A machine learning-based method, device, and medium for automated automotive styling.

Abstract

This application discloses a machine learning-based automated automotive styling method, device, and medium, relating to the field of electronic digital data processing. The method involves obtaining component description information from a preliminary automotive styling sketch to obtain components to be assembled; arranging and combining these components according to their types and size ratios to obtain several component combinations for a complete automotive styling, and generating several models of the vehicle to be styling based on surface modeling; detecting each model to identify those with areas requiring optimization and identifying the areas and anomalies; determining the target optimization points and their actual coordinates, and calculating the deviation of the areas requiring optimization based on ideal coordinates, to determine compensation values ​​for the areas requiring optimization; adjusting the areas requiring optimization based on the compensation values ​​to obtain optimized models of the vehicle to be styling, and rendering these optimized models to generate a 3D model of the optimized vehicle, thus achieving automated automotive styling.

Description

Technical Field

[0001] This application relates to the field of image recognition technology, and in particular to a machine learning-based method, device and medium for automated automotive styling. Background Technology

[0002] As consumers increasingly demand personalized and differentiated car exteriors, design teams need to quickly generate and evaluate multiple design options to ensure that the final model appeals to the target consumer group. However, the traditional car styling design process is cumbersome and time-consuming, typically requiring designers to gradually construct the car's shape through multiple stages such as hand-drawing, modeling, and rendering. This not only demands high levels of professional skills from designers but also limits the diversity and innovation of design options and significantly extends the product's timeline from design to market. Summary of the Invention

[0003] This application provides a machine learning-based automotive automated styling method, device, and medium to solve the aforementioned technical problems.

[0004] On one hand, embodiments of this application provide a machine learning-based automated vehicle styling method, including:

[0005] Obtain component description information from the initial sketches of the car design, and obtain the assembly parts of the car design based on the component description information; wherein, the component description information includes the region to which each car component belongs, the shape outline of the component, the geometric features of the component, and the size ratio between each component;

[0006] Based on the types of parts to be assembled and the size ratio between each part, the parts to be assembled are arranged and combined to obtain several parts combinations corresponding to the complete car shape, and several cars to be shaped are generated according to the curved surface shape.

[0007] Each car to be styled is inspected to identify those with areas that need optimization, and the anomaly types corresponding to those areas are identified; the anomaly types include irregular styling and incompatible specifications.

[0008] Based on the anomaly type, the target point to be optimized and the actual coordinate information of the target point to be optimized are determined in the part to be optimized. Combined with the ideal coordinate information of the target point to be optimized, the degree of deviation of the part to be optimized is calculated, so as to determine the compensation value corresponding to the part to be optimized based on the degree of deviation.

[0009] The parts to be optimized are adjusted according to the compensation value to obtain the optimized car to be styled. The optimized car to be styled is then rendered to dynamically display the corresponding 3D model, thereby realizing automated car styling.

[0010] In one implementation of this application, obtaining component description information from a preliminary automotive styling sketch specifically includes:

[0011] Obtain a pre-determined preliminary sketch of the car's design, and input the preliminary sketch into an image recognition model for analysis to determine multiple shapes in the preliminary sketch;

[0012] For each of the multiple shapes, the edge of the corresponding shape in the preliminary manuscript is identified by an edge detection algorithm, and the shape outline in the preliminary manuscript is segmented based on the shape edge and a graph segmentation algorithm.

[0013] Based on the degree of matching between the segmented shape contour and the preset car parts, the region to which the car parts to which the shape contour belongs is determined; wherein, the region to which the parts belong includes: the front of the car, the rear of the car, the roof of the car, and the body of the car.

[0014] Based on the preliminary manuscript, the size ratio between the components corresponding to the design intent is determined, and the geometric features of the automotive component are calculated by combining the shape outline and the region to which the component belongs.

[0015] In one implementation of this application, obtaining the automotive-shaped assembly parts based on the component description information specifically includes:

[0016] Based on the preset mapping relationship, determine multiple selectable accessories corresponding to the area to which the component belongs in the preset accessory style library;

[0017] Based on the shape outline in the component description information, determine whether there is a target component to be assembled corresponding to the area to which the component belongs from the selectable components;

[0018] If so, then obtain at least one target component to be assembled corresponding to the shape outline;

[0019] If not, then based on the region to which the component belongs in the component description information and the feature points in the shape outline, the corresponding assembly part is constructed, and based on the geometric features of the component in the component description information, the size of the constructed assembly part is adjusted to generate the target assembly part.

[0020] In one implementation of this application, the components to be assembled are arranged and combined according to the types of components and the size ratio between the components to be assembled, resulting in several component combinations corresponding to the complete car shape. Then, based on the curved surface shape, several cars to be shaped corresponding to these component combinations are generated. Specifically, this includes:

[0021] All parts to be assembled are classified, the type of each part is determined, and the detailed information of the parts to be assembled is determined in the preset part style library; wherein, the part types include at least: body, door, window, rearview mirror and tire, and the detailed information of the parts includes size, shape, curvature and curvature direction;

[0022] Based on the characteristics of the car's styling, the types of accessories required for each car to be styling and the number of accessories required for each type of accessory are determined. Combined with the size ratio between the components in the component description information and the detailed accessory information of each component to be assembled, all components to be assembled are arranged and combined to generate several complete car shapes corresponding to several accessory combinations.

