Ceramic incense burner design method and system based on kansei engineering

Abstract

This application discloses a ceramic incense burner design method and system based on Kansei Engineering, belonging to the field of neural network technology. The method includes: acquiring and preprocessing ceramic incense burner samples to establish a morphological analysis table, which includes morphological feature vectors of multiple ceramic incense burners; establishing multiple positive adjectives based on user review text; performing cloud model parameter matrix operations on the multiple positive adjectives to obtain sentiment word weight vectors and generate a Kansei evaluation matrix table; inputting the morphological analysis table into a preset TCN-Transformer-GRU model to obtain the optimal product morphological design combination; and inputting the optimal product morphological design combination into a preset stable diffusion model to obtain a ceramic incense burner rendering. This invention efficiently expresses design information through cross-modal generation technology and uses quantitative methods for evaluation and optimization, resulting in more realistic and refined ceramic incense burner images.

Description

Technical Field

[0001] This application belongs to the field of network security technology, specifically relating to a design method and system for ceramic incense burners based on Kansei Engineering. Background Technology

[0002] Ceramic incense burners, as an important symbol of traditional Chinese culture, have a development history spanning various important periods in Chinese history. They hold significant artistic and collectible value, and their evolving forms reflect the development and changes in ancient Chinese society and culture, varying greatly due to the influence of time, culture, form, and craftsmanship. With the rapid advancement of industrialization and the evolution of modern lifestyles, people's aesthetic concepts and cultural needs are also undergoing continuous changes.

[0003] However, existing technologies have the following problems: First, user needs are not accurately captured, and it is difficult to quantify users' detailed preferences for ceramic incense burners, resulting in low realism of the generated ceramic incense burner images; Second, the mapping efficiency of emotion and design features is low, traditional machine learning models are insufficient in capturing non-linear relationships and ignore temporal emotional changes, resulting in low refinement of the generated ceramic incense burner images. Summary of the Invention

[0004] The purpose of this application is to provide a ceramic incense burner design method and system based on kinetic engineering, which can solve the problem of low realism and refinement of ceramic incense burner images generated by existing technologies.

[0005] To solve the above-mentioned technical problems, this application is implemented as follows:

[0006] In a first aspect, embodiments of this application provide a ceramic incense burner design method based on kinetic engineering, the method comprising:

[0007] Obtain and preprocess ceramic incense burner samples to establish a morphological analysis table, which includes morphological feature vectors of multiple ceramic incense burners.

[0008] Multiple positive adjectives were created based on user comment text;

[0009] Multiple positive adjectives are used to perform cloud model parameter matrix operations to obtain sentiment word weight vectors, and a Kansei evaluation matrix table is generated based on the sentiment word weight vectors and morphological analysis table.

[0010] Input the Kansei evaluation matrix into the preset TCN-Transformer-GRU model to obtain the optimal product form design combination;

[0011] The optimal product form design combination is input into the preset stable diffusion model to obtain the ceramic incense burner rendering. The stable diffusion model consists of a sketch encoder layer, a diffusion layer, a deconvolution decoder layer, and a style transfer layer connected in sequence.

[0012] As an optional implementation of the first aspect of this application, the process of obtaining sentiment word weight vectors by performing cloud model parameter matrix operations on multiple positive adjectives includes:

[0013] Multiple positive adjectives and a preset user rating matrix are input into the cloud model to calculate cloud model parameters and membership degree generation, and the cloud model parameter matrix is ​​obtained.

[0014] The entropy weight vector of sentiment words is obtained by calculating the cloud model parameter matrix and the preset user rating matrix using the entropy weight method.

[0015] As an optional implementation of the first aspect of this application, the process of inputting the Kansei evaluation matrix table into a preset TCN-Transformer-GRU model to obtain the optimal product form design combination includes:

[0016] The Kansei evaluation matrix table is input into the TCN layer for feature extraction to obtain the temporal feature tensor.

[0017] The temporal feature tensor is input into the Transformer for multi-head attention mechanism and feedforward neural network processing to obtain the globally optimized feature tensor;

[0018] The global optimization feature tensor is input into the GRU for gated recurrent and fully connected layer mapping to obtain multiple sentiment word weight vectors;

[0019] Obtain the sample index of the highest sentiment word among multiple sentiment word weight vectors, and obtain the optimal product form design combination based on the sample index.

[0020] As an optional implementation of the first aspect of this application, the process of inputting the Kansei evaluation matrix table into the TCN layer for feature extraction and obtaining the temporal feature tensor includes:

[0021] Serialize the Kansei evaluation matrix table to obtain the Kansei sequence;

[0022] The Kansei sequence is dilated and convolved based on the kernel weights and a preset dilation factor to obtain the temporal feature tensor.

