Tea space color matching method and device based on user preference, equipment and storage medium

Abstract

The invention discloses a tea space color matching method, device and equipment based on user preferences and a storage medium, and relates to the technical field of intelligent design, and the tea space color matching method based on the user preferences comprises the following steps: obtaining a tea ceremony type label, a tea ware picture, a tea space intention picture, an artistic conception text description and other multi-modal data input by a user; analyzing and generating a tea culture semantic vector, and extracting key features in the image by using an improved convolutional neural network to generate an image feature vector; and then, fusing the two into a user preference vector through a multi-modal attention mechanism, and finally, calculating the similarity between the vector and a color matching scheme feature vector in a preset database, and determining a most matched target tea space color matching scheme. According to the method, the personalized color matching scheme conforming to tea space culture attributes can be generated according to user interests and preferences, and efficient and accurate personalized recommendation is realized.

Description

Technical Field

[0001] This application relates to the field of intelligent design technology, and in particular to a method, apparatus, device, and storage medium for color matching in tea spaces based on user preferences. Background Technology

[0002] Tea spaces not only serve as a means of tea tasting, but also embody the spirit of tea ceremony, personal temperament, and spatial aesthetics. The effectiveness of atmosphere creation directly impacts the user's tea-tasting experience and state of mind. However, most users struggle to accurately translate their understanding of tea culture, personality traits, and ideal aesthetics into a practical color scheme, necessitating an intelligent color matching method that can "understand people's hearts and reflect culture."

[0003] Currently, the industry generally adopts three methods: (1) Designers make proposals manually based on their experience, and users modify them through repeated communication; (2) Batch recommendations based on fixed style templates (such as "New Chinese Style" and "Japanese Wabi-sabi"); (3) Using general color matching software (Adobe Color, Pinterest color palette, etc.) to simply combine hue and brightness.

[0004] The existing practices have the following problems: (1) High subjectivity: Designers' personal aesthetics are easily misaligned with users' expectations, resulting in high communication costs; (2) Generalization and template-based: Fixed style templates ignore the differences in users' deep interests, leading to a "one-size-fits-all" approach; (3) Cultural absence: General tools lack a systematic understanding of tea culture, material semantics, and spatial artistic conception, and color schemes are disconnected from the cultural attributes of tea spaces; (4) Single data dimension: Traditional practices are difficult to process multimodal information such as text descriptions, preference images, and interest tags at the same time, and cannot capture users' complex and implicit aesthetic demands. Therefore, how to generate personalized color schemes that conform to the cultural attributes of tea spaces based on users' interests and preferences has become an urgent problem to be solved.

[0005] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0006] The purpose of this application is to provide a method, apparatus, device, and storage medium for color matching in tea spaces based on user preferences, aiming to solve the technical problem of how to generate personalized color schemes that conform to the cultural attributes of tea spaces according to user interests and preferences.

[0007] To achieve the above objectives, this application proposes a color scheme method for tea spaces based on user preferences, the method comprising: Acquire multimodal data of tea space input by the user, including tea ceremony type tags, tea utensils images, tea space imagery images, and text descriptions of the tea space's artistic conception; The tea ceremony type tags and the text descriptions of the tea space artistic conception are parsed to generate a tea culture semantic vector. The tea space feature extraction model extracts the glaze color gradation, weaving texture and light and shadow layer features from the tea ware image and the tea space intention image, and generates image feature vectors. The tea space feature extraction model is an improvement of the convolutional neural network. The user preference vector is obtained by fusing the tea culture semantic vector and the image feature vector through a multimodal attention mechanism. Calculate the similarity between the user preference vector and the feature vectors of each color scheme in the preset tea space database, and determine the target tea space color scheme from the preset tea space database based on the similarity.

[0008] In one embodiment, the tea space feature extraction model includes a convolutional layer, a residual module, an object detection head, a texture analysis branch, an optical processing branch, and a fully connected layer; The step of extracting glaze color gradation, woven texture, and light and shadow layer features from the tea ware image and the tea space concept image using a tea space feature extraction model to generate an image feature vector includes: The convolutional layer identifies key areas of the glaze in the tea ware image; The pixel matrix of the key region of the glaze is analyzed by the residual module to obtain the hue gradient feature tensor. The object detection head identifies the woven structure area and the translucent material area in the tea space concept image; The density feature vector of the woven structure region is extracted through the texture analysis branch; The transmittance characteristic value of the light-transmitting material region is calculated through the optical processing branch; The image feature vector is generated by fusing the hue gradient feature tensor, the density feature vector, and the transmittance feature value through the fully connected layer.

[0009] In one embodiment, the steps for constructing the tea space feature extraction model include: Construct a residual module, an object detection head, a texture analysis branch, and an optical processing branch; Obtain a convolutional neural network, which includes an input layer, a convolutional layer, a pooling layer, and a fully connected layer; The pooling layer is replaced with the residual module, and the object detection head is added after the input layer. The texture analysis branch and the optical processing branch are added after the object detection head to obtain the tea space feature extraction model.

[0010] In one embodiment, the steps of constructing the residual module, object detection head, texture analysis branch, and optical processing branch include: Residual modules are constructed based on 3×3 convolutional kernel sets and skip connection channels; An object detection head is constructed based on a third-order anchor frame matrix and a ROI bilinear alignment layer. A texture analysis branch is constructed based on the complex frequency domain filter and the linear density statistics layer; An optical processing branch is constructed based on the transmission calculation unit.

[0011] In one embodiment, the step of fusing the tea culture semantic vector and the image feature vector through a multimodal attention mechanism to obtain the user preference vector includes: The tea culture semantic vector is used as the query vector, and the image feature vector is used as the key vector and value vector; Perform a dot product operation on the query vector and the key vector to obtain an attention score matrix, and normalize the attention score matrix to obtain the cross-modal association weight distribution; The value vector is weighted and summed according to the cross-modal association weight distribution to generate a fused feature vector; The fused feature vector is modeled using a gated loop unit to obtain spatial context continuity features, which are then mapped to user preference vectors.

[0012] In one embodiment, the step of parsing the tea ceremony type label and the text description of the tea space's artistic conception to generate a tea culture semantic vector includes: Natural language processing was used to identify ritual characteristic keywords in the tea ceremony type tags, as well as the frequency of words related to tranquility, naturalness, and history in the text description of the tea space's artistic conception. Based on the aforementioned ritual characteristic keywords, spatial element symbols in a preset tea culture imagery library are matched; The spatial tranquility index, material naturalness index, and traditional weight index are calculated based on the frequency of the tranquility word, the frequency of the naturalness word, and the frequency of the historicalness word. The spatial element symbols, the spatial tranquility index, the material naturalness index, and the traditional weight index are encoded to generate a tea culture semantic vector.