[0023] For each of the several component combinations, based on the detailed component information, the component parameters corresponding to each component to be assembled in the component combination are determined, and the component parameters of adjacent components in the same component combination are adjusted according to the connection method between adjacent components; wherein, the component parameters include at least the length, width, height of the vehicle body and the curvature of the curve;

[0024] Using a surface modeling algorithm, multiple automotive surfaces corresponding to the component combination are generated based on the adjusted component parameters. These multiple automotive surfaces are then spliced ​​together to obtain the corresponding automotive model to be modeled.

[0025] In one implementation of this application, each car to be styled is inspected to identify cars with parts that need optimization, and the anomaly type corresponding to the parts that need optimization is identified, specifically including:

[0026] For each car to be styled, the geometric features of the car to be styled are extracted from the 3D model of the car to be styled, and a curvature map corresponding to the 3D model is generated; wherein, the geometric features of the car to be styled include: curvature, radius, angle and slope;

[0027] Based on the curvature changes in the curvature diagram, the transition regions in the three-dimensional model are identified, and in conjunction with the transition regions, it is determined whether the continuity of the same component surface in the three-dimensional model is abnormal.

[0028] If yes, then it is determined that the car to be styled has parts that need to be optimized, and the anomaly type corresponding to the parts that need to be optimized is unsmooth styling; if no, then it is determined whether the connection between adjacent parts is matched based on the accessory parameters in the car to be styled and the size ratio between each part.

[0029] In the case of connection mismatch, it is determined that the car to be styled has a part that needs to be optimized, and the anomaly type corresponding to the part to be optimized is specification mismatch; wherein, connection matching is used to determine whether the gap between the parts matches the preset gap and whether adjacent parts are aligned.

[0030] In one implementation of this application, the target point to be optimized and its actual coordinate information are determined based on the anomaly type. Then, the deviation degree of the part to be optimized is calculated by combining the ideal coordinate information of the target point to be optimized. Specifically, this includes:

[0031] For parts to be optimized where the anomaly type is irregular shape, the target anomaly region of the part to be optimized is identified in the transition region and surface inflection point corresponding to the part to be optimized based on the curvature change in the curvature diagram, and the target point to be optimized is determined in the target anomaly region.

[0032] The ideal coordinate information of the target optimization point is determined based on the component description information, and the actual coordinate information of the target optimization point is measured in the three-dimensional model of the car to be modeled using a coordinate measuring machine.

[0033] Determine the difference between the actual coordinate information and the ideal coordinate information in each dimension of the 3D model, calculate the sum of squares of the differences in each dimension, and take the square root of the sum of squares in each dimension to obtain the degree of deviation of the part to be optimized.

[0034] Based on the actual coordinate information, ideal coordinate information, and preset compensation factor for each dimension, the deviation component corresponding to the degree of deviation in each dimension is calculated.

[0035] In one implementation of this application, the target point to be optimized and its actual coordinate information are determined based on the anomaly type. Then, the deviation degree of the part to be optimized is calculated by combining the ideal coordinate information of the target point to be optimized. Specifically, this includes:

[0036] For the parts to be optimized that are of the abnormal type of specification mismatch, the actual dimensions of the parts to be optimized and the parts connected to the parts to be optimized are obtained through the three-dimensional model of the car to be modeled.

[0037] Based on the component description information, determine the size ratio between the part to be optimized and the connected part, and combine the actual dimensions of the part to be optimized and the connected part to determine whether the part to be optimized needs to be sized.

[0038] If yes, then calculate the deviation between the actual size and the ideal size of the part to be optimized; if no, then use a coordinate measuring machine to measure the actual coordinate information of the connection point between the part to be optimized and the connected part in the three-dimensional model of the car to be modeled.

[0039] Determine the ideal coordinates of the connection point between the part to be optimized and the connected part, and calculate the degree of deviation of the connection point based on the actual coordinates and the ideal coordinates.

[0040] In one implementation of this application, the optimized car to be styled is rendered to dynamically display the corresponding 3D model of the optimized car to be styled, specifically including:

[0041] The optimized car to be styled is exported to the rendering engine, and the material textures corresponding to each accessory of the car to be styled are selected in the rendering engine in combination with the car styling display intent corresponding to the component description information; wherein, the material textures are used to simulate the paint color and metallic luster;

[0042] In the rendering engine, a lighting environment is selected for the car to be modeled, the lighting environment is used to simulate a real scene, and the car to be modeled is rendered according to preset rendering parameters;

[0043] An animation is created for the car to be modeled, the corresponding animation sequence is rendered by the rendering engine, and the rendered animation is encoded so as to be dynamically displayed in video format; wherein, the animation is used to simulate the movement of the car.

[0044] On the other hand, embodiments of this application also provide a machine learning-based automated automotive styling device, the device comprising:

[0045] At least one processor;

[0046] And, a memory communicatively connected to the at least one processor;

[0047] The memory stores instructions that can be executed by the at least one processor, which enable the at least one processor to perform a machine learning-based automated vehicle styling method as described above.

[0048] On the other hand, embodiments of this application also provide a non-volatile computer storage medium storing computer-executable instructions, which, when executed, implement the aforementioned machine learning-based automated vehicle styling method.

