Data analysis-based visual communication design evaluation system and method

Abstract

The application relates to the field of computer-aided design and visual computing technology, and provides a visual communication design evaluation system and method based on data analysis. The method comprises the following steps: obtaining original design data, extracting a spatial layout feature matrix, a color distribution feature vector and a semantic content feature set; calling a design specification knowledge base based on design purpose information; performing simulation attention distribution analysis and design compliance analysis to generate corresponding features; constructing a communication efficiency prediction model to generate a comprehensive communication efficiency score and an initial design evaluation report; performing joint analysis of interference effects to obtain a comprehensive interference effect quantitative evaluation result; based on the result, optimizing and correcting the efficiency of the initial design evaluation report, and outputting a final design evaluation report integrating spatial layout and visual attribute optimization suggestions. The application accurately quantifies the interference effect through joint analysis of spatial layout and semantic attributes, thereby improving the accuracy and reliability of design communication efficiency evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer-aided design and visual computing, and particularly relates to a visual communication design evaluation system and method based on data analysis. BACKGROUND

[0002] Visual communication design is an important medium for efficient information transmission through effective organization and presentation of visual elements. In the fields of advertising, brand promotion, and public information dissemination, the quality of visual communication design directly affects the efficiency of information dissemination and the cognitive effect of the audience.

[0003] Traditional design evaluation mainly relies on the experience and aesthetic intuition of designers, and is carried out through manual review or focus group discussion. This evaluation method is greatly influenced by the subjective preferences, professional backgrounds, and cognitive differences of the reviewers, resulting in a lack of objective and unified standardized system for evaluation results. With the development of digital design tools, the efficiency of the design process has been significantly improved, but there are still obvious shortcomings in the quantitative evaluation of design quality.

[0004] Current design evaluation methods based on digital means mainly focus on aesthetic analysis, such as color matching, layout composition, and detection of basic visual attributes. These methods still have significant technical limitations in practical applications: on the one hand, existing technologies are difficult to quantitatively analyze the complex interactions between design elements, especially the synergistic influence mechanism between spatial layout and semantic attributes; on the other hand, traditional evaluation methods cannot accurately identify and quantify the interference effect of design elements on the core information transmission path, resulting in a large deviation between the evaluation results and the actual transmission effect. This limitation is particularly prominent in design scenarios with high information density or scattered audience attention, severely affecting the accuracy and effectiveness of design optimization. SUMMARY

[0005] In order to overcome the technical defects of strong subjectivity and single evaluation dimension in existing visual communication design evaluation, the present application provides a visual communication design evaluation system and method based on data analysis, which accurately quantifies the interference characteristics of design elements on information transmission through joint analysis of spatial layout features and visual semantic attributes, effectively improving the accuracy and reliability of design evaluation.

[0006] In order to achieve the above purpose, the present application adopts the following technical solutions:

[0007] In a first aspect, the present application provides a visual communication design evaluation method based on data analysis, comprising the following steps:

[0008] S1: Obtain the original design data of the visual communication design to be evaluated, perform feature analysis, and obtain a spatial layout feature matrix, a color distribution feature vector, and a semantic content feature set;

[0009] S2: identifying design purpose information of the visual communication design based on the original design data, and calling a corresponding design specification knowledge base based on the design purpose information;

[0010] S3: performing simulation attention distribution analysis and design compliance analysis based on the spatial layout feature matrix, the color distribution feature vector, the semantic content feature set, and the design specification knowledge base, to generate simulation attention distribution features and design compliance features;

[0011] S4: constructing a communication effectiveness prediction model based on the simulation attention distribution features and the design compliance features, to generate a comprehensive communication effectiveness score and an initial design evaluation report;

[0012] S5: performing joint analysis of the spatial layout feature matrix and the semantic content feature set based on the design purpose information and the design specification knowledge base, to obtain a comprehensive interference effect quantitative evaluation result;

[0013] S6: performing effectiveness optimization correction on the initial design evaluation report based on the comprehensive interference effect quantitative evaluation result, and outputting a final design evaluation report containing quantitative improvement data and executable suggestions.

