A design method and system based on deep learning

Abstract

The application relates to the computer technical field, in particular to a design method and system based on deep learning. The method comprises the following steps: sequentially performing feature extraction on static three-dimensional body shape data and dynamic riding posture data to obtain static three-dimensional body shape features and dynamic riding posture features; performing feature fusion on the static three-dimensional body shape features and the dynamic riding posture features according to a preset proportion to obtain fusion features; inputting the fusion features into a preset deep learning model to obtain personalized customized bicycle parameters of a current user; determining a three-dimensional model of the personalized customized bicycle based on the personalized customized bicycle parameters; and when the three-dimensional model of the personalized customized bicycle does not satisfy personalized constraint conditions of the bicycle, re-executing the step of obtaining the static three-dimensional body shape data and the dynamic riding posture data of the current user. Through the above configuration mode, the application can accurately improve the generation quality of a user personalized design scheme.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a design method and system based on deep learning. Background Technology

[0002] Traditional bicycle design and manufacturing typically employs a standardized, mass-production model, making it difficult to meet the increasingly diverse and personalized needs of consumers. While some customized bicycle services exist in the market, they mostly rely on human experience and simple rule matching, resulting in low design efficiency and accuracy, failing to fully explore users' potential needs and generate high-quality personalized design solutions. With the development of deep learning technology, which has achieved significant results in areas such as image generation and pattern recognition, new ideas and methods have been provided for personalized bicycle design.

[0003] Based on this, the present invention proposes a design method and system based on deep learning to solve the above-mentioned technical problems. Summary of the Invention

[0004] This invention describes a deep learning-based design method and system that can accurately improve the quality of user-personalized design solutions.

[0005] According to a first aspect, the present invention provides a deep learning-based design method, the method comprising:

[0006] Obtain the current user's static 3D body shape data and dynamic cycling posture data;

[0007] The static three-dimensional body shape data and the dynamic cycling posture data are sequentially subjected to feature extraction to obtain static three-dimensional body shape features and dynamic cycling posture features.

[0008] The static three-dimensional body shape features and the dynamic riding posture features are fused according to a preset ratio to obtain fused features;

[0009] The fused features are input into a preset deep learning model to obtain the personalized bicycle parameters for the current user.

[0010] Based on the parameters of the personalized bicycle, a three-dimensional model of the personalized bicycle is determined;

[0011] If the 3D model of the personalized bicycle does not meet the personalized constraints of the bicycle, the step "obtain the current user's static 3D body shape data and dynamic riding posture data" is executed again.

[0012] According to a second aspect, the present invention provides a deep learning-based design system, comprising:

[0013] The acquisition unit is configured to acquire the current user's static 3D body shape data and dynamic cycling posture data;

[0014] The first data processing unit is configured to extract features from the static three-dimensional body shape data and the dynamic riding posture data in sequence to obtain static three-dimensional body shape features and dynamic riding posture features.

[0015] The second data processing unit is configured to fuse the static three-dimensional body shape features and the dynamic riding posture features according to a preset ratio to obtain fused features;

[0016] The third data processing unit is configured to input the fused features into a preset deep learning model to obtain the personalized bicycle parameters of the current user.

[0017] The fourth data processing unit is configured to determine a three-dimensional model of the personalized bicycle based on the personalized bicycle parameters.

[0018] The fifth data processing unit is configured to re-execute the step "obtaining the current user's static three-dimensional body shape data and dynamic riding posture data" when the three-dimensional model of the personalized bicycle does not meet the personalized constraints of the bicycle.

[0019] Thirdly, embodiments of this specification also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the method described in any embodiment of this specification.

[0020] Fourthly, embodiments of this specification also provide a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the methods described in any embodiment of this specification.

