Product appearance optimization method and device, terminal equipment and storage medium

Abstract

The invention is suitable for the technical field of data processing, and provides a product appearance optimization method and device, terminal equipment and a storage medium, the method screens out a plurality of core perceptual vocabularies through subjective weight values and objective weight values of perceptual vocabularies, so that the core perceptual vocabularies can more accurately reflect perceptual cognition of a user on the product appearance, and the user experience is improved. Besides, an initial neural network model is trained by designing a model training set constructed by combined features and core perceptual vocabularies to obtain a perceptual evaluation model, pre-selected appearance design features are input into the perceptual evaluation model to obtain a corresponding perceptual evaluation value, then the appearance design of the target product is adjusted according to the perceptual evaluation value, and the perceptual evaluation result is obtained. According to the method, the appearance design of the product by the user can be predicted through the perceptual evaluation value output by the perceptual evaluation model, so that the appearance of the product is improved in time when the perceptual evaluation value is low, and the efficiency of optimizing the appearance of the product is improved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly to a method, device, terminal device and storage medium for optimizing product appearance. Background Art

[0002] With the intensification of market competition and the improvement of consumers' requirements for product appearance design, the appearance of products has become one of the important factors in consumers' purchase decisions. Especially in the household appliance industry, such as in the design process of consumer products like air purifiers, air conditioners, smart speakers, etc., the perceptual value of product appearance directly affects consumers' emotional experience and purchase decisions. Traditional product appearance design methods mainly rely on designers' experience and subjective judgments, unable to fully consider consumers' emotional needs and lacking a systematic optimization process. Although traditional kansei engineering methods can theoretically guide design, in practical applications, due to the lack of effective optimization means, it is often difficult to accurately quantify consumers' perceptual needs.

[0003] As a design method that combines consumers' emotional needs with product design elements, kansei engineering has been widely used in the optimization of product appearance, color, material, etc. However, in practice, the application of kansei engineering often faces the following challenges: First, the selection of kansei words and the determination of weights still rely mostly on experts' subjective judgments, lacking objective and effective evaluation methods; Second, during the design optimization process, it is often impossible to accurately capture the complex relationship between design parameters and consumers' emotional needs, resulting in low design optimization efficiency; Third, traditional kansei engineering methods usually rely on manual adjustment and are difficult to achieve large-scale and diversified design optimization. Therefore, how to accurately and comprehensively identify and optimize consumers' emotional needs in product appearance design and improve the efficiency of product appearance optimization has become a technical problem to be solved urgently. Summary of the Invention

[0004] Embodiments of this application provide a method, device, terminal device and storage medium for optimizing product appearance, which can solve the technical problems in the prior art that product appearance design cannot accurately and comprehensively identify and optimize consumers' emotional needs and the product appearance optimization efficiency is low.

[0005] In a first aspect, embodiments of this application provide a method for optimizing product appearance, including: Obtain a plurality of kansei words related to the appearance design of a target product, and construct a judgment matrix corresponding to each kansei word according to the relative importance scores of each kansei word in the plurality of kansei words; When the judgment matrix passes the consistency test, determine the subjective weight value of the corresponding kansei word according to the judgment matrix; Construct an evaluation matrix according to the kansei word scores of each sample product on each kansei word; Determine the objective weight values of the respective perceptual words according to the evaluation matrix; Screen out a plurality of core perceptual words from the respective perceptual words according to the subjective weight values and objective weight values of the respective perceptual words; Construct a model training set according to the multiple design combination features of the target product and the multiple core perceptual words; Train an initial neural network model through the model training set to obtain a perceptual evaluation model, where the initial neural network model is composed of a genetic algorithm and a backpropagation neural network; Input the preselected appearance design features into the perceptual evaluation model to obtain the perceptual evaluation value corresponding to the preselected appearance design features; Adjust the preselected appearance design features according to the perceptual evaluation value to optimize the appearance of the target product.

[0006] Further, in the case where the judgment matrix passes the consistency test, determining the subjective weight value of the corresponding perceptual word according to the judgment matrix includes: Determine the preliminary weights of the respective perceptual words according to the product of the elements in each row of the judgment matrix; Perform normalization processing on the preliminary weights of the respective perceptual words to obtain the initial weight values corresponding to the respective perceptual words; Determine the maximum eigenvalue corresponding to the judgment matrix according to the initial weight values corresponding to the respective perceptual words and the judgment matrix; Determine the consistency index of the judgment matrix according to the maximum eigenvalue corresponding to the judgment matrix and the matrix order; Determine the adjustment coefficient of the judgment matrix according to the matrix order of the judgment matrix; Determine the consistency ratio of the judgment matrix according to the consistency index and adjustment coefficient of the judgment matrix; When the consistency ratio is less than the preset ratio, use the initial weight value as the subjective weight value of the corresponding perceptual word.

[0007] Further, the determining the objective weight values of the respective perceptual words according to the evaluation matrix includes: Determine the score mean and score standard deviation of each perceptual word according to the evaluation matrix; Determine the coefficient of variation of each perceptual word according to the score mean and score standard deviation of each perceptual word; Perform normalization processing on the coefficient of variation of each perceptual word to obtain the objective weights of the respective perceptual words.

[0008] Further, screening out multiple core perceptual words from the various perceptual words according to the subjective weight values and objective weight values of the various perceptual words includes: Determining the coupling degree of the various perceptual words according to the subjective weight values and objective weight values of the various perceptual words; Determining the coupling coordination degree of each perceptual word according to the coupling degree of the various perceptual words and a preset adjustment factor; Determining the perceptual words with a coupling coordination degree greater than a preset coupling coordination degree as core perceptual words.

[0009] Further, constructing a model training set according to the multiple design combination features of the target product and the multiple core perceptual words includes: Obtaining the perceptual values corresponding to the multiple design combination features of the target product on each core word; Encoding each design combination feature to obtain an encoded feature; Constructing a model training set according to the encoded feature and the corresponding perceptual value.

[0010] Further, the genetic algorithm is used to optimize the weight values and threshold values of the initial neural network model.

[0011] Further, screening out multiple core perceptual words from the various perceptual words according to the subjective weight values and objective weight values of the various perceptual words further includes: Obtaining the subjective score set and objective score set related to the target product and multiple perceptual words; Calculating the comprehensive weight value of each perceptual word according to the subjective score set, the objective score set, the subjective weight value, and the objective weight value through a comprehensive weight calculation formula; Determining the perceptual words with a comprehensive weight value greater than a preset weight threshold as core perceptual words; Among them, the comprehensive weight calculation formula is: In the formula, is the comprehensive weight value of the th perceptual word; is a preset coefficient, and its value range is between [0, 1], and ; is the subjective weight value of the th perceptual word; is the subjective score of the th user; is the rd objective score; is the subjective score set; is the objective score set; is the average score value of the subjective score set; is the average score value of the objective score set; is the variance of the subjective score set; is the variance of the objective score set; is the minimum score value in the subjective score set and the objective score set; is the maximum score value in the subjective score set and the objective score set.