[0023] As an optional implementation of the first aspect of this application, the process of inputting a temporal feature tensor into a Transformer for multi-head attention mechanism and feedforward neural network processing to obtain a globally optimized feature tensor includes:

[0024] The temporal feature tensor is input into the Transformer for multi-head attention mechanism and feedforward neural network processing to obtain the globally optimized feature tensor. The encoder in the Transformer includes multiple identical decoder layers. Each decoder layer includes multiple multi-head attention layers and feedforward neural network layers. The number of decoder layers is equal to the number of encoder layers. Each decoder layer and encoder layer contains residual connection layers and normalization layers.

[0025] As an optional implementation of the first aspect of this application, the process of inputting the globally optimized feature tensor into a GRU for gated loop and fully connected layer mapping to obtain multiple sentiment word weight vectors includes:

[0026] The global optimization feature tensor is input into the update gate to obtain the update gate vector at the current time step;

[0027] The candidate state vector for the current time step is obtained based on the globally optimized feature tensor.

[0028] The output vectors for multiple current time steps are obtained based on the update gate vector and candidate state vector;

[0029] The output vectors of multiple current time steps are input into a fully connected layer for vector transformation to obtain multiple sentiment word weight vectors.

[0030] As an optional implementation of the first aspect of this application, the specific process of inputting the optimal product form design combination into a preset stable diffusion model to obtain a ceramic incense burner rendering includes:

[0031] A sketch of the ceramic incense burner's shape was obtained based on the morphological analysis table;

[0032] The sketch is input into the sketch encoder layer of the preset stable diffusion model for convolution and vector processing to obtain the latent space feature tensor;

[0033] The latent space feature tensor is input into the diffusion layer for forward diffusion, reverse denoising, and adaptive attention mechanism processing to obtain the denoised latent space tensor;

[0034] The denoised latent space tensor is input into the deconvolution decoder layer for deconvolution processing to obtain the initial rendering image of the ceramic incense burner.

[0035] The initial rendering of the ceramic incense burner is input into the style transfer layer for style feature fusion processing to obtain the final rendering of the ceramic incense burner.

[0036] Secondly, embodiments of this application provide a ceramic incense burner design system based on kinetic engineering, the system comprising:

[0037] The morphological analysis table creation module is used to acquire and preprocess ceramic incense burner samples to create morphological analysis tables.

[0038] The positive adjective generation module is used to generate multiple positive adjectives based on user comment text.

[0039] The Kansei evaluation matrix generation module is used to perform cloud model parameter matrix operations on multiple positive adjectives to obtain sentiment word weight vectors, and generate Kansei evaluation matrix tables based on sentiment word weight vectors and morphological analysis tables.

[0040] The TCN-Transformer-GRU module is used to input the Kansei evaluation matrix into the preset TCN-Transformer-GRU model to obtain the optimal product form design combination.

[0041] The stable diffusion module is used to input the optimal product form design combination into the preset stable diffusion model to obtain the ceramic incense burner rendering.

[0042] Thirdly, embodiments of this application provide an electronic device, which includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor. When the program or instructions are executed by the processor, they implement the steps of the method described in the first aspect.

[0043] Fourthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0044] Compared with existing technologies, this invention proposes a creative design method for ceramic incense burners based on kansen mechanics. This method combines cloud models with entropy weighting, organically unifying the fuzziness of emotional needs with the objectivity of weight allocation, enabling the scientific and precise capture of users' Kansei needs. This approach has significant advantages in handling diverse and complex design tasks, providing strong theoretical and technical support for emotion-driven innovative design. Furthermore, this invention proposes a TCN-Transformer-GRU model, which complements traditional methods in temporal modeling, global feature extraction, and detail optimization. This model not only solves the problem of insufficient nonlinear feature capture in traditional methods but also significantly improves the accuracy and stability of the mapping between user emotions and morphological features, providing a new approach and practical tool for emotion-driven creative design. Finally, this invention proposes a stable diffusion model to transform the mapping results of user emotions and design features into high-quality two-dimensional renderings, achieving an intuitive output from data to design, significantly reducing manual intervention steps, shortening the design cycle, and improving the design efficiency of ceramic incense burners. Attached Figure Description

[0045] Figure 1 This is a flowchart of a ceramic incense burner design method based on kinetic engineering provided in the first embodiment of this application;

[0046] Figure 2 This is a flowchart of the Transformer model provided in the first embodiment of this application;

[0047] Figure 3 This is a flowchart of the GRU model provided in the first embodiment of this application;

[0048] Figure 4 This is a flowchart illustrating the stable diffusion model provided in the first embodiment of this application;

[0049] Figure 5 This is a dataset of ceramic incense burner images provided in the first embodiment of this application;

[0050] Figure 6 This is a morphological analysis table provided in the first embodiment of this application;

[0051] Figure 7 This is a network relationship diagram between adjectives provided in the first embodiment of this application;

[0052] Figure 8 This is an internal structural diagram of a ceramic incense burner design system based on Kansei Engineering, provided in the second embodiment of this application. Detailed Implementation

[0053] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0054] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0055] The following description, in conjunction with the accompanying drawings, details a ceramic incense burner design method and system based on kinetic engineering, provided by this application, through specific embodiments and application scenarios.