[0049] This application provides a machine learning-based automated automotive styling method, device, and medium, which offers at least the following advantages:

[0050] By acquiring component descriptions from initial car design sketches, including the region, shape, geometric features, and size ratios of each component, the system can accurately determine the parts to be assembled. This significantly improves the efficiency of converting creative sketches into actual car models, reducing the tedious and error-prone process of manually selecting and matching parts. Based on the types of parts and their precise size ratios, the components are arranged and combined to generate multiple combinations and corresponding car models, fully exploring the possibilities of different combinations. This helps designers find the most innovative and practical car designs among numerous options. Each car model is automatically detected to identify any anomalies. Optimizing the parts and their corresponding anomaly types allows for the timely detection of irregularities affecting the aesthetics of the design, or assembly problems caused by incompatible specifications. This avoids rework due to design flaws in subsequent production stages, saving significant time and costs. By identifying target optimization points based on anomaly types and calculating the degree of deviation using actual and ideal coordinate information, compensation values ​​can be determined. This enables timely fine-tuning of the parts to be optimized, ensuring the adjusted car styling meets design expectations and improving the overall quality and refinement of the car's design. Finally, by rendering the optimized car into a 3D model, the design results become more intuitive, facilitating observation of the car's styling from various angles and allowing for the early detection of potential problems. Attached Figure Description

[0051] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0052] Figure 1 A flowchart illustrating a machine learning-based automated vehicle styling method provided in this application embodiment;

[0053] Figure 2 This is a schematic diagram of the internal structure of an automated automotive styling device based on machine learning, provided as an embodiment of this application. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0055] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0056] Figure 1 This is a flowchart illustrating a machine learning-based automated vehicle styling method provided in an embodiment of this application.

[0057] The analysis method involved in the embodiments of this application can be implemented by a terminal device or a server, and this application does not impose any special limitations on it. For ease of understanding and description, the following embodiments are all described in detail using a server as an example.

[0058] It should be noted that the server can be a single device or a system composed of multiple devices, i.e., a distributed server. This application does not make any specific limitations on this.

[0059] like Figure 1 As shown in the figure, an embodiment of this application provides a machine learning-based automated vehicle styling method, including:

[0060] 101. Obtain the component description information from the preliminary car design sketch, and obtain the car design parts to be assembled based on the component description information.

[0061] The component description information includes the region to which each automotive component belongs, the component's shape outline, the component's geometric features, and the size ratio between the components.

[0062] In one embodiment of this application, obtaining component description information from a preliminary automotive styling sketch specifically includes:

[0063] Obtain a pre-determined preliminary sketch of the car's design, and input the preliminary sketch into an image recognition model for analysis to determine multiple shapes in the preliminary sketch;

[0064] For each of the multiple shapes, the edge detection algorithm is used to identify the corresponding shape edge in the preliminary manuscript, and the shape outline in the preliminary manuscript is segmented based on the shape edge and a graph segmentation algorithm.

[0065] Based on the degree of matching between the segmented shape contour and the preset car parts, the region to which the car part to which the shape contour belongs is determined; wherein, the region to which the part belongs includes: the front of the car, the rear of the car, the roof of the car, and the body of the car.

[0066] Based on the preliminary sketches, the size ratios between the components corresponding to the design intent are determined, and the geometric features of the automotive components are calculated by combining the shape outlines and the regions to which the components belong.

[0067] In one embodiment, a preliminary manuscript is used as the input image and imported into a pre-trained image recognition model. This model employs deep learning technology and is capable of recognizing basic shapes and features in the image. The model preprocesses the input image, including denoising and contrast enhancement, to improve the accuracy of subsequent shape recognition.

[0068] Then, edge detection algorithms, such as Canny edge detection, are used to process the preprocessed image to identify the edges of various shapes in the manuscript. Edge detection algorithms determine the edge positions of shapes by calculating the gradient changes of pixel values ​​in the image. Graph segmentation algorithms, such as GrabCut, are then used to further process the identified edges to segment the shape outlines in the manuscript. Graph segmentation algorithms achieve accurate shape segmentation by analyzing the differences in color, texture, and other features between the edges and the background.

[0069] Next, the segmented shape contours are matched against a pre-defined automotive parts library. This library contains shape contour templates for various automotive parts, along with their corresponding part names and regions, such as the front, rear, roof, and body. By calculating the similarity between the shape contours and the library templates, such as using a shape context matching algorithm, the automotive part corresponding to each shape contour and its corresponding region are determined.

[0070] By combining the shape contour and the region information to which the component belongs, the geometric features of each automotive component are calculated. It should be noted that the geometric features in this embodiment include the component's dimensions (length, width, height, etc.), shape (circle, rectangle, curve, etc.), and relative positional relationship with other components. Finally, the parsed component description information (including component name, region, geometric features, etc.) is output in a structured format.

[0071] In one embodiment of this application, obtaining the automotive-shaped assembly parts based on component description information specifically includes:

[0072] Based on the preset mapping relationship, determine multiple selectable accessories corresponding to the area to which the component belongs in the preset accessory style library;

[0073] Based on the shape outline in the component description information, determine whether there is a target component to be assembled corresponding to the area to which the component belongs from the selectable components;

[0074] If so, then obtain at least one target component to be assembled corresponding to the shape outline;

[0075] If not, the corresponding assembly part is constructed based on the region to which the part belongs and the feature points in the shape outline in the part description information. The size of the constructed assembly part is adjusted according to the geometric features of the part in the part description information to generate the target assembly part.

[0076] In one embodiment, a pre-defined mapping relationship has been established in the automotive design project. This relationship defines the correspondence between the region to which a component belongs (such as the front, rear, roof, body, etc.) and the components in a pre-defined component style library. The pre-defined component style library contains 3D models of various automotive components, which are categorized and stored according to the different regions to which the components belong.

[0077] Based on a preset mapping relationship, multiple candidate parts corresponding to the region to which the component belongs are searched in a preset part style library. For example, if the region to which the component belongs is the front of a vehicle, all parts related to the front of the vehicle are selected as candidate parts from the part style library. The shape outline in the component description information is matched with the candidate parts. The matching process can be achieved by calculating the similarity between the shape outline and the part outline, and the similarity can be calculated using a shape context matching algorithm or other shape matching algorithms. If there is a candidate part that completely matches or is highly similar to the shape outline, it is identified as the target component to be assembled.