[0013] In one embodiment, the step of calculating the similarity between the user preference vector and the feature vectors of each color scheme in a preset tea space database, and determining the target tea space color scheme from the preset tea space database based on the similarity, includes: Iterate through the various color schemes in the preset tea space database; The feature vector of the color scheme is generated through natural language processing, the tea space feature extraction model, and the multimodal attention mechanism. Calculate the cosine similarity between the user preference vector and the feature vector; Arrange all the color schemes in descending order of their cosine similarity, and select the color scheme with the highest cosine similarity as the target tea space color scheme.

[0014] Furthermore, to achieve the above objectives, this application also proposes a tea space color scheme device based on user preferences, the device comprising: The data acquisition module is used to acquire multimodal data of tea space input by the user. The multimodal data of tea space includes tea ceremony type tags, tea utensils pictures, tea space imagery pictures and text descriptions of tea space artistic conception. The semantic parsing module is used to parse the tea ceremony type tags and the text description of the tea space artistic conception to generate a tea culture semantic vector. The feature extraction module is used to extract the glaze color level, weaving texture and light and shadow layer features from the tea ware image and the tea space intention image through the tea space feature extraction model, and generate image feature vectors. The tea space feature extraction model is obtained by improving the convolutional neural network. The fusion module is used to fuse the tea culture semantic vector and the image feature vector through a multimodal attention mechanism to obtain a user preference vector; The scheme determination module is used to calculate the similarity between the user preference vector and the feature vectors of each color scheme in the preset tea space database, and to determine the target tea space color scheme from the preset tea space database based on the similarity.

[0015] Furthermore, to achieve the above objectives, this application also proposes a tea space color matching device based on user preferences, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the tea space color matching method based on user preferences as described above.

[0016] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the user preference-based tea space color scheme method described above.

[0017] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the user preference-based tea space color scheme method described above.

[0018] One or more technical solutions proposed in this application have at least the following technical effects: First, multimodal data on tea spaces input by users is acquired, including tea ceremony type tags, tea ware images, tea space imagery, and textual descriptions of the tea space's atmosphere. This process, through multimodal data collection, comprehensively captures users' explicit interests and implicit aesthetic preferences, providing a rich information foundation for generating personalized color schemes. Second, the tea ceremony type tags and tea space imagery descriptions are parsed to generate tea culture semantic vectors. This step uses natural language processing technology to transform textual information into computable semantic features, accurately extracting users' preferences for the tea ceremony's atmosphere and cultural ambiance. Next, an improved convolutional neural network model extracts glaze gradations, woven textures, and light and shadow layers from tea ware images and tea space imagery, generating image feature vectors. This process utilizes deep learning technology to transform visual information into quantifiable features, providing a visual basis for color schemes. Finally, a multimodal attention mechanism is used to fuse the tea culture semantic vectors and image feature vectors to obtain the user preference vector. This fusion process dynamically aligns semantic and visual features, ensuring that the generated preference vector encompasses both cultural atmosphere and personal style. Finally, the similarity between the user preference vector and the feature vectors of each color scheme in the preset tea space database is calculated, and the target tea space color scheme is determined based on the similarity. This process uses cosine similarity to quickly retrieve the most matching color scheme, which can generate personalized color schemes that conform to the cultural attributes of the tea space based on the user's interests and preferences, thus achieving efficient and accurate personalized recommendations. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating an embodiment of the user-preference-based tea space color scheme method of this application. Figure 2 This is a schematic diagram of the module structure of the tea space feature extraction model provided in Embodiment 1 of the user preference-based tea space color matching method of this application; Figure 3 This is a flowchart illustrating Embodiment 2 of the user-preference-based tea space color scheme method of this application; Figure 4A simplified flowchart illustrating the user preference-based tea space color scheme method provided in Embodiment 2 of this application; Figure 5 This is a schematic diagram of the module structure of the tea space color matching device based on user preferences, as described in an embodiment of this application. Figure 6 This is a schematic diagram of the hardware operating environment involved in the user-preference-based tea space color matching method in the embodiments of this application.

[0022] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0023] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0024] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0025] It should be noted that the executing entity of this application embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of realizing the above functions, such as a tea space color matching system. The following uses a tea space color matching system as an example to describe this embodiment and the following embodiments.

[0026] Based on this, embodiments of this application provide a tea space color scheme method based on user preferences, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the user-preference-based color scheme for tea spaces in this application.

[0027] In this embodiment, the tea space color scheme method based on user preferences includes steps S10 to S50: Step S10: Obtain multimodal data of the tea space input by the user. The multimodal data of the tea space includes tea ceremony type tags, tea utensils images, tea space imagery images, and text descriptions of the tea space's artistic conception.

[0028] It should be noted that the "Tea Ceremony Type" tag refers to multiple interest keywords focused on the tea culture dimension, such as "Kung Fu Tea," "Sweet Tea," "Matcha," "Tea Utensil Appreciation," and "Zen and Tea as One," used to summarize the user's inclinations in tea ceremony schools, tea preferences, tea-related activities, or spiritual orientations. "Tea Utensil Images" refers to images of utensils directly related to tea ceremonies uploaded by users or selected from the system's image library, including teapots, teacups, tea scoops, tea caddies, tea stoves, and tea trays, used to reflect the user's aesthetic preferences for materials, colors, shapes, and textures. "Tea Space Concept Images" refers to visual materials uploaded by users or selected from the system's image library that express the overall atmosphere or partial scene of their ideal tea room. These can be images of actual tea rooms, natural landscapes, oriental-style interiors, or scenes with negative space and light and shadow, used to convey their intuitive preferences for spatial scale, light, color, and style. The text description of the artistic conception of a tea space refers to a short text written by the user in natural language, describing the atmosphere or emotional experience of an ideal tea space, such as "like a quiet tea mountain cottage" or "a corner full of wabi-sabi beauty". It is used to present the abstract artistic conception and emotional appeal in the user's mind that is difficult to express directly with images.

[0029] Understandably, the tea space color scheme system first presents users with selectable tea ceremony type tags such as "Kung Fu Tea / Sweet Tea / Matcha," "Tea Utensil Appreciation," and "Zen and Tea in Harmony" for them to choose from. Then, it opens an upload portal or image library for users to browse, allowing them to upload photos of their favorite tea utensils to match tea space scenes that reflect their ideal atmosphere. Figure 1 The system allows users to submit their requests and enter descriptive phrases such as "I want a quiet tea mountain cabin," thus completing the collection of tags, tea ware images, concept images, and text descriptions all at once.