[0012] In a second aspect, an embodiment of the present application provides a product appearance optimization device, including: An acquisition module, configured to acquire a plurality of perceptual words related to the appearance design of a target product, and construct a judgment matrix corresponding to each perceptual word according to the relative importance scores of each perceptual word in the plurality of perceptual words; A first determination module, configured to determine a subjective weight value corresponding to a perceptual word according to the judgment matrix when the judgment matrix passes the consistency test; A first construction module, configured to construct an evaluation matrix according to the perceptual word scores of each sample product on each perceptual word; A second determination module, configured to determine an objective weight value of each perceptual word according to the evaluation matrix; A screening module, configured to screen out a plurality of core perceptual words from each perceptual word according to the subjective weight value and the objective weight value of each perceptual word; A second construction module constructs a model training set according to the multiple design combination features of the target product and the multiple core perceptual words; A training module, configured to train an initial neural network model through the model training set to obtain a perceptual evaluation model, where the initial neural network model is composed of a genetic algorithm and a backpropagation neural network; An input module, configured to input preselected appearance design features into the perceptual evaluation model to obtain a perceptual evaluation value corresponding to the preselected appearance design features; An adjustment module, configured to adjust the preselected appearance design features according to the perceptual evaluation value to optimize the appearance of the target product.

[0013] In a third aspect, an embodiment of the present application provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method in the first aspect above is implemented.

[0014] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method in the first aspect above is implemented.

[0015] It can be understood that the beneficial effects of the second to fourth aspects above can be referred to the relevant descriptions in the first aspect above, and will not be elaborated here.

[0016] The beneficial effects of the embodiments of the present application compared with the prior art are as follows: By the subjective weight value and objective weight value of perceptual words, multiple core perceptual words are screened out, so that the core perceptual words can more accurately reflect the user's perceptual cognition of the product appearance. In addition, by training the initial neural network model with the model training set constructed by the design combination features and core perceptual words, a perceptual evaluation model is obtained. The preselected appearance design features are input into the perceptual evaluation model to obtain the corresponding perceptual evaluation value, and then the appearance design of the target product is adjusted according to the perceptual evaluation value. It is possible to predict the user's perception of the product appearance through the perceptual evaluation value output by the perceptual evaluation model, so that when the perceptual evaluation value is low, the product appearance can be improved in time, and the efficiency of product appearance optimization is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 is the implementation flowchart of the first embodiment of a product appearance optimization method provided by the embodiments of the present application; Figure 2 is the implementation flowchart of the second embodiment of a product appearance optimization method provided by the embodiments of the present application; Figure 3 is the structural block diagram of a product appearance optimization device provided by the embodiments of the present application; Figure 4 is the structural block diagram of a product appearance optimization device provided by the embodiments of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] In the following description, specific details such as specific system structures and technologies are proposed for the purpose of illustration rather than limitation, so as to thoroughly understand the embodiments of the present application. However, those skilled in the art should clearly understand that the present application can also be implemented in other embodiments without these specific details. In other cases, the detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present application.

[0020] It should be understood that when used in the specification of this application and the appended claims, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their combinations.

[0021] It should also be understood that the term "and / or" used in the specification of this application and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0022] As used in the specification of this application and the appended claims, the term "if" may be interpreted, depending on the context, as "when", "once", "in response to determining", or "in response to detecting". Similarly, the phrase "if determined" or "if [the described condition or event] is detected" may be interpreted, depending on the context, as meaning "once determined", "in response to determining", "once [the described condition or event] is detected", or "in response to detecting [the described condition or event]".

[0023] In addition, in the description of the specification of this application and the appended claims, the terms "first", "second", "third", etc. are only used for differential description and should not be construed as indicating or implying relative importance.

[0024] Reference to "one embodiment" or "some embodiments" etc. described in the specification of this application means that a specific feature, structure or characteristic described in connection with that embodiment is included in one or more embodiments of this application. Thus, statements such as "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments", etc. that appear in different places in this specification do not necessarily all refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "comprising", "including", "having" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0025] Please refer to Figure 1 , Figure 1 which shows the implementation flowchart of the first embodiment of a product appearance optimization method provided by the embodiment of this application, including: Step S10: Obtain a plurality of perceptual words related to the appearance design of the target product, and construct a judgment matrix corresponding to each of the perceptual words according to the relative importance scores of the perceptual words in the plurality of perceptual words.

[0026] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication, and program running functions, or an electronic device, product appearance optimization device, computer, tablet computer, etc. that can implement the above functions. Hereinafter, taking the product appearance optimization device as an example, this embodiment and the following embodiments will be illustrated by way of example.

[0027] It can be understood that the target product can be a product whose design appearance needs to be optimized. The emotional vocabulary can be a vocabulary used to reflect the emotional needs of consumers for the product's appearance design, such as: modern, fashionable, technological, high-end, warm, and elegant, etc. The relative weight score can be the score given by experts after pairwise comparison of the emotional vocabulary, which is used to represent the relative importance between each emotional vocabulary. The judgment matrix can be a matrix constructed with the relative importance scores as elements. The relative weight scores between each emotional vocabulary and other emotional vocabularies are used as the elements of a row of the judgment matrix.

[0028] Step S20: When the judgment matrix passes the consistency test, determine the subjective weight value of the corresponding emotional vocabulary according to the judgment matrix.

[0029] It can be understood that the consistency of the matrix usually appears in the analytic hierarchy process and is used to measure the rationality of the judgment matrix. If the judgment matrix passes the consistency test, it is determined that the judgment matrix is reasonable.

[0030] Step S30: Construct an evaluation matrix according to the emotional vocabulary scores of each sample product on each emotional vocabulary.

[0031] It can be understood that the evaluation matrix contains the emotional vocabulary scores of all product samples. Suppose there are n sample products and m emotional vocabularies. Each item x ij in the matrix represents the score of the i th sample product on the j th emotional vocabulary.

[0032] Step S40: Determine the objective weight value of each emotional vocabulary according to the evaluation matrix.

[0033] It can be understood that the objective weight value can be a value used to objectively reflect the weight of the emotional vocabulary.

[0034] Step S50: Screen out multiple core emotional vocabularies from each emotional vocabulary according to the subjective weight value and objective weight value of each emotional vocabulary.

[0035] It can be understood that the core emotional words can be the selected words that can more accurately reflect the emotions of consumers. The comprehensive weight value of each emotional word is determined according to the subjective weight value and the objective weight value, and multiple core words are selected from each emotional word according to the size of the comprehensive weight value.

[0036] Step S60: Construct a model training set according to the multiple design combination features of the target product and the multiple core emotional words.

