An intelligent screening method for automotive styling based on a measurement aesthetic entropy evaluation model

Abstract

The present invention discloses an intelligent screening method for automobile styling based on a measurement aesthetic entropy evaluation model, which relates to the technical field of automobile design. The method includes: collecting automobile styling image data, performing preprocessing operations such as grayscale processing, background clipping, and normalization processing on the images, and constructing a line drawing atlas for calculating the beauty degree value of automobile styling; calculating the beauty degree value of automobile styling based on the measurement aesthetic entropy evaluation model; using a pre-trained ResNet50 convolutional neural network model for training, taking the mean square error MSE as the loss function, and adopting an Adam optimizer for parameter update until the loss function converges; finally, intelligently screening the automobile styling scheme according to the trained model. The present invention realizes the objective and quantitative evaluation of automobile styling schemes, improves the efficiency and accuracy of automobile design, provides new design tools and methods for automobile designers, and helps to promote the innovation and development of intelligent automobile design.

Description

Technical Field

[0001] The present invention relates to the technical field of automotive design, and particularly to an intelligent screening method for automotive styling based on a metrological aesthetic entropy evaluation model. Background Art

[0002] With the rapid development of the automotive industry, automotive design has gradually shifted from traditional computer-aided design to intelligent design. The traditional automotive design process usually relies on the experience and intuition of designers, lacking scientific and quantitative evaluation criteria, resulting in low design efficiency and strong subjectivity. In recent years, the rapid development of artificial intelligence technology has brought new opportunities to automotive design. In particular, the excellent performance of convolutional neural networks (CNNs) in image recognition and classification has made it an important tool for metrological aesthetic convolutional neural networks.

[0003] However, most existing intelligent aesthetic evaluation methods for automotive styling are based on simple visual feature extraction and classification, lacking an objective measurement of the beauty degree of automotive styling. Aesthetic measurement theory and information entropy theory provide a theoretical basis for quantitatively evaluating automotive styling. Aesthetic measurement theory describes beauty through mathematical models, while information entropy theory is used to evaluate the orderliness and complexity of information. Combining these two theories can more comprehensively evaluate the aesthetic value of automotive styling. Summary of the Invention

[0004] The purpose of the present invention is to propose an intelligent screening method for automotive styling based on a metrological aesthetic entropy evaluation model. By combining aesthetic measurement theory, information entropy theory, and complexity theory, a metrological aesthetic entropy evaluation model is constructed to achieve scientific quantification and objective evaluation of the beauty degree of automotive styling.

[0005] To achieve the above purpose, the present invention proposes an intelligent screening method for automotive styling based on a metrological aesthetic entropy evaluation model, and the specific steps are as follows:

[0006] S1. Collect and preprocess automotive styling image data, and construct an automotive styling image dataset for metrological aesthetic convolutional neural networks and a line drawing atlas for calculating the beauty degree value of automotive styling;

[0007] S2. Construct a metrological aesthetic entropy evaluation model based on the Birkhoff aesthetic measurement theory model, entropy evaluation theory model, and complexity theory model, and calculate the beauty degree value of automotive styling;

[0008] S3. Construct a metrological aesthetic convolutional neural network model based on the metrological aesthetic entropy evaluation;

[0009] S4. Verify the metrological aesthetic convolutional neural network model.

[0010] Preferably, in S1, the preprocessing operation includes grayscale processing, background clipping batch processing, normalization processing, automotive styling structure line extraction processing, and noise reduction processing.

[0011] Preferably, in S2, the specific calculation steps of the automotive styling beauty value are as follows:

[0012] S21. Measure the entropy value redundancy of the automotive three-dimensional styling line based on the entropy evaluation model, and calculate the automotive styling aesthetic order;

[0013] S22. Calculate the complexity of the automotive styling aesthetics based on the complexity theory model;

[0014] S23. Based on the measurement aesthetic entropy evaluation model, comprehensively calculate the automotive styling beauty value by combining the automotive styling aesthetic order and the automotive styling aesthetic complexity.

[0015] Preferably, in S21, the calculation formula of the automotive styling aesthetic order is as follows:

[0016] ;

[0017] Among them, is the automotive styling aesthetic order, is the morphological order of each view, is the weight value of each view, and h is the four views of the automotive styling;

[0018] The morphological order of each view, and its calculation formula is as follows:

[0019] ;

[0020] Among them, is the information entropy of the structure line , is the weight value of the styling structure line, is the automotive styling curve function, is the th point on the structure line, is the ratio of the derivative value of the point on the structure line to the sum of the derivative values of all points of the th differential on the structure line, , which can satisfy ; is the number of points with different absolute values of derivatives on the curve, is an integer, is an integer, is an integer.

