Specifically, first, a sensor group including a displacement sensor, a pressure sensor, and a vision sensor is installed on the front and rear pillars of the target sunroof to monitor key parameters during the attachment process in real time. Then, based on the attachment path of the target foam, the control parameters of the automatic attachment device are parsed to generate a set of device attachment control parameters to guide the device to perform the attachment operation. During the attachment process, the sensor group collects data such as the fitting accuracy, attachment pressure, and path deviation between the foam and the sunroof surface in real time to form an attachment monitoring data stream. Subsequently, the monitoring data is analyzed through attachment quality analysis to determine the attachment quality deviation amount, such as problems like inaccurate fitting or uneven pressure. Based on the deviation amount, simulation optimization analysis is used to optimize the device attachment control parameter set and the target foam attachment path to generate optimized foam attachment parameters.

[0070] Finally, the optimized parameters are applied to the automatic attachment device to achieve optimized control of the foam attachment to the front and rear pillars of the target sunroof. Through this closed-loop feedback mechanism, the high precision, high quality, and high efficiency of the foam attachment process are ensured, and problems of deviation and defect that may occur during the attachment process are solved.

[0071] Furthermore, in the method provided by the application embodiment, obtaining the attachment monitoring data stream further includes:

[0072] A sensor group is installed on the front and rear pillars of the target sunroof, the sensor group includes a displacement sensor, a pressure sensor, and a vision sensor; based on the attachment path of the target foam, the control parameters of the automatic attachment device are parsed to obtain a set of device attachment control parameters; the automatic attachment device executes the set of device attachment control parameters to perform automatic foam attachment to the front and rear pillars of the target sunroof, and at the same time, the sensor group is used to monitor the foam attachment process in real time to obtain the attachment monitoring data stream.

[0073] In the embodiment of the present application, a sensor group is installed at a predetermined position on the front and rear pillars of the target sunroof, specifically including a displacement sensor (such as a laser displacement sensor or a linear encoder), a pressure sensor (such as a piezoelectric sensor or a strain gauge sensor), and a vision sensor (such as an industrial camera or a 3D vision scanner). The displacement sensor monitors the position deviation between the foam and the sunroof surface in real time through laser ranging or encoder feedback; the pressure sensor measures the pressure applied during the attachment process through the piezoelectric effect or strain gauge; the vision sensor collects the surface image of the foam through a high-resolution industrial camera or a 3D scanner.

[0074] Next, according to the target foam attachment path (including the area attachment path and the connection path), the control parameters of the automatic attachment device (such as a six-axis robotic arm or a numerically controlled attachment machine) are parsed. Through a kinematic model (such as a forward and inverse kinematic model), the path information is converted into a set of device attachment control parameters that the device can execute, including parameters such as movement speed, acceleration, attachment pressure, and tool attitude. Specifically, first, the joint angles or movement trajectories of the device are calculated based on the path coordinate points; second, smooth movement instructions are generated in combination with the movement characteristics of the device (such as maximum speed and acceleration limits); finally, the attachment pressure parameters are integrated with the movement instructions to form a complete set of device attachment control parameters.

[0075] Subsequently, the automatic foam attachment to the front and rear pillars of the target sunroof is performed by the automatic attachment device according to the set of device attachment control parameters. The automatic attachment device performs foam attachment according to the set of device attachment control parameters through a motion control system (such as a servo motor and an encoder). During the attachment process, the sensor group continuously collects the fitting data between the foam and the sunroof surface to form an attachment monitoring data stream.

[0076] Furthermore, in the method provided by the application embodiment, the introduction of the closed-loop feedback mechanism for optimizing the foam attachment based on the attachment monitoring data stream further includes:

[0077] Performing attachment quality analysis on the attachment monitoring data stream to determine the attachment quality deviation amount; performing simulation optimization analysis on the set of device attachment control parameters and the target foam attachment path based on the attachment quality deviation amount to obtain optimized foam attachment parameters, and using the optimized foam attachment parameters to perform optimized control of the foam attachment to the front and rear pillars of the target sunroof.