[0037] It can be understood that first, the emotional value of each design feature combination on each core emotional word is determined, and the design combination feature is used as the input data, and the emotional value is used as the prediction target to construct a model training set.

[0038] Step S70: Train the initial neural network model through the model training set to obtain an emotional evaluation model, where the initial neural network model is composed of a genetic algorithm and a backpropagation neural network.

[0039] Step S80: Input the preselected appearance design features into the emotional evaluation model to obtain the emotional evaluation value corresponding to the preselected appearance design features.

[0040] Step S90: Adjust the preselected appearance design features according to the emotional evaluation value to optimize the appearance of the target product.

[0041] It can be understood that if the emotional evaluation value corresponding to the preselected appearance design feature is less than the preset emotional evaluation value, it is determined that the preselected appearance design feature does not meet the design requirements, and the appearance of the target product needs to be optimized; otherwise, it is determined that the preselected appearance design feature meets the design requirements.

[0042] The method provided in this embodiment screens out multiple core emotional words through the subjective weight value and the objective weight value of the emotional words, so that the core emotional words can more accurately reflect the user's emotional perception of the product appearance. In addition, the initial neural network model is trained through the model training set constructed by the design combination features and the core emotional words to obtain an emotional evaluation model. The preselected appearance design features are input into the emotional evaluation model to obtain the corresponding emotional evaluation value, and then the appearance design of the target product is adjusted according to the emotional evaluation value. It is possible to predict the user's perception of the product appearance through the emotional evaluation value output by the emotional evaluation model, so as to improve the efficiency of product appearance optimization in a timely manner when the emotional evaluation value is low.

[0043] In some alternative implementations, step S20 can be implemented through the following steps: Determine the preliminary weights of each perceptual word according to the product of the elements in each row of the judgment matrix; perform normalization processing on the preliminary weights of each perceptual word to obtain the initial weight values corresponding to each perceptual word; determine the maximum eigenvalue of the judgment matrix according to the initial weight values corresponding to each perceptual word and the judgment matrix; determine the consistency index of the judgment matrix according to the maximum eigenvalue of the judgment matrix and the matrix order; determine the adjustment coefficient of the judgment matrix according to the matrix order of the judgment matrix; determine the consistency ratio of the judgment matrix according to the consistency index and the adjustment coefficient of the judgment matrix; when the consistency ratio is less than the preset ratio, use the initial weight value as the subjective weight value of the corresponding perceptual word.

[0044] It can be understood that in this embodiment, the analytic hierarchy process is used to determine the subjective weight value of the perceptual word. The analytic hierarchy process is an effective multi-criteria decision-making analysis method, especially suitable for dealing with qualitative and quantitative factors in complex decision-making problems. In this embodiment, the analytic hierarchy process is used to determine the importance of each perceptual word in the appearance design of the target product, and the weight of each perceptual word in the design is quantified through expert assignment and the analytic hierarchy process.

[0045] In an example, for instance, if the target product is an air purifier, the process of determining the subjective weight value using the analytic hierarchy process is as follows: 1. Construct an evaluation matrix: First, the importance of perceptual words is judged by an expert group. Based on the role of each perceptual word in the appearance design of the air purifier, pairwise comparisons are made and a judgment matrix is constructed. During this process, experts use the scale values from 1 to 9 and their reciprocals to represent the relative importance between each perceptual word according to their own experience and design requirements. The definitions of the scale values are shown in Table 1.

[0046] Table 1 Meanings of Judgment Matrix Scale Values 2. Solve the weight vector: After obtaining the judgment matrix, the next step is to solve the weight vector of the judgment matrix. Usually, the geometric mean method is used to solve the weight. First, calculate the product of each row of elements: In the formula: is the product of the i th row; is the element in the i th row and j th column of the judgment matrix; m represents the number of perceptual words.

[0047] 3. Calculate the geometric mean: Next, based on the product results of each row, calculate the geometric mean of each row, which will be used as the preliminary weight of the corresponding perceptual word. The specific formula is as follows: In the formula, is the initial weight of the i th perceptual word.

[0048] 4. Normalize the weights: To obtain the relative weights of each perceptual word, it is necessary to normalize the above geometric mean. The normalization formula is as follows: In the formula, is the initial weight value of the i th perceptual word.

[0049] 5. Calculate the maximum eigenvalue of the judgment matrix: In the analytic hierarchy process, consistency check is one of the key steps to judge whether the judgment matrix is reasonable. First, it is necessary to calculate the maximum eigenvalue of the judgment matrix. According to the preliminary weights obtained, the formula for calculating the maximum eigenvalue is: In the formula, is the vector obtained by multiplying the judgment matrix by the preliminary weights.

[0050] 6. Consistency check: In the analytic hierarchy process, to ensure the reasonableness of the judgment matrix, a consistency check is required. The consistency check is carried out by calculating the consistency index. The formula for calculating the consistency index CI is as follows: In the formula, n is the order of the judgment matrix.

[0051] 7. Calculate the consistency ratio: Next, a final judgment needs to be made through the consistency ratio (CR). The formula for CR is: According to Table 2, look up and determine the value of the adjustment coefficient RI in the table according to the order n of the judgment matrix. When CR ≤ 0.1, it is considered that the consistency of the judgment matrix is good; otherwise, the judgment matrix needs to be re-evaluated and adjusted.

[0052] Table 2 In some alternative implementations, step S40 can be implemented through the following steps: Determine the mean score and standard deviation of each perceptual word according to the evaluation matrix; Determine the coefficient of variation of each perceptual word according to the mean score and standard deviation of each perceptual word; Perform normalization processing on the coefficient of variation of each perceptual word to obtain the objective weight of each perceptual word.

[0053] It can be understood that in this embodiment, the objective weight of perceptual words is determined by the coefficient of variation method. The coefficient of variation method is an objective method for determining weights and is widely used in multi-criteria decision-making and weight allocation in product design. This method evaluates the importance of each factor by calculating the degree of dispersion (variability) of the data. Generally, the greater the variability of a factor, the greater its impact on the overall system. In this embodiment, the coefficient of variation method is used to determine the objective weight of each perceptual word in the appearance design of the target product.

[0054] In an example, the steps to determine the objective weight of each perceptual word by the coefficient of variation method are as follows: 1. Construct an evaluation matrix: First, an evaluation matrix needs to be constructed, which contains the perceptual word scores of all product samples. Suppose there are d air purifier product samples and m perceptual words. Each item in the matrix represents the score of the r th sample on the s th perceptual word.

[0055] In the formula: r = 1, 2, 3, …, d; s = 1, 2, 3, …, m .

[0056] 2. Calculate the coefficient of variation: For each perceptual word, first calculate the mean of its scores and the standard deviation . The coefficient of variation is calculated by taking the ratio of the standard deviation to the mean, measuring the relative dispersion degree of this perceptual word. The specific calculation is as in formula (8): In the formula: represents the mean score of the s th perceptual word, d is the number of samples, is the score of the r th sample on the s th perceptual word.