[0014] According to the above technical solution, the step of obtaining the original design data of the visual communication design to be evaluated and performing feature analysis comprises:

[0015] The original design data is processed by parsing and format unification to identify all visual elements and calculate their geometric properties, and a spatial layout feature matrix is constructed. The parsing and format unification processing refers to uniformly parsing the original data from structured design files or raster images. The structured design files extract visual element objects by parsing their internal format specifications, and the raster images identify visual element regions based on a deep learning-based instance segmentation model. All parsed elements are converted into a unified standardized internal data representation form. The visual elements refer to basic visual units that constitute the design draft, including basic graphics, text blocks, embedded images, and icon symbols. The construction of the spatial layout feature matrix is achieved by calculating the center point coordinates, width height dimensions, and pixel area of the bounding box of each visual element, organizing these geometric properties of all elements into feature vectors, and combining them into a multi-dimensional matrix. The core purpose of this step is to establish a standardized layout description system by uniformly measuring different sources and formats of design elements in the same geometric space, providing accurate and calculable data basis for subsequent objective evaluation of the rationality, orderliness of the layout, and the spatial relationship between elements, thereby overcoming the instability caused by subjective perception differences in manual evaluation.

[0016] The original design data is preprocessed and converted to an HSV color space, color statistical features and color entropy values of the image are calculated, and a color distribution feature vector is formed. The preprocessing includes converting the design draft pixel data from an RGB color space to an HSV color space. The color statistical features include a main color tone extracted by a clustering algorithm, a color contrast calculated, and a color channel variance. The color entropy value is obtained by constructing a color histogram after quantizing the color channel value and calculating the information entropy. The color distribution feature vector is a one-dimensional array formed by splicing all the color statistical features and color entropy values in a predetermined order. The HSV color space is selected for conversion because it is more consistent with the human color perception mode, can effectively separate the hue, saturation and brightness information, and makes the color analysis more semantically meaningful. By calculating the main color tone, contrast, channel variance and color entropy, the method can quantitatively describe the color tendency, visual impact and color complexity of the design, providing a key basis for objectively evaluating the harmony and expressiveness of color matching, and avoiding the limitations of subjective aesthetic judgment.

[0017] The original design data is preprocessed and converted to an HSV color space, color statistical features and color entropy values of the image are calculated, and a color distribution feature vector is formed. The preprocessing includes converting the design draft pixel data from an RGB color space to an HSV color space. The color statistical features include a main color tone extracted by a clustering algorithm, a color contrast calculated, and a color channel variance. The color entropy value is obtained by constructing a color histogram after quantizing the color channel value and calculating the information entropy. The color distribution feature vector is a one-dimensional array formed by splicing all the color statistical features and color entropy values in a predetermined order. The HSV color space is selected for conversion because it is more consistent with the human color perception mode, can effectively separate the hue, saturation and brightness information, and makes the color analysis more semantically meaningful. By calculating the main color tone, contrast, channel variance and color entropy, the method can quantitatively describe the color tendency, visual impact and color complexity of the design, providing a key basis for objectively evaluating the harmony and expressiveness of color matching, and avoiding the limitations of subjective aesthetic judgment.

[0018] According to the above technical solution, the step of performing simulated attention distribution analysis and generating a simulated attention distribution feature includes:

[0019] The spatial layout feature matrix and the color distribution feature vector are processed by a pre-trained deep visual saliency network to simulate visual attention and generate an initial attention heat map. A deep neural network is used to simulate the human eye attention mechanism.

[0020] The design specification knowledge base is extracted through layout guide criteria, a core information area is determined based on design purpose information, and an expected attention distribution is generated.

[0021] The initial attention heat map and the expected attention distribution are processed by weight assignment and pixel-level weighted summation, wherein the initial attention heat map is assigned a first weight and the expected attention distribution is assigned a second weight, to form a final attention distribution that conforms to the design conveying characteristics. By weighted fusion of "what can actually be seen" and "what should be seen", the method ingeniously balances visual rules and design intentions. This processing method makes the final evaluation standard respect the objective rules of human vision and emphasize the requirements for design functionality, so as to obtain a more comprehensive and more instructive attention distribution comprehensive evaluation.

[0022] The final attention distribution is calculated by statistical characteristics to obtain the average attention intensity, the attention distribution standard deviation and the attention distribution entropy value; wherein the average attention intensity is obtained by calculating the arithmetic mean of the final attention distribution, the attention distribution standard deviation is obtained by calculating the square root of the average of the sum of squares of the difference between all values in the final attention distribution and the average value, and the attention distribution entropy value is obtained by first normalizing the final attention distribution into a probability distribution and then calculating the information entropy thereof;

[0023] The average attention intensity, the attention distribution standard deviation and the attention distribution entropy value are processed by feature combination to form a three-dimensional simulated attention distribution feature in a preset order. Multiple statistical indexes are combined into a comprehensive feature vector, aiming to construct a multidimensional signal that can comprehensively describe the attention distribution characteristics, so as to facilitate the overall analysis and interpretation of the subsequent prediction model, thereby realizing the accurate prediction of the design conveying performance.