[0021] According to the deep learning-based design method and system provided by this invention, static 3D body shape data and dynamic riding posture data of the current user are collected in real time through a multi-source sensor array and 3D scanning technology. Feature extraction algorithms from the field of deep learning are used to process the static 3D body shape data and dynamic riding posture data respectively. Specifically, for static 3D body shape data, a fully connected neural network architecture is used to extract static 3D body shape features through multi-layer nonlinear transformation; for dynamic riding posture data, a hybrid model of Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN) is used to effectively capture time-series information and local features, thereby obtaining dynamic riding posture features. Based on preset weight parameters, a weighted fusion operation is performed on the static 3D body shape features and dynamic riding posture features to obtain a fused feature vector containing ergonomics and motion characteristics. This fused feature vector is used as input and fed into a preset deep learning regression model built on a Multilayer Perceptron (MLP). Through the forward propagation operation of the model, the personalized customized bicycle parameters of the current user are output, which cover key design elements such as geometric dimensions and component specifications. Using computer-aided design (CAD) software, a 3D digital model of a personalized bicycle is constructed based on its parameters. After construction, the 3D digital model is comprehensively evaluated based on a pre-defined set of personalized constraints. If the evaluation shows that the 3D digital model fails to meet any of the aforementioned personalized constraints, an iterative process of data acquisition and model construction is triggered, repeating the step of "collecting the current user's static 3D body shape data and dynamic riding posture data through a multi-source sensor array and 3D scanning technology" until a 3D digital model that meets all personalized constraints is generated. Through this closed-loop design process, this invention can significantly improve the quality and accuracy of user-personalized design schemes, ensuring that the customized bicycle achieves an optimal balance in terms of ergonomics, functional performance, and manufacturing feasibility. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart illustrating a deep learning-based design method according to one embodiment is shown;

[0024] Figure 2 A schematic block diagram of a deep learning-based design system according to one embodiment is shown. Detailed Implementation

[0025] The solution provided by the present invention will now be described with reference to the accompanying drawings.

[0026] Figure 1 This diagram illustrates a flowchart of a deep learning-based design method according to one embodiment. It is understood that this method can be executed by any system, device, platform, or cluster of devices with computing and processing capabilities. Figure 1 As shown, the method includes:

[0027] Step 100: Obtain the current user's static 3D body shape data and dynamic cycling posture data;

[0028] Step 102: Extract features from the static 3D body shape data and the dynamic riding posture data in sequence to obtain static 3D body shape features and dynamic riding posture features;

[0029] Step 104: Fuse the static 3D body shape features and dynamic riding posture features according to a preset ratio to obtain fused features;

[0030] Step 106: Input the fused features into the preset deep learning model to obtain the personalized bicycle parameters for the current user;

[0031] Step 108: Based on the parameters of the personalized bicycle, determine the three-dimensional model of the personalized bicycle;

[0032] Step 110: When the 3D model of the personalized bicycle does not meet the personalized constraints of the bicycle, repeat the step "obtain the current user's static 3D body shape data and dynamic riding posture data".

[0033] In this embodiment, a multi-source sensor array and 3D scanning technology are used to collect the current user's static 3D body shape data and dynamic riding posture data in real time. Feature extraction algorithms from the field of deep learning are employed to process the static 3D body shape data and dynamic riding posture data respectively. Specifically, for the static 3D body shape data, a fully connected neural network architecture is used to extract static 3D body shape features through multi-layer nonlinear transformations; for the dynamic riding posture data, a hybrid model of Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN) is used to effectively capture time-series information and local features, thereby obtaining dynamic riding posture features. Based on preset weight parameters, a weighted fusion operation is performed on the static 3D body shape features and dynamic riding posture features to obtain a fused feature vector containing ergonomics and motion characteristics. This fused feature vector is used as input and fed into a preset deep learning regression model built on a Multilayer Perceptron (MLP). Through the forward propagation operation of the model, the personalized bicycle parameters of the current user are output. These parameters cover key design elements such as geometric dimensions and component specifications. Using computer-aided design (CAD) software, a 3D digital model of the personalized bicycle is constructed based on the personalized bicycle parameters. After construction, the 3D digital model is comprehensively evaluated based on a pre-defined set of personalized constraints. If the evaluation results show that the 3D digital model fails to meet any of the aforementioned personalized constraints, the iterative process of data acquisition and model building is triggered, and the step of "acquiring the current user's static 3D body shape data and dynamic riding posture data through a multi-source sensor array and 3D scanning technology" is repeated until a 3D digital model that meets all personalized constraints is generated. Through the above closed-loop design process, this invention can significantly improve the generation quality and design accuracy of user-personalized design schemes, ensuring that customized bicycles achieve an optimal balance in terms of human adaptability, functional performance, and manufacturing feasibility.

[0034] In one embodiment of the present invention, the personalized constraints include frame stiffness constraints and joint angle constraints.