[0021] Preferably, the weight value of the styling structure line assigns weights to the automotive styling structure line using a Likert nine-point scale, and its calculation formula is as follows:

[0022] ;

[0023] Among them, is the average score of each modeling structure line, is the total number of modeling structure lines in each view.

[0024] Preferably, in S22, the formula of the complexity theory model is as follows:

[0025] ;

[0026] Among them, is the aesthetic complexity of the car styling, is the simplest order of the car styling, is the dissipation of the information energy of the car styling, is the slow growth rate.

[0027] Preferably, in S23, the formula of the measurement aesthetic entropy evaluation model is as follows:

[0028] ;

[0029] Among them, is the beauty value of the car styling.

[0030] Preferably, in S3, the measurement aesthetic convolutional neural network model uses the pre-trained ResNet50 network as the basic model and uses the mean square error MSE as the loss function.

[0031] Preferably, the calculation formula of the loss function is as follows:

[0032] ;

[0033] Among them, is the total number of samples, , is the true label data of the beauty value of the th sample, , is the predicted data of the beauty value of the th sample.

[0034] Preferably, the measurement aesthetic convolutional neural network model uses the Adam optimizer to update the model parameters, and the model training parameters include the number of iterations epoch, the batch size batch_size and the learning rate.

[0035] Therefore, the present invention proposes an intelligent screening method for car styling based on the measurement aesthetic entropy evaluation model, and its beneficial effects are as follows:

[0036] (1) By combining the aesthetic measurement theory, information entropy theory, and complexity theory, the present invention constructs a measurement aesthetic entropy evaluation model, realizes the scientific quantification and objective evaluation of the beauty degree of automobile styling, and improves the scientificity and reliability of automobile design.

[0037] (2) The present invention uses a convolutional neural network (CNN) to intelligently screen automobile styling schemes, greatly improving the screening efficiency of automobile styling schemes, accelerating the design cycle, and reducing the design cost.

[0038] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic flow chart of an intelligent screening method for automobile styling based on a measurement aesthetic entropy evaluation model of the present invention;

[0040] Figure 2 It is a schematic diagram of the effect of the line drawing processing process of an intelligent screening method for automobile styling based on a measurement aesthetic entropy evaluation model of the present invention;

[0041] Figure 3 It is a distribution diagram of the corresponding positions of automobile styling evaluation indexes of an intelligent screening method for automobile styling based on a measurement aesthetic entropy evaluation model of the present invention;

[0042] Figure 4 It is the ResNet50 network structure of an intelligent screening method for automobile styling based on a measurement aesthetic entropy evaluation model of the present invention;

[0043] Figure 5 It is a schematic diagram of a questionnaire for evaluating the beauty degree of an automobile styling scheme of an intelligent screening method for automobile styling based on a measurement aesthetic entropy evaluation model of the present invention;

[0044] Figure 6 It is a schematic diagram of the pseudo code for training the ResNet50 model in an intelligent screening method for automobile styling based on a measurement aesthetic entropy evaluation model of the present invention;

[0045] Figure 7 It is a schematic diagram of the pseudo code of the measurement aesthetic convolutional neural network model in an intelligent screening method for automobile styling based on a measurement aesthetic entropy evaluation model of the present invention;

[0046] Figure 8 It is a convergence broken line graph of the loss function in an intelligent screening method for automobile styling based on a measurement aesthetic entropy evaluation model of the present invention;

[0047] Figure 9 It is a schematic diagram of the preprocessed picture of the validation data set in the embodiment of an intelligent screening method for automobile styling based on a measurement aesthetic entropy evaluation model of the present invention. Detailed implementation manners

[0048] To make the technical solutions, advantages and objectives of the present invention clearer, the technical solutions of the embodiments of the present invention will be described clearly and completely below. The described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the described embodiments of the present invention without creative efforts fall within the protection scope of this application.

[0049] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meaning understood by those with ordinary skills in the field to which the present invention pertains.

[0050] As Figure 1 shown, the present invention provides an intelligent screening method for automobile styling based on a measurement aesthetic entropy evaluation model. The specific steps are as follows:

[0051] S1. Collection and preprocessing of automobile styling image data

[0052] For the collection of automobile styling samples, a crawler program was written based on the Python language to obtain automobile styling samples from the network platform of Dongchedi (www.dongchedi.com), and image data (front view, rear view, side view, top view) of different brand sedans, SUVs and sports cars produced after 2003 were selected.