[0030] Step S20: Parse the tea ceremony type label and the text description of the tea space artistic conception to generate a tea culture semantic vector.

[0031] It should be noted that the tea culture semantic vector refers to the high-dimensional numerical representation of the tea ceremony type tags selected by the user and the text description of the tea space artistic conception entered, which is mapped by natural language processing technology. This vector condenses the user's deep semantic preferences for tea ceremony schools, tea activities, spatial artistic conception and emotional atmosphere.

[0032] As an example, the step of parsing the tea ceremony type tag and the text description of the tea space's artistic conception to generate a tea culture semantic vector includes: identifying ritual feature keywords in the tea ceremony type tag, and the frequency of words related to tranquility, naturalness, and history in the text description of the tea space's artistic conception through natural language processing; matching spatial element symbols in a preset tea culture imagery library based on the ritual feature keywords; calculating a spatial tranquility index, a material naturalness index, and a traditional weight index based on the frequency of words related to tranquility, naturalness, and history; and encoding the spatial element symbols, the spatial tranquility index, the material naturalness index, and the traditional weight index to generate a tea culture semantic vector.

[0033] Natural Language Processing (NLP) refers to computer technology that uses algorithms such as word segmentation, part-of-speech tagging, named entity recognition, and sentiment analysis to perform structured semantic analysis of text. Ritual characteristic keywords refer to high-frequency words in tea ceremony type tags or text descriptions that characterize the tea ceremony process, etiquette actions, or scene configuration, such as "tea preparation," "tea ceremony," "main brewing position," and "ten tea-making techniques." Tranquility frequency refers to the relative frequency of occurrence of words expressing a quiet atmosphere, such as "secluded," "tranquil," and "ethereal," in the text description of the tea space's atmosphere. Naturalness frequency refers to the relative frequency of occurrence of words embodying natural materials, such as "logo," "bamboo," "stone," and "moss," in the text description of the tea space's atmosphere. Historical frequency refers to the relative frequency of occurrence of words embodying historical accumulation, such as "Song Dynasty style," "ancient style," "years," and "mottled," in the text description of the tea space's atmosphere. The pre-built tea culture imagery database refers to a semantic dictionary pre-constructed by the system that maps tea ceremony vocabulary to corresponding spatial element symbols. Spatial element symbols refer to visual or materialized symbols in the tea culture imagery database that correspond to key ritual characteristics, such as "low tea table," "bamboo curtain," and "dry landscape rock arrangement." The spatial tranquility index is a numerical indicator reflecting the degree of quietness of a space, calculated based on the frequency of words related to tranquility. The calculation formula is as follows: in, This refers to the spatial tranquility index. These are preset weight parameters. This refers to the frequency of words indicating quietness.

[0034] The Material Naturalness Index is a numerical indicator reflecting the proportion of natural materials in a space, calculated based on the frequency of words related to naturalness. The calculation formula is as follows: in, This refers to the material's naturalness index. These are preset weight parameters. It refers to the frequency of words related to natural feelings.

[0035] The Traditional Profoundness Index is a numerical indicator reflecting the depth of spatial history and culture, calculated based on the frequency of words related to historical sentiment. The calculation formula is as follows: in, This refers to the traditional weight index. These are preset weight parameters. It refers to the frequency of words related to historical sentiment.

[0036] First, the tea space color scheme system utilizes NLP's word segmentation and entity recognition modules to segment tea ceremony type tags word by word and compare them against a pre-set tea vocabulary list. It extracts ritualistic keywords such as "Kung Fu tea," "main brewing position," and "incense burning." Based on this, it retrieves corresponding spatial element symbols from a tea culture imagery database, such as "low tea table, bamboo curtain, dry landscape stone set," ensuring precise alignment between symbols and ritual scenes. Second, the system performs sentiment dictionary matching and word frequency statistics on the textual descriptions of the tea space's atmosphere. It accumulates the frequency of occurrences of words like "quiet" and "ethereal" to obtain the frequency of tranquility, words like "wood," "bamboo," and "stone" to obtain the frequency of naturalness, and words like "Song Dynasty style," "mottled," and "time" to obtain the frequency of historical feel. Then, it normalizes these three types of word frequencies using a pre-set linear or non-linear weighting formula and maps them to the 0-1 range, calculating the space's tranquility index, material naturalness index, and traditional weight index. This quantifies the user's strength or weakness in their demand for atmosphere, materials, and a sense of history. Finally, the system encodes the obtained spatial element symbols into discrete one-hot vectors or embedded vectors, concatenates the three indices as continuous values ​​into the same dimension, and obtains a tea culture semantic vector of uniform length after standardization, which retains both the category information of the symbols and the intensity information of the emotions.

[0037] Step S30: Extract the glaze color level, weaving texture and light and shadow layer features from the tea ware image and the tea space intention image using the tea space feature extraction model to generate image feature vectors. The tea space feature extraction model is obtained by improving the convolutional neural network.

[0038] It should be noted that the tea space feature extraction model refers to a deep convolutional network specifically optimized for visual elements of tea spaces. Its structure adds attention branches to glaze color gradation, woven texture, and lighting levels to the conventional convolutional neural network (CNN), used to capture visual features highly correlated with the tea culture atmosphere from images of teaware and tea spaces. Glaze color gradation refers to the continuous color variation range of the glaze surface of teaware under light, including subtle transitions in hue, saturation, and brightness, reflecting the unique texture of the clay, glaze, and firing process. Woven texture refers to the surface patterns formed by the interweaving of fabrics such as tea mats, carpets, curtains, or bamboo blinds with natural materials, embodying the material's texture and the warmth of handcraftsmanship. Lighting levels refer to the contrast and gradation of light and shadow shaped by the main light, fill light, and reflected light, creating a tranquil or warm tea room atmosphere. Image feature vectors are fixed-length numerical sequences output after model convolution, pooling, and fully connected processing, condensing quantifiable visual information about glaze color gradation, woven texture, and lighting levels, which can be used for subsequent similarity calculations.

[0039] As an example, the tea space feature extraction model includes a convolutional layer, a residual module, an object detection head, a texture analysis branch, an optical processing branch, and a fully connected layer. The step of extracting glaze color gradation, woven texture, and light and shadow layer features from the tea ware image and the tea space concept image using the tea space feature extraction model to generate an image feature vector includes: identifying key glaze regions in the tea ware image through the convolutional layer; analyzing the pixel matrix of the key glaze regions through the residual module to obtain a hue gradient feature tensor; identifying woven structure regions and translucent material regions in the tea space concept image through the object detection head; extracting the density feature vector of the woven structure region through the texture analysis branch; calculating the transmittance feature value of the translucent material region through the optical processing branch; and fusing the hue gradient feature tensor, the density feature vector, and the transmittance feature value through the fully connected layer to generate an image feature vector.