[0056] Example 1

[0057] Please see Figure 1 The flowchart below shows a ceramic incense burner design method based on kinetic engineering proposed in the first embodiment of this application. The method includes steps S1 to S5.

[0058] Step S1: Obtain and preprocess ceramic incense burner samples to establish a morphological analysis table, which includes morphological feature vectors of multiple ceramic incense burners.

[0059] Specifically, the process of acquiring and preprocessing ceramic incense burner samples to establish a morphological analysis table includes: collecting and filtering different ceramic incense burner images from books and the internet, ultimately retaining a dataset of 100 representative ceramic incense burner images with uniform background size. Based on this, the form of the ceramic incense burners is further subdivided into 5 parts and 6 type features, and a morphological analysis table is established. Figure 6 The table shows the morphological analysis.

[0060] Step S2: Create multiple positive adjectives based on user comment text.

[0061] Specifically, user comment text is collected through web crawlers. After data cleaning, word segmentation, and part-of-speech tagging, the online comment text data is tagged with part-of-speech tags to obtain all adjectives. Invalid or similar adjectives are removed and categorized. The network relationships between adjectives are listed, and the TF-IDF value of each adjective is obtained and sorted from high to low according to the value. The top seven adjectives with the highest TF-IDF values ​​are selected as positive adjectives. Figure 7 A network diagram showing the relationships between adjectives.

[0062] Step S3: Perform cloud model parameter matrix operations on multiple positive adjectives to obtain sentiment word weight vectors, and generate a Kansei evaluation matrix table based on the sentiment word weight vectors and the morphological analysis table.

[0063] Specifically, cloud model parameter matrix for:

[0064] ;

[0065] in, This indicates the user evaluations collected in the survey on seven emotional terms, namely, the types of design options or the number of user evaluations. This represents the 7 high-value sentiment words selected through TF-IDF filtering. Each word serves as an evaluation indicator, and the sentiment score given by users to each sample constitutes the indicator value.

[0066] Furthermore, based on the cloud model parameter matrix, the expected value, average absolute deviation, and hyperentropy value of the user's subjective evaluation sample data are obtained, among which...

[0067] Expected value The mathematical expression is:

[0068] ;

[0069] The average of absolute deviations The mathematical expression is:

[0070] ;

[0071] hyperentropy The mathematical expression is:

[0072] ;

[0073] Furthermore, based on the expected value The average of absolute deviations and hyperentropy Obtaining standardized proportions Its mathematical expression is:

[0074] = + + ;

[0075] Furthermore, the information entropy of the indicator is obtained based on the standardized proportions, and its mathematical expression is:

[0076] ;

[0077] In the above formula, Indicates the first j The information entropy of the indicator. Represents the normalization coefficient. Indicates the first The sample at the th The standardized proportion of the indicators Represents the cloud model parameter matrix Elements within, This represents the expected average value.

[0078] Furthermore, based on the information entropy of the indicators Obtaining information effect value Its mathematical expression is:

[0079] ;

[0080] Furthermore, based on the information effect value, the first... Item indicator weights Its mathematical expression is:

[0081] ;

[0082] Furthermore, according to the first Obtaining the sentiment word weight vector from the indicator weights of the items Its mathematical expression is:

[0083] ;

[0084] Furthermore, the weight vector of each sentiment word is... Combine them to construct the Kansei evaluation matrix table. ].

[0085] Step S4: Input the Kansei evaluation matrix into the preset TCN-Transformer-GRU model to obtain the optimal product form design combination.

[0086] Specifically, the process of obtaining the optimal product form design combination includes: using the Kansei evaluation matrix table. The input is fed into the TCN layer for feature extraction to obtain a temporal feature tensor; the temporal feature tensor is then fed into a Transformer for multi-head attention and feedforward neural network processing to obtain a globally optimized feature tensor; the globally optimized feature tensor is then fed into a GRU for gated recurrent loops and fully connected layer mapping to obtain multiple sentiment word weight vectors; finally, the sample index with the highest sentiment word weight vector is obtained, and the optimal product form design combination is obtained based on the sample index. Figure 3 This represents the GRU model diagram.

[0087] Among them, the Kansei evaluation matrix table The specific steps for inputting the data into the TCN layer for feature extraction and obtaining the temporal feature tensor include: inputting the weighted score data containing 7 sentiment words in the Kansei evaluation matrix table and the morphological feature vector of the ceramic incense burner (based on 5 types of design feature encoding generated from the morphological analysis table) into the dilated causal convolutional layer in the TCN layer for temporal feature extraction, capturing the local dependency relationship between sentiment and morphological features, generating a high-dimensional feature sequence, and obtaining the temporal feature tensor. The dimension of the temporal feature tensor is the product of the number of samples and the dimension of the hidden layer.

