Three-dimensional modeling method, system and equipment based on path lofting and medium

Abstract

The invention relates to the technical field of three-dimensional modeling, and discloses a three-dimensional modeling method, system and equipment based on path lofting and a medium. The method comprises the following steps: acquiring path data and section contour data; sequentially inputting the path data and the section contour data into a pre-constructed first neural network for feature extraction to obtain local curve features and local contour features; the local curve features and the local contour features are sequentially input into a pre-constructed second neural network for fitting, and a lofting path and a section contour are obtained; and carrying out lofting on the section contour along the lofting path to generate an initial three-dimensional model, and adjusting the initial three-dimensional model to obtain a target three-dimensional model. According to the method, three-dimensional modeling is carried out based on path lofting and a deep learning algorithm, so that the workload of manual adjustment can be reduced, and the modeling efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional modeling, and in particular to a three-dimensional modeling method, system, equipment and medium based on path lofting. Background Art

[0002] 3D modeling technology can support path lofting to a certain extent, but traditional modeling methods often rely on manual adjustments, which are inefficient and difficult to achieve precise and consistent shape changes. Current modeling tools focus on overall shape generation or texture optimization, but fail to fully incorporate the advantages of path lofting.

[0003] Therefore, there is an urgent need for a 3D modeling method that can efficiently and accurately create complex shapes that change along the path. Summary of the Invention

[0004] In order to solve the above technical problems, in a first aspect, the present invention provides a three-dimensional modeling method based on path lofting, comprising:

[0005] Obtain path data and cross-section profile data;

[0006] Inputting the path data and the cross-sectional profile data into a pre-built first neural network in sequence for feature extraction to obtain local curve features and local profile features;

[0007] Inputting the local curve features and the local contour features into a pre-built second neural network in sequence for fitting, thereby obtaining a lofting path and a cross-sectional contour;

[0008] The cross-sectional profile is lofted along the lofting path to generate an initial three-dimensional model, and the initial three-dimensional model is adjusted to obtain a target three-dimensional model.

[0009] In an optional embodiment, lofting the cross-sectional contour along the lofting path to generate an initial three-dimensional model, and adjusting the initial three-dimensional model to obtain a target three-dimensional model, includes: lofting the cross-sectional contour along the lofting path to generate an initial three-dimensional model through a deep learning model, and adjusting the initial three-dimensional model to obtain a target three-dimensional model.

[0010] In an optional embodiment, the method further includes:

[0011] Constructing a neural network model, wherein the neural network model includes the first neural network and the second neural network:

[0012] Build an initial neural network model;

[0013] Obtain training samples and divide them into training set and validation set according to preset ratios;

[0014] Calculating the loss between the predicted value and the true value through forward propagation of the training set; calculating the gradient through a backpropagation algorithm based on the loss, and updating the parameter values of the parameters of the initial neural network model through a preset optimization algorithm;

[0015] During the training process, the model performance is evaluated by the loss corresponding to the validation set. When the loss corresponding to the validation set meets the preset conditions, the constructed neural network model is obtained.

[0016] In an optional embodiment, the loss is determined according to a mean square error loss function.

[0017] In an optional embodiment, the updating of the parameter values of the parameters of the initial neural network model by using a preset optimization algorithm includes:

[0018] The parameter values of the parameters of the initial neural network model are updated through an adaptive moment estimation optimization algorithm and a bias correction mechanism.

[0019] In an optional embodiment, obtaining training samples and dividing the training samples into a training set and a validation set according to a preset ratio includes:

[0020] A training sample is obtained, the training sample is preprocessed to obtain a preprocessed training sample, and the preprocessed training sample is divided into a training set and a validation set according to a preset ratio.

[0021] In an optional embodiment, the method further includes:

[0022] Displaying on a display interface the process of lofting the cross-sectional contour along the lofting path to generate an initial three-dimensional model, and displaying the target three-dimensional model;

[0023] When adjusting the loft path, section profile or target 3D model in the interactive editing interface, the target 3D model is regenerated.

[0024] In a second aspect, the present invention provides a three-dimensional modeling system based on path lofting, the three-dimensional modeling system based on path lofting comprising:

[0025] A first processing module is used to obtain path data and cross-sectional profile data;

[0026] A second processing module is used to input the path data and the cross-sectional profile data into a pre-built first neural network in sequence to perform feature extraction, thereby obtaining local curve features and local profile features;

[0027] A third processing module is used to sequentially input the local curve features and the local contour features into a pre-built second neural network for fitting, so as to obtain a lofting path and a cross-sectional contour;

[0028] The fourth processing module is configured to loft the cross-sectional contour along the lofting path to generate an initial three-dimensional model, and adjust the initial three-dimensional model to obtain a target three-dimensional model.

[0029] In a third aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to execute the three-dimensional modeling method based on path lofting of the above-mentioned first aspect or any corresponding embodiment thereof.

[0030] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein a single computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the three-dimensional modeling method based on path lofting of the above-mentioned first aspect or any corresponding embodiment thereof.

[0031] In a fifth aspect, the present invention provides a computer program product comprising computer instructions for enabling a computer to execute the three-dimensional modeling method based on path lofting according to the first aspect or any corresponding embodiment thereof.

[0032] The technical solution provided by the present invention has the following technical effects:

[0033] The technical solution of the embodiment of the present invention utilizes a first neural network to extract features from path data and cross-sectional profile data, enabling in-depth exploration of local curve features and local profile features in the data. Compared to traditional methods that can only provide relatively broad feature summaries, neural networks are able to learn the complex patterns and nuances of the data. For example, when designing industrial products with complex surfaces, it is possible to accurately capture features such as changes in surface curvature and irregular contour fluctuations, providing an accurate data foundation for subsequent modeling, thereby significantly improving the shape accuracy of the three-dimensional model.