[0040] A convolutional layer is a neural network layer that extracts local spatial features by sliding convolutions on the input image using learnable filters. A residual module is a network unit that introduces cross-layer identity mappings into a standard convolution stack, mitigating gradient vanishing and enhancing feature reuse capabilities. An object detection head is a detection sub-network that slides an anchor box on a feature map and outputs category and location information. A texture analysis branch is an additional convolutional channel specifically designed to capture the statistical properties of fabric or rough surface textures. An optical processing branch is an additional computational channel that calculates the transmission or reflection properties of light based on an optical physics model. A fully connected layer is a neural network layer that flattens multidimensional features and maps them to a fixed-dimensional output through a weight matrix. A glaze key region refers to the local pixel region in a teaware image that is identified as the glaze and has the most representative color changes. A pixel matrix is ​​a numerical matrix composed of the color values ​​of each pixel in an image arranged in rows and columns. A hue gradient feature tensor is a multidimensional numerical array that records the continuous spatial variation of hue within the glaze key region. The woven structure region refers to the localized areas in the conceptual image of a tea space that are detected as having interwoven textures such as bamboo mats, linen, and rattan. The translucent material region refers to the localized areas in the conceptual image of a tea space that are detected as having permeable textures such as rice paper, gauze curtains, and glass. The density feature vector is a numerical sequence that quantifies the density, directional distribution, and repetition period of the woven structure region's texture. The transmittance feature value is a single-valued or low-dimensional vector characterizing the proportion of light transmitted through the translucent material region under specific lighting conditions.

[0041] First, the tea space color matching system feeds images of tea utensils into a convolutional layer. It then uses 64 3×3 convolutional kernels with a stride of 1 to generate a 128×128 feature map, as shown in the following formula: in, This refers to the tea ware image matrix (a three-dimensional tensor H×W×C), where H is the image height (number of pixel rows), W is the image width (number of pixel columns), and C is the number of channels (input feature map). This refers to the convolution kernel parameters (four-dimensional tensor k). h ×k w ×C×N), k h This refers to the kernel height, k. w The kernel width, N refers to the number of convolution kernels (the number of output feature map channels). It refers to the bias vector. This refers to the ReLU activation function. This refers to the output feature map (including information on key areas of the glaze).

[0042] Then, a 1×1 convolution is used to reduce the dimensionality to 32 channels and followed by Sigmoid activation to obtain a probability heatmap between 0 and 1. The bounding rectangles of the top three largest connected components are then cropped at a threshold of 0.75 as the key glaze regions. This accurately identifies the areas with the richest glaze color and reduces background interference. Next, these 128×128 key glaze regions are uniformly resized to 64×64, input into four sets of residual modules (each set consisting of three 3×3 convolutions + skip connections), and output as a 64×8×8 hue gradient feature tensor, as shown in the following formula: , in, This refers to the input features (from the convolutional layer). This refers to the residual function (containing two convolutional layers): ), This refers to the identity mapping. It refers to the hue gradation feature tensor.

[0043] Meanwhile, the object detection head generates 1024 anchor boxes on a 512×512 tea space intention image using an 8×8 grid. After filtering with an IoU (Intersection over Union) of 0.5 and NMS (Non-Maximum Suppression), it selects 1–2 anchor boxes each from the woven structure region and the translucent material region with the highest confidence. The texture analysis branch performs 3 more layers of 3×3 convolution and global average pooling on the cropped 64×64 woven region to obtain a 128-dimensional density feature vector. The optical processing branch samples a 32×32 grid in the translucent region, estimates the transmittance per pixel using 256 rays projected by Monte Carlo, and takes the average value as a single-value transmittance feature value. In this way, both texture and optical information are quantified, which can compensate for atmospheric details other than color. Finally, the 64×8×8 tensor is flattened into 4096 dimensions, and concatenated with a 128-dimensional density vector and a 1-dimensional transmittance value to form a 4225-dimensional long vector. After passing through two fully connected layers (512→256) and being activated by ReLU and then normalized by L2, a 256-dimensional image feature vector is output, which simultaneously contains glaze color, texture and light and shadow features.

[0044] Please refer to Figure 2 , Figure 2This diagram illustrates the module structure of the tea space feature extraction model provided in Embodiment 1 of the user-preference-based tea space color matching method of this application. After receiving data through the input layer, the model enters a dual-channel input stage, processing tea ware images and tea space concept images separately. For the tea ware images, the data flows through convolutional layers and residual modules, ultimately outputting a hue gradient feature tensor. For the tea space concept image, the model first performs region segmentation using an object detection head, dividing it into woven structure regions and translucent material regions. Then, the data from the woven structure regions undergoes texture analysis to extract density feature vectors, while the data from the translucent material regions undergoes optical processing to calculate transmittance feature values. All these feature tensors, feature vectors, and feature values ​​are integrated in a fully connected layer, outputting a 256-dimensional image feature vector, providing a foundation for subsequent multimodal feature fusion and user preference vector generation.

[0045] As an example, the construction steps of the tea space feature extraction model include: constructing a residual module, an object detection head, a texture analysis branch, and an optical processing branch; obtaining a convolutional neural network, which includes an input layer, a convolutional layer, a pooling layer, and a fully connected layer; replacing the pooling layer with the residual module, adding the object detection head after the input layer, and adding the texture analysis branch and the optical processing branch after the object detection head to obtain the tea space feature extraction model.

[0046] As an example, the steps of constructing the residual module, object detection head, texture analysis branch, and optical processing branch include: constructing the residual module based on a 3×3 convolution kernel group and skip connection channels; constructing the object detection head based on a third-order anchor frame matrix and a ROI bilinear alignment layer; constructing the texture analysis branch based on a complex frequency domain filter and a line density statistics layer; and constructing the optical processing branch based on a transmission calculation unit.