[0035] In one embodiment of the present invention, the joint angle constraint is constructed using the following formula:

[0036]

[0037] In the formula, For knee joint angle, This represents the lower limit of the knee joint angle. This represents the upper limit of the knee joint angle. For users with long legs, The height of the seat tube. For the thickness of the foot pedal, For the thickness of the shoe sole, The crank length, The angle of the crank relative to the horizontal direction. The rate of change of the crank angle. Let be the acceleration due to the crank angle, and t be the time. This is a fatigue effect function related to crank angle and time.

[0038] In this embodiment, by constructing biomechanically based joint angle constraints, the generation quality and design accuracy of personalized user design schemes can be effectively improved. This constraint, based on ergonomic principles, comprehensively considers the range of motion and movement trajectory of key joints such as the knee, hip, and shoulder joints during cycling. By establishing precise mathematical models and parameterized constraints, joint angles are limited to a comfortable and efficient range. In the specific implementation process, firstly, in-depth analysis is performed on the user's static three-dimensional body shape data and dynamic cycling posture data to extract key features related to joint angles. Then, these features are integrated into the personalized design process as important constraints to optimize and adjust design parameters such as the bicycle's geometric dimensions and component layout. In this way, it is ensured that the customized bicycle, in terms of human adaptability, minimizes the risk of fatigue and injury during cycling; in terms of functional performance, it ensures optimal handling and power transmission efficiency; and in terms of manufacturing feasibility, it balances the manufacturability and cost-effectiveness of the design scheme, thereby achieving an optimal balance in terms of human adaptability, functional performance, and manufacturing feasibility.

[0039] In one embodiment of the present invention, the frame stiffness constraint is constructed using the following formula:

[0040]

[0041] In the formula, For the torsional stiffness of the frame, For integration variables, The torsional moment varies with time. The angle of twist changes over time. The torsional damping coefficient is... The rate of change of the torsion angle, To the torsional moment of inertia, The acceleration due to the torsional angle, For direction coefficients, This is the minimum torsional stiffness requirement.

[0042] In this embodiment, by constructing a multi-dimensional frame stiffness constraint system, the generation quality and design accuracy of personalized user design schemes can be significantly improved. This system, based on the principles of materials mechanics and structural dynamics, establishes precise mathematical constraints and parameterized threshold standards at three levels: static stiffness, dynamic stiffness, and fatigue stiffness. In the specific implementation process, the user's personalized needs, such as riding habits, weight distribution, and riding road conditions, are first quantitatively analyzed. Combined with the bicycle's usage scenarios (such as road racing, mountain biking, and urban commuting), the required stiffness indicators for the frame under different working conditions are determined. During the design process, finite element analysis (FEA) technology is used to perform mechanical simulation of the frame, verifying in real time whether the frame stiffness meets the constraint requirements. By continuously adjusting the frame's geometric dimensions, material distribution, and structural details, its lightweight design is optimized while ensuring sufficient load-bearing capacity and deformation resistance. In this way, in terms of ergonomics, the frame provides stable and comfortable support for the user during riding, reducing riding fatigue caused by frame deformation; in terms of functionality, the bicycle's handling response speed and power transmission efficiency are improved to meet the needs of different riding scenarios; and in terms of manufacturing feasibility, the design scheme is ensured to meet the requirements of the production process, reducing manufacturing costs and production difficulty, ultimately achieving the optimal balance of ergonomics, functionality, and manufacturing feasibility in customized bicycles.

[0043] In one embodiment of the present invention, static three-dimensional body shape data includes height, weight, leg length, and ischial tuberosity distance; dynamic cycling posture data includes cadence, center of gravity distribution, and joint range of motion; personalized bicycle parameters include the dimensions of bicycle components.

[0044] In this embodiment, the static three-dimensional body shape data acquisition system covers key anthropometric parameters such as height, weight, leg length, and ischial tuberosity distance, achieving millimeter-level precision data acquisition through high-precision three-dimensional scanning equipment. Dynamic cycling posture data includes biomechanical parameters such as cadence, center of gravity distribution, and joint range of motion, monitored in real-time using multi-axis inertial sensors and motion capture systems. The generated personalized bicycle parameters focus on the precise dimensional specifications of bicycle components, including but not limited to frame geometry (such as head tube length and top tube tilt), handlebar width, seat tube height, and crank length, providing a quantitative design basis for achieving high-precision personalized bicycle customization.