[0053] For the preprocessing of the sample set, the present invention calculates the beauty degree of automobile styling through a measurement aesthetic entropy evaluation model. The content does not involve color beauty degree. Therefore, the sample set is preprocessed with grayscale, and a standardized view is constructed to eliminate the interference of redundant factors and construct a standardized data set for the training and optimization of the convolutional neural network. The standardized data set after grayscale preprocessing is further processed to construct a line drawing atlas for calculating the beauty degree value of automobile styling. The specific process is as follows:

[0054] First, perform grayscale preprocessing. Import the sample set images into the background clipping website (ydtai.com / remove-background) in batches according to the automobile model for background clipping; import the clipped images into the Photoshop software for batch grayscale processing, and set the bit depth value to 8; perform unified standardized graying processing on the image background (R: 138, G: 138, B: 138), and uniformly cover the license plate areas in the front and rear views with gray rectangles (R: 202, G: 202, B: 202).

[0055] Then, construct a standardized view, place the front, rear, side, and top views of the car in a 1181px × 1378px, 100dpi-sized image. The front and rear views are at the top of the image, the side and top views have the front of the car facing left, the side view is in the middle, and the top view is at the bottom. The image sizes of the four views of the car's shape are unified based on the vehicle height, and the margins of the figures are uniformly set as follows: the top, bottom, left, and right margins are 39px, and the spacing between each view is 39px.

[0056] Finally, further process the standardized dataset after grayscale preprocessing, construct a line drawing atlas for calculating the aesthetic value of the car's shape, implement the Sobel operator program through MATLAB (an image processing tool for edge detection that can identify the boundaries and contours in the car's shape image), batch extract the shape structure lines of the front, rear, side, and top views of the car's shape, import the line drawing atlases of the four views into Photoshop software, smooth and connect the structure curves of the car's shape, and perform denoising processing on the processed line drawings according to the list of car shape structure lines proposed by Li Yanlong et al. as shown in Table 1; the effect during the line drawing processing (taking NIO ET5T as an example) is as Figure 2 shown, and map the car shape structure lines to the corresponding views, as Figure 3 shown. The processed line drawing atlas is used to calculate the aesthetic value of the car's shape, and the aesthetic value is used as the label for the convolutional neural network dataset.

[0057] Table 1 List of car shape structure lines

[0058]

[0059] S2. Construct a measurement aesthetic entropy evaluation model based on the Birkhoff aesthetic measurement theory model, entropy evaluation theory model, and complexity theory model to calculate the aesthetic value of the car's shape.

[0060] First, calculate the entropy redundancy of the car shape structure lines. Import the car shape line drawing into Rhino8 software, and use the curve tool to construct the car shape structure lines proportionally (due to the high symmetry of the car shape, only half of the shape structure lines of the front view, rear view, and top view of the car shape are selected based on the central symmetry axis), and use the derivative calculation method based on the parametric representation of the curve (Grasshopper plug-in) to calculate the derivative of the j th structure line. The specific method steps are as follows: Select and import the j th structure line into the Curve component; use the Divide Distance component to differentially divide the structure line at equal intervals to obtain discrete points; use the Derivatives component to calculate the derivatives of each point on the structure line; use the Deconstruct Vector component to split the derivative vector into x and yComponent (two-dimensional curve), calculate the derivatives of each node, use the Grasshopper plug-in to calculate the derivatives of each discrete point on the automotive styling structure line, and after importing the derivative data into the Excel software, use the data analysis tool in the software according to the formula to calculate the redundancy of the information entropy of the j root structure line of the automotive styling. The data analysis tool is based on the formula:

[0061] .

[0062] Secondly, calculate the weight value of the styling structure line , and sum up the aesthetic order of the automotive styling.

[0063] As shown in Table 2, use the Likert nine-point scale combined with expert evaluation (the expert team mainly consists of professors and associate professors from the School of Design Arts, including 5 automotive design experts, 8 industrial design experts, 4 ergonomics experts, 3 aesthetic theory experts, and 5 vehicle engineering experts) to carry out the weight assignment of the automotive styling structure line. First, collect expert evaluation data; then, calculate the average score of each styling structure line; finally, normalize the importance scores of each styling structure line. The normalization formula is:

[0064] ;

[0065] Among them, is the weight value of the styling structure line of each view, is the average score of each styling structure line, and n is the total number of styling structure lines of each view.