[0057] In the formula: represents the sThe standard deviation of a perceptual word reflects the degree of dispersion of the perceptual word among all samples.

[0058] In the formula: is the s coefficient of variation of the nth perceptual word.

[0059] 3. Calculate the weights: For the convenience of comparison and analysis, it is necessary to normalize the coefficient of variation to obtain the objective weights of each perceptual word . The normalized weights represent the relative importance of each perceptual word in the overall design. The calculation is as shown in formula (11): In some optional implementation manners, the step S50 can be implemented through the following steps: Determine the coupling degree of each perceptual word according to the subjective weight value and the objective weight value of each perceptual word; Determine the coupling coordination degree of each perceptual word according to the coupling degree of each perceptual word and a preset adjustment factor; Determine the perceptual words with a coupling coordination degree greater than the preset coupling coordination degree as the core perceptual words.

[0060] It can be understood that in this embodiment, the coupling coordination degree method is used to calculate the coupling coordination degree of each perceptual word. The coupling coordination degree method is an analysis method used to measure the degree of interaction and coordination among various factors in a system, and is usually used for multi-factor comprehensive evaluation. In this embodiment, the coupling coordination degree method is applied to screen the optimal perceptual words in the appearance design of the target product. Based on the subjective weight and the objective weight, the coordination among each perceptual word is evaluated, and those words that play a key role in the design can be identified. Words with a high coordination degree indicate a high degree of matching between subjective and objective factors, and these words can better reflect the emotional needs of consumers, and are thus used for the optimization of the appearance design of the target product.

[0061] In an example, the steps of calculating the coupling coordination degree of perceptual words using the coupling coordination degree method are as follows: 1. Calculate the coupling degree: According to the concept of "coupling" in physics, the coupling degree is used to describe the degree of interdependence among multiple systems. In this embodiment, the subjective weight and the objective weight are regarded as two systems, and the coupling degree between them is calculated. The calculation formula of the coupling degree is: In the formula: respectively represent the weight values of each system. For the subjective weight and the objective weight obtained in this embodiment, they are simplified into two systems, and the coupling degree between them is calculated using formula (13): In the formula: and They represent the subjective weight and the objective weight respectively.

[0062] 2. Construct a coupling coordination degree model: After obtaining the coupling degree , it is necessary to further evaluate the coordination between systems. The coordination degree is used to measure the coordination of systems, that is, whether the interaction between different systems can reach the optimal state. Therefore, in this embodiment, a coupling coordination degree model is constructed to quantify the coordination between the subjective weight and the objective weight. It is represented by the coupling coordination degree D , and the specific calculation formula is: In the formula: C is the coupling degree, T is a preset adjustment factor, that is, a pre-set adjustment factor.

[0063] In this embodiment, since the subjective weight and the objective weight are equally important for the final design result, the preset adjustment factor is defined as: In the formula: α = β = 0.5. In this way, the preset adjustment factor can be expressed as the balanced weighting of the two weights.

[0064] In some alternative implementation manners, the step S60 can be implemented through the following steps: obtaining the perceptual values corresponding to multiple design combination features of the target product on each core vocabulary; encoding each design combination feature to obtain encoded features; and constructing a model training set according to the encoded features and the corresponding perceptual values.

[0065] In some alternative implementation manners, the genetic algorithm is used to optimize the weights and thresholds of the initial neural network model.

[0066] In an example, this embodiment combines the genetic algorithm (GA) and the backpropagation neural network (BPNN) to optimize the product appearance design. Through the global search ability of the genetic algorithm, multiple design parameters in the design space are optimized to find the optimal design combination features. At the same time, by using the powerful learning ability of the backpropagation neural network, through the trained model, the complex mapping relationship between the design combination features and the perceptual values is captured to achieve accurate perceptual demand prediction and design optimization. Finally, by combining the genetic algorithm and the BP neural network, an optimization model that can accurately predict the emotional needs of consumers is established.

[0067] The backpropagation neural network is a commonly used supervised learning algorithm. Through the training of the sample set, the weights and thresholds in the network are continuously adjusted to make the network output results as close as possible to the true values. In this method, BPNN is used to predict the perceptual values of the air purifier appearance design solutions, aiming to optimize the appearance design through training the model to make it more in line with the emotional needs of consumers. BPNN consists of three main layers: the input layer, the hidden layer, and the output layer. The input layer receives the input data (in this embodiment, it is the design combination features of the air purifier appearance design), the hidden layer processes the data through a non-linear activation function, and the output layer generates the final prediction result (i.e., the perceptual value of the design solution). The input layer : Each input node represents a design feature and is passed into the network for processing. The hidden layer : After the neurons in the hidden layer perform weighted summation and then pass through the activation function calculation, the value of each neuron is output. The output layer : The output layer generates the final predicted value, that is, the perceptual value of the air purifier appearance design. BPNN adjusts the weights and thresholds of each layer through the backpropagation algorithm, that is and . In addition, each neuron in the hidden layer and the output layer has its own threshold, and the magnitudes of the weights and thresholds determine the accuracy of the output result.

[0068] The genetic algorithm is an optimization algorithm that simulates natural selection and genetic mechanisms and has strong global search capabilities. In this embodiment, the genetic algorithm is used to optimize the weights and thresholds of BPNN to accelerate the training process and improve the accuracy of the model. GA helps to select the best combination of weights and thresholds, thereby reducing the training time of BPNN and improving the prediction accuracy. GA and BPNN are combined to form the GA-BP model, and its main steps are as follows: 1. Initialize the population: The first step of the genetic algorithm is to initialize the population. In this embodiment, each individual in the population represents a combination of weights and thresholds of a BPNN. The chromosome of each individual consists of the weights between the hidden layer and the input layer, the hidden layer threshold, the weights between the hidden layer and the output layer, and the output layer threshold. Each chromosome is encoded as a real number encoding, representing each weight and threshold in the BPNN network.

[0069] 2. Fitness evaluation: Fitness evaluation is the core step in the genetic algorithm. Define the fitness function as the reciprocal of the mean square error (MSE): Where: q is the number of training samples; and represent the actual value and the predicted value of the i th output node.

[0070] 3. Selection operation: The selection operation is used to determine which individuals enter the next generation based on the fitness values of the individuals. The commonly used selection method is Roulette Wheel Selection, where the selection probability is proportional to the fitness: 4. Crossover operation: The crossover operation simulates the process of genetic recombination by exchanging the genes of parental individuals to generate new individuals. The goal of the crossover operation is to produce better offspring by combining the advantages of two parents. In this embodiment, the crossover operation adopts a real-number crossover method, where the chromosomes of the parental individuals are exchanged at random positions to generate new offspring individuals.