[0047] Skip connection channels refer to identity mapping paths in the residual module that directly add the input feature map across layers to the convolutional output, used to alleviate gradient vanishing and preserve low-level details. A third-order anchor box matrix refers to a matrix with an area of ​​128 in the object detection head. 2 256 2 512 2A 3×H×W tensor, composed of three sets of anchor boxes at different scales generated by pixels, is used to cover the woven structures and translucent materials of varying sizes in the tea space image. The ROI bilinear alignment layer refers to a pooling layer in the object detection head that maps candidate region coordinates to the feature map and samples them as a fixed-size 7×7 using bilinear interpolation, ensuring that position-sensitive features are not distorted with scale changes. The complex frequency domain filter refers to the operational unit within the texture analysis branch that transforms the input feature map to the frequency domain using a two-dimensional Fourier transform, then uses learnable ring and fan masks to weight and filter different frequency energies, and finally inversely transforms it back to the spatial domain, highlighting the periodic and directional information of the woven texture. The linear density statistical layer refers to the statistical layer in the texture analysis branch that accumulates edge responses and calculates the mean and variance along the four directions of 0°, 45°, 90°, and 135°, quantizing the texture density into a 128-dimensional vector. The transmission calculation unit refers to the unit within the optical processing branch that estimates the transmittance of 256 optical paths and takes the average value based on the RGB values ​​of the input light-transmitting material region and the pre-stored material refractive index table, and outputs a single-value transmittance feature.

[0048] First, the tea space color scheme system uses 64 3×3 convolutional kernels connected in series with ReLU, followed by another 64 3×3 convolutional kernels, and inserts identity jump connection channels between the two layers, stacking 4 sets of such structures to obtain residual modules; then, 128 2 256 2 512 2 Nine anchor boxes at three scales are written as third-order tensors and encapsulated with a ROI bilinear alignment layer to form an object detection head, responsible for outputting candidate boxes and confidence scores. Then, a complex frequency domain filter (ring + fan mask) and a linear density statistics layer (four-way edge accumulation) are concatenated and encapsulated into a texture analysis branch, and a transmission calculation unit (256 ray tracing mean values) is independently encapsulated into an optical processing branch. Then, a conventional CNN pre-trained on ImageNet is loaded, retaining only its input layer, convolutional layer, pooling layer, and fully connected layer. Finally, the original pooling layer is replaced with a residual module, which is connected in parallel after the input layer to the object detection head. The texture analysis branch and optical processing branch are then connected in parallel after the detection head. The three outputs and the residual module output are fed into the original fully connected layer, thus completing the construction of the tea space feature extraction model.

[0049] As an example, the training steps of the tea space feature extraction model include: constructing a training dataset based on labeled tea ware images and tea space images, wherein the tea ware images are labeled with glaze color gradient values, and the tea space images are labeled with braid density and transmittance; defining a material-aware loss function based on glaze color loss term, texture density loss term, and transmittance loss term; loading ResNet34 pre-trained weights onto the residual module of the tea space feature extraction model, and loading pre-calibrated Gabor filter parameters with a wavelength of 0.3 mm onto the texture analysis branch of the tea space feature extraction model; and training the tea space feature extraction model using a backpropagation algorithm based on the training dataset until the value of the material-aware loss function is less than a preset loss threshold.

[0050] The training dataset consists of manually labeled images of teaware and tea spaces. The teaware images include glaze color gradient values, while the tea space images include braid density and transmittance. The glaze color gradient value is calculated by taking the L channel of the Lab color space as a reference and then selecting a 5×5 sampling grid on the glaze surface of the teaware, calculating the average ΔL value between adjacent pixels as the continuous variation of the glaze color gradient. The braid density is the average value measured manually and recorded within the woven structure area of ​​the tea space image, measured in units of the number of warp and weft yarns per centimeter. Transmittance is the proportion of light passing through the material, expressed as a floating-point number in the range of 0–1, measured by an integrating sphere measuring instrument under a standard D65 light source on the labeled translucent area of ​​the tea space image. The glaze color loss term is the weighted sum of the L2 errors between the model-predicted glaze color gradient values ​​and the manually labeled values, accounting for 60% of the total loss weight. The texture density loss term is the weighted sum of the L1 errors between the model-predicted braid density and the manually labeled values, accounting for 30% of the total loss weight. The transmittance loss term refers to the weighted sum of the L2 errors between the model's predicted transmittance and the measured transmittance, accounting for 10% of the total loss weight. The material-aware loss function is a single scalar loss formed by linearly combining the glaze color loss term, texture density loss term, and transmittance loss term with weights of 60%, 30%, and 10%, respectively, used for backpropagation optimization. The ResNet34 pre-training weights refer to the convolutional kernel parameters of the ResNet34 network trained on ImageNet; the system directly loads them into the residual module to accelerate convergence. The pre-calibrated Gabor filter parameters refer to the Gabor kernel orientation, frequency, and bandwidth parameters obtained through calibration board experiments at a wavelength of 0.3 mm; the system uses these as the initial weights for the texture analysis branch. The preset loss threshold is the target value of the material-aware loss function set during training; training automatically stops when the validation set loss falls below this value.

[0051] First, the tea space color matching system inputs the glaze color gradient values ​​of 5000 tea ware images along with the line density and transmittance of 5000 tea space images into TFRecord, dividing the training, validation, and test sets in an 8:1:1 ratio. Then, the three errors are weighted and summed at 60%, 30%, and 10% respectively to obtain the material-aware loss function, allowing the network to prioritize learning glaze details while also considering texture and transmittance. Next, the residual module directly loads the ResNet34 weights pre-trained on ImageNet. The 8-direction Gabor kernels of the texture analysis branch are initialized with a 0.3 mm wavelength parameter, while the remaining branches are randomly initialized with Xavier kernels. This preserves general texture priors while avoiding cold starts. Finally, the system is trained for 200 epochs using Adam, an initial learning rate of 1e-4, a batch size of 32, and cosine annealing. The loss is evaluated on the validation set every 5 epochs. Early stopping is triggered when the material-aware loss drops below 0.012, and the weights are saved, completing the training of the tea space feature extraction model.

[0052] Step S40: The tea culture semantic vector and the image feature vector are fused through a multimodal attention mechanism to obtain the user preference vector.

[0053] It should be noted that the multimodal attention mechanism refers to a computational framework that uses cross-modal attention weights to dynamically align and weight-fuse the semantic vector of tea culture and the image feature vector in the same high-dimensional space. The user preference vector is a single numerical sequence generated after multimodal attention fusion. It comprehensively carries the user's semantic understanding of the artistic conception of tea ceremony and the image preference for the visual style of tea space, and can be directly used for subsequent color scheme matching.

[0054] Understandably, the system first takes the tea culture semantic vector and the image feature vector as input, respectively. Through query, key-value mapping in the multimodal attention mechanism, it calculates the relevance weights between the two vectors. Then, it weights the image feature vector according to these weights, adds the weighted image feature vector to the tea culture semantic vector, and obtains a user preference vector that integrates user text description and image preferences, which is used for subsequent color scheme recommendations.