[0088] To capture long-term dependencies, the TCN layer of this invention employs dilated convolution. By increasing the spacing between elements in the convolution kernel, dilated convolution expands the receptive field of the convolutional layer, enabling the network to cover longer input sequences without significantly increasing computational complexity. Dilated convolution introduces a dilation factor. This is used to increase the spacing between convolutional kernel weights, thereby expanding the receptive field of the convolutional layer and capturing long-term dependencies. The operation of dilated convolution can be represented as:

[0089] ;

[0090] in, This represents the expansion factor, which can cover a longer range of input sequences without significantly increasing the number of parameters. This represents the output of the dilated convolution. Represents the sentiment word weight vector. Indicates the width of the convolution kernel. Represents the dynamic weights of emotional features.

[0091] Obtain the output of dilated convolution Then, the output of the dilated convolution is... Perform residual processing to obtain temporal feature tensors. Its mathematical expression is:

[0092] ；

[0093] in, Represents the temporal feature tensor. This represents the activation function operation. This represents the output of the convolutional layer. This represents the input to the residual layer.

[0094] Specifically, this invention inputs the temporal feature tensor into a Transformer to perform multi-head attention mechanism and feedforward neural network processing to obtain a globally optimized feature tensor. The encoder in the Transformer includes multiple identical decoder layers. Each decoder layer includes multiple layers composed of multi-head attention layers and feedforward neural network layers. The number of decoder layers is equal to the number of encoder layers. Each decoder layer and encoder layer contains a residual connection layer and a normalization layer.

[0095] The Transformer model of this invention consists of two main parts: an encoder on the left and a decoder on the right. The encoder consists of six identical layers, each containing two sub-layers: a multi-head self-attention mechanism and a fully connected feedforward neural network. The decoder replicates this structure, also with six identical layers; however, each layer in the decoder contains three sub-layers. The first two sub-layers are similar to those in the encoder, while the third sub-layer focuses on the attention mechanism between the encoding and decoding processes. Residual connections and layer normalization are applied in each sub-layer of the Transformer. The Transformer model of this invention employs an attention mechanism that includes three vectors: a query vector, a key vector, and a value vector. The query vector represents the feature matrix of the query, the key vector represents the feature matrix of the key, and the value vector represents the feature matrix of the value. Each weight is determined by normalization using a softmax function and then used to generate the output. Figure 2 This represents the workflow diagram of the Transformer model.

[0096] When the temporal feature tensor is input into the Transformer for multi-head attention mechanism and feedforward neural network processing to obtain the globally optimized feature tensor, the following formula is used for calculation:

[0097] ;

[0098] in, This represents the global optimization feature tensor. This represents the dimension of the key vector in each head. Represents the query vector. Represents the key vector. Indicates matrix transpose. express Activation function.

[0099] Furthermore, the process of obtaining the sample index with the highest sentiment word weight vector among multiple sentiment word weight vectors and obtaining the optimal product form design combination based on the sample index includes: inputting the global optimization feature tensor into a GRU for processing through gated recurrent units and fully connected layers to obtain the sentiment word weight vector; updating the hidden state step by step according to the sequence step size to model the dynamic mapping relationship between sentiment and design features and mapping the final hidden state to the sentiment value score (range 0~1) to quantify the matching degree between user sentiment needs and design features and obtain the sentiment fit of each design combination; finally, the optimal design combination is screened, selecting the sample index with the highest sentiment value from the sentiment word weight vector, and outputting the optimal form design combination by encoding the combination of the 5 types of design features in the form analysis table.

[0100] Step S5: Input the optimal product form design combination into the preset stable diffusion model to obtain the ceramic incense burner rendering. The stable diffusion model consists of a sketch encoder layer, a diffusion layer, a deconvolution decoder layer, and a style transfer layer connected in sequence.

[0101] Specifically, the process of obtaining a ceramic incense burner rendering includes: based on the optimal form design combination and corresponding form analysis table, drawing a sketch of the ceramic incense burner form based on the optimal form design combination; inputting the sketch (based on the optimal form design combination) and prompt words into the sketch encoding module of the stable diffusion model; processing through a convolutional encoder and conditional latent code vector to output a latent space feature tensor, fusing sketch visual features with semantic information of design parameters; inputting the latent space feature tensor into the diffusion process module for forward diffusion, inverse denoising, and adaptive attention mechanisms, performing noise iteration and detail generation, outputting a denoised latent space tensor; inputting the denoised latent space tensor into the decoding rendering module for deconvolutional decoder, applying a style transfer layer and an enhanced super-resolution generative adversarial network to perform latent space reconstruction and image synthesis, outputting a ceramic incense burner rendering. The workflow diagram of the stable diffusion model provided by this invention can be found in [link to relevant documentation]. Figure 4 As shown.