[0045] In one embodiment of the present invention, after determining the three-dimensional model of the personalized bicycle based on the personalized bicycle parameters, the method further includes:

[0046] Obtain the user's historical heart rate data during cycling;

[0047] Based on the user's historical heart rate data during cycling, determine the user's historical heart rate fluctuation curve during cycling.

[0048] Based on the heart rate fluctuation curve, determine the seat cushion comfort adjustment parameters;

[0049] Based on the seat comfort adjustment parameters, the seat comfort in the 3D model of the personalized custom bicycle is adjusted.

[0050] In this embodiment, after constructing a 3D model based on personalized bicycle parameters, the design optimization process is further expanded. Specifically, wearable devices and built-in bicycle sensors collect the user's historical heart rate data, which covers real-time heart rate changes throughout the ride. Subsequently, time series analysis is used to process the collected heart rate data. Through preprocessing steps such as data smoothing and outlier removal, a heart rate fluctuation curve is constructed, visually presenting the dynamic trend of heart rate changes over time and riding intensity. Based on this heart rate fluctuation curve, combined with fatigue assessment models and comfort measurement standards in the biomedical field, characteristic parameters such as heart rate fluctuation amplitude and the proportion of high-frequency fluctuations are analyzed in depth. Through regression analysis or machine learning algorithms, a mapping relationship between heart rate fluctuation characteristics and saddle comfort is established, thereby determining saddle comfort adjustment parameters, including but not limited to key indicators such as saddle shape curvature parameters, filling hardness coefficient, and surface material friction coefficient. Finally, the determined seat comfort adjustment parameters are fed back into the 3D model design stage of the personalized custom bicycle. Using the parametric modeling function of computer-aided design (CAD) software, the seat geometry, material properties, etc. in the 3D model are adjusted and optimized in a targeted manner to ensure that the custom bicycle provides users with a riding experience that is more in line with physiological characteristics and comfort needs during actual riding, thus realizing a complete closed loop from parameter design to user experience optimization.

[0051] In one embodiment of the present invention, after determining the three-dimensional model of the personalized bicycle based on the personalized bicycle parameters, the method further includes:

[0052] Obtain the user's historical cycling route data;

[0053] Based on the user's historical cycling route data, determine the friction adjustment coefficient of the wheels;

[0054] Based on the friction adjustment coefficient of the wheels, the tread pattern of the tires in the 3D model of a personalized custom bicycle is adjusted.

[0055] In this embodiment, after constructing the 3D model based on the personalized bicycle parameters, the design optimization process is further refined. Specifically, the user's historical riding route data is obtained through the riding recording device and the Geographic Information System (GIS) data interface. This data includes multi-dimensional information such as route terrain features (slope, roughness), road surface materials (asphalt, cement, gravel, etc.), and environmental temperature and humidity. Subsequently, based on the historical route data, a road surface friction characteristic analysis model is used, combined with empirical formulas for friction coefficients under different road surface materials and environmental conditions, to construct a route-friction coefficient mapping relationship. Through data mining and machine learning algorithms, the correlation between historical route characteristics and the actual friction requirements of the wheels is analyzed to determine the wheel friction adjustment coefficient suitable for the user's riding habits and commonly encountered road conditions. This coefficient comprehensively reflects the tire's required grip, wear resistance, and other performance indicators. Finally, the determined wheel friction adjustment coefficient is fed back to the 3D model design stage. With the help of computer-aided design (CAD) and simulation analysis software, the tire tread pattern in the personalized bicycle 3D model is parametrically adjusted. The tread depth, shape (block, longitudinal, mixed patterns) and arrangement density are optimized based on the friction adjustment coefficient to ensure that the tire can provide sufficient grip to ensure riding safety under the user's usual riding conditions, while also taking into account rolling resistance and service life, so as to achieve a precise match between the performance of the customized bicycle and the user's actual riding needs.