[0034] The second neural network fits local features to generate lofting paths and cross-sectional profiles that better meet actual requirements. Traditional fitting methods are often based on fixed mathematical formulas and are difficult to adapt to complex and changing design requirements. However, neural networks have powerful nonlinear fitting capabilities and can adaptively adjust the fitting method based on different input features, reducing fitting errors and ensuring that the generated lofting paths and cross-sectional profiles are more aligned with the design intent, further improving model accuracy.

[0035] Two neural networks automatically complete the feature extraction and fitting process, eliminating the tedious manual work involved in traditional modeling. In traditional 3D modeling, designers must manually analyze data, extract features, and perform fitting, which is not only time-consuming and labor-intensive but also prone to human error. This solution, however, leverages the efficient computing power of neural networks to rapidly process input data, significantly reducing modeling preparation time and improving overall work efficiency.

[0036] After obtaining an accurate lofting path and cross-sectional profile, lofting operations along the path can quickly generate an initial 3D model. This path-based lofting method has clear calculation logic and efficient algorithm implementation, which can quickly build the basic framework of the model, saving a lot of time for subsequent fine-tuning.

[0037] Because neural networks can process a wide variety of complex path and cross-sectional profile data, this solution can adapt to diverse design needs. Whether it's a simple regular geometry or a complex free-form surface, modeling can be performed by inputting different path and profile data. For example, in architectural design, a variety of curves and profiles can be flexibly input to generate unique 3D building models based on different architectural styles and site conditions.

[0038] By adjusting the initial three-dimensional model generated based on precise feature extraction and fitting, a target three-dimensional model with higher smoothness can be obtained.

[0039] The present invention applies neural networks to 3D modeling of path lofting. Compared with traditional modeling methods based on geometric algorithms, neural networks have stronger learning ability and adaptability and can handle more complex 3D models.

[0040] Although path data and cross-sectional profile data in different fields have different characteristics, neural networks can achieve efficient and accurate modeling and better adaptability by learning the characteristics of these data. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in related technologies, the following briefly introduces the drawings required for use in the specific embodiments or related technical descriptions. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0042] Figure 1 1 is a flow chart of a three-dimensional modeling method based on path lofting according to an embodiment of the present invention;

[0043] Figure 2 is a schematic diagram of a lofting path according to an embodiment of the present invention;

[0044] Figure 3 is a schematic cross-sectional profile diagram of an embodiment of the present invention;

[0045] Figure 4 is a schematic diagram of an initial three-dimensional model of an embodiment of the present invention;

[0046] Figure 5 is a schematic diagram of a target three-dimensional model according to an embodiment of the present invention;

[0047] Figure 6 1 is a schematic structural diagram of a three-dimensional modeling system based on path lofting according to an embodiment of the present invention;

[0048] Figure 7 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0050] The technical solution of the present invention combines traditional path lofting technology with advanced AI deep learning algorithms for 3D modeling. By selecting a specific path and cross-sectional profile, and utilizing an optimized 3D modeling design platform plug-in and its deep learning model, it automatically generates complex and detailed surfaces or objects that change along the path. By adjusting parameters such as the shape mode, number of cross sections, and number of path segments, and combining AI's automatic optimization of shape, texture, and detail, users can efficiently and flexibly create highly realistic and creative 3D models. At the same time, it supports real-time preview and editing functions, further enhancing the user experience and modeling efficiency.

[0051] While traditional 3D modeling technologies support path lofting capabilities to a certain extent, they often lack sufficient flexibility and precision, especially when dealing with complex shapes and details. Furthermore, traditional modeling methods often rely on manual adjustments and control points, which is inefficient and makes it difficult to achieve precise and consistent shape changes. While some AI-based modeling tools exist, they often focus on overall shape generation or texture optimization, failing to fully incorporate the advantages of path lofting. Therefore, there is an urgent need for a 3D modeling method that can efficiently and accurately create complex shapes that change along a path, while incorporating AI deep learning algorithms for shape, texture, and detail optimization.

[0052] The purpose of the present invention is to solve the problems of low efficiency, limited functionality, lack of flexibility and precision in creating complex shapes that change along a path in related technologies based on a three-dimensional modeling method that combines path lofting with AI deep learning.

[0053] An embodiment of the present invention provides an embodiment of a three-dimensional modeling method based on path lofting. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer device such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0054] Figure 1 It is a flow chart of a three-dimensional modeling method based on path lofting according to an embodiment of the present invention.

[0055] like Figure 1 As shown, an embodiment of the present invention provides a three-dimensional modeling method based on path lofting, and the three-dimensional modeling method based on path lofting includes:

[0056] S101: Acquire path data and cross-section profile data.

[0057] In this embodiment, path data and cross-sectional profile data can be obtained by extracting data from the path lofting. The data is represented by a series of points, line segments, or cross sections, which form a continuous curve and surface in three-dimensional space. The coordinates of these points, line segments, or cross sections are used as input data for the neural network. For example, when the object to be modeled is a car part, the designer uses the software's drawing tools to accurately draw the car's lines and contours. These lines can be used as path data when constructing the three-dimensional model. For example, when designing the car's waistline or roof curve, the coordinates of the control points of these curves and the curve equation constitute the path data. At the same time, when designing automotive parts, the various two-dimensional graphics drawn by the designer, such as the cross-sectional view of the engine block and the outline of the car door, can be used as cross-sectional profile data after processing.

[0058] Digital Model Construction: When building a digital car model, engineers create a skeleton model. The lines within the skeleton model represent the main structural elements of the vehicle, and these lines serve as path data. For various internal components, such as transmission gears and seat frames, the cross-sectional data of the 3D models constructed in the design software is the cross-sectional profile data. By sectioning the digital model, cross-sectional profile data at different positions and angles can be obtained.