[0102] The core idea of ​​the stable diffusion model used in this invention for rendering ceramic incense burners is to simulate the data generation process, gradually adding noise and then learning the reverse process to remove the noise, thereby restoring the original data. The stable diffusion model generates data by gradually introducing and removing noise in the reverse process. This process includes two main stages: forward diffusion and reverse diffusion, both of which are parameterized Markov chains. In the forward diffusion stage, the model gradually adds noise to the data, transforming it from an ordered state to an almost random noisy state. This process follows a predetermined path, ultimately converting the data into a Gaussian noise distribution. (Depending on time...) Result calculation time As a result, with With the addition of Gaussian noise, the image will become closer to pure noise. For the original data, each step of the diffusion process adds Gaussian noise to the data obtained in the previous step according to a formula, the mathematical expression of which is:

[0103] ;

[0104] ;

[0105] in, Indicates time step Data, Indicates from time step 1 to time step 2. Data sequence, Indicates time step The noise figure is used to control the noise level. Indicates given initial data At that time, from time step 1 to T The joint probability distribution of the diffusion sequences, Indicates time step Data, The identity matrix is ​​used to describe the covariance structure of noise.

[0106] Furthermore, back diffusion is the core of the stable diffusion model of this invention, and the mathematical expression for back diffusion is:

[0107] ;

[0108] in, Indicates time step Data, Represents the noise scheduling coefficient. Represents the noise scheduling coefficient. This represents the initial data.

[0109] During the backdiffusion process, the neural network is trained to adapt to this distribution. This invention employs a U-net network structure, where the reverse process is defined as a Markov chain, consisting of a series of Gaussian distributions parameterized by the neural network, mathematically expressed as:

[0110] ;

[0111] ;

[0112] in, Indicates starting from the initial data The entire sequence of noisy data, The initial distribution for back diffusion is usually set as a standard Gaussian distribution. , For the backdiffusion conditional probability parameterized in the neural network, This represents the mean of the neural network predictions. Let represent the covariance matrix.

[0113] The training portion of the diffusion model of this invention consists of the following steps: First, an input image is captured, and then from 1 to... The first step is to randomly select a time point within a given time frame. The second step is to extract noise samples from a standard normal distribution. The ultimate goal is to optimize the model parameters to reduce the difference between the noise predicted by the model and the actual sampled noise.

[0114] This invention focuses on ceramic incense burners for the Chinese market and further explains the method of this invention.

[0115] A ceramic incense burner image dataset was constructed by searching various shopping websites, books, and museums for relevant images. Images of "ceramic incense burners" were collected from search engines on major shopping platforms and museums using keywords, and duplicate images were filtered. Then, Photoshop was used to remove the background and add a white background. The images were then uniformly adjusted to 600x600mm using the UU online tool and converted to JPG format. Finally, a high-quality dataset containing 200 ceramic incense burner images was successfully constructed. Morphological analysis was then used to decompose the shapes of these ceramic incense burners into lids, necks, handles, bodies, and feet, classifying them into six types. Figure 5 A dataset representing images of ceramic incense burners.

[0116] Taking the appearance and design reviews of ceramic incense burners as the research object, this study aims to obtain more accurate emotional needs from customers regarding ceramic incense burners. Using web crawling technology, this research crawled customer reviews of ceramic incense burner products from e-commerce platforms such as Taobao, JD.com, and Tmall. Based on word segmentation and part-of-speech tagging methods, considering computational efficiency and the presence of some text repetition, 7111 text data entries were obtained, totaling 703334 words. The data was then classified and processed according to part-of-speech. First, the data was categorized into four parts of speech: nouns, adjectives, verbs, and adverbs. Then, TF-IDF values ​​were calculated for the crawled reviews, and the top seven emotional words were selected, such as: exquisite, special, light, cute, solemn, meticulous, and beautiful. As shown in Table 1, the weights of these words were calculated using the cloud model-entropy weight method. The frequency of each word in the dataset was statistically analyzed and converted into a membership degree (between 0 and 1). For each emotional word, a curve showing the change in its membership degree with a certain attribute (such as rating) was plotted, as shown in Figure 1. Each subgraph represents the change in membership degree of an emotion term. Next, for each emotion term, the entropy value of its membership degree distribution is calculated (the smaller the entropy value, the more concentrated the membership degree distribution of the term, and the more certain the information). The difference coefficient is 1 minus the entropy value. Based on the difference coefficient, the weight of each emotion term is calculated (the larger the weight, the more important the term is in the evaluation). Finally, the weight curves of all emotion terms are plotted (to visually demonstrate the importance of each term). The top seven emotion terms ranked in this invention are shown in Table 1.