[0078] In the embodiment of the present application, first, the attachment quality of the attachment monitoring data stream is analyzed. The attachment monitoring data stream includes data collected by displacement sensors, pressure sensors, and vision sensors. First, the data collected by the displacement sensors is analyzed to calculate the position deviation between the foam and the front and rear pillars of the sunroof during the attachment process, and the position deviation amount is obtained, specifically, the offset distance between the actual attachment position of the foam and the target position. Secondly, the data collected by the pressure sensors is processed to evaluate whether the attachment pressure is evenly distributed and whether there is a situation of excessive or too little pressure, and the pressure deviation amount is obtained. The specific evaluation method is as follows: The pressure values at different positions during the attachment process are collected by the pressure sensors, and the standard deviation and average value of the pressure distribution are calculated. If the standard deviation exceeds the preset threshold (for example, ±10%), it is determined that the pressure distribution is uneven, and at the same time, it is checked whether there is a situation of excessive or too little local pressure, and the pressure deviation amount is recorded to avoid foam deformation or poor bonding. Finally, the image data collected by the vision sensors is processed, and traditional image processing techniques (such as grayscale conversion, edge detection, and threshold segmentation) are used to identify defects such as bubbles and wrinkles on the surface of the foam, and the area and position of the defects are calculated, and the defect deviation amount is obtained, specifically, the percentage of the defect area in the total area of the foam and the offset distance between the defect position and the target position. Through this process, the attachment quality deviation amount is obtained.

[0079] Next, the attachment quality deviation amount is input into the finite element analysis (FEA) simulation model to optimize the analysis of the device attachment control parameter set (including movement speed, acceleration, attachment pressure, and tool posture) and the target foam attachment path. The specific steps include establishing a geometric model of the foam and the front and rear pillars of the sunroof in the FEA software, defining the material properties (such as the elastic modulus, Poisson's ratio, and density of the foam, as well as the rigidity and surface characteristics of the front and rear pillars of the sunroof), inputting the current device attachment control parameters and the target foam attachment path into the simulation model, running the simulation to simulate the attachment process of the foam under different parameters, analyzing the stress distribution, deformation situation, and attachment effect of the foam, evaluating whether the attachment quality meets the requirements according to the simulation results, focusing on whether the stress distribution is uniform, whether the foam undergoes excessive deformation, and whether there are bonding defects, and optimizing the attachment process by iteratively adjusting parameters such as movement speed, attachment pressure, and tool posture to ensure uniform stress distribution during the foam attachment process, reduce deformation and defects, and finally output the optimized device attachment control parameters and the target foam attachment path as the optimized foam attachment parameters.

[0080] Load the optimized foam attachment parameters into the control system of the automatic attachment device. Through the motion control system (such as servo motors and encoders), perform foam attachment on the front and rear pillars of the target sunroof according to the optimized parameters and paths. First, input the optimized motion speed, attachment pressure, and tool posture parameters into the servo controller, configure the device operation parameters, and start the device to perform the attachment operation according to the optimized attachment path and parameters to ensure the precise fit of the foam to the front and rear pillars of the sunroof. At the same time, collect data during the attachment process in real time through displacement sensors, pressure sensors, and vision sensors to form a new attachment monitoring data stream, and feedback the monitoring data to the data processing unit for comparison with the target value. If a deviation is found, further fine-tune the device parameters to ensure continuous optimization of the attachment quality.