[0056] The foregoing has described specific embodiments of the invention. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0057] According to another embodiment, the present invention provides a design system based on deep learning. Figure 2 A schematic block diagram of a deep learning-based design system according to one embodiment is shown. It will be understood that this system can be implemented using any system, device, platform, or cluster of devices with computing and processing capabilities. Figure 2 As shown, the system includes: an acquisition unit 200, a first data processing unit 202, a second data processing unit 204, a third data processing unit 206, a fourth data processing unit 208, and a fifth data processing unit 210. The main functions of each component are as follows:

[0058] The acquisition unit is configured to acquire the current user's static 3D body shape data and dynamic cycling posture data;

[0059] The first data processing unit is configured to extract features from the static three-dimensional body shape data and the dynamic riding posture data in sequence to obtain static three-dimensional body shape features and dynamic riding posture features.

[0060] The second data processing unit is configured to fuse the static three-dimensional body shape features and the dynamic riding posture features according to a preset ratio to obtain fused features;

[0061] The third data processing unit is configured to input the fused features into a preset deep learning model to obtain the personalized bicycle parameters of the current user.

[0062] The fourth data processing unit is configured to determine a three-dimensional model of the personalized bicycle based on the personalized bicycle parameters.

[0063] The fifth data processing unit is configured to re-execute the step "obtaining the current user's static three-dimensional body shape data and dynamic riding posture data" when the three-dimensional model of the personalized bicycle does not meet the personalized constraints of the bicycle.

[0064] In one embodiment of the present invention, the personalized constraints include frame stiffness constraints and joint angle constraints.

[0065] In one embodiment of the present invention, the joint angle constraint is constructed by the following formula:

[0066]

[0067] In the formula, For knee joint angle, This represents the lower limit of the knee joint angle. This represents the upper limit of the knee joint angle. For users with long legs, The height of the seat tube. For the thickness of the foot pedal, For the thickness of the shoe sole, The crank length, The angle of the crank relative to the horizontal direction. The rate of change of the crank angle. Let be the acceleration due to the crank angle, and t be the time. This is a fatigue effect function related to crank angle and time.

[0068] In one embodiment of the present invention, the frame stiffness constraint is constructed by the following formula:

[0069]

[0070] In the formula, For the torsional stiffness of the frame, For integration variables, The torsional moment varies with time. The angle of twist changes over time. The torsional damping coefficient is... The rate of change of the torsion angle, To the torsional moment of inertia, The acceleration due to the torsional angle, For direction coefficients, This is the minimum torsional stiffness requirement.

[0071] In one embodiment of the present invention, the static three-dimensional body shape data includes height, weight, leg length, and ischial tuberosity distance; the dynamic cycling posture data includes cadence, center of gravity distribution, and joint range of motion; and the personalized bicycle parameters include the dimensions of the bicycle components.

[0072] In one embodiment of the present invention, after determining the three-dimensional model of the personalized bicycle based on the personalized bicycle parameters, the method further includes:

[0073] Obtain the user's historical heart rate data during cycling;

[0074] Based on the user's historical heart rate data, determine the user's historical heart rate fluctuation curve during cycling.

[0075] Based on the heart rate fluctuation curve, determine the seat cushion comfort adjustment parameters;

[0076] Based on the seat comfort adjustment parameters, the seat comfort in the three-dimensional model of the personalized bicycle is adjusted.

[0077] In one embodiment of the present invention, after determining the three-dimensional model of the personalized bicycle based on the personalized bicycle parameters, the method further includes:

[0078] Obtain the user's historical cycling route data;

[0079] Based on the user's historical cycling route data, the friction adjustment coefficient of the wheels is determined;

[0080] Based on the friction adjustment coefficient of the wheel, the tread pattern of the tire in the three-dimensional model of the personalized bicycle is adjusted.

[0081] According to another embodiment, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed in a computer, causes the computer to perform a combination Figure 1 The method described.

[0082] According to another embodiment, an electronic device is also provided, including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, it implements a combination... Figure 1 The method described.

[0083] The various embodiments in this invention 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 system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0084] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in this invention can be implemented using hardware, software, firmware, or any combination thereof. When implemented in software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium.