[0078] If no matching component is found among the available components, a new component needs to be constructed based on the component description information. This construction process can be achieved using 3D modeling software, building a basic component shape based on the region the component belongs to and feature points in its shape profile. After construction, the dimensions of the constructed component are adjusted according to the component's geometric features in the component description information. This adjustment can be achieved through scaling, stretching, and other operations to ensure the generated component meets design requirements. After the matching or construction and adjustment steps, the final target component that meets the design requirements is generated. These components can be stored as 3D models and used in subsequent automotive assembly simulations, rendering demonstrations, and other stages.

[0079] 102. Based on the types of accessories to be assembled and the size ratio between each component, arrange and combine the accessories to be assembled to obtain several accessory combinations corresponding to the complete car shape, and generate several cars to be shaped based on the curved surface shape corresponding to several accessory combinations.

[0080] In one embodiment of this application, the components to be assembled are arranged and combined according to the types of components and the size ratio between the components to be assembled, resulting in several component combinations corresponding to a complete car shape. Then, based on the curved surface shape, several cars to be shaped corresponding to these component combinations are generated. Specifically, this includes:

[0081] All parts to be assembled are classified, the type of each part is determined, and the detailed information of the parts to be assembled is determined in the preset part style library. The part types include at least: body, door, window, rearview mirror and tire. The detailed information of the parts includes size, shape, curvature and curvature direction.

[0082] Based on the characteristics of the car's styling, the types of parts required for each car to be styling and the number of parts required for each type of part are determined. Combined with the size ratio between the parts in the part description information and the detailed information of each part to be assembled, all the parts to be assembled are arranged and combined to generate several complete car shapes and several parts combinations.

[0083] For each of the several component combinations, based on the detailed component information, determine the component parameters corresponding to each component to be assembled in the component combination, and adjust the component parameters of adjacent components in the same component combination according to the connection method between adjacent components; wherein, the component parameters include at least the length, width, height of the vehicle body and the curvature of the curve;

[0084] Using a surface modeling algorithm, multiple automotive surfaces corresponding to the component combinations are generated based on the adjusted component parameters. These multiple automotive surfaces are then spliced ​​together to obtain the corresponding automotive model to be modeled.

[0085] In one embodiment, classifying all components to be assembled according to their function and location facilitates the orderly combination and parameter adjustment of subsequent components. This includes, but is not limited to, components such as the vehicle body, doors, windows, rearview mirrors, and tires. In a preset component style library, for each component to be assembled, its detailed information is searched and determined, including key parameters such as dimensions (length, width, height, etc.), shape, surface curvature, and surface curvature direction.

[0086] Based on the characteristics of the vehicle's styling, the types and quantities of parts required for each vehicle to be styled are determined. Combining the size ratios between components in the component description information, as well as the detailed component information for each part to be assembled, all components are arranged and combined. Through algorithm optimization, several component combinations corresponding to complete vehicle styling that meet the design requirements are generated.

[0087] For each component assembly, the initial component parameters for each component to be assembled are determined based on the component details. These parameters include, but are not limited to, the length, width, height of the vehicle body, and the curvature of curves. Based on the connection method between adjacent components (such as hinged connections, bolted connections, etc.), the component parameters of adjacent components within the same assembly are adjusted to ensure smooth connections between components and to meet actual assembly requirements.

[0088] Surface modeling algorithms (such as NURBS surface algorithms and subdivision surface algorithms) are applied to generate multiple automotive surfaces corresponding to each component assembly based on the adjusted component parameters. These surfaces should possess high smoothness and continuity to ensure that the final generated car model has a realistic appearance and texture. The generated multiple automotive surfaces are then stitched together to obtain the corresponding 3D model of the car to be modeled. The stitching process may involve smooth transitions between surfaces to ensure the coherence and consistency of the entire car model.

[0089] 103. Inspect each car to be styled to identify those with parts that need optimization and identify the anomaly types corresponding to those parts; among them, the anomaly types include irregular styling and incompatible specifications.

[0090] In one embodiment of this application, each car to be styled is inspected to identify cars with parts that need optimization, and the anomaly type corresponding to the parts to be optimized is identified, specifically including:

[0091] For each car to be styled, the geometric features of the car to be styled are extracted from the 3D model of the car to be styled, and the curvature map corresponding to the 3D model is generated; wherein, the geometric features of the car to be styled include: curvature, radius, angle and slope.

[0092] Based on the curvature changes in the curvature diagram, the transition regions in the 3D model are identified, and combined with the transition regions, it is determined whether the continuity of the same component surface in the 3D model is abnormal.

[0093] If yes, then it is determined that there are parts of the car to be styled that need to be optimized, and the anomaly type corresponding to the parts to be optimized is styling irregularity; if no, then it is determined whether the connection between adjacent parts is matched based on the accessory parameters of the car to be styled and the size ratio between each part.

[0094] In the case of connection mismatch, it is determined that there are parts of the car to be styled that need to be optimized, and the anomaly type corresponding to the parts to be optimized is specification mismatch; among them, connection matching is used to determine whether the gap between the parts matches the preset gap and whether adjacent parts are aligned.

[0095] In one embodiment, for each 3D model of the car to be modeled, its geometric features, including key parameters such as curvature, radius, angle, and slope, are extracted using specialized 3D modeling software or algorithms. These features reflect the shape, contour, and surface variations of the car model. Based on the extracted geometric features, a curvature map corresponding to the 3D model is generated. The curvature map is a visualization tool that can intuitively show the curvature changes of the model's surface, aiding in subsequent analysis and optimization.