[0071] 5. Mutation operation: The mutation operation is a local search strategy in the genetic algorithm, which introduces new features by randomly modifying the genes of some individuals. The mutation operation helps to maintain the diversity of the population and avoid falling into local optimal solutions.

[0072] 6. Determining the number of hidden layer nodes: The performance of the BPNN is closely related to the number of nodes in the hidden layer. Too many nodes in the hidden layer may lead to overfitting, while too few may lead to underfitting. Therefore, it is crucial to select an appropriate number of hidden layer nodes. The selection of the number of nodes is often calculated through an empirical formula: In the formula: j, n, k is the number of nodes in the hidden layer, input layer, and output layer; c is a constant, and the range is [0, 10].

[0073] 7. Optimizing the weights and thresholds of the BPNN: By obtaining the output results of the hidden layer, multiplying them by the weights and adding the thresholds, and then calculating the activation function in combination with the purelin function, the final output layer results are obtained. The sum of the squared errors between the output layer results and the actual sample output values is used as the fitness value to guide the optimization of the genetic algorithm. The specific calculation formula is as follows: In the formula: , are the output data of the hidden layer and output layer, is the input data, are the weights and thresholds of each layer, and tansig and purelin are the node transfer functions.

[0074] In the optional implementation manners of the embodiments of the present application, the above step S50 may include the following steps S501 - S503. Among them, Figure 2 is the implementation flowchart of the second embodiment of the product appearance optimization method provided by the embodiment of the present application.

[0075] Step S501: Obtain the subjective score set and the objective score set related to multiple perceptual words for the target product.

[0076] It can be understood that the subjective score set can be a set composed of the scoring values of different users on the degree of conformity of the product appearance described by each perceptual word. For example, the subjective score set is , where is the score of the i th user on a certain perceptual word. The objective score set can be a set composed of the scoring values on the perceptual words obtained from aspects such as the product design database and market research data. For example, the objective score set is , where is the i th objective score value.

[0077] Step S502: Calculate the comprehensive weight values of each perceptual word according to the subjective score set, the objective score set, the subjective weight value, and the objective weight value through the comprehensive weight calculation formula.

[0078] Step S503: Determine the core perceptual words as the perceptual words whose comprehensive weight values are greater than the preset weight threshold.

[0079] Among them, the comprehensive weight calculation formula is: (20) In the formula, is the comprehensive weight value of the th perceptual word; is a preset coefficient, and its value range is between [0, 1], and ; is the subjective weight value of the th perceptual word; is the subjective score of the th user; is the th objective score; is the subjective score set; is the objective score set; is the average score value of the subjective score set; is the average score value of the objective score set; is the variance of the subjective score set; is the variance of the objective score set; is the minimum score value in the subjective score set and the objective score set; is the maximum score value in the subjective score set and the objective score set.

[0080] It can be understood that determining the comprehensive weight value through the above method has the following advantages: 1. Comprehensive utilization of multi-dimensional data When selecting core emotional words by traditional methods, subjective judgment is often relied on alone, or only some objective factors are simply considered. In this embodiment, subjective evaluation data and objective data are comprehensively incorporated. Among them, the subjective weight value reflects the direct feelings of users towards the product appearance, while the objective weight value provides support from objective aspects such as the market and design. At the same time, the ratio of the sum of user scores to the sum of objective data, as well as the ratio of the number of users to the number of objective data, are incorporated into the formula, further connecting the subjective and objective data, so that the comprehensive weight value can more comprehensively and three-dimensionally reflect the characteristics of the product appearance. 2. Dynamic balance mechanism: The preset coefficient in the formula is not only limited in the value range, but can also be trained based on historical product appearance description data through machine learning algorithms. This feature enables the formula to flexibly adjust the weights of subjective and objective factors and other indicators when facing different types of products, ensuring accurate calculation of the comprehensive weight value in different scenarios. For example, for fashion products, the influence of the subjective weight value can be appropriately increased; while for functional products, the proportion of the objective weight value is increased. 3. Accurately reflect data characteristics: In this embodiment, by introducing statistics such as the mean, variance, minimum value, and maximum value, the information behind the data is deeply explored. The mean reflects the central tendency of the data, the variance reflects the degree of dispersion of the data, and the extreme values highlight the boundary conditions of the data. Through a series of operations, these statistics cooperate with each other to more accurately reflect the internal characteristics of the subjective and objective data, as well as the differences and connections between them, thereby improving the accuracy of the comprehensive weight value. 4. Improve the adaptability of the model: Since the preset coefficient can be optimized through machine learning, this formula has strong adaptability. As new product appearance data continues to accumulate, the model can continuously learn and adjust to adapt to market changes and the evolution of user needs. This dynamic optimization mechanism enables this formula to play a good role in the appearance descriptions of different industries and different types of products, effectively avoiding the limitations caused by fixed parameters.

[0081] In an example, taking the appearance design of an air purifier as an example for illustration, the specific implementation method is as follows. Determine emotional words: First, collect emotional words related to the appearance design of the air purifier through literature research, expert interviews, and market analysis. Secondly, organize the collected emotional words into cards, one card for each word, and mark its source (such as literature, interview, or market analysis). Finally, organize design experts, consumer representatives, and researchers to conduct group discussions, and using the principles of the KJ method, freely combine and classify these emotional words. In this process, the participants classify the emotional words into different categories according to the similarity and relevance of the words, and assign a descriptive label to each category. These labels represent the core emotional characteristics expressed by the emotional words in the design.

[0082] During the classification process, the research team conducted an in-depth analysis of each category, removing redundant, overly abstract, or irrelevant words. Finally, 8 core emotional words that can better reflect consumers' emotional needs were selected. These words will serve as the key inputs in the subsequent research to guide the optimization direction of product appearance design. The specific content is shown in Table 3.

[0083] Table 3 Emotional Word Classification Subjective Weight Analysis: To determine the importance of emotional words related to the appearance design of air purifiers in consumers' emotional needs, the Analytic Hierarchy Process (AHP) was used to analyze the subjective weights of the 8 emotional words. For example, 10 design experts (8 professors majoring in industrial design and 2 product design engineers) were invited to make pairwise comparisons based on the importance of these emotional words to construct a judgment matrix. Through expert scoring and correction using the maximum improvement direction algorithm, a judgment matrix was constructed. The weights were calculated according to formulas (1)-(6) and consistency tests were performed. The specific content is shown in Tables 4-5. It can be seen that the CR values of all judgment matrices are less than 0.1.

[0084] Table 4 Consistency Test Results Table 5 Emotional Word Weights Objective Weight Analysis: The coefficient of variation method can effectively evaluate the importance of each emotional word among actual consumers by analyzing the dispersion degree of consumers' scores, thus providing an objective basis for assigning weights to emotional words. To conduct objective weight analysis, in this embodiment, first, 50 air purifier samples with the highest sales volume on the online shopping platform were collected. These samples represent air purifier products that are relatively popular in the market and have high consumer attention. Next, 20 consumers were invited to evaluate the appearance design of these 50 products. Each consumer scored the appearance design of each air purifier on each emotional word according to their intuitive feelings and preferences. The scoring standard was 1-9 points, where 1 means "completely does not conform" and 9 means "completely conforms". Finally, the average emotional evaluation value of each emotional word was calculated, and the weights were calculated according to formulas (7)-(11). The specific results are shown in Table 6.