[0081] In the embodiments of the present application, in summary, the embodiments of the present application at least have the following technical effects:

[0082] This application uses a three-dimensional scanning device to collect and obtain a three-dimensional shape data set of the front and rear pillars of the target sunroof from multiple perspectives, performs denoising processing and coordinate transformation on the three-dimensional shape data set to obtain a shape space data set of the front and rear pillars of the sunroof; performs three-dimensional reconstruction and scoring optimization based on the shape space data set of the front and rear pillars of the sunroof to build a space model of the front and rear pillars of the sunroof; obtains foam material information and thickness information, and performs profiling attachment simulation on the space model of the front and rear pillars of the sunroof based on the foam material information and thickness information according to the sunroof foam attachment standard to generate a foam attachment model for the front and rear pillars; performs surface feature recognition and attachment path planning based on the foam attachment model for the front and rear pillars to determine the target foam attachment path; performs foam attachment monitoring on the front and rear pillars of the target sunroof by executing the target foam attachment path through an automatic attachment device to obtain an attachment monitoring data stream, and introduces a closed-loop feedback mechanism to optimize foam attachment based on the attachment monitoring data stream. The present invention solves the technical problem of low foam attachment accuracy in the prior art and achieves the technical effect of improving the accuracy of foam attachment through three-dimensional scanning modeling, profiling attachment simulation, automatic attachment execution, and closed-loop feedback optimization.

[0083] It should be noted that the above sequence of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above description of specific embodiments of this specification has been made. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0084] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0085] This specification and the accompanying drawings are merely illustrative of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A data modeling optimization method for automatic contour attachment of front and rear pillar foam of a sunroof, characterized in that: The method comprises: Using a three-dimensional scanning device to acquire a three-dimensional shape data set of the front and rear pillars of the target skylight from multiple viewing angles, performing denoising and coordinate transformation on the three-dimensional shape data set to obtain a spatial data set of the front and rear pillar shapes of the skylight; Based on the spatial data set of the shape of the front and rear columns of the skylight, three-dimensional reconstruction and score optimization are performed to build a spatial model of the front and rear columns of the skylight; Acquire foam material information and thickness information, perform contour attachment simulation on the front and rear pillar space model of the sunroof based on the foam material information and thickness information according to the sunroof foam attachment standard, and generate the front and rear pillar foam attachment model; Based on the front and rear pillar foam attachment model, surface feature recognition and attachment path planning are performed to determine a target foam attachment path; Performing foam attachment monitoring on the front and rear pillars of the target sunroof by executing the target foam attachment path through an automatic attachment device to obtain an attachment monitoring data stream, and introducing a closed-loop feedback mechanism to optimize foam attachment based on the attachment monitoring data stream; The method of constructing the space model of the front and rear columns of the skylight includes: Extracting feature points and performing multi-dimensional description on the shape spatial data set of the front and rear columns of the skylight to obtain a feature point set of the front and rear columns of the skylight and a feature point multi-dimensional description sub-set; Reconstructing the three-dimensional model of the skylight front and rear pillar shape space data set based on the skylight front and rear pillar feature point set and the feature point multidimensional descriptor set to generate an initial front and rear pillar space model; Acquire a model space data set, perform alternating training based on the model space data set, and obtain a model generation adversarial network, wherein the model generation adversarial network includes a model generator and a model scorer; Based on the model generator and the model scorer, the initial front and rear pillar space model is iteratively scored and optimized to build a skylight front and rear pillar space model; The method of generating the front and rear column foam attachment model comprises: Determining foam attachment evaluation indicators according to the sunroof foam attachment standard, wherein the foam attachment evaluation indicators include position deviation threshold, overlap limit, void ratio, and smoothness; According to the foam attachment evaluation index, setting contour attachment parameters; Mapping the foam material information and thickness information to the front and rear pillar space model of the sunroof according to the contour attachment parameters to perform contour attachment simulation optimization, and generate the front and rear pillar foam attachment model; The step of determining the target foam attachment path includes: Performing surface feature recognition based on the front and rear pillar foam attachment model to obtain a front and rear pillar foam surface feature set, wherein the front and rear pillar foam surface feature set includes shape contour, curvature change, and concavity and convexity; Segmenting the front and rear pillar foam attachment model into attachment regions according to the front and rear pillar foam surface feature set to obtain N front and rear pillar foam attachment regions; An attachment path planning target is obtained, and based on the attachment path planning target, an attachment path optimization is performed on the N front and rear pillar foam attachment areas to determine a target foam attachment path.