[0096] By utilizing curvature variation information in the curvature map, transition regions in the 3D model are identified. These transition regions are typically located at the boundaries of different surfaces or components, and are areas prone to irregularities or breaks in the automotive styling. Based on the information from these transition regions, the continuity of surfaces of the same component in the 3D model is checked. By comparing parameters such as curvature variations and radius differences between adjacent surfaces, the continuity and smoothness of the surfaces are determined. If continuity anomalies are found on surfaces of the same component (such as abrupt curvature changes or inconsistent radii), it is determined that there are areas in the automotive model to be optimized, and the anomaly type is irregular styling.

[0097] After confirming the continuity of the curved surface, the connection between adjacent parts is verified based on the component parameters and size ratios of the parts in the car to be modeled. Connection matching verification includes two aspects: first, checking whether the gaps between components match the preset gaps to ensure that there are no excessively large or small gaps during assembly; second, checking whether adjacent components are aligned to ensure that the assembled car model has a neat appearance and stable structure. If a mismatch is found between adjacent components (such as excessive gaps or misaligned components), it is determined that there are parts in the car to be modeled that need optimization, and the anomaly type is specification mismatch.

[0098] For the identified areas requiring optimization and their anomalies, corresponding optimization plans were developed. For areas with uneven shapes, improvements could be made by adjusting surface parameters and adding transition surfaces; for areas with incompatible specifications, it was necessary to redesign component dimensions and adjust assembly methods. The optimization plans were then applied to the 3D model, and steps such as geometric feature extraction, curvature map generation, transition region identification, continuity checks, and connection matching verification were performed again to ensure that the optimized car model met design requirements and possessed high manufacturing feasibility.

[0099] 104. Based on the anomaly type, determine the target point to be optimized and its actual coordinate information in the part to be optimized. Combine this with the ideal coordinate information of the target point to be optimized to calculate the degree of deviation of the part to be optimized, and determine the corresponding compensation value of the part to be optimized based on the degree of deviation.

[0100] In one embodiment of this application, the target point to be optimized and its actual coordinate information are determined based on the anomaly type. Then, the deviation degree of the part to be optimized is calculated by combining this with the ideal coordinate information of the target point. Specifically, this includes:

[0101] For parts to be optimized where the anomaly type is irregular shape, the target anomaly region of the part to be optimized is identified in the transition region and surface inflection point of the part to be optimized based on the curvature change in the curvature map, and the target point to be optimized is determined in the target anomaly region.

[0102] The ideal coordinates of the target points to be optimized are determined based on the component description information, and the actual coordinates of the target points to be optimized are measured in the three-dimensional model of the car to be modeled using a coordinate measuring machine.

[0103] Determine the difference between the actual coordinate information and the ideal coordinate information in each dimension of the 3D model, calculate the sum of squares of the differences in each dimension, and take the square root of the sum of squares in each dimension to obtain the degree of deviation of the part to be optimized.

[0104] Based on the actual coordinate information, ideal coordinate information, and preset compensation factor for each dimension, the deviation component corresponding to the degree of deviation in each dimension is calculated.

[0105] In one embodiment, for the part to be optimized where the anomaly type is irregular shape, the target anomaly region is first meticulously identified in the transition region and surface inflection points based on the curvature change information in the curvature map. These regions typically exhibit characteristics such as abrupt curvature changes, surface discontinuities, or excessive inflection points. Within the target anomaly region, the target points to be optimized are further determined. These points are usually key locations causing irregular shape, such as surface inflection points and curvature extrema.

[0106] Based on the component description information, the ideal coordinates of the target points to be optimized are determined. Using a high-precision coordinate measuring machine (CMM), the actual coordinates of the target points to be optimized are accurately measured in the 3D model of the vehicle to be modeled. The CMM provides high-precision spatial coordinate measurements, ensuring data accuracy.

[0107] Determine the differences between the actual and ideal coordinate information along each dimension (X, Y, Z axes) of the 3D model. These differences reflect the deviation between the actual and designed positions of the part to be optimized in space. Calculate the sum of squares of the differences in each dimension and take the square root of the sum to obtain the overall degree of deviation of the part to be optimized. This step uses the concept of Euclidean distance, which can intuitively reflect the deviation of the part to be optimized in 3D space.

[0108] Based on the actual coordinates, ideal coordinates, and preset compensation factors for each dimension, the deviation component corresponding to the degree of deviation in each dimension is calculated. The preset compensation factor is an adjustment coefficient pre-set based on design experience and manufacturing tolerances, used to adjust the calculation results of the deviation component to better reflect reality. Calculating the deviation component helps designers gain a deeper understanding of the deviation of the part to be optimized in each dimension.

[0109] Based on the calculation results of the deviation degree and deviation components, a detailed optimization plan is formulated. The optimization plan may include adjusting surface parameters, adding transition surfaces, and modifying component dimensions. The optimization plan is applied to the 3D model, and curvature diagram analysis, coordinate measuring machine (CMM) measurement, and deviation degree calculation are performed again to ensure that the optimized car model meets design requirements and has high manufacturing feasibility. This accurately identifies and quantifies the target abnormal areas and their deviation degree of the parts to be optimized, which not only improves the accuracy and quality of the car design but also ensures that the final generated car model has high manufacturing feasibility and market competitiveness.

[0110] In one embodiment of this application, the target point to be optimized and its actual coordinate information are determined based on the anomaly type. Then, the deviation degree of the part to be optimized is calculated by combining this with the ideal coordinate information of the target point. Specifically, this includes:

[0111] For parts that are not compatible with specifications and require optimization, the actual dimensions of the parts to be optimized and the parts connected to them are obtained from the 3D model of the car to be modeled.