[0117] Table 1: Top Seven Emotional Vocabulary

[0118]

[0119] One hundred high-quality images were selected from an existing image database of ceramic incense burners as samples. Weights were calculated based on the steps outlined above. The top three sentiment words with the highest weights were combined with the samples for a questionnaire survey. One hundred participants with a design background were invited to rate the images on a five-step Likert scale. The average values ​​were used to construct the Kansei evaluation matrix, as shown in Table 2.

[0120] Table 2 Kansei Evaluation Matrix

[0121]

[0122] The TCN and Transformer branches are merged through an additive layer, then concatenated with the final output of the GRU branch, and finally mapped to the output through a fully connected layer. The model training optimizer selected in this invention is Adam, a widely used optimization algorithm in deep learning that has demonstrated good performance and applicability in various machine learning tasks. The initial learning rate is 0.001, decreasing in stages to 0.1 times every 20 training epochs to balance convergence speed and stability. The L2 regularization coefficient is 0.0001, and the gradient threshold is 1 to prevent gradient explosion. The maximum number of training epochs selected in this model is 30. After each training epoch, the data is randomly shuffled to enhance generalization ability. Data normalization is performed as a preprocessing step; both input and output are normalized to [0,1]. After each training epoch, the data is randomly shuffled to eliminate dimensional differences.

[0123] This invention utilizes a deep neural network prediction model to calculate all possible combinations of optimal morphological features for ceramic incense burners, totaling 7776. By performing Kansei prediction on these combinations, the features most closely matching the emotional labels were identified as "lightweight," "delicate," and "unique." Specifically, the numerical combination [6,6,4,6,5] was associated with the highest score for "lightweight," the combination [1,1,1,1,1] achieved the highest score for "delicate," and the combination [6,5,6,5,6] performed best in the "unique" emotional score. These predicted combinations provide designers with new design ideas and, through scientific data calculation, reduce the complexity of the design process, greatly improving design efficiency. Furthermore, this method also enables traditional handicrafts to better meet customers' emotional and aesthetic needs, bringing new vitality to the future development of handicrafts.

[0124] Designers sketched the predicted optimal shape combination and incorporated it into the stable diffusion model proposed in this invention. A combination of a large model and auxiliary models was then used to generate a complete ceramic incense burner image. The generated image is elegant and unique in shape, with smooth and lively lines, giving the ceramic incense burner a more vivid visual effect. Finally, artisans, inspired by these creative images, collaborated through a series of processes including kneading, shaping, coloring, and firing to create ceramic incense burner products that customers would love. Based on the prediction results, selecting the ceramic incense burner combination with the highest score leads to the following conclusions:

[0125] The "lightweight" form evokes a sense of harmony and ease. For example, the body's lines possess a graceful, flowing beauty, while the design of the handles conveys a sense of solemnity and dignity. Furthermore, the proportions of each part are perfectly balanced; the height of the body, the size of the handles, and the dimensions of the lid are all harmoniously integrated, without any sense of bulkiness. This lightweight design not only aligns with modern aesthetics' pursuit of simplicity and agility but also imbues the ceramic incense burner with a unique character. The handles, lid, and body of this ceramic incense burner are meticulously crafted. The lines of the handles are smooth and natural, seamlessly connecting with the body; the edges of the lid are neat and smooth, fitting perfectly with the opening of the burner. These meticulous designs reflect the craftsman's rigorous standards for quality and pursuit of detail, giving the entire incense burner an exquisite and elegant quality. The "unique" form breaks away from the conventional shapes of traditional ceramic incense burners. Its body adopts a unique cylindrical design, replacing the common round or square shapes, creating a striking visual effect. This innovative design not only gives the incense burner a unique visual effect but also adds a touch of modern style. For example, the combination of straight lines and curves on the body creates both the simplicity and elegance of straight lines and the softness and grace of curves. The traditional shape of the handles echoes the modern body, forming a striking contrast and fusion. The six best combinations with the highest scores for each adjective were selected and sketched to generate a realistic 2D rendering of the ceramic product.

[0126] The beneficial effects of the ceramic incense burner design method based on Kansei Engineering proposed in this invention are as follows: First, by establishing a Kansei evaluation matrix, this invention organically unifies the ambiguity of emotional needs with the objectivity of weight allocation, enabling the scientific and accurate capture of users' Kansei needs. This method has significant advantages in handling design tasks with diverse and complex user needs, providing strong theoretical and technical support for emotion-driven innovative design. Second, the TCN-Transformer-GRU model proposed in this invention complements each other in terms of time modeling, global feature extraction, and detail optimization. It not only solves the problem of insufficient nonlinear feature capture in traditional methods but also significantly improves the accuracy and stability of the mapping between user emotions and morphological features, providing a new approach and practical tool for emotion-driven creative design. Third, this invention utilizes a stable diffusion model to transform the mapping results of user emotions and design features into high-quality two-dimensional renderings, achieving intuitive output from data to design, significantly reducing manual intervention steps, shortening the design cycle, and improving design efficiency.