[0024] According to the above technical solution, the step of performing compliance analysis and generating design compliance features comprises:

[0025] The visual element space layout feature matrix is compared and analyzed with the grid system and safety margin specifications in the design specification knowledge base to calculate the element alignment degree, the spacing uniformity and the core area blank proportion, and to generate a score of the specification compliance degree of the layout. This process accurately compares the actual layout of the design with the professional grid system and safety margin specifications, and converts the original layout aesthetics depending on the designer's feeling into objective indexes such as alignment degree and uniformity. This provides a unified and reproducible standard for evaluating the professionalism and orderliness of the layout design, and significantly reduces the evaluation deviation caused by subjective differences of the evaluators.

[0026] This method combines the primary and secondary colors represented by the color distribution feature vectors with standard brand color palettes from the design specification knowledge base. Through similarity calculation and contrast analysis, along with accessibility standard threshold verification, a weighted fusion is performed to generate a color system compliance score. This approach not only focuses on the accuracy of brand color reproduction, ensuring visual consistency through color difference calculation, but also mandates accessibility standard verification to guarantee the recognizability of color combinations for color-blind, color-weak, and other special populations. This dual verification mechanism ensures that the design, beyond aesthetics, meets the rigid requirements of brand management and information accessibility.

[0027] The font type encoding in the semantic content feature set is matched and verified against the list of authorized standard fonts in the design specification knowledge base. Simultaneously, the consistency of the font size hierarchy system is checked, generating a font usage compliance score. This step systematically reviews the legality and logic of font usage. Font matching verification avoids copyright risks, while font size hierarchy consistency detection ensures that the text information has a clear and consistent visual logic, helping to guide the audience to smoothly read and understand the primary and secondary relationships of the information, thereby improving the professionalism of the design and the efficiency of information delivery.

[0028] The scores for compliance with layout specifications, color scheme specifications, and font usage specifications are integrated using a weighted fusion algorithm to generate a design compliance feature. This algorithm combines compliance scores from three different dimensions—layout, color, and font—into a single comprehensive feature, allowing a concise indicator to represent the overall degree of design compliance across multiple specifications. This provides a holistic quantitative input for the subsequent communication effectiveness prediction model regarding design "compliance," avoiding the limitations of single-dimensional compliance judgments.

[0029] According to the above technical solution, the steps of constructing a communication effectiveness prediction model and generating a comprehensive communication effectiveness score include:

[0030] The simulated attention distribution features and the design compliance features are combined to form a composite feature vector. By combining attention features reflecting visual guidance effectiveness with compliance features reflecting professional standards, the aim is to describe the design's potential communication capabilities from two core dimensions: "attractiveness" and "standardization." This multi-feature fusion strategy provides the model with more comprehensive information, enabling it to learn more complex underlying patterns that determine design effectiveness, laying the foundation for accurate comprehensive predictions.

[0031] The combined feature vectors are processed by a pre-trained neural network model, which includes a feature crossover layer and a multilayer perceptron. The feature crossover layer uses a self-attention mechanism to capture the interaction between simulated attention distribution features and various sub-features in the design compliance features, and calculates their weights. The advantage of using a neural network structure with a feature crossover layer is that the model can automatically discover and learn the complex nonlinear relationships and interaction effects between attention features and compliance features. For example, it can intelligently determine under what compliance background a specific attention distribution pattern contributes more to the final effectiveness, thereby achieving a more refined and intelligent evaluation of overall communication effectiveness, surpassing the limitations of simple linear weighted models.

[0032] A comprehensive communication effectiveness score is obtained from the output of the neural network model. Through forward propagation calculation of the neural network model, a quantitative comprehensive communication effectiveness score is finally output. This score reduces the complex attributes of the design across multiple dimensions to an intuitive and comparable scalar value, greatly facilitating the rapid ranking of design solutions, quality classification, and decision-making on whether they meet the standards, providing a clear target orientation for design optimization and iteration.

[0033] According to the above technical solution, the steps for generating the initial design evaluation report include:

[0034] The self-attention weight matrix in the feature cross layer is analyzed for feature importance. The top N features with the highest weights associated with the comprehensive communication effectiveness score are selected from the sub-features included in the simulated attention distribution features and design compliance features, respectively, to generate a list of key design features. N is a positive integer determined according to preset rules, and its value ranges from 3 to 6.

[0035] The comprehensive communication effectiveness score, the list of key design features and their analysis are integrated with the optimization cases in the design specification knowledge base through the report generation module to generate an initial design evaluation report containing the comprehensive communication effectiveness score, the list of key design features and their analysis.