[0085] Table 6 Consumer Evaluation Matrix and Objective Weight Values Sensory Vocabulary Screening and Coupling Coordination Analysis: By combining subjective and objective weight analysis and using the coupling coordination degree method to evaluate sensory vocabulary, the core sensory vocabulary that best meets consumers' emotional needs is screened out. The core of the coupling coordination degree method lies in calculating the coupling degree (C), coordination index (T), and final coupling coordination degree (D) between subjective and objective weights. These indicators can quantify the coordination of each sensory vocabulary in the appearance design. According to the magnitude of the D value, the matching degree of each sensory vocabulary can be intuitively evaluated, and through the grading of the coupling coordination degree, their effectiveness in the design can be clarified from "serious imbalance" to "high-quality coordination", as shown in Table 7. According to the analysis of formulas (12)-(15), the four sensory vocabulary of modern, simple, fashionable, and technological show high coordination between subjective and objective evaluations, so they are finally retained as the core sensory vocabulary, as shown in Table 8. This screening result provides a solid theoretical basis for the subsequent optimization of the air purifier appearance design.

[0086] Table 7 Coupling Coordination Degree Grading Standard Table 8 Calculation Results of Subjective and Objective Weight Coupling Coordination Degree Product Component Coding and Morphological Analysis: To systematically analyze the design morphology of the air purifier, this embodiment uses morphological analysis to deconstruct the appearance of the sample product and summarize its key components and morphological characteristics. By sorting through market samples, six key modeling elements of the top contour (A), bottom contour (B), front contour (C), operation panel contour (D), air inlet contour (E), and air outlet contour (F) are extracted, and their morphological factors are classified and coded, as shown in Table 9. In the coding rule, the modeling elements use A-F as the major category identifiers, and different sub-item morphological factors are distinguished by discrete data 1-5. These numbers not only systematize the morphological feature description but also serve as the training set for the input end of the neural network.

[0087] Table 9 Air Purifier Morphological Component Coding Determination of Input Layer Indicators: According to Table 10, 30 air purifier samples are extracted, and the main design elements are numerically coded. The design elements of the air purifier mainly consist of six modules: the top contour (A), bottom contour (B), front contour (C), operation panel contour (D), air inlet contour (E), and air outlet contour (F). These 6 modules are used as the input layer indicators of the neural network model, so the number of input layer neuron nodes is set to 6. And the quantified values of 4 groups of sensory vocabulary are used as the indicators of the output layer of the neural network model. Therefore, the number of output layer neurons is 4.

[0088] To make the data of the input layer and the output layer comparable, this method standardizes the input layer indicators and uses the min-max normalization method to normalize the data, as shown in Table 10. After normalization, the values of each input indicator are mapped to the interval from 0 to 1, effectively eliminating the differences between different magnitudes, thus avoiding network training errors caused by magnitude differences and ensuring the accuracy and stability of the neural network model.

[0089] Table 10 Design Element Encoding and Perceptual Evaluation Values Construction and Training of the GA-BP Model: In this embodiment, a hybrid optimization model (GA-BP) that combines a BP neural network and a genetic algorithm is constructed using a PyTorch-based framework. The topological structure of the BPNN is defined as follows: The number of neurons in the input layer is 6, representing six main morphological features of the air purifier design; the number of neurons in the hidden layer is 10, and the ReLU activation function is used to introduce non-linear mapping. The number of nodes is defined by an empirical formula, where the value is 4 to balance the model complexity and generalization ability; the output layer contains 4 neurons, corresponding to the normalized evaluation values of four key perceptual words in the air purifier design, and a linear activation function is used. The network parameters are mapped to gene vectors through real number encoding. For a model with 6×10 + 10×4 + 10 + 4 = 114 parameters, each individual is represented as a 114-dimensional vector , and the encoding rule is , where and are the weight matrices of the hidden layer and the output layer respectively, and are the bias terms. Next, GA optimization is adopted. First, the population is initialized, and individuals are randomly generated to form the initial population, and each gene value follows a uniform distribution , ensuring the diversity of the initial parameter space. Then, fitness evaluation is carried out, where the parameter m is the number of training samples, 25; the roulette wheel selection strategy is continued to retain individuals with high fitness while maintaining the population diversity. In crossover and mutation, in single-point crossover, the cut point is randomly selected with a probability to exchange the gene segments of the parent generation and generate offspring individuals. In Gaussian mutation, the offspring genes are perturbed with a probability to avoid premature convergence. The maximum number of evolution generations is set. When the change rate of the fitness of the optimal individual is lower than for 10 consecutive generations Or terminate the optimization when reaching the maximum generation. The change curve of the MSE of the optimal individual during the training process shows that the MSE drops rapidly in the initial stage (generations 1 - 10), indicating that the GA effectively explores the high - fitness region; in the middle stage (generations 10 - 30), it enters the local optimization stage and the convergence rate slows down; in the later stage (after generation 30), it tends to be stable, proving that the algorithm reaches the global optimum and meets the industrial design requirements. The model is constructed by PyTorch 1.12.1, the genetic algorithm is implemented using custom operators, and all experiments are completed on an NVIDIA RTX 3090 GPU, with the single - training time ≤ 20 seconds.

[0090] Design practice and verification: In this embodiment, the appearance form of the air purifier is disassembled into six basic constituent elements, and each element is further subdivided into multiple form factors. By combining different form factors, a total of 2880 possible design combinations are finally obtained (4×3×5×4×3×4 = 2880). After encoding all these different design combinations, they are imported into the input layer of the constructed GA - BP neural network model for prediction. The form factor combinations with the highest predicted values under the modern, simple, fashionable, and technological indicators are A1, B3, C2, D2, E2, F4, with a predicted value of 0.904; A2, B1, C3, D1, E3, F2, with a predicted value of 0.873; A4, B3, C5, D3, E1, F4, with a predicted value of 0.951; A2, B2, C2, D4, E1, F1, with a predicted value of 0.892. Design schemes are made according to the experimental results of the GA - BP user perceptual evaluation prediction model. The GA - BP model combines the genetic algorithm with the BP neural network to solve the local - optimum problem of traditional neural networks in the multi - dimensional parameter space and enhances the global optimization ability of the model. The genetic algorithm provides the global search ability, while the BP neural network can capture the complex relationships in the data through multi - level non - linear mapping. The application of this method in the form design of air purifier products significantly improves the accuracy of perceptual value prediction and verifies its practical application potential in product design optimization.