[0059] Virtual Reality / Augmented Reality Design Assistance: Some automotive design companies are beginning to adopt virtual reality / augmented reality technology to assist in design. Designers use gestures, controllers, and other devices to directly draw the exterior and interior structure of a car in a virtual environment. The system records the drawing trajectory and shape information, which can be converted into path data and cross-sectional profile data. For example, if a designer draws the overall outline of a car in virtual reality, the drawing trajectory is the path data, and the cross-sectional shape obtained by virtually cutting the virtual model is the cross-sectional profile data.

[0060] In this embodiment, the path data is represented by a series of points or line segments, which form a continuous curve in three-dimensional space. The path data includes the coordinates of the points or line segments, and a continuous curve can be formed based on the path data, such as Figure 2 shown.

[0061] In this embodiment, the cross-sectional data is represented by a series of shapes or contours that form cross sections on the path, such as Figure 3 The cross-sectional profile data includes the coordinates of the cross-sectional profile.

[0062] Path selection: Users can select one or more curves as paths in the 3D modeling environment, supporting multiple types such as straight lines, curves, polygons, etc.

[0063] Section Profile Selection: Users can select one or more faces or edges as section profiles. Both closed shapes and open line segments are supported as section profiles, and multiple section profiles can be combined. Path data and section profile data can be determined by user selection.

[0064] S102: Inputting the path data and the cross-sectional profile data into a pre-built first neural network in sequence to perform feature extraction, thereby obtaining local curve features and local profile features.

[0065] In this embodiment, the local curve features may include curvature, direction, and the like.

[0066] Path curvature variation: In path data, paths are not simple straight lines or regular curves; their curvature varies at different locations. For example, in the design of a car's waistline, the degree of curvature varies at different locations. The first neural network can accurately capture these curvature variations, such as whether the curve is gradual or abrupt. This is crucial for generating a natural and smooth lofting path, and determines the shape and direction of the 3D model along the path.

[0067] Local Direction: A path has a specific direction in 3D space. Local curve features contain information about the path's direction within a local area. For example, when constructing a complex building model, the path of the structure may have multiple direction variations. Local direction features captured by the neural network can clearly determine whether the path is horizontal, vertical, or inclined, as well as the direction of the change. This helps accurately generate a layout path that meets design requirements.

[0068] Local contour features may include: shape, size, corner details, symmetry, etc.

[0069] Edge details: The edges of a cross-sectional profile often contain rich details, such as sharp corners and smooth curves. For example, a cross-sectional view of an automobile engine block may have details such as slots and protrusions for mounting components. The first neural network can extract these edge details, allowing the generated cross-sectional profile to accurately represent these details and ensure the accuracy of the 3D model.

[0070] Local shape characteristics: The local shapes of cross-sectional profiles vary, and local profile features capture these characteristics. For example, the outline of a car door may have local concave or convex shapes. Extracting these shape characteristics allows the generated cross-sectional profile to better match the actual design and improve the shape accuracy of the 3D model.

[0071] S103: Inputting the local curve features and the local contour features into a pre-built second neural network in sequence for fitting, and obtaining a lofting path and a cross-sectional contour.

[0072] In this embodiment, the first neural network is a convolutional neural network (CNN), and the second neural network is a recurrent neural network (RNN). Both the local curve feature and the local contour feature are feature sequences.

[0073] In this embodiment, the first neural network (convolutional neural network) can be used to perform feature extraction on the path data to obtain local curve features, and to perform feature extraction on the cross-sectional profile data to obtain local profile features.

[0074] Then, the local curve features are input into the second neural network (recurrent neural network) to generate a smooth and continuous lofting path. The local contour features are input into the second neural network (recurrent neural network) to generate a smooth and continuous cross-sectional contour.

[0075] In this embodiment, the first neural network uses a convolutional neural network, which uses the convolution kernel of the convolution layer to slide convolution on the path data and cross-sectional profile data, which can accurately capture local features. For path data, key features such as path curvature changes and local direction can be extracted. For cross-sectional profile data, edge details of the contour and local shape characteristics can be obtained. This refined local feature extraction provides a rich and accurate information foundation for subsequent model generation, improving the neural network model's ability to model complex shapes.

[0076] In this embodiment, the second neural network utilizes a recurrent neural network, which possesses memory properties and can effectively process sequential data. When processing local curve features to generate a lofting path, the RNN incorporates information from previous moments to avoid sudden changes or unnatural bends in the path. When processing cross-sectional contours, it also ensures contour continuity, making the 3D model more geometrically reasonable and improving the model's geometric accuracy.

[0077] In this embodiment, the combination of CNN and RNN can better process different types of path and cross-sectional profile data. Whether it's simple geometric shape data or complex, irregular data, it can effectively extract features and generate reasonable lofting paths and cross-sectional profiles. Even if the data contains some noise or incompleteness, the neural network can learn and process it to generate a relatively reasonable model, enhancing adaptability and robustness to data changes.

[0078] In this embodiment, the parallel computing capabilities of CNNs enable rapid processing of large amounts of data, extracting local features and saving time for subsequent processing. When processing sequential data, RNNs can efficiently utilize historical information, reducing unnecessary computations and improving the efficiency of generating lofting paths and cross-sectional profiles. Working together, these two approaches enhance the efficiency of the entire 3D modeling process, reducing modeling time and improving work efficiency while ensuring model quality.

[0079] S104: Stake out the cross-section profile along the stakeout path to generate an initial three-dimensional model, and adjust the initial three-dimensional model to obtain a target three-dimensional model.