[0055] Step S50: Calculate the similarity between the user preference vector and the feature vectors of each color scheme in the preset tea space database, and determine the target tea space color scheme from the preset tea space database based on the similarity.

[0056] It should be noted that the preset tea space database refers to a large collection of high-quality tea space color schemes pre-collected and stored by the system. Each scheme has been verified by professional designers or classic cases, covering primary colors, secondary colors, accent colors, and related material information, and is accompanied by cultural style tags. The feature vector of each color scheme refers to the vector obtained after performing semantic parsing, image extraction, and multimodal attention fusion on the tea ceremony tags, artistic conception text, tea ware images, and spatial intention images of each color scheme in the database in a manner completely consistent with the user's experience. Similarity refers to the cosine similarity between the user's preference vector and the feature vector of the color scheme; the closer the value is to 1, the higher the degree of matching between the two in terms of color, material, and cultural artistic conception. The target tea space color scheme refers to the Top N (pre-set number) color schemes selected from the preset tea space database that best match the user's preference vector after being sorted by similarity, which are used as the final personalized tea space color scheme recommended to the user.

[0057] As an example, the step of calculating the similarity between the user preference vector and the feature vectors of each color scheme in the preset tea space database, and determining the target tea space color scheme from the preset tea space database based on the similarity, includes: traversing each color scheme in the preset tea space database; generating the feature vector of the color scheme through natural language processing, the tea space feature extraction model, and the multimodal attention mechanism; calculating the cosine similarity between the user preference vector and the feature vector; arranging all the color schemes in descending order of the cosine similarity, and selecting the color scheme with the highest cosine similarity as the target tea space color scheme.

[0058] Cosine similarity refers to the cosine value of the angle between two vectors of equal length in a multidimensional space. The value ranges from -1 to 1. The closer the value is to 1, the more consistent the directions are. It is used to measure the degree of matching between user preference vectors and color scheme feature vectors.

[0059] First, the tea space color scheme system sequentially reads each color scheme record from the preset tea space database, packaging its "primary color + secondary color + accent color + material + mood description" text and accompanying example image into a processing unit. Then, natural language processing is applied to the text to obtain a semantic vector, and a tea space feature extraction model is applied to the example image to obtain an image vector. These two vectors are then fused into a feature vector for the color scheme using a multimodal attention mechanism. This entire process is identical to the method used to generate the user preference vector, ensuring dimensional alignment. Next, the system calculates the cosine similarity between the user preference vector and this feature vector using a single matrix multiplication and writes the result into a sorting table. After traversal, quicksort is used to sort all color schemes in descending order of similarity, and the top record is directly selected as the target tea space color scheme, completing the personalized recommendation.

[0060] This embodiment provides a tea space color scheme method based on user preferences. First, it acquires multimodal data of the tea space input by the user, including tea ceremony type tags, tea ware images, tea space imagery, and textual descriptions of the tea space's atmosphere. This process, through multimodal data acquisition, comprehensively captures the user's explicit interests and implicit aesthetic preferences, providing a rich information foundation for generating personalized color schemes. Second, it parses the tea ceremony type tags and tea space imagery text descriptions to generate tea culture semantic vectors. This step uses natural language processing technology to transform textual information into computable semantic features, accurately extracting the user's preferences for the tea ceremony's atmosphere and cultural ambiance. Next, it extracts glaze gradations, woven textures, and light and shadow layers from the tea ware images and tea space imagery using an improved convolutional neural network model, generating image feature vectors. This process utilizes deep learning technology to transform visual information into quantifiable features, providing a visual basis for the color scheme. Then, it fuses the tea culture semantic vectors and image feature vectors through a multimodal attention mechanism to obtain the user preference vector. This fusion process dynamically aligns semantic and visual features, ensuring that the generated preference vector encompasses both cultural atmosphere and personal style. Finally, the similarity between the user preference vector and the feature vectors of each color scheme in the preset tea space database is calculated, and the target tea space color scheme is determined based on the similarity. This process uses cosine similarity to quickly retrieve the most matching color scheme, which can generate personalized color schemes that conform to the cultural attributes of the tea space based on the user's interests and preferences, thus achieving efficient and accurate personalized recommendations.

[0061] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 , Figure 3 This is a flowchart illustrating the second embodiment of the user-preference-based tea space color scheme method of this application. Step S40 of the user-preference-based tea space color scheme method includes steps S41 to S44: Step S41: Use the tea culture semantic vector as the query vector and the image feature vector as the key vector and value vector.

[0062] It should be noted that the query vector refers to the vector obtained by linearly mapping the semantic vector of tea culture, used to actively retrieve relevant information in the attention mechanism. The key vector refers to the index vector obtained by linearly mapping the image feature vector, used to calculate the attention weights with the query vector. The value vector refers to the content vector obtained by linearly mapping the image feature vector, which will be weighted and summed and output after the attention weights are determined.

[0063] Step S42: Perform a dot product operation on the query vector and the key vector to obtain an attention score matrix, and normalize the attention score matrix to obtain the cross-modal association weight distribution.

[0064] It should be noted that the attention score matrix is ​​the original numerical matrix obtained by taking the dot product of the query vector and the key vector element by element position. Each element reflects the initial correlation strength between the query and the corresponding key. The cross-modal association weight distribution is a probability matrix obtained by performing softmax normalization on the attention score matrix along the key dimension. Each row represents the weighted proportion of the query to all keys, which is used to guide the weighted fusion of value vectors.

[0065] Understandably, the tea space color matching system first performs matrix multiplication on the 256-dimensional query vector obtained by linearly mapping the tea culture semantic vector through W_q and the 256-dimensional key vector obtained by linearly mapping the image feature vector through W_k, producing a dot product result of shape 1×N, i.e., the attention score matrix. Then, softmax is applied along the N dimensions, and each score is divided by the temperature coefficient of 0.1 and then exponentially normalized to obtain a 1×N probability distribution, which serves as the cross-modal association weight distribution for the next step of weighting the vector.

[0066] Step S43: The value vector is weighted and summed according to the cross-modal association weight distribution to generate a fused feature vector.

[0067] It should be noted that the fused feature vector refers to a single 256-dimensional vector obtained by weighting and summing the value vectors using the cross-modal association weight distribution as coefficients. This vector carries key information of both tea culture semantics and visual imagery, and is used to generate user preference vectors in the future.

[0068] Understandably, the system multiplies the cross-modal association weight distribution element-wise with the value vector and then accumulates them along the key dimension to output a 256-dimensional fused feature vector.

[0069] Step S44: The fused feature vector is modeled using a gated loop unit to obtain spatial context continuity features, and the spatial context continuity features are mapped to user preference vectors.