2. The data modeling optimization method for automatic contour attachment of front and rear pillars of a sunroof according to claim 1, characterized in that: The obtaining of the spatial data set of the shapes of the front and rear columns of the skylight comprises: Identifying noise variation rules of the three-dimensional shape data set to obtain noise data characteristic information, wherein the noise data characteristic information includes noise type, noise intensity, noise distribution and noise variation trend; Selecting a data filter according to the noise data characteristic information, and using the data filter to perform denoising filtering on the three-dimensional shape data set to obtain a usable three-dimensional shape data set; Constructing a camera coordinate system based on the three-dimensional scanning device, mapping the available three-dimensional shape data set to the camera coordinate system, and obtaining a three-dimensional coordinate set of the front and rear columns; A foam attachment coordinate system is customized according to the foam attachment requirements, and the three-dimensional coordinate sets of the front and rear pillars are transformed based on the foam attachment coordinate system to obtain the shape space data set of the front and rear pillars of the sunroof.

3. The data modeling optimization method for automatic contour attachment of front and rear pillars of a sunroof according to claim 1, characterized in that: The generating of the initial front and rear column space model comprises: Based on the feature point multidimensional descriptor set, similarity calculation is performed on the feature point set of the front and rear pillars of the skylight to obtain a feature point similarity set; Performing feature point registration on the front and rear pillar feature point set of the sunroof according to the feature point similarity set to obtain a front and rear pillar matching feature point set; Performing plane projection and mesh conversion on the front and rear column matching feature point set to obtain a front and rear column triangular mesh; The three-dimensional model is reconstructed by mapping the shape space data set of the front and rear columns of the skylight to the front and rear column triangle meshes to generate the initial front and rear column space model.

4. The data modeling optimization method for automatic contour attachment of front and rear pillars of a sunroof according to claim 1, characterized in that: The method of constructing the space model of the front and rear columns of the skylight includes: Using the model scorer to score the effect of the initial front and rear column space model to obtain front and rear column model effect score information; Feeding back the front and rear column model effect score information to the input end of the model generator for optimization generation to obtain the front and rear column optimized space model; Based on the model scorer and the model scorer, the front and rear pillar optimization space model is iteratively scored and tuned to build the front and rear pillar space model of the sunroof.

5. The data modeling optimization method for automatic contour attachment of front and rear pillars of a sunroof according to claim 1, characterized in that: The step of determining the target foam attachment path includes: Based on the attachment path planning target, respectively optimizing the attachment paths of the N front and rear pillar foam attachment areas to obtain N front and rear pillar foam attachment paths; The distribution position information of the N front and rear pillar foam attachment areas is used as a constraint condition, and the regional connection path optimization is performed on the N front and rear pillar foam attachment paths to determine the target foam attachment path.

6. The data modeling optimization method for automatic contour attachment of front and rear pillars of a sunroof according to claim 1, characterized in that: The step of obtaining the attached monitoring data stream includes: Installing a sensor group on the front and rear pillars of the target sunroof, wherein the sensor group includes a displacement sensor, a pressure sensor and a visual sensor; Based on the target foam attachment path, the control parameters of the automatic attachment device are analyzed to obtain a device attachment control parameter set; The automatic attaching device executes the device attaching control parameter set to automatically attach foam to the front and rear pillars of the target sunroof, and the sensor group is used to monitor the foam attaching process in real time to obtain the attachment monitoring data stream.

7. The data modeling optimization method for automatic contour attachment of front and rear pillars of a sunroof according to claim 6, characterized in that: The closed-loop feedback mechanism is introduced to optimize the foam attachment based on the attachment monitoring data stream, including: Performing an attachment quality analysis on the attachment monitoring data stream to determine an attachment quality deviation; Based on the attachment quality deviation, the device attachment control parameter set and the target foam attachment path are simulated and optimized to obtain foam optimization attachment parameters, and the foam optimization attachment parameters are used to perform foam attachment optimization control on the front and rear pillars of the target sunroof.

Images