[0080] In this embodiment, the cross-section profile can be lofted along the lofting path according to preset parameters to generate an initial three-dimensional model. The preset parameters include shape mode, number of cross sections, number of path segments, etc. The initial three-dimensional model is as follows: Figure 4 As shown, the target three-dimensional model is as follows Figure 5 shown.

[0081] The shape mode refers to the geometric transformation rules followed when the cross-section changes along the path during the path lofting process, which determines how the shape of the three-dimensional model changes at different positions. In the linear gradient mode, the cross-section gradually transitions linearly from the starting point to the end point on the path. For example, when creating a cylinder from thick to thin, using the linear gradient mode, the size of the cross-section will shrink linearly along the path. The nonlinear gradient mode allows for more complex shape changes, such as gradients based on specific functions (such as sine functions and exponential functions), which can achieve unique bending and twisting effects. When designing art sculptures, the nonlinear gradient mode can be used to produce irregular but orderly changes on the surface of the model. In addition, there is a symmetry mode that makes the model symmetrical on both sides with the path as the axis of symmetry. It is often used to create symmetrical objects such as bottles and dumbbells.

[0082] Number of Sections: This refers to the number of cross-sectional profiles placed along a path during path lofting. Increasing the number of sections can make a 3D model more detailed and accurately represent the object's shape. The number of sections can be determined based on the complexity of the model. For simple geometric models, a smaller number of sections may be sufficient, improving modeling efficiency.

[0083] Number of Path Segments: The number of path segments is the number of line segments that a path is divided into, reflecting the degree of subdivision of the path. A higher number of path segments means that the path is divided more finely, providing more anchor points for cross sections during path lofting, making the model's changes in the path direction more precise. When creating a curved pipe model, increasing the number of path segments can make the pipe's curves smoother and more natural. The number of path segments also affects the model's topology and data volume. Increasing the number of path segments will make the model's topology more complex, and the data volume will increase accordingly. Therefore, the number of path segments can also be determined based on the complexity of the 3D model or object.

[0084] In an optional embodiment, S104 comprises: lofting the cross-sectional contour along the lofting path to generate an initial three-dimensional model, and adjusting the initial three-dimensional model to obtain a target three-dimensional model, specifically comprising: lofting the cross-sectional contour along the lofting path through a deep learning model to generate an initial three-dimensional model, and adjusting the initial three-dimensional model to obtain a target three-dimensional model.

[0085] In this embodiment, the AI deep learning model is constructed based on a deep learning algorithm of a convolutional neural network, which is trained based on a large amount of three-dimensional model data and can automatically optimize the shape, texture and details of the deep learning model. The model can automatically generate a three-dimensional model with high realism and creativity based on the lofting path and cross-sectional profile selected by the user, as well as preset parameters. The deep learning model is used to optimize the shape, texture, details, etc. of the initial three-dimensional model to obtain the final target three-dimensional model. The training process of the deep learning model in this embodiment can refer to the training process of the neural network described below and will not be repeated here.

[0086] When obtaining a training set for a deep learning model, a large amount of diverse 3D model data is collected from multiple sources, such as models from different fields such as automobiles, buildings, and industrial products. The training set can be obtained from open source model libraries, model files exported from professional design software, 3D scan data of actual products, etc. Ensure that the data covers a variety of shapes, textures, and detailed features to improve the model's generalization ability. In this embodiment, the specific training process for the deep learning model can be carried out using conventional methods in the field.

[0087] This invention combines path lofting technology with advanced AI deep learning models to meet user needs for personalized 3D model generation. Users can automatically generate surfaces or objects that follow a specified path, based on optimized lofting paths and cross-sectional profiles, and set preset parameters such as shape mode, number of cross sections, and number of path segments. This includes shape adjustment, texture generation, and detail enhancement, significantly improving the model's sophistication and realism.

[0088] In an optional embodiment, the three-dimensional modeling method based on path lofting further includes:

[0089] Construct a neural network model, which includes a first neural network and a second neural network:

[0090] Build an initial neural network model.

[0091] Obtain training samples and divide them into training set and validation set according to the preset ratio.

[0092] The loss between the predicted and true values is calculated by forward propagation of the training set. Based on this loss, the gradient is calculated using the backpropagation algorithm, and the parameters of the initial neural network model are updated using a pre-defined optimization algorithm. The pre-defined optimization algorithm is an adaptive moment estimation optimization algorithm and a bias correction mechanism.

[0093] During training, the model performance is evaluated using the validation set loss. A successful neural network model is obtained when the validation set loss meets a pre-set condition. This condition can be that the loss does not decrease for a pre-set number of consecutive rounds. This number can be set and modified based on actual needs, for example, 10 rounds.

[0094] In this embodiment, as an example, obtaining a training sample specifically includes:

[0095] Data is collected from data sources to construct training samples. Data sources include: automobile design companies and automobile manufacturing plants.

[0096] Automotive design companies collect a large amount of 3D model data for various car brands and models. These models accurately record the details of each body part during the design phase, including the control points of the waistline. For example, they collect 3D model data for multiple car brands and models, covering different types such as sedans, SUVs, and sports cars.

[0097] Automotive manufacturing plants: Obtain scan data of actual car bodies on the production line. Laser scanning equipment scans the car bodies in all directions during production, generating 3D point cloud data containing waistline information. This data reflects the specific shape and position of the waistline in actual production, complementing the 3D model data.

[0098] Three-dimensional model data can be collected from automobile design companies, and scan data can be collected from automobile manufacturing plants, and training samples can be obtained based on the three-dimensional model data and scan data.

[0099] Specifically, the collected 3D model data can be preliminarily sorted out:

[0100] The collected 3D model data in different formats are uniformly converted, for example, into a common 3D model format to facilitate subsequent processing.