[0127] Example 2

[0128] Please see Figure 8 The image shows a ceramic incense burner design system based on kinetic engineering, as proposed in the second embodiment of this application. The system includes:

[0129] The morphological analysis table creation module is used to acquire and preprocess ceramic incense burner samples to create morphological analysis tables.

[0130] The positive adjective generation module is used to generate multiple positive adjectives based on user comment text.

[0131] The Kansei evaluation matrix generation module is used to perform cloud model parameter matrix operations on multiple positive adjectives to obtain sentiment word weight vectors, and generate Kansei evaluation matrix tables based on sentiment word weight vectors.

[0132] The TCN-Transformer-GRU module is used to input the Kansei evaluation matrix into the preset TCN-Transformer-GRU model to obtain the optimal product form design combination.

[0133] The stable diffusion module is used to input the optimal product form design combination into the preset stable diffusion model to obtain the ceramic incense burner rendering.

[0134] The beneficial effects of the ceramic incense burner design system based on kinesiology provided by this invention are as follows: This invention utilizes a Kansei evaluation matrix generation module to perform cloud model parameter matrix operations on multiple positive adjectives to obtain sentiment word weight vectors, and generates a Kansei evaluation matrix table based on these sentiment word weight vectors. This organically unifies the ambiguity of sentiment needs with the objectivity of weight allocation, enabling the scientific and accurate capture of users' Kansei needs. This method has significant advantages in handling design tasks with diverse and complex user needs, providing strong theoretical and technical support for emotion-driven innovative design. Furthermore, this invention utilizes a TCN-Transformer-GRU module to input a morphological analysis table into a preset TCN-Transformer-GRU model to obtain the optimal product morphological design combination. This achieves complementary advantages in time modeling, global feature extraction, and detail optimization, solving the problem of insufficient nonlinear feature capture in traditional methods and significantly improving the accuracy and stability of the mapping between user emotions and morphological features, providing a new approach and practical tool for emotion-driven creative design. The present invention also uses a stable diffusion module to input the optimal product form design combination into a preset stable diffusion model to obtain a ceramic incense burner rendering. This enables an intuitive output from data to design, significantly reducing the steps of manual intervention, shortening the design cycle, and improving design efficiency.

[0135] The ceramic incense burner design system based on Kansei Engineering in this application embodiment can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device or a non-mobile electronic device. For example, mobile electronic devices can be mobile phones, tablets, laptops, PDAs, in-vehicle electronic devices, wearable devices, ultra-mobile personal computers (UMPCs), netbooks, or personal digital assistants (PDAs), etc., while non-mobile electronic devices can be servers, network-attached storage (NAS), personal computers (PCs), televisions (TVs), ATMs, or self-service machines, etc. This application embodiment does not impose specific limitations.

[0136] The ceramic incense burner design system based on Kansei Engineering in this application embodiment can be a device with an operating system. This operating system can be Android, iOS, or other possible operating systems; this application embodiment does not specifically limit it.

[0137] This application provides a ceramic incense burner design system based on kinesiology, which can achieve... Figures 1 to 7 The various processes of a ceramic incense burner design method based on kinetic engineering in the method embodiment are not described in detail here to avoid repetition.

[0138] Optionally, this application embodiment also provides an electronic device, including a processor, a memory, and a program or instructions stored in the memory and executable on the processor. When the program or instructions are executed by the processor, they implement the various processes of the above-described embodiment of a ceramic incense burner design method based on kinetic engineering, and can achieve the same technical effect. To avoid repetition, they will not be described again here.

[0139] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described embodiment of a ceramic incense burner design method based on kinetic engineering, and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0140] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0141] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0142] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0143] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A design method for ceramic incense burners based on kinetic engineering, characterized in that, include: A morphological analysis table is established by acquiring and preprocessing ceramic incense burner samples, and the morphological analysis table includes morphological feature vectors of multiple ceramic incense burners. Multiple positive adjectives were created based on user comment text; The cloud model parameter matrix operation is performed on multiple positive adjectives to obtain sentiment word weight vectors, and a Kansei evaluation matrix table is generated based on the sentiment word weight vectors and the morphological analysis table. The Kansei evaluation matrix is ​​input into a preset TCN-Transformer-GRU model to obtain the optimal product form design combination, including: The Kansei evaluation matrix table is input into the TCN layer for feature extraction to obtain the temporal feature tensor. The temporal feature tensor is input into a Transformer to perform multi-head attention mechanism and feedforward neural network processing to obtain a globally optimized feature tensor; The global optimization feature tensor is input into a GRU for gated loop and fully connected layer mapping to obtain multiple sentiment word weight vectors; Obtain the sample index of the highest sentiment word among multiple sentiment word weight vectors, and obtain the optimal product form design combination based on the sample index; The optimal product form design combination is input into a preset stable diffusion model to obtain a ceramic incense burner rendering. The stable diffusion model consists of a sketch encoder layer, a diffusion layer, a deconvolution decoder layer, and a style transfer layer connected in sequence.