[0112] Based on the component description information, determine the size ratio between the part to be optimized and the connected parts, and combine the actual dimensions of the part to be optimized and the connected parts to determine whether the part to be optimized needs to be sized.

[0113] If yes, then calculate the deviation between the actual size and the ideal size of the part to be optimized; if no, then use a coordinate measuring machine to measure the actual coordinate information of the connection point between the part to be optimized and the connected parts in the three-dimensional model of the car to be modeled.

[0114] Determine the ideal coordinates of the connection points between the part to be optimized and the connected parts, and calculate the degree of deviation of the connection points based on the actual coordinates and the ideal coordinates.

[0115] In one embodiment, firstly, the actual dimensions of the part to be optimized and its connected parts are accurately obtained using a 3D model of the vehicle to be designed. This includes key dimensional parameters such as length, width, and height. Next, based on the component description information, the size ratio between the part to be optimized and its connected parts is determined. These ratios are typically determined in the early stages of design and serve as a benchmark for subsequent design and manufacturing. A comprehensive analysis is then performed, combining the actual dimensions and size ratios of the part to be optimized and its connected parts, to determine whether dimensional adjustments are needed. If the actual dimensions do not conform to the preset ratios or do not match the dimensions of the connected parts, then dimensional adjustments are required.

[0116] If it is determined that the part to be optimized requires dimensional adjustment, the deviation between the actual and ideal dimensions of the part to be optimized is further calculated. This can be achieved by comparing the actual dimensions with the preset ideal dimensions (usually derived from the CAD model from the initial design phase or verified design data). The deviation can be calculated using various methods, such as directly calculating the difference or calculating the relative error. The specific method chosen depends on factors such as design requirements and manufacturing tolerances.

[0117] If it is determined that the part to be optimized does not require dimensional adjustment, or if further verification of the accuracy of the connection points is needed after dimensional adjustment, a high-precision coordinate measuring machine (CMM) is used to measure the actual coordinates of the connection points between the part to be optimized and the connected parts in the 3D model of the vehicle to be modeled. Simultaneously, based on the component description information or design data, the ideal coordinates of the connection points are determined. These ideal coordinates typically represent the expected positions from the initial design stage. Based on the actual and ideal coordinates, the degree of deviation of the connection points is calculated. This can be achieved by comparing the differences between the two coordinate points in various dimensions (X, Y, Z axes).

[0118] 105. Adjust the parts to be optimized according to the compensation value to obtain the optimized car to be modeled, and render the optimized car to generate the corresponding 3D model, thus realizing the automated modeling of the car.

[0119] Specifically, in one embodiment of this application, the optimized vehicle to be styled is rendered to dynamically display the corresponding 3D model of the optimized vehicle to be styled, specifically including:

[0120] The optimized car to be modeled is exported to the rendering engine, and the material textures corresponding to each part of the car to be modeled are selected in the rendering engine based on the car modeling display intention corresponding to the part description information; among them, the material textures are used to simulate the paint color and metallic luster.

[0121] In the rendering engine, a lighting environment is selected for the car to be modeled. The lighting environment simulates a real scene, and the car to be modeled is rendered according to the preset rendering parameters.

[0122] Animations are created for the car to be modeled, and the corresponding animation sequence is rendered using a rendering engine. The rendered animation is then encoded to be dynamically displayed in video format; the animation is used to simulate the movement of the car.

[0123] In one embodiment, after optimizing and adjusting the car to be styled, it is exported to a format recognizable by the rendering engine (such as OBJ, FBX, etc.) and imported into the rendering engine. Based on the car's styling intent corresponding to the component description information, appropriate material textures are selected for each component of the car in the rendering engine. These material textures should accurately simulate the paint color, metallic sheen, and appearance characteristics of other materials such as plastic and glass. When selecting material textures, the reflection, refraction, and shadow effects between different materials must also be considered to ensure the realism and three-dimensionality of the rendering results.

[0124] In the rendering engine, select a suitable lighting environment for the car to be modeled. The lighting environment should be able to simulate the lighting conditions in a real-world scene, including natural and artificial light. By adjusting parameters such as the position, intensity, and color of the light source, different visual effects and atmospheres can be created. At the same time, the impact of lighting on material textures must also be considered to ensure the accuracy and realism of the rendering results.

[0125] The vehicle to be modeled is rendered based on preset rendering parameters (such as resolution, sampling quality, anti-aliasing, etc.). The rendering process generates high-quality image or video frames to showcase the vehicle's appearance and details. During rendering, it is necessary to closely monitor the progress and results, and adjust rendering parameters and material textures as needed to ensure rendering quality and efficiency.

[0126] Animations are created for the car to be modeled to simulate its movement. Animations can include simple rotations and translations, as well as complex driving scenes and interactive effects. The corresponding animation sequence is then rendered in the rendering engine. The rendering process generates a series of consecutive images or video frames to showcase the car's dynamic effects.

[0127] The rendered animation is encoded and converted into common video formats (such as MP4, AVI, etc.). The encoding process should ensure video quality and smoothness. The encoded animation is then dynamically displayed through video playback software or an online platform. The display process can include features such as loop playback and slow-motion playback, allowing customers to gain a more comprehensive understanding of the car's appearance and performance.

[0128] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide a machine learning-based automated automotive styling device, the structure of which is as follows: Figure 2 As shown.