[0101] The 3D model data is stored in a classified manner and labeled according to information such as car brand, model, design version, etc. for easy management and retrieval.

[0102] Obtaining scan data from an automotive manufacturing plant:

[0103] Scanning equipment deployment and commissioning: Install laser scanning equipment, such as 3D laser scanners, on the production lines of automotive manufacturing plants. Ensure the accuracy and stability of the scanning equipment, and commission and calibrate the equipment to accurately capture 3D point cloud data of the vehicle body.

[0104] Data Collection Process: During vehicle production, perform a full-scale scan of the vehicle body. Ensure that the scanning angle covers all areas of the vehicle body, particularly the waistline. Record the scan data, including the acquisition time, vehicle model, and production batch, to facilitate subsequent data tracing and analysis.

[0105] Data preprocessing: De-noise the collected 3D point cloud data to remove noise points generated during the scanning process and improve data quality. Perform point cloud data registration and splicing, fusing point cloud data scanned from different angles to obtain a complete vehicle body point cloud model.

[0106] Data fusion and feature extraction:

[0107] 1) Data Fusion

[0108] Aligning 3D model data with scanned data: Using a fixed coordinate system for the vehicle as a reference, align the 3D model data obtained from the automotive design company with the scanned data from the automotive manufacturing plant. Accurate data alignment can be achieved through methods such as feature matching and iterative closest point algorithms.

[0109] Data fusion processing: The aligned 3D model data and scan data are fused to generate comprehensive data containing both design information and actual production information. For example, the control point information in the 3D model data can be associated with the point cloud information in the scan data to form a more complete vehicle body data model.

[0110] 2) Feature extraction:

[0111] Waistline feature extraction: Extract waistline feature information from the fused comprehensive data, including waistline position, curvature, length, etc. Edge detection algorithms, curve fitting algorithms, and other technologies can be used to extract waistline features.

[0112] The control point information of the waistline is further processed, such as calculating the distance and angle between the control points, as the feature vector of the training sample.

[0113] Extraction of other relevant features: In addition to waistline features, other relevant features of the vehicle body can also be extracted, such as body length, width, height, door position, etc. These features can be used as auxiliary information to improve the accuracy of the training model.

[0114] Training sample construction:

[0115] 1) Sample annotation:

[0116] Define labeling rules: Define labeling rules for training samples based on actual needs. For example, waistline classification can be labeled based on waistline shape, style, etc. The overall characteristics of the vehicle body can be labeled based on model, brand, etc.

[0117] Manual or automatic labeling: For some complex features, manual labeling can be used to ensure labeling accuracy. For some simple features, automatic labeling algorithms can be developed to improve labeling efficiency.

[0118] 2) Sample Storage: Store training samples in a format suitable for machine learning models, such as CSV, HDF5, etc. When storing training samples, both their feature vectors and annotation information should be saved to facilitate subsequent model training and evaluation.

[0119] In an optional embodiment, the loss is determined according to a mean square error loss function.

[0120] In an optional embodiment, the technical solution for constructing a neural network model, wherein the parameter values of the parameters of the initial neural network model are updated by a preset optimization algorithm, specifically includes:

[0121] The parameter values of the parameters of the initial neural network model are updated through an adaptive moment estimation optimization algorithm and a bias correction mechanism.

[0122] In an optional embodiment, the technical solution for building a neural network model includes obtaining training samples and dividing the training samples into a training set and a validation set according to a preset ratio, specifically including:

[0123] Obtain training samples, preprocess the training samples to obtain preprocessed training samples, and divide the preprocessed training samples into a training set and a validation set according to a preset ratio. The preset ratio can be set and modified as needed, for example, 8:2.

[0124] Data preprocessing:

[0125] Extract waistline-related data: Extract waistline-related points from 3D model data and scanned data manually or using simple graphics processing tools. For example, based on annotation information in a car design, select the series of 3D points that constitute the waistline.

[0126] Formatting: Organize the extracted waistline points into an array that meets the requirements, ensuring that each sample has a shape of [N, 3]. Arrange these 3D points in a certain order for subsequent processing.

[0127] Clean up anomalies: Carefully examine the waistline point set and remove any points that significantly deviate from the normal waistline trend. For example, there may be isolated points in the scan data caused by equipment errors. These anomalies should be removed.

[0128] Data augmentation: Perform simple transformations on the waistline point set. For example, slightly translate some waistline point sets to simulate the changes in waistlines under different assembly positions. Rotate some waistline point sets by small angles to increase data diversity.

[0129] Training process:

[0130] Model construction:

[0131] CNN layer: Builds a simple convolutional neural network. It consists of two convolutional layers and two pooling layers. The first convolutional layer uses a 3x3 convolution kernel to extract local waistline features, followed by downsampling through a 2x2 max pooling layer. The second convolutional layer also uses a 3x3 convolution kernel to further extract features, and then downsamples through a pooling layer.

[0132] The RNN layer passes the features output by the CNN layer to the Long Short Term Memory (LSTM) layer. Two LSTM layers, each containing 64 hidden units, are set up to capture the continuity and overall trend of the waistline. Finally, a fully connected layer outputs the predicted waistline point coordinates.

[0133] Training steps:

[0134] Prepare data: Divide the preprocessed waistline data into a training set and a validation set, generally in a ratio of 8:2.

[0135] Setting parameters: Select the mean square error loss function, use the Adam optimizer, and set the initial learning rate to 0.001.

[0136] Training begins: The training set is fed into the neural network model in batches of 32 samples. The neural network model performs forward propagation, calculating the loss between the predicted values and the true values, and then updates the model parameters through backpropagation.