[0036] According to the above technical solution, the steps for jointly analyzing the interference effects of the spatial layout feature matrix and the semantic content feature set include:

[0037] By calculating the spatial relationship between any two visual elements and then using a standardized coordinate system to determine the ratio of their area ratio product to the square of the Euclidean distance between their center points, the spatial interference intensity is obtained. This calculation model, based on the principles of visual perception, quantifies the degree of mutual interference between design elements due to their spatial positional relationships.

[0038] By performing attribute difference analysis on any two semantic elements, the weighted sum of their absolute differences across three dimensions—semantic importance weight, text length, and font size—is calculated. This weighted sum is then multiplied by an adjustment factor based on the semantic content similarity between the two elements to obtain the semantic interference strength. This step aims to assess the risk of information hierarchy confusion arising from attribute differences and content relevance between different text elements.

[0039] The spatial interference intensity and semantic interference intensity are linearly combined and weighted according to a preset weight to form a comprehensive interference matrix. By fusing the interference intensities of two different dimensions, spatial and semantic, into a unified comprehensive interference matrix according to preset weights, this method constructs a mathematical model that comprehensively describes all types of mutual interference relationships among design elements.

[0040] The comprehensive interference matrix is ​​processed by spectral analysis to extract the sum of the first k largest eigenvalues ​​of the matrix to obtain the global interference index. Here, k is the number of eigenvalues ​​corresponding to the first time the cumulative contribution rate of the sum of all eigenvalues ​​in the matrix reaches a preset threshold, and the minimum value of k is 1, and the maximum value is the order of the matrix. At the same time, the square root of the sum of squares of all elements in the matrix is ​​calculated to obtain the average interference intensity.

[0041] The global interference index and average interference intensity are constructed using feature vectors to generate a two-dimensional feature vector containing the global interference index and average interference intensity, thus obtaining a quantitative evaluation result of the comprehensive interference effect.

[0042] By constructing two key indicators into a two-dimensional feature vector, the complex design interference effects can be characterized by a concise and informative mathematical form.

[0043] According to the above technical solution, the step of optimizing and correcting the initial design evaluation report based on the quantitative evaluation results of comprehensive interference effects includes:

[0044] The global interference index and the comprehensive communication effectiveness score are constructed using a multi-objective optimization function, forming an optimization function with the objectives of minimizing the global interference index and maximizing the comprehensive communication effectiveness score.

[0045] The constraints set in the design specification knowledge base are processed by a constrained optimization algorithm. Under the condition of satisfying the constraints, iterative optimization is performed to obtain the optimized solution set.

[0046] Design modification schemes are extracted from the optimized solution set, including suggestions for adjusting the spatial layout and suggestions for optimizing the visual attributes of color and font.

[0047] The optimized design parameters are recalculated to obtain the optimized comprehensive communication effectiveness score and comprehensive interference effect quantitative evaluation results. The effectiveness improvement value and interference reduction value before and after optimization are used as the basis for correction. An optimization effect comparison chapter and a specific implementation guidance module are added to the initial design evaluation report, and the final design evaluation report containing quantitative improvement data and executable suggestions is output.

[0048] Secondly, this application provides a visual communication design evaluation system based on data analysis, comprising:

[0049] The data acquisition and feature extraction module is used to acquire the original design data of the visual communication design to be evaluated, and to perform parsing and format unification processing in order to extract the spatial layout feature matrix of visual elements, calculate the color distribution feature vector, and extract the semantic content feature set.

[0050] The Attention and Compliance Analysis module is used to simulate visual attention distribution based on spatial layout feature matrix and color distribution feature vector to generate simulated attention distribution features, and to perform design compliance analysis based on design specification knowledge base to generate design compliance features.

[0051] The communication effectiveness prediction module is used to construct a communication effectiveness prediction model based on the simulated attention distribution characteristics and design compliance characteristics, and generate a comprehensive communication effectiveness score.

[0052] The interference effect analysis module calculates the spatial and semantic interference intensity between elements based on the spatial layout feature matrix and semantic content feature set. It quantifies the comprehensive interference effect by forming a comprehensive interference matrix and performing spectral analysis. This module serves as a deep diagnostic tool for the system, specifically designed to detect and quantify the mutual interference and conflicts within design elements. Through precise mathematical models, it reveals deep-seated problems that are not easily detected intuitively but severely impact information transmission efficiency, providing crucial diagnostic evidence and clear directions for improvement in design optimization.