[0066] Table 2 Likert nine-point scale for index importance scoring

[0067]

[0068] Combined with the formula:

[0069] ,

[0070] Use the Excel software to sum the weighted redundancy of the information entropy of each structure line to calculate the aesthetic order of each view of the automotive styling.

[0071] Then calculate the aesthetic complexity of the automotive styling. The calculation formula is as follows:

[0072] ;

[0073] is the dimension of the vector space of the automotive styling structure line. Based on the principle of the simplest order of complexity, calculate the order of the automotive appearance form from the front view, rear view, side view, and top Figure 4 views. Therefore ; Since the present invention only involves the beauty measurement of the automobile appearance and does not involve the iteration and improvement of the automobile appearance, therefore , .

[0074] Finally, calculate the beauty value of the automobile shape according to the measurement aesthetic entropy evaluation model, and its formula is as follows:

[0075] ;

[0076] Among them, is the importance weight of each view. The Likert scale is adopted and weighted in combination with expert evaluation.

[0077] S3. Construct a measurement aesthetic convolutional neural network model based on the measurement aesthetic entropy evaluation and conduct training.

[0078] 1. Data standardization processing. Before the start of training, performing data standardization processing on the input data is a necessary step to ensure the effective convergence of the model. For the automobile beauty evaluation task, in order to avoid the influence of the scale difference of the input values on the training effect of the convolutional neural network, the following normalization processing formula is adopted:

[0079] ;

[0080] This method aims to ensure the balanced input of data and prevent the adverse impact of small numerical differences on the model performance. By scaling each input value within the range of the square of its maximum value, we can reduce the influence of the data scale on the training stability of the model.

[0081] 2. Initial setting of the model. As Figure 4 shown, the present invention uses the pre-trained ResNet50 network as the basic model. During the model training process, the mean square error (MSE) is used as the loss function to measure the difference between the predicted value and the true value, where the predicted value is the output of the model and the true value is the calculated value of the measurement aesthetic entropy evaluation model; the Adam optimizer is selected to update the model parameters, the epoch is set to 300, the batch_size is set to 32, and the learning rate is set to 0.0001. The detailed configuration of the model hyperparameters is shown in Table 3.

[0082] Table 3 Model hyperparameter configuration table

[0083]

[0084] 3. Model training evaluation and update. Input data into the model for forward propagation and backward propagation. Forward propagation calculates the output predicted value of the model, and the backward propagation algorithm calculates the gradient of the model parameters through the loss function until the loss function converges; the optimizer adjusts and updates the model parameters according to the direction and magnitude of the gradient.

[0085] 4. Model Saving and Deployment. When the training process reaches the expected convergence effect, save the trained model weights and structure for subsequent automated calculation and application of the beauty degree evaluation task.

[0086] S4. Verification of the Quantitative Aesthetic Convolutional Neural Network Model.

[0087] Use the undergraduate classroom assignments of automotive styling design taught by the author as verification data. Through high-quality screening, select 9 hand-drawn automotive styling design schemes as basic samples, and combine AIGC technology to assist in generating intelligent automotive styling design schemes, so as to construct a verification dataset for the quantitative aesthetic entropy evaluation model. Specific process: Apply the Stable Diffusion algorithm, import the automotive styling Lora model selected on the liblib (www.liblib.art) website into the Stable Diffusion model library, select the Chilloutmix-Ni-pruned-fp16-fix model in combination with its webui interface, adjust the prompt words and related parameters, randomly generate a preliminary automotive styling scheme, run controlnet to upload the view of the hand-drawn automotive styling design scheme, and control the form to further generate a relatively optimized scheme. Select the ideal scheme and automatically iterate and refine it under the 4x-ultrasharp algorithm of the high-resolution repair plug-in to complete the scheme design.

[0088] Use the verification dataset to verify the quantitative aesthetic convolutional neural network model and organize the verification data. Specific process: Run the intelligent screening program for automotive styling schemes through PyTorch, save the verification dataset to the image folder to be evaluated, use the cd command in the terminal window to switch to the directory where the quantitative aesthetic convolutional neural network model is located and run it, and enter the python test.py command in the terminal window to run the program and output the results.