[0070] It should be noted that a gated recurrent unit (GRU) is a recurrent neural network structure containing update and reset gates, used to capture long-term dependencies in sequential data. Spatial atmosphere continuity features refer to the implicit state vectors obtained after processing the fused feature vectors through a GRU, reflecting the smooth transition of the tea space atmosphere over time or context.

[0071] Understandably, the system feeds the 256-dimensional fused feature vector as a single-step input into a gated recurrent unit with 128 hidden units. The GRU calculates the current input and the hidden state from the previous time step through update and reset gates, and outputs a new 128-dimensional hidden state. After processing for three consecutive steps, the hidden state of the last step is taken as the spatial mood continuity feature. Then, a fully connected layer from 128 to 256 with ReLU activation is used to map the hidden state into a user preference vector.

[0072] This embodiment first uses the tea culture semantic vector as the query vector and the image feature vector as the key and value vectors, respectively. An attention score matrix is ​​obtained through dot product operations. This process quantifies the correlation between text and image features, providing a weighting basis for subsequent feature fusion. Next, the attention score matrix is ​​normalized to obtain a cross-modal association weight distribution. This step ensures that the importance of different modal features is reasonably allocated during fusion. Then, the value vectors are weighted and summed according to the cross-modal association weight distribution to generate a fused feature vector. This step achieves a deep fusion of tea culture semantics and visual features, ensuring that the generated feature vector encompasses both cultural connotations and visual details. Finally, a gated recurrent unit is used to perform temporal feature modeling on the fused feature vector to obtain spatial context continuity features, which are then mapped to a user preference vector. This process captures continuous changes in context, further enhancing the expressive power of the user preference vector and ultimately achieving accurate personalized color scheme recommendations.

[0073] For example, to help understand the implementation process of the user preference-based tea space color scheme method obtained by combining this embodiment with the above embodiment one, please refer to... Figure 4 , Figure 4 A simplified flowchart illustrating a user-preference-based color scheme for tea spaces is provided, specifically: User input includes tea ceremony type tags, descriptive text, images of tea utensils, and images of tea space concepts. The system first processes the tea ceremony type tags and descriptive text through a tea culture semantic parsing module, generating a tea culture semantic vector. Simultaneously, the images of tea utensils and tea space concepts are fed into a feature extraction model, which includes convolutional layers, residual modules, an object detection head, a texture analysis branch, and an optical processing branch, to extract hue gradation features, density features, and transmittance features, generating an image feature vector. Next, a multimodal attention mechanism uses the query vector (tea culture semantic vector), key vector, and value vector (image feature vector) to fuse these features, generating a user preference vector. The user preference vector is then compared with the feature vectors of color schemes in a pre-set tea space database to calculate similarity and determine the target tea space color scheme. The entire process, through deep learning technology, automates and intelligently recommends personalized color schemes from user input.

[0074] This application also provides a color scheme device for a tea space based on user preferences; please refer to [reference needed]. Figure 5 The user-preference-based tea space color scheme device includes: Data acquisition module 10 is used to acquire multimodal data of tea space input by the user. The multimodal data of tea space includes tea ceremony type tags, tea utensils pictures, tea space imagery pictures and tea space mood text descriptions. Semantic parsing module 20 is used to parse the tea ceremony type label and the tea space artistic conception text description to generate a tea culture semantic vector; Feature extraction module 30 is used to extract glaze color level, weaving texture and light and shadow layer features from the tea ware image and the tea space intention image through the tea space feature extraction model, and generate image feature vector. The tea space feature extraction model is obtained by improving the convolutional neural network. The fusion module 40 is used to fuse the tea culture semantic vector and the image feature vector through a multimodal attention mechanism to obtain a user preference vector; The scheme determination module 50 is used to calculate the similarity between the user preference vector and the feature vectors of each color scheme in the preset tea space database, and to determine the target tea space color scheme from the preset tea space database based on the similarity.

[0075] The user-preference-based tea space color scheme device provided in this application, employing the user-preference-based tea space color scheme method described in the above embodiments, can solve the technical problem of how to generate personalized color schemes that conform to the cultural attributes of tea spaces based on user interests and preferences. Compared with the prior art, the beneficial effects of the user-preference-based tea space color scheme device provided in this application are the same as those of the user-preference-based tea space color scheme method provided in the above embodiments, and other technical features in the user-preference-based tea space color scheme device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0076] This application provides a user-preference-based tea space color matching device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the user-preference-based tea space color matching method in Embodiment 1 above.

[0077] The following is for reference. Figure 6This document illustrates a structural schematic diagram of a user-preference-based tea space color matching device suitable for implementing embodiments of this application. The user-preference-based tea space color matching device in this application embodiment may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 6 The tea space color scheme device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0078] like Figure 6 As shown, the user-preference-based tea space color matching device may include a processing unit 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in ROM (Read Only Memory) 1002 or a program loaded from storage device 1003 into RAM (Random Access Memory) 1004. RAM 1004 also stores various programs and data required for the operation of the user-preference-based tea space color matching device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. Input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, LCDs (Liquid Crystal Displays), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the user-preference-based tea space color matching device to wirelessly or wiredly communicate with other devices to exchange data. Although a user-preference-based tea space color matching device with various systems is shown in the figure, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0079] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0080] The user-preference-based tea space color scheme device provided in this application, employing the user-preference-based tea space color scheme method described in the above embodiments, solves the technical problem of how to generate personalized color schemes that conform to the cultural attributes of tea spaces based on user interests and preferences. Compared with the prior art, the beneficial effects of the user-preference-based tea space color scheme device provided in this application are the same as those of the user-preference-based tea space color scheme method provided in the above embodiments, and other technical features in this user-preference-based tea space color scheme device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0081] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0082] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0083] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the user-preference-based tea space color matching method in the above embodiments.

[0084] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory or Flash Memory), optical fibers, CD-ROM (CD-Read Only Memory), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0085] The aforementioned computer-readable storage medium may be included in a user-preference-based tea space color matching device; or it may exist independently and not assembled into a user-preference-based tea space color matching device.

[0086] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by a user-preference-based tea space color matching device, the user-preference-based tea space color matching device: acquires multimodal tea space data input by the user, the multimodal tea space data including tea ceremony type tags, tea ware images, tea space intention images, and tea space mood text descriptions; parses the tea ceremony type tags and the tea space mood text descriptions to generate a tea culture semantic vector; extracts glaze color levels, woven textures, and light and shadow layer features from the tea ware images and the tea space intention images using a tea space feature extraction model, generating an image feature vector, the tea space feature extraction model being an improved convolutional neural network; fuses the tea culture semantic vector and the image feature vector using a multimodal attention mechanism to obtain a user preference vector; calculates the similarity between the user preference vector and the feature vectors of various color schemes in a preset tea space database, and determines a target tea space color scheme from the preset tea space database based on the similarity.