[0129] Figure 2 This is a schematic diagram of the internal structure of a machine learning-based automated automotive styling device, provided as an embodiment of this application. Figure 2 As shown, the device includes:

[0130] At least one processor;

[0131] And, a memory that is communicatively connected to at least one processor;

[0132] The memory stores instructions that can be executed by at least one processor, and the instructions, when executed by at least one processor, enable at least one processor to:

[0133] Obtain component description information from the initial sketches of the car design, and obtain the assembly parts of the car design based on the component description information; wherein, the component description information includes the region to which each car component belongs, the shape outline of the component, the geometric features of the component, and the size ratio between each component;

[0134] Based on the types of parts to be assembled and the size ratio between each part, the parts to be assembled are arranged and combined to obtain several parts combinations corresponding to the complete car shape, and several cars to be shaped are generated according to the curved surface shape.

[0135] Each car to be styled is inspected to identify those with areas that need optimization, and the anomaly types corresponding to these areas are identified. Anomaly types include styling irregularities and incompatible specifications.

[0136] Based on the anomaly type, determine the target point to be optimized and its actual coordinate information in the part to be optimized. Combined with the ideal coordinate information of the target point to be optimized, calculate the degree of deviation of the part to be optimized, and determine the corresponding compensation value of the part to be optimized based on the degree of deviation.

[0137] The parts to be optimized are adjusted according to the compensation value to obtain the optimized car to be modeled. The optimized car to be modeled is then rendered to generate a corresponding 3D model, thus realizing automated car modeling.

[0138] This application also provides a non-volatile computer storage medium storing computer-executable instructions, which, when executed, can:

[0139] Obtain component description information from the initial sketches of the car design, and obtain the assembly parts of the car design based on the component description information; wherein, the component description information includes the region to which each car component belongs, the shape outline of the component, the geometric features of the component, and the size ratio between each component;

[0140] Based on the types of parts to be assembled and the size ratio between each part, the parts to be assembled are arranged and combined to obtain several parts combinations corresponding to the complete car shape, and several cars to be shaped are generated according to the curved surface shape.

[0141] Each car to be styled is inspected to identify those with areas that need optimization, and the anomaly types corresponding to these areas are identified. Anomaly types include styling irregularities and incompatible specifications.

[0142] Based on the anomaly type, determine the target point to be optimized and its actual coordinate information in the part to be optimized. Combined with the ideal coordinate information of the target point to be optimized, calculate the degree of deviation of the part to be optimized, and determine the corresponding compensation value of the part to be optimized based on the degree of deviation.

[0143] The parts to be optimized are adjusted according to the compensation value to obtain the optimized car to be modeled. The optimized car to be modeled is then rendered to generate a corresponding 3D model, thus realizing automated car modeling.

[0144] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0145] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A machine learning-based automated vehicle styling method, characterized in that, The method includes: Obtain component description information from the initial sketches of the car design, and obtain the assembly parts of the car design based on the component description information; wherein, the component description information includes the region to which each car component belongs, the shape outline of the component, the geometric features of the component, and the size ratio between each component; Based on the types of parts to be assembled and the size ratio between each part, the parts to be assembled are arranged and combined to obtain several parts combinations corresponding to the complete car shape, and several cars to be shaped are generated according to the curved surface shape. Each car to be styled is inspected to identify those with areas that need optimization, and the anomaly types corresponding to those areas are identified; the anomaly types include irregular styling and incompatible specifications. Based on the anomaly type, the target point to be optimized and the actual coordinate information of the target point to be optimized are determined in the part to be optimized. Combined with the ideal coordinate information of the target point to be optimized, the degree of deviation of the part to be optimized is calculated, so as to determine the compensation value corresponding to the part to be optimized based on the degree of deviation. The parts to be optimized are adjusted according to the compensation value to obtain the optimized car to be shaped, and the optimized car to be shaped is rendered to dynamically display the three-dimensional model corresponding to the optimized car to be shaped, thereby realizing automated car styling. Based on the component description information, the automotive-shaped parts to be assembled are obtained, specifically including: Based on the preset mapping relationship, determine multiple selectable accessories corresponding to the area to which the component belongs in the preset accessory style library; Based on the shape outline in the component description information, determine whether there is a target component to be assembled corresponding to the area to which the component belongs from the selectable components; If so, then obtain at least one target component to be assembled corresponding to the shape outline; If not, then based on the region to which the component belongs in the component description information and the feature points in the shape outline, the corresponding assembly part is constructed, and based on the geometric features of the component in the component description information, the size of the constructed assembly part is adjusted to generate the target assembly part.

2. The machine learning-based automated vehicle styling method according to claim 1, characterized in that, Obtain component description information from the initial automotive styling sketches, specifically including: Obtain a pre-determined preliminary sketch of the car's design, and input the preliminary sketch into an image recognition model for analysis to determine multiple shapes in the preliminary sketch; For each of the multiple shapes, the edge of the corresponding shape in the preliminary manuscript is identified by an edge detection algorithm, and the shape outline in the preliminary manuscript is segmented based on the shape edge and a graph segmentation algorithm. Based on the degree of matching between the segmented shape contour and the preset car parts, the region to which the car parts to which the shape contour belongs is determined; wherein, the region to which the parts belong includes: the front of the car, the rear of the car, the roof of the car, and the body of the car. Based on the preliminary manuscript, the size ratio between the components corresponding to the design intent is determined, and the geometric features of the automotive component are calculated by combining the shape outline and the region to which the component belongs.