[0137] Monitor and evaluate: During training, regularly evaluate the model's performance using the validation set. If the loss on the validation set does not decrease for 10 consecutive epochs, reduce the learning rate or stop training.

[0138] Save model: After training, save the neural network model with the best performance for subsequent waistline search tasks.

[0139] Loss function: The mean square error (MSE) between the lofted paths and cross-section profiles generated by the neural network and the actual paths is used as the loss function to calculate the loss and evaluate the model performance.

[0140] The loss function is defined as the mean squared error of the difference between the actual path and the predicted path. For each sample point, the square of the difference between the predicted coordinates and the actual coordinates is calculated, and then the squared differences are averaged over all sample points.

[0141] The loss function formula is as follows:

[0142] MSE=(1 / N)*Σ[(x_pred_i-x_true_i) 2 +(y_pred_i-y_true_i) 2 +(z_pred_i-z_true_i) 2 ].

[0143] Where MSE represents the loss, N is the number of sample points, x_pred_i, y_pred_i, z_pred_i are the x, y, z coordinates of the i-th point of the predicted path, and x_true_i, y_true_i, z_true_i are the x, y, z coordinates of the i-th point of the actual path.

[0144] Optimization algorithm:

[0145] Adam optimization algorithm is used to optimize the parameters of the neural network model. t =β1·m t-1 +(1-β1)·g t and Calculate the first-order moment mt and the second-order moment vt, and correct them to accurately estimate the learning rate, according to Update parameters and adaptively adjust the learning rate to accelerate convergence and improve performance. The specific methods are as follows:

[0146] The Adam optimization algorithm is used to optimize the parameters of the neural network model.

[0147] Calculation of first-order and second-order moments:

[0148] First-order moment m t =β1·m t-1 +(1-β1)·g t .

[0149] Where β1 is the attenuation coefficient of the first-order moment, which is set to 0.9. t-1 is the first-order moment of the previous time step, g t is the gradient at the current time step.

[0150] Second moment:

[0151] Where β2 is the attenuation coefficient of the second-order moment, which is set to 0.999. t-1 is the second moment of the previous time step, is the square of the gradient at the current time step.

[0152] Bias correction:

[0153] Since mt and vt are both set to 0 during initialization, their values will tend to be 0 in the early stages of training, leading to inaccurate learning rate estimation. To eliminate this bias, the Adam algorithm introduces a bias correction mechanism. The bias correction formula is as follows:

[0154] Bias correction of first-order moments

[0155] Bias correction of second-order moments

[0156] Among them, vt are the corrected first-order moment and second-order moment, and By correcting the deviation, mt and vt can more accurately reflect the changes in gradients in the early stages of training, thereby improving the accuracy of learning rate estimation.

[0157] Parameter update:

[0158] formula:

[0159] Among them, θ t is the parameter value at the current time step, θ t-1 is the parameter value at the previous time step, η is the learning rate, are the modified first-order and second-order moments, respectively, and ∈ is a small constant (such as 1e-8) to prevent the denominator from being zero. This formula indicates that the parameter update is calculated based on the ratio of the modified first-order and second-order moments, multiplied by the learning rate η. In this way, the Adam algorithm can adaptively adjust the learning rate of each parameter, thereby accelerating convergence and improving model performance.

[0160] As another example, the training steps may include:

[0161] Data preparation:

[0162] Collect a large amount of 3D model path lofting data, including original and manually optimized paths, section deformations, and examples of different section alignments.

[0163] Perform data preprocessing, such as normalization and denoising, to ensure data consistency and stability.

[0164] Normalization: A data preprocessing technique called normalization converts data of different dimensions or ranges to the same scale. This helps eliminate the effects of data dimension and improves model stability and accuracy. This function uses the Min-Max normalization method, which calculates the maximum and minimum values of the model data and then linearly maps each data point to the range [0, 1]. This method is suitable for scenarios where the data range is known and needs to be normalized to a fixed range. The calculation formula is:

[0165] Xnorm=X-Xmin / Xmax-Xmin.

[0166] Among them, X is the original data, Xmin is the minimum value of the data, Xmax is the maximum value of the data, and Xn orm is the normalized data.

[0167] Denoising: Denoising is another important step in data preprocessing, aiming to remove noise and outliers from the data to improve data quality and model performance. This function uses linear interpolation to fill missing values, thereby reducing noise. Interpolation estimates the value of unknown data points using known data points. This method uses a straight line connecting two known data points to estimate the value of any point between them. This method is simple and intuitive and works well for most smoothly varying data.

[0168] Principle: Linear interpolation assumes that the change between two known data points is linear, that is, the rate of change (slope) is constant. The slope between the known data points is calculated and the equation of the line is used to estimate the value of the unknown data point.

[0169] The data is divided into training, validation, and test sets to evaluate the performance of the model and prevent overfitting.

[0170] Model construction:

[0171] Build a model based on the selected neural network structure.

[0172] Set the parameters of the input layer, hidden layer, and output layer.

[0173] Define the loss function and optimization algorithm.

[0174] Model training:

[0175] Use the training set data to train the model.

[0176] During the training process, the performance of the model is evaluated through a loss function, and the parameters of the model are updated using an optimization algorithm.

[0177] Regularly use validation data to evaluate the generalization ability of the model and adjust training parameters to avoid overfitting.

[0178] Model evaluation and optimization:

[0179] Use the test set data to evaluate the performance of the model.

[0180] Compare the differences between the paths generated by the model and the actual paths.

[0181] Fine-tune the model based on the evaluation results, such as adjusting the network structure, increasing the training set, etc.

[0182] Training parameter settings:

[0183] The following is an example of training parameter settings combined with the 3D model path lofting requirements:

[0184] Neural network structure:

[0185] Input layer: accepts serialized data of model data, such as coordinate sequences of points, line segments, and contour surfaces.