[0087] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including LAN (Local Area Network) or WAN (Wide Area Network)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0088] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0089] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0090] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described user-preference-based tea space color scheme method. This solves the technical problem of how to generate personalized color schemes that conform to the cultural attributes of tea spaces based on user interests and preferences. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the user-preference-based tea space color scheme method provided in the above embodiments, and will not be repeated here.

[0091] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the user preference-based tea space color scheme method described above.

[0092] The computer program product provided in this application can solve the technical problem of how to generate personalized color schemes that conform to the cultural attributes of tea spaces based on user interests and preferences. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the user preference-based tea space color scheme method provided in the above embodiments, and will not be repeated here.

[0093] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A color scheme method for tea spaces based on user preferences, characterized in that, The method includes: Acquire multimodal data of tea space input by the user, including tea ceremony type tags, tea utensils images, tea space imagery images, and text descriptions of the tea space's artistic conception; The tea ceremony type tags and the text descriptions of the tea space artistic conception are parsed to generate a tea culture semantic vector. The tea space feature extraction model extracts the glaze color gradation, weaving texture and light and shadow layer features from the tea ware image and the tea space intention image, and generates image feature vectors. The tea space feature extraction model is an improvement of the convolutional neural network. The user preference vector is obtained by fusing the tea culture semantic vector and the image feature vector through a multimodal attention mechanism. Calculate the similarity between the user preference vector and the feature vectors of each color scheme in the preset tea space database, and determine the target tea space color scheme from the preset tea space database based on the similarity.

2. The method as described in claim 1, characterized in that, The tea space feature extraction model includes convolutional layers, residual modules, object detection heads, texture analysis branches, optical processing branches, and fully connected layers; The step of extracting glaze color gradation, woven texture, and light and shadow layer features from the tea ware image and the tea space concept image using a tea space feature extraction model to generate an image feature vector includes: The convolutional layer identifies key areas of the glaze in the tea ware image; The pixel matrix of the key region of the glaze is analyzed by the residual module to obtain the hue gradient feature tensor. The object detection head identifies the woven structure area and the translucent material area in the tea space concept image; The density feature vector of the woven structure region is extracted through the texture analysis branch; The transmittance characteristic value of the light-transmitting material region is calculated through the optical processing branch; The image feature vector is generated by fusing the hue gradient feature tensor, the density feature vector, and the transmittance feature value through the fully connected layer.

3. The method as described in claim 1, characterized in that, The steps for constructing the tea space feature extraction model include: Construct a residual module, an object detection head, a texture analysis branch, and an optical processing branch; Obtain a convolutional neural network, which includes an input layer, a convolutional layer, a pooling layer, and a fully connected layer; The pooling layer is replaced with the residual module, and the object detection head is added after the input layer. The texture analysis branch and the optical processing branch are added after the object detection head to obtain the tea space feature extraction model.

4. The method as described in claim 3, characterized in that, The steps for constructing the residual module, object detection head, texture analysis branch, and optical processing branch include: Residual modules are constructed based on 3×3 convolutional kernel sets and skip connection channels; An object detection head is constructed based on a third-order anchor frame matrix and a ROI bilinear alignment layer. A texture analysis branch is constructed based on the complex frequency domain filter and the linear density statistics layer; An optical processing branch is constructed based on the transmission calculation unit.

5. The method as described in claim 1, characterized in that, The step of fusing the tea culture semantic vector and the image feature vector through a multimodal attention mechanism to obtain the user preference vector includes: The tea culture semantic vector is used as the query vector, and the image feature vector is used as the key vector and value vector; Perform a dot product operation on the query vector and the key vector to obtain an attention score matrix, and normalize the attention score matrix to obtain the cross-modal association weight distribution; The value vector is weighted and summed according to the cross-modal association weight distribution to generate a fused feature vector; The fused feature vector is modeled using a gated loop unit to obtain spatial context continuity features, which are then mapped to user preference vectors.

6. The method as described in claim 1, characterized in that, The step of parsing the tea ceremony type label and the text description of the tea space artistic conception to generate a tea culture semantic vector includes: Natural language processing was used to identify ritual characteristic keywords in the tea ceremony type tags, as well as the frequency of words related to tranquility, naturalness, and history in the text description of the tea space's artistic conception. Based on the aforementioned ritual characteristic keywords, spatial element symbols in a preset tea culture imagery library are matched; The spatial tranquility index, material naturalness index, and traditional weight index are calculated based on the frequency of the tranquility word, the frequency of the naturalness word, and the frequency of the historicalness word. The spatial element symbols, the spatial tranquility index, the material naturalness index, and the traditional weight index are encoded to generate a tea culture semantic vector.

7. The method according to any one of claims 1 to 6, characterized in that, The step of calculating the similarity between the user preference vector and the feature vectors of each color scheme in the preset tea space database, and determining the target tea space color scheme from the preset tea space database based on the similarity, includes: Iterate through the various color schemes in the preset tea space database; The feature vector of the color scheme is generated through natural language processing, the tea space feature extraction model, and the multimodal attention mechanism. Calculate the cosine similarity between the user preference vector and the feature vector; Arrange all the color schemes in descending order of their cosine similarity, and select the color scheme with the highest cosine similarity as the target tea space color scheme.

8. A color scheme device for tea spaces based on user preferences, characterized in that, The device includes: The data acquisition module is used to acquire multimodal data of tea space input by the user. The multimodal data of tea space includes tea ceremony type tags, tea utensils pictures, tea space imagery pictures and text descriptions of tea space artistic conception. The semantic parsing module is used to parse the tea ceremony type tags and the text description of the tea space artistic conception to generate a tea culture semantic vector. The feature extraction module is used to extract the glaze color level, weaving texture and light and shadow layer features from the tea ware image and the tea space intention image through the tea space feature extraction model, and generate image feature vectors. The tea space feature extraction model is obtained by improving the convolutional neural network. The fusion module is used to fuse the tea culture semantic vector and the image feature vector through a multimodal attention mechanism to obtain a user preference vector; The scheme determination module is used to calculate the similarity between the user preference vector and the feature vectors of each color scheme in the preset tea space database, and to determine the target tea space color scheme from the preset tea space database based on the similarity.

9. A tea space color matching device based on user preferences, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the user preference-based tea space color scheme method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the tea space color matching method based on user preferences as described in any one of claims 1 to 7.

Images