2. The ceramic incense burner design method based on kinetic engineering according to claim 1, characterized in that, The process of obtaining sentiment word weight vectors by performing cloud model parameter matrix operations on multiple positive adjectives includes: Multiple positive adjectives and a preset user rating matrix are input into the cloud model to calculate cloud model parameters and membership degree generation, thereby obtaining the cloud model parameter matrix. The cloud model parameter matrix and the preset user rating matrix are used to calculate the entropy weight method to obtain the sentiment word weight vector.

3. The ceramic incense burner design method based on kinetic engineering according to claim 2, characterized in that, The process of inputting the Kansei evaluation matrix table into the TCN layer for feature extraction and obtaining the temporal feature tensor includes: The Kansei evaluation matrix table is serialized to obtain the Kansei sequence; The Kansei sequence is subjected to dilated convolution operation based on the convolution kernel weights and a preset dilation factor to obtain the temporal feature tensor.

4. The ceramic incense burner design method based on kinetic engineering according to claim 2, wherein the process of inputting the temporal feature tensor into a Transformer for multi-head attention mechanism and feedforward neural network processing to obtain the globally optimized feature tensor includes: The temporal feature tensor is input into the Transformer for multi-head attention mechanism and feedforward neural network processing to obtain the globally optimized feature tensor. The encoder in the Transformer includes multiple identical decoder layers. Each decoder layer includes multiple multi-head attention layers and feedforward neural network layers. The number of decoder layers is equal to the number of encoder layers. Each decoder layer and encoder layer contains a residual connection layer and a normalization layer.

5. The ceramic incense burner design method based on kinetic engineering according to claim 2, characterized in that, The process of inputting the global optimization feature tensor into a GRU for gated recurrent loops and fully connected layer mapping to obtain multiple sentiment word weight vectors includes: The global optimization feature tensor is input into the update gate to obtain the update gate vector for the current time step; The candidate state vector for the current time step is obtained based on the global optimization feature tensor. Based on the update gate vector and the candidate state vector, obtain the output vectors of multiple current time steps; The output vectors of the multiple current time steps are input into the fully connected layer for vector transformation processing to obtain multiple sentiment word weight vectors.

6. The ceramic incense burner design method based on kinetic engineering according to claim 1, characterized in that, The specific process of inputting the optimal product form design combination into a preset stable diffusion model to obtain a ceramic incense burner rendering includes: A sketch of the ceramic incense burner's shape was obtained based on the aforementioned morphological analysis table. The sketch is input into the sketch encoder layer of a preset stable diffusion model for convolution and vector processing to obtain the latent spatial feature tensor. The latent space feature tensor is input into the diffusion layer for forward diffusion, reverse denoising, and adaptive attention mechanism processing to obtain the denoised latent space tensor; The denoised latent space tensor is input into the deconvolution decoder layer for deconvolution processing to obtain the initial rendering image of the ceramic incense burner. The initial rendering of the ceramic incense burner is input into the style transfer layer for style feature fusion processing to obtain the ceramic incense burner rendering.

7. A ceramic incense burner design system based on kinetic engineering, characterized in that, The system includes: The morphological analysis table creation module is used to acquire and preprocess ceramic incense burner samples to create morphological analysis tables. The positive adjective generation module is used to generate multiple positive adjectives based on user comment text. The Kansei evaluation matrix generation module is used to perform cloud model parameter matrix operations on multiple positive adjectives to obtain sentiment word weight vectors, and generate a Kansei evaluation matrix based on the sentiment word weight vectors and the morphological analysis table. The TCN-Transformer-GRU module is used to input the Kansei evaluation matrix table into a preset TCN-Transformer-GRU model to obtain the optimal product form design combination, including: The Kansei evaluation matrix table is input into the TCN layer for feature extraction to obtain the temporal feature tensor. The temporal feature tensor is input into a Transformer to perform multi-head attention mechanism and feedforward neural network processing to obtain a globally optimized feature tensor; The global optimization feature tensor is input into a GRU for gated loop and fully connected layer mapping to obtain multiple sentiment word weight vectors; Obtain the sample index of the highest sentiment word among multiple sentiment word weight vectors, and obtain the optimal product form design combination based on the sample index; The stable diffusion module is used to input the optimal product form design combination into a preset stable diffusion model to obtain a rendering of the ceramic incense burner.

8. An electronic device, characterized in that, It includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein when the program or instructions are executed by the processor, they implement the steps of a ceramic incense burner design method based on kinetic engineering as described in any one of claims 1-6.

9. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of a ceramic incense burner design method based on kinetic engineering as described in any one of claims 1-6.

Images