[0186] Example: To represent a simple semicircular curve (centered at the origin and with a radius of 5), in a two-dimensional plane, we can discretize this curve by taking points at a certain angle interval (here we take an angle interval of 10 degrees as an example). The serialized coordinate sequence is similar to the following:

[0187] [(5.0, 0.0), (4.829629131445341, 1.710100716628348), (4.330127018922193, 2.5), ...].

[0188] Hidden layer: A two-layer LSTM network is used, with each layer containing 128 neurons.

[0189] Output layer: Output optimized model data, such as coordinate sequences of points, line segments, and contour surfaces.

[0190] Optimizer: Select Adam optimizer.

[0191] Initial learning rate: 0.001.

[0192] Learning rate decay: After every 10 epochs of training, the learning rate decays to 0.5 times the original value.

[0193] Loss function:

[0194] Main loss function: MSE.

[0195] Additional loss term: regularization term for path curvature change, with weight 0.1.

[0196] Batch size and number of iterations:

[0197] Batch size: 32.

[0198] Number of iterations: 100 epochs.

[0199] In an optional embodiment, the three-dimensional modeling method based on path lofting further includes:

[0200] The display interface shows the process of generating an initial 3D model by lofting the cross-section profile along the lofting path, and displays the target 3D model, which is a 3D object.

[0201] When adjusting the loft path, section profile or target 3D model in the interactive editing interface, the target 3D model is regenerated.

[0202] The display interface and interactive editing interface allow real-time preview and editing. When adjusting the parameters of the lofting path and cross-section profile, users can immediately see the effect of the regenerated object based on the adjusted parameters. At the same time, it supports editing functions such as moving, rotating, and scaling the generated objects, and AI-assisted editing suggestions help users adjust and optimize models more efficiently.

[0203] The generated target 3D model can be exported to common 3D file formats (such as OBJ, STL, etc.), and supports importing lofting paths and cross-section profiles from external files. The display interface and interactive editing interface are the interfaces of the 3D modeling design platform.

[0204] The beneficial effects of the present invention include:

[0205] Combining path lofting technology and AI deep learning algorithms, it achieves efficient and accurate generation of complex shapes.

[0206] It supports a variety of cross-section profiles and path type combinations, as well as real-time preview and editing functions, providing greater flexibility.

[0207] The main innovative points of the technical solution of the present invention are:

[0208] 1. Intelligent path and section optimization:

[0209] Automatic path optimization: AI deep learning algorithms can automatically identify the curvature and continuity of the path, smooth the path, reduce redundancy and mutation points in the path, and thus improve the accuracy and smoothness of the lofted object.

[0210] Intelligent Section Deformation: Based on the curvature and direction of the path, AI algorithms automatically adjust the shape and size of the section, ensuring that the lofted object maintains a consistent appearance and detail along the path. This intelligent deformation technology significantly reduces manual adjustment workload and improves modeling efficiency.

[0211] 2. Real-time preview and interactive editing:

[0212] Real-time Preview: Path lofting and AI deep learning-based 3D modeling methods enable real-time preview, meaning users can instantly see the effects of lofted objects as they adjust parameters. This real-time feedback mechanism helps users quickly iterate and optimize designs.

[0213] Interactive Editing: Provides a user-friendly interactive editing interface that allows users to modify and adjust paths, sections, and lofted objects in real time. At the same time, AI algorithms can automatically adjust relevant parameters based on user editing operations to ensure the continuity and consistency of lofted objects.

[0214] 3. High-precision modeling and detail optimization:

[0215] High-precision modeling: AI deep learning algorithms can handle more complex paths and cross-sectional shapes, achieving high-precision modeling. This helps meet application scenarios with high model accuracy requirements, such as medical image analysis and product design.

[0216] Detail Optimization: AI algorithms are used to optimize the details of lofted objects, such as edge smoothness and chamfers. This optimization can enhance the visual effect and realism of the model, making it more in line with actual application needs.

[0217] 4. Scalability and customization:

[0218] Scalability: The AI deep learning-based path lofting and 3D modeling methods offer excellent scalability and seamless integration with other 3D modeling tools and features. This helps users select the appropriate tools and features based on their specific needs, achieving a more efficient and flexible modeling process.

[0219] It should be noted that the contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

[0220] This embodiment also provides a three-dimensional modeling system based on path lofting. A single system is used to implement the above-mentioned embodiments and optional implementations. Details already described are not repeated here. As used below, the term "module" may refer to a combination of software and / or hardware that implements a predetermined function. Although the systems described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0221] Figure 6 It is a structural diagram of a three-dimensional modeling system based on path lofting according to an embodiment of the present invention.

[0222] The present invention provides a three-dimensional modeling system based on path lofting, such as Figure 6 As shown, the 3D modeling system based on path lofting includes:

[0223] The first processing module 11 is used to obtain path data and cross-sectional profile data.

[0224] The second processing module 12 is used to input the path data and the cross-sectional profile data into a pre-built first neural network in sequence to perform feature extraction, thereby obtaining local curve features and local profile features.

[0225] The third processing module 13 is used to sequentially input the local curve features and the local contour features into a pre-built second neural network for fitting, so as to obtain a lofting path and a cross-sectional contour.

[0226] The fourth processing module 14 is configured to loft the cross-sectional contour along the lofting path to generate an initial three-dimensional model, and adjust the initial three-dimensional model to obtain a target three-dimensional model.

[0227] In an optional embodiment, the fourth processing module 14 is specifically configured to loft the cross-sectional contour along the lofting path through a deep learning model to generate an initial three-dimensional model, and adjust the initial three-dimensional model to obtain a target three-dimensional model.