[0053] The report generation and optimization module is used to generate an initial design evaluation report based on the comprehensive communication effectiveness score, and to optimize and revise the initial design evaluation report based on the comprehensive interference effect quantitative evaluation results, and output a final design evaluation report containing quantitative improvement data and actionable suggestions.

[0054] Thirdly, this application provides an electronic device, including: a processor and a memory, wherein the memory stores a computer program that can be called by the processor, and the processor executes a visual communication design evaluation method based on data analysis by calling the computer program stored in the memory.

[0055] Fourthly, this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform a data analysis-based visual communication design evaluation method.

[0056] Compared with the prior art, this application has the following advantages and beneficial effects:

[0057] This application achieves accurate prediction of design communication effectiveness by jointly analyzing the spatial layout characteristics and visual semantic attributes of design elements and constructing a communication effectiveness prediction model based on simulated attention distribution characteristics and design compliance characteristics. It also accurately quantifies the comprehensive interference effect through joint analysis of interference effects, identifies key interference factors, and thus identifies and eliminates communication obstacles in advance during the design stage. Finally, based on the quantitative evaluation results of the comprehensive interference effect, a multi-objective optimization algorithm is used to optimize and correct the design scheme, significantly improving the accuracy of visual communication design evaluation and the effectiveness of optimization implementation. Attached Figure Description

[0058] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0059] Figure 1 This is a schematic diagram of the overall process of a visual communication design evaluation method based on data analysis provided in an embodiment of this application;

[0060] Figure 2 This is a structural schematic diagram of the detailed feature extraction process provided in the embodiments of this application;

[0061] Figure 3 This is a schematic diagram of the simulated attention distribution analysis process provided in the embodiments of this application;

[0062] Figure 4 This is a schematic diagram of the design compliance analysis process provided in the embodiments of this application;

[0063] Figure 5 This is a schematic diagram of the communication effectiveness prediction model structure provided in the embodiments of this application;

[0064] Figure 6 This is a schematic diagram of the structure of a visual communication design evaluation system based on data analysis provided in an embodiment of this application;

[0065] Figure 7 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0066] The technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solution of this application, rather than limitations thereof. In the absence of conflict, the embodiments and technical features in the embodiments can be combined with each other.

[0067] Please see Figure 1 , Figure 1 This is a schematic diagram of the overall process of a visual communication design evaluation method based on data analysis provided in an embodiment of this application, which specifically includes the following steps:

[0068] S1: Obtain the original design data of the visual communication design to be evaluated, perform feature analysis on the original design data, and obtain the spatial layout feature matrix, color distribution feature vector, and semantic content feature set.

[0069] The feature analysis steps include:

[0070] All visual elements in the raw data are identified. For structured design documents, visual element objects are extracted by parsing their internal format specifications. For raster images, visual element regions are identified using a deep learning-based instance segmentation model. For each identified visual element, its position coordinates, width, height, area percentage, and aspect ratio in a standardized coordinate system are calculated. These geometric attributes of all visual elements are then arranged in order to construct a spatial layout feature matrix.

[0071] The preprocessed image data is converted to the HSV color space. The average and standard deviation of pixel values ​​in the hue channel, saturation channel, and lightness channel are calculated for each pixel. The color entropy value of the image is calculated by constructing a color histogram after quantizing the color channel values ​​and calculating its information entropy. The average value of the hue channel, the standard deviation of the hue channel, the average value of the saturation channel, the standard deviation of the saturation channel, the average value of the lightness channel, the standard deviation of the lightness channel, and the color entropy value are concatenated in this order to form a seven-dimensional color distribution feature vector.

[0072] All text elements in the design are identified and extracted, and optical character recognition (OCR) technology is used to obtain the text content and its positional information. For each text element, its character length and font size are recorded, and its font type is represented by digital encoding. The font size, font style, and vertical position coordinates of the text elements are quantified. The font style is quantified by assigning a first coefficient to regular fonts, a second coefficient to bold fonts, a third coefficient to italic fonts, and a fourth coefficient to bold and italic fonts, wherein the second, third, and fourth coefficients are all greater than the first coefficient. The vertical position coordinates are quantified by mapping the coordinate values ​​to a continuous score from the top to the bottom of the design screen, with values ​​decreasing in descending order. The quantified font size, font style, and vertical position coordinate scores are multiplied by their corresponding weighting factors and then summed to obtain the semantic importance weight score. The character length, font size, digitally encoded font type, and semantic importance weight score of all text elements are used to construct a corresponding feature tuple for each text element, and all feature tuples of text elements are organized in list form to form a semantic content feature set.

[0073] S2: Identify the design purpose information