[0089] Conduct a questionnaire survey on the beauty degree evaluation of automotive styling schemes using the verification dataset. The specific process is as follows:

[0090] The first step is questionnaire design; as Figure 5 shown, the information designed in the questionnaire survey mainly includes: the gender and occupation of the subjects (including: students, teachers), the automotive styling scheme number and picture, and the beauty degree level (Likert nine-level scale beauty degree: extremely ugly, very ugly, ugly, a bit ugly, neutral, a bit beautiful, beautiful, very beautiful, extremely beautiful).

[0091] The second step is to clarify the subjects; determine 150 teachers and students in the School of Design Art as the subjects.

[0092] In the third step, distribute a questionnaire on the beauty evaluation of car styling solutions to the subjects as a reward, and require each subject to evaluate the beauty of each styling solution based on their true feelings.

[0093] In the fourth step, screen out invalid questionnaires (including missing data, uniform answers, and those not filled out as required), organize the survey data, and analyze the data reliability using Cronbach's Alpha coefficient. 0 indicates non - credible; 1 indicates completely credible; less than 0.6 indicates poor reliability; greater than 0.7 indicates good reliability. The classic Cronbach's Alpha formula is:

[0094] ;

[0095] Among them, represents the Cronbach's Alpha value, represents the number of solutions in the questionnaire, is the variance of the th solution in the questionnaire, is the variance of the total scores of all solutions for each respondent.

[0096] In the fifth step, calculate the beauty evaluation scores for each car styling design solution and sort them from high to low according to these scores.

[0097] Finally, compare and analyze the verification results of the measurement aesthetic convolutional neural network model with the results of the questionnaire on the beauty evaluation of car styling solutions to verify the accuracy of the intelligent screening of the measurement aesthetic convolutional neural network model. Specific method: Use the Spearman rank - correlation coefficient test to verify the results (- 1 indicates a completely negative correlation, 0 indicates no correlation, and 1 indicates a completely positive correlation). The formula for the Spearman rank - correlation coefficient is:

[0098] ;

[0099] Among them: is the Spearman rank - correlation coefficient, is the rank difference of each pair of observed values (that is, for each calculate the difference in its ranks in the two variables), is the total number of observed values.

[0100] Example 1

[0101] 1. Collection and pre - processing of car styling samples

[0102] According to the method of collecting and preprocessing automotive styling samples in S1, a total of 10,504 styling pictures of 2,626 automobiles were obtained. All vehicle models were randomly divided into a training set, a test set, and a validation set at a ratio of 8:1:1: among them, the training set included 2,100 vehicle models, the test set included 263 vehicle models, and the validation set included 263 vehicle models, and there was no intersection between the divided data sets. The automotive styling pictures in the data set were integrated by vehicle model unit, and gray-scale preprocessing and normalization processing were carried out to construct a standardized data set (2,626 pictures) for convolutional neural network training and optimization; the standardized data set after gray-scale preprocessing was further processed to construct a line drawing data set (2,626 groups) for calculating the beauty degree value of automotive styling.

[0103] 2. Calculation of the beauty degree of automotive styling

[0104] Combined with the view weights, the weights of the automotive styling structure lines of each view were calculated as shown in Table 4; the importance weights of the four views of the automotive styling are shown in Table 5; finally, the beauty degree values of the automotive styling of 2,626 vehicle models were calculated.

[0105] Table 4 Importance weights of automotive styling structure lines

[0106]

[0107] Table 5 Importance weights of the four views of automotive styling

[0108]

[0109] 3. Construction and training of the measurement aesthetic convolutional neural network model

[0110] As Figures 6 - 7 shown, a measurement aesthetic convolutional neural network model was constructed using the pre-trained ResNet50 network as the basic model. As Figure 8 shown, during the model training process, the mean squared error (MSE) was used as the loss function to train the model until the loss function converged.

[0111] 4. Verification and analysis of the training results of the measurement aesthetic convolutional neural network model

[0112] First of all, a verification data set for the measurement aesthetic convolutional neural network model was constructed, and preprocessing was designed and carried out in combination with the method in S1. As Figure 9As shown below. Secondly, the verification dataset is used to verify the measurement aesthetic convolutional neural network model, and the result data is output, as shown in Table 6. Thirdly, the verification dataset is used to conduct a questionnaire survey on the beauty degree evaluation of the automotive styling scheme, and reliability analysis is carried out, and the relevant data is sorted out, as shown in Table 7. Finally, a comparative analysis is carried out on the verification results of the measurement aesthetic convolutional neural network model and the questionnaire survey results of the beauty degree evaluation of the automotive styling scheme. The analysis results are shown in Table 8. The Spearman rank correlation coefficient of the two rankings is 0.9667, which proves that the results of the intelligent screening of the measurement aesthetic convolutional neural network model are accurate, and further shows that the measurement aesthetic entropy evaluation model is reliable.