[0085] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. A design method based on deep learning, characterized in that, The method includes: Obtain the current user's static 3D body shape data and dynamic cycling posture data; The static three-dimensional body shape data and the dynamic cycling posture data are sequentially subjected to feature extraction to obtain static three-dimensional body shape features and dynamic cycling posture features. The static three-dimensional body shape features and the dynamic riding posture features are fused according to a preset ratio to obtain fused features; The fused features are input into a preset deep learning model to obtain the personalized bicycle parameters for the current user. Based on the parameters of the personalized bicycle, a three-dimensional model of the personalized bicycle is determined; If the 3D model of the personalized bicycle does not meet the personalized constraints of the bicycle, the step "obtain the current user's static 3D body shape data and dynamic riding posture data" is executed again. The personalized constraints include frame stiffness constraints and joint angle constraints; The joint angle constraint is constructed using the following formula: ， In the formula, For knee joint angle, This represents the lower limit of the knee joint angle. This represents the upper limit of the knee joint angle. For users with long legs, The height of the seat tube. For the thickness of the foot pedal, For the thickness of the shoe sole, The crank length, The angle of the crank relative to the horizontal direction. The rate of change of the crank angle. Let be the acceleration due to the crank angle, and t be the time. The fatigue effect function is related to crank angle and time; The frame stiffness constraint is constructed using the following formula: ， In the formula, For the torsional stiffness of the frame, For integration variables, The torsional moment varies with time. The angle of twist changes over time. The torsional damping coefficient is... The rate of change of the torsion angle, To the torsional moment of inertia, The acceleration due to the torsional angle, For direction coefficients, Minimum torsional stiffness requirement; The static three-dimensional body data includes height, weight, leg length, and ischial tuberosity distance; the dynamic cycling posture data includes cadence, center of gravity distribution, and joint range of motion; the personalized bicycle parameters include the dimensions of the bicycle components.

2. The method according to claim 1, characterized in that, After determining the 3D model of the personalized bicycle based on the personalized bicycle parameters, the process further includes: Obtain the user's historical heart rate data during cycling; Based on the user's historical heart rate data, determine the user's historical heart rate fluctuation curve during cycling. Based on the heart rate fluctuation curve, determine the seat cushion comfort adjustment parameters; Based on the seat comfort adjustment parameters, the seat comfort in the three-dimensional model of the personalized bicycle is adjusted.

3. The method according to claim 1, characterized in that, After determining the 3D model of the personalized bicycle based on the personalized bicycle parameters, the process further includes: Obtain the user's historical cycling route data; Based on the user's historical cycling route data, the friction adjustment coefficient of the wheels is determined; Based on the friction adjustment coefficient of the wheel, the tread pattern of the tire in the three-dimensional model of the personalized bicycle is adjusted.

4. A design system based on deep learning, characterized in that, include: The acquisition unit is configured to acquire the current user's static 3D body shape data and dynamic cycling posture data; The first data processing unit is configured to extract features from the static three-dimensional body shape data and the dynamic riding posture data in sequence to obtain static three-dimensional body shape features and dynamic riding posture features. The second data processing unit is configured to fuse the static three-dimensional body shape features and the dynamic riding posture features according to a preset ratio to obtain fused features; The third data processing unit is configured to input the fused features into a preset deep learning model to obtain the personalized bicycle parameters of the current user. The fourth data processing unit is configured to determine a three-dimensional model of the personalized bicycle based on the personalized bicycle parameters. The fifth data processing unit is configured to re-execute the step "obtaining the current user's static three-dimensional body shape data and dynamic riding posture data" when the three-dimensional model of the personalized bicycle does not meet the personalized constraints of the bicycle. The personalized constraints include frame stiffness constraints and joint angle constraints; The joint angle constraint is constructed using the following formula: ， In the formula, For knee joint angle, This represents the lower limit of the knee joint angle. This represents the upper limit of the knee joint angle. For users with long legs, The height of the seat tube. For the thickness of the foot pedal, For the thickness of the shoe sole, The crank length, The angle of the crank relative to the horizontal direction. The rate of change of the crank angle. Let be the acceleration due to the crank angle, and t be the time. The fatigue effect function is related to crank angle and time; The frame stiffness constraint is constructed using the following formula: ， In the formula, For the torsional stiffness of the frame, For integration variables, The torsional moment varies with time. The angle of twist changes over time. The torsional damping coefficient is... The rate of change of the torsion angle, To the torsional moment of inertia, The acceleration due to the torsional angle, For direction coefficients, Minimum torsional stiffness requirement; The static three-dimensional body data includes height, weight, leg length, and ischial tuberosity distance; the dynamic cycling posture data includes cadence, center of gravity distribution, and joint range of motion; the personalized bicycle parameters include the dimensions of the bicycle components.

5. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed in a computer, causes the computer to perform the method described in any one of claims 1-3.

Images