[0091] Model performance and design - scheme verification: To verify the prediction ability and practical application effect of the GA - BP neural network model in air purifier design, this method compares the model prediction results with the actual consumer evaluation results. The predicted perceptual values are compared with the actual perceptual means as shown in Table 11, and the relative errors are calculated. The prediction accuracy of the model is relatively high, and the relative errors of all core perceptual words are within 3%. These results prove the effectiveness and reliability of the GA - BP neural network model constructed by this method in air purifier appearance design.

[0092] Table 11 Prediction comparison results To verify the accuracy and practical effect of the design scheme proposed by this method, 30 consumers were invited to conduct a perceptual evaluation of the best design scheme. The content of the questionnaire covered the performance of four core perceptual words, namely modern, simple, fashionable, and technological, in the design scheme. The average values of the evaluation results were 0.913, 0.826, 0.964, and 0.854 respectively. It can be seen that the predicted evaluation values differ very little, and the error range is controlled within ±0.5. This result fully verifies the high consistency between the prediction made through the GA-BP neural network model and the actual consumer feedback, demonstrating the reliability and practicality of this method in product design optimization. Based on this evaluation method, enterprises can carry out the design and research and development of new products more accurately, avoid unnecessary trial-and-error processes, better meet the emotional needs of consumers, reduce R & D risks, and improve market competitiveness. This research result has built a closer bridge between designers and consumers, promoted the effective transformation of consumers' emotional needs, and thus promoted the successful launch of products.

[0093] This embodiment proposes a research method that combines the complex mapping relationship between user perception and product form. Taking an air purifier as an example, it explores how to optimize the appearance design through an effective perceptual word weight assignment method, form analysis, and the GA-BP neural network model. Through the comparative analysis of subjective and objective weight assignment methods and the combination of AHP and CV coefficients, the superiority of this method in actual design is further verified. On this basis, this embodiment can accurately predict user preferences, improve the perceptual value of the design scheme, and provide a theoretical basis and practical reference for the design of other products.

[0094] The main contributions of this embodiment are reflected in method innovation and application value. First, by combining subjective expert assignment and consumer feedback data and adopting a subjective and objective weight assignment method that combines the AHP method and the CV coefficient method, a more stable and scientific basis is provided for weighting perceptual words. Second, by using the GA-BP neural network model, the global optimization ability of the genetic algorithm is combined with the nonlinear mapping ability of the BP neural network to achieve precise optimization of the product appearance design. Compared with traditional methods, the model of this research shows strong advantages in the optimization ability of the multi-dimensional parameter space, effectively solves the problem of local optimum, and improves the accuracy and reliability of the prediction results. Please refer to Figure 3 , Figure 3 which is the structural block diagram of a product appearance optimization device 400 provided by an embodiment of the present application. In this embodiment, each module included in the product appearance optimization device is used to execute Figure 1 - Figure 2 the corresponding steps in the corresponding embodiment. For details, please refer to Figure 1 - Figure 2 and Figure 1 - Figure 2 the relevant descriptions in the corresponding embodiment. For the sake of illustration, only the parts related to this embodiment are shown. SeeFigure 3 , the product appearance optimization device 400 includes: An acquisition module 401, configured to acquire a plurality of perceptual words related to the appearance design of a target product, and construct a judgment matrix corresponding to each of the perceptual words according to the relative importance scores of the perceptual words in the plurality of perceptual words; A first determination module 402, configured to determine a subjective weight value corresponding to a perceptual word according to the judgment matrix when the judgment matrix passes a consistency test; A first construction module 403, configured to construct an evaluation matrix according to the perceptual word scores of each sample product on each of the perceptual words; A second determination module 404, configured to determine an objective weight value of each of the perceptual words according to the evaluation matrix; A screening module 405, configured to screen out a plurality of core perceptual words from the perceptual words according to the subjective weight values and objective weight values of the perceptual words; A second construction module 406, configured to construct a model training set according to a plurality of design combination features of the target product and the plurality of core perceptual words; A training module 407, configured to train an initial neural network model through the model training set to obtain a perceptual evaluation model, where the initial neural network model is composed of a genetic algorithm and a backpropagation neural network; An input module 408, configured to input preselected appearance design features into the perceptual evaluation model to obtain a perceptual evaluation value corresponding to the preselected appearance design features; An adjustment module 409, configured to adjust the preselected appearance design features according to the perceptual evaluation value to optimize the appearance of the target product.

[0095] Through the subjective weight values and objective weight values of the perceptual words, the device screens out a plurality of core perceptual words, enabling the core perceptual words to more accurately reflect the user's perceptual perception of the product appearance. In addition, by training the initial neural network model with the model training set constructed by the design combination features and the core perceptual words to obtain a perceptual evaluation model, inputting the preselected appearance design features into the perceptual evaluation model to obtain the corresponding perceptual evaluation value, and then adjusting the appearance design of the target product according to the perceptual evaluation value, it is possible to predict the user's perception of the product appearance through the perceptual evaluation value output by the perceptual evaluation model, so as to timely improve the product appearance when the perceptual evaluation value is low, improving the efficiency of product appearance optimization.

[0096] It should be understood that Figure 3 in the structural block diagram of the product appearance optimization device shown, each module is used to execute Figure 1 - Figure 2 the respective steps in the corresponding embodiments, and for Figure 1 - Figure 2The steps in the corresponding embodiments have been explained in detail in the above embodiments. For details, please refer to Figure 1 - Figure 2 and Figure 1 - Figure 2 the relevant descriptions in the corresponding embodiments, which will not be elaborated here.

[0097] Figure 4 FIG. is a structural block diagram of a terminal device provided in another embodiment of the present application. As Figure 4 shown, the terminal device 500 in this embodiment includes: a processor 501, a memory 502, and a computer program 503 stored in the memory 502 and executable on the processor 501, such as a program for optimizing the product appearance method. When the processor 501 executes the computer program 503, the steps in each embodiment of the above-mentioned various product appearance optimization methods are implemented, such as Figure 1 the steps S10 to S90 shown. Alternatively, when the processor 501 executes the computer program 503, the functions of each module in the above Figure 3 corresponding embodiments are implemented. For details, please refer to Figure 3 the relevant descriptions in the corresponding embodiments, which will not be elaborated here.

[0098] Exemplarily, the computer program 503 can be divided into one or more units. One or more units are stored in the memory 502 and executed by the processor 501 to complete the present application. One or more units can be a series of computer program instruction segments capable of performing specific functions, and these instruction segments are used to describe the execution process of the computer program 503 in the terminal device 500.

[0099] The terminal device may include, but is not limited to, a processor 501 and a memory 502. Those skilled in the art can understand that Figure 4 this is only an example of the terminal device 500, and does not constitute a limitation on the terminal device 500. It may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, the turntable terminal device may further include an input / output terminal device, a network access terminal device, a bus, etc.