[0113] Table 6 Output Results of the Intelligent Screening Program of the Measurement Aesthetic Convolutional Neural Network

[0114]

[0115] Table 7 Questionnaire Survey Results of the Automotive Styling Scheme

[0116]

[0117] Table 8 Comparison of the Ranking Results of the Beauty Degree Values of the Automotive Styling Scheme

[0118]

[0119] Therefore, an intelligent screening method for automotive styling based on the measurement aesthetic entropy evaluation model proposed by the present invention constructs a measurement aesthetic entropy evaluation model by combining the aesthetic measurement theory, information entropy theory and complexity theory, realizes the scientific quantification and objective evaluation of the beauty degree of automotive styling, and improves the beauty degree and efficiency of automotive styling design.

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent screening method for automobile styling based on a quantitative aesthetic entropy evaluation model, characterized in that: The specific steps are as follows: S1. Car styling image data collection and preprocessing, constructing a car styling image dataset for measuring aesthetic convolutional neural networks and a line drawing atlas for calculating car styling beauty values; S2. Based on the Birkhoff aesthetic measurement theory model, entropy evaluation theory model and complexity theory model, a quantitative aesthetic entropy evaluation model is constructed to calculate the beauty value of automobile styling; S3. Construct a quantitative aesthetic convolutional neural network model based on quantitative aesthetic entropy evaluation; S4, validation of the convolutional neural network model for measuring aesthetics; In S2, the specific calculation steps of the automobile styling beauty value are as follows: S21. Measure the entropy redundancy of the three-dimensional modeling lines of the automobile based on the entropy evaluation model and calculate the aesthetic order of automobile modeling; S22. Calculate the aesthetic complexity of automobile styling based on the complexity theory model; S23, calculating the beauty value of automobile styling based on the quantitative aesthetic entropy evaluation model and comprehensively considering the automobile styling aesthetic order and the automobile styling aesthetic complexity; In S21, the automobile styling aesthetic order calculation formula is as follows: ； in, To create an aesthetic order for car styling, For each view's morphological order, is the weight value for each view, h is the four views of the car shape; The morphological order of each view , and its calculation formula is as follows: ； in, For structural lines The information entropy of is the weight value of the modeling structure line, is the car modeling curve function, For the Points on the structure line The derivative value of The weight of the sum of the derivative values ​​at all points of the subdifferential, is the number of points on the curve with different absolute values ​​of derivatives, , is an integer, is an integer, is an integer; In S22, the formula of the complexity theory model is as follows: ； in, For the aesthetic complexity of car styling, The simplest order for car styling, For the dissipation of car styling information energy, A slow growth rate.

2. According to claim 1, a method for intelligently screening automobile shapes based on a quantitative aesthetic entropy evaluation model is characterized in that: In S1, the preprocessing operation includes grayscale processing, background clipping batch processing, standardization processing, automobile modeling structure line extraction processing and noise reduction processing.

3. The method for intelligently screening automobile shapes based on the quantitative aesthetic entropy evaluation model according to claim 1 is characterized in that: The modeling structure line weight value The Likert nine-level scale is used to weight the car styling structure line, and the calculation formula is as follows: ； in, is the average score of each modeling structure line, The total number of shape construction lines for each view.

4. The method for intelligently screening automobile shapes based on the quantitative aesthetic entropy evaluation model according to claim 1 is characterized in that: In S23, the quantitative aesthetic entropy evaluation model formula is as follows: ； in, The beauty value of the car's styling.

5. The method for intelligently screening automobile shapes based on the quantitative aesthetic entropy evaluation model according to claim 1 is characterized in that: In S3, the quantitative aesthetic convolutional neural network model uses the pre-trained ResNet50 network as the basic model and the mean square error MSE as the loss function.

6. The method for intelligently screening automobile shapes based on the quantitative aesthetic entropy evaluation model according to claim 5 is characterized in that: The calculation formula of the loss function is as follows: ； in, is the total number of samples, , For the The true label data of the beauty value of samples, , For the The beauty value prediction data of samples.

7. The method for intelligently screening automobile shapes based on the quantitative aesthetic entropy evaluation model according to claim 5 is characterized in that: The quantitative aesthetic convolutional neural network model uses the Adam optimizer to update model parameters, and the model training parameters include the number of iterations epoch, the batch size batch_size and the learning rate.

Images