3. The machine learning-based automated vehicle styling method according to claim 1, characterized in that, Based on the types of components to be assembled and the size ratios between them, the components are arranged and combined to obtain several component combinations corresponding to the complete car shape. Then, based on the curved surface shapes, several car models corresponding to these component combinations are generated, specifically including: All parts to be assembled are classified, the type of each part is determined, and the detailed information of the parts to be assembled is determined in the preset part style library; wherein, the part types include at least: body, door, window, rearview mirror and tire, and the detailed information of the parts includes size, shape, curvature and curvature direction; Based on the characteristics of the car's styling, the types of accessories required for each car to be styling and the number of accessories required for each type of accessory are determined. Combined with the size ratio between the components in the component description information and the detailed accessory information of each component to be assembled, all components to be assembled are arranged and combined to generate several complete car shapes corresponding to several accessory combinations. For each of the several component combinations, based on the detailed component information, the component parameters corresponding to each component to be assembled in the component combination are determined, and the component parameters of adjacent components in the same component combination are adjusted according to the connection method between adjacent components; wherein, the component parameters include at least the length, width, height of the vehicle body and the curvature of the curve; Using a surface modeling algorithm, multiple automotive surfaces corresponding to the component combination are generated based on the adjusted component parameters. These multiple automotive surfaces are then spliced ​​together to obtain the corresponding automotive model to be modeled.

4. The machine learning-based automated vehicle styling method according to claim 1, characterized in that, Each car to be styled is inspected to identify those with areas requiring optimization, and the anomaly type corresponding to these areas is identified, specifically including: For each car to be styled, the geometric features of the car to be styled are extracted from the 3D model of the car to be styled, and a curvature map corresponding to the 3D model is generated; wherein, the geometric features of the car to be styled include: curvature, radius, angle and slope; Based on the curvature changes in the curvature diagram, the transition regions in the three-dimensional model are identified, and in conjunction with the transition regions, it is determined whether the continuity of the same component surface in the three-dimensional model is abnormal. If yes, then it is determined that the car to be styled has parts that need to be optimized, and the anomaly type corresponding to the parts that need to be optimized is unsmooth styling; if no, then it is determined whether the connection between adjacent parts is matched based on the accessory parameters in the car to be styled and the size ratio between each part. In the case of connection mismatch, it is determined that the car to be styled has a part that needs to be optimized, and the anomaly type corresponding to the part to be optimized is specification mismatch; wherein, connection matching is used to determine whether the gap between the parts matches the preset gap and whether adjacent parts are aligned.

5. The machine learning-based automated vehicle styling method according to claim 1, characterized in that, Based on the anomaly type, determine the target point to be optimized and its actual coordinates within the area to be optimized. Then, combine this with the ideal coordinates of the target point to be optimized to calculate the degree of deviation of the area to be optimized. Specifically, this includes: For parts to be optimized where the anomaly type is irregular shape, the target anomaly region of the part to be optimized is identified in the transition region and surface inflection point corresponding to the part to be optimized based on the curvature change in the curvature diagram, and the target point to be optimized is determined in the target anomaly region. The ideal coordinate information of the target optimization point is determined based on the component description information, and the actual coordinate information of the target optimization point is measured in the three-dimensional model of the car to be modeled using a coordinate measuring machine. Determine the difference between the actual coordinate information and the ideal coordinate information in each dimension of the 3D model, calculate the sum of squares of the differences in each dimension, and take the square root of the sum of squares in each dimension to obtain the degree of deviation of the part to be optimized. Based on the actual coordinate information, ideal coordinate information, and preset compensation factor for each dimension, the deviation component corresponding to the degree of deviation in each dimension is calculated.

6. The machine learning-based automated vehicle styling method according to claim 1, characterized in that, Based on the anomaly type, determine the target point to be optimized and its actual coordinates within the area to be optimized. Then, combine this with the ideal coordinates of the target point to be optimized to calculate the degree of deviation of the area to be optimized. Specifically, this includes: For the parts to be optimized that are of the abnormal type of specification mismatch, the actual dimensions of the parts to be optimized and the parts connected to the parts to be optimized are obtained through the three-dimensional model of the car to be modeled. Based on the component description information, determine the size ratio between the part to be optimized and the connected part, and combine the actual dimensions of the part to be optimized and the connected part to determine whether the part to be optimized needs to be sized. If yes, then calculate the deviation between the actual size and the ideal size of the part to be optimized; if no, then use a coordinate measuring machine to measure the actual coordinate information of the connection point between the part to be optimized and the connected part in the three-dimensional model of the car to be modeled. Determine the ideal coordinates of the connection point between the part to be optimized and the connected part, and calculate the degree of deviation of the connection point based on the actual coordinates and the ideal coordinates.

7. The machine learning-based automated vehicle styling method according to claim 1, characterized in that, The optimized car to be styled is rendered to dynamically display the corresponding 3D model, specifically including: The optimized car to be styled is exported to the rendering engine, and the material textures corresponding to each accessory of the car to be styled are selected in the rendering engine in combination with the car styling display intent corresponding to the component description information; wherein, the material textures are used to simulate the paint color and metallic luster; In the rendering engine, a lighting environment is selected for the car to be modeled, the lighting environment is used to simulate a real scene, and the car to be modeled is rendered according to preset rendering parameters; An animation is created for the car to be modeled, the corresponding animation sequence is rendered by the rendering engine, and the rendered animation is encoded so as to be dynamically displayed in video format; wherein, the animation is used to simulate the movement of the car.

8. A machine learning-based automated automotive styling device, characterized in that, The device includes: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform a machine learning-based automated vehicle styling method as described in any one of claims 1-7.

9. A non-volatile computer storage medium storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed, they implement a machine learning-based automated vehicle styling method as described in any one of claims 1-7.

Images