[0100] The so-called processor 501 may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0101] The memory 502 may be an internal storage unit of the terminal device 500, such as the hard disk or memory of the terminal device 500. The memory 502 may also be an external storage terminal device of the terminal device 500, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc. equipped on the terminal device 500. Further, the memory 502 may also include both the internal storage unit of the terminal device 500 and the external storage terminal device. The memory 502 is used to store computer programs and other programs and data required by the turntable terminal device. The memory 502 may also be used to temporarily store data that has been output or is to be output.

[0102] In addition, in each embodiment of the present application, each functional unit may be integrated in a processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit. The above integrated unit may be implemented in the form of hardware or in the form of a software functional unit.

[0103] When an integrated module is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium can be non-volatile or volatile. Based on such an understanding, to implement all or part of the processes in the above-mentioned embodiment methods of this application, it can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable storage medium can include: any entity or device capable of carrying computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0104] The above embodiments are only used to illustrate the technical solutions of this application, rather than to limit them; although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of this application, and should all be included in the protection scope of this application.

Claims

1. A method for optimizing product appearance, characterized in that: The method comprises: Acquire a plurality of perceptual words related to the appearance design of the target product, and construct a judgment matrix corresponding to each perceptual word according to a relative importance score of each perceptual word in the plurality of perceptual words; When the judgment matrix passes the consistency test, determining the subjective weight value of the corresponding perceptual vocabulary according to the judgment matrix; Constructing an evaluation matrix according to the perceptual vocabulary scores of each sample product on each perceptual vocabulary; Determining the objective weight value of each perceptual word according to the evaluation matrix; According to the subjective weight value and the objective weight value of each of the perceptual words, a plurality of core perceptual words are selected from the perceptual words; Constructing a model training set according to the multiple design combination features of the target product and the multiple core perceptual words; The initial neural network model is trained by the model training set to obtain a perceptual evaluation model, wherein the initial neural network model is composed of a genetic algorithm and a back propagation neural network; Inputting the preselected design features into the perceptual evaluation model to obtain perceptual evaluation values ​​corresponding to the preselected design features; The preselected appearance design features are adjusted according to the perceptual evaluation values ​​to optimize the appearance of the target product.

2. The method according to claim 1, characterized in that When the judgment matrix passes the consistency test, determining the subjective weight value of the corresponding perceptual vocabulary according to the judgment matrix includes: Determining the preliminary weight of each perceptual word according to the element product of each row of elements in the judgment matrix; Normalizing the initial weights of the perceptual words to obtain initial weight values ​​corresponding to the perceptual words; According to the initial weight values ​​and judgment matrix corresponding to each perceptual word, the maximum eigenvalue corresponding to the judgment matrix is ​​determined; Determining the consistency index of the judgment matrix according to the maximum eigenvalue value and the matrix order corresponding to the judgment matrix; Determining an adjustment coefficient of the judgment matrix according to the matrix order of the judgment matrix; Determining the consistency ratio of the judgment matrix according to the consistency index and the adjustment coefficient of the judgment matrix; When the consistency ratio is less than a preset ratio, the initial weight value is used as the subjective weight value of the corresponding perceptual vocabulary.

3. The method according to claim 1, characterized in that Determining the objective weight value of each perceptual word according to the evaluation matrix includes: According to the evaluation matrix, determining the mean score and standard deviation of each perceptual vocabulary; Determining the coefficient of variation of each perceptual vocabulary according to the score mean and score standard deviation of each perceptual vocabulary; The coefficient of variation of each of the perceptual words is normalized to obtain the objective weight of each of the perceptual words.

4. The method according to claim 1, characterized in that The method of selecting a plurality of core perceptual words from the perceptual words according to the subjective weight value and the objective weight value of the perceptual words comprises: Determining the coupling degree of each perceptual word according to the subjective weight value and the objective weight value of each perceptual word; Determining the coupling coordination degree of each perceptual word according to the coupling degree of each perceptual word and the preset adjustment factor; The inductive words whose coupling coordination degree is greater than the preset coupling coordination degree are determined as core inductive words.

5. The method according to claim 1, characterized in that The step of constructing a model training set according to the multiple design combination features of the target product and the multiple core perceptual words includes: Obtaining the perceptual values ​​corresponding to the multiple design combination features of the target product on each core vocabulary; Encode each design combination feature to obtain a coded feature; A model training set is constructed based on the encoding features and the corresponding perceptual values.

6. The method according to any one of claims 1 to 5, characterized in that: The genetic algorithm is used to optimize the weights and thresholds of the initial neural network model.

7. The method according to any one of claims 1 to 5, characterized in that: The step of selecting a plurality of core perceptual words from the perceptual words according to the subjective weight value and the objective weight value of the perceptual words further includes: Obtaining a subjective score set and an objective score set related to the target product and a plurality of perceptual words; According to the subjective scoring set, the objective scoring set, the subjective weight value and the objective weight value, a comprehensive weight value of each perceptual vocabulary is calculated by a comprehensive weight calculation formula; Determine the perceptual words whose comprehensive weight value is greater than the preset weight threshold as core perceptual words; The comprehensive weight calculation formula is: In the formula, For the The comprehensive weight value of the perceptual words; is a preset coefficient, the value range is between [0,1], and ; For the The subjective weight of each perceptual word; For the A user's subjective rating; For the An objective rating; is a subjective rating set; is an objective scoring set; is the average score of the subjective rating set; is the average score of the objective rating set; is the variance of the subjective rating set; is the variance of the objective score set; is the minimum score value in the subjective score set and the objective score set; is the maximum score value in the subjective score set and the objective score set.

8. A product appearance optimization device, characterized in that: The device comprises: An acquisition module, used to acquire a plurality of perceptual words related to the appearance design of the target product, and construct a judgment matrix corresponding to each perceptual word according to the relative importance score of each perceptual word in the plurality of perceptual words; A first determination module is used to determine the subjective weight value of the corresponding perceptual vocabulary according to the judgment matrix when the judgment matrix passes the consistency test; The first construction module is used to construct an evaluation matrix according to the perceptual vocabulary scores of each sample product on each perceptual vocabulary; A second determination module is used to determine the objective weight value of each perceptual word according to the evaluation matrix; A screening module, used for screening out a plurality of core perceptual words from the perceptual words according to the subjective weight value and the objective weight value of the perceptual words; A second construction module is used to construct a model training set according to the multiple design combination features of the target product and the multiple core perceptual words; A training module, used to train the initial neural network model through the model training set to obtain a perceptual evaluation model, wherein the initial neural network model is composed of a genetic algorithm and a back propagation neural network; An input module, used to input the pre-selected design features into the perceptual evaluation model to obtain a perceptual evaluation value corresponding to the pre-selected design features; An adjustment module is used to adjust the preselected appearance design features according to the perceptual evaluation value to optimize the appearance of the target product.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images