[0228] In an optional embodiment, the path-based 3D modeling system further includes: a modeling module for constructing a neural network model, wherein the neural network model includes a first neural network and a second neural network:

[0229] The modeling module includes:

[0230] The first processing unit is used to construct an initial neural network model.

[0231] The second processing unit is used to obtain training samples and divide the training samples into a training set and a validation set according to a preset ratio.

[0232] The third processing unit is configured to calculate the loss between the predicted value and the true value by forward propagation of the training set, calculate the gradient based on the loss by backpropagation algorithm, and update the parameter values of the parameters of the initial neural network model by using a preset optimization algorithm.

[0233] The fourth processing unit is used to evaluate the model performance through the loss corresponding to the validation set during the training process, and obtain the constructed neural network model when the loss corresponding to the validation set meets the preset conditions.

[0234] In an optional embodiment, the loss is determined according to a mean square error loss function.

[0235] In an optional embodiment, the third processing unit in the modeling module is specifically used to update the parameter values of the parameters of the initial neural network model through an adaptive moment estimation optimization algorithm and a deviation correction mechanism.

[0236] In an optional embodiment, the second processing unit in the modeling module is specifically used to obtain training samples, preprocess the training samples to obtain preprocessed training samples, and divide the preprocessed training samples into a training set and a validation set according to a preset ratio.

[0237] In an optional embodiment, the path-based 3D modeling system further includes:

[0238] The display interface is used to display the process of generating an initial three-dimensional model by lofting the cross-section profile along the lofting path, and to display the target three-dimensional model.

[0239] Interactive editing interface for regenerating the target 3D model when adjustments are made to the loft path, section profile, or target 3D model.

[0240] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0241] The path lofting-based three-dimensional modeling system in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0242] The present invention also provides a computer device. Figure 7 , Figure 7 Schematic diagram of the hardware structure of the computer device according to the embodiment of the present invention. Figure 7 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In an optional embodiment, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Equally, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor device). Figure 7 A processor 10 is taken as an example.

[0243] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0244] The memory 20 stores instructions that can be executed by at least one processor 10, so as to enable at least one processor 10 to execute the method shown in the above embodiment.

[0245] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating device, an application required for at least one function. The data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In an optional embodiment, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0246] The memory 20 may include volatile memory, such as random access memory. The memory may also include non-volatile memory, such as flash memory, a hard disk, or a solid-state drive. The memory 20 may also include a combination of the above types of memory.

[0247] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0248] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or downloaded through a network and originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc. Further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0249] A portion of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0250] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A three-dimensional modeling method based on path lofting, characterized in that: include: Obtain path data and cross-section profile data; Inputting the path data and the cross-sectional profile data into a pre-built first neural network in sequence for feature extraction to obtain local curve features and local profile features; Inputting the local curve features and the local contour features into a pre-built second neural network in sequence for fitting, thereby obtaining a lofting path and a cross-sectional contour; The cross-sectional profile is lofted along the lofting path to generate an initial three-dimensional model, and the initial three-dimensional model is adjusted to obtain a target three-dimensional model.

2. The method according to claim 1, characterized in that The step of lofting the cross-sectional contour along the lofting path to generate an initial three-dimensional model and adjusting the initial three-dimensional model to obtain a target three-dimensional model comprises: lofting the cross-sectional contour along the lofting path to generate an initial three-dimensional model through a deep learning model, and adjusting the initial three-dimensional model to obtain a target three-dimensional model.

3. The method according to claim 1, characterized in that The method further comprises: Constructing a neural network model, wherein the neural network model includes the first neural network and the second neural network: Build an initial neural network model; Obtain training samples and divide them into training set and validation set according to preset ratios; Calculating the loss between the predicted value and the true value through forward propagation of the training set; calculating the gradient through a backpropagation algorithm based on the loss, and updating the parameter values of the parameters of the initial neural network model through a preset optimization algorithm; During the training process, the model performance is evaluated by the loss corresponding to the validation set. When the loss corresponding to the validation set meets the preset conditions, the constructed neural network model is obtained.

4. The method according to claim 3, characterized in that The loss is determined according to a mean square error loss function.

5. The method according to claim 3, characterized in that The parameter values of the parameters of the initial neural network model are updated by a preset optimization algorithm, including: The parameter values of the parameters of the initial neural network model are updated through an adaptive moment estimation optimization algorithm and a bias correction mechanism.

6. The method according to claim 3, characterized in that The obtaining of training samples and dividing the training samples into a training set and a validation set according to a preset ratio includes: A training sample is obtained, the training sample is preprocessed to obtain a preprocessed training sample, and the preprocessed training sample is divided into a training set and a validation set according to a preset ratio.

7. The method according to claim 1, characterized in that The method further comprises: Displaying on a display interface the process of lofting the cross-sectional contour along the lofting path to generate an initial three-dimensional model, and displaying the target three-dimensional model; When adjusting the loft path, section profile or target 3D model in the interactive editing interface, the target 3D model is regenerated.

8. A three-dimensional modeling system based on path lofting, characterized in that: include: A first processing module is used to obtain path data and cross-sectional profile data; A second processing module is used to input the path data and the cross-sectional profile data into a pre-built first neural network in sequence to perform feature extraction, thereby obtaining local curve features and local profile features; A third processing module is used to sequentially input the local curve features and the local contour features into a pre-built second neural network for fitting, so as to obtain a lofting path and a cross-sectional contour; The fourth processing module is configured to loft the cross-sectional contour along the lofting path to generate an initial three-dimensional model, and adjust the initial three-dimensional model to obtain a target three-dimensional model.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the three-dimensional modeling method based on path lofting according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the three-dimensional modeling method based on path lofting according to any one of claims 1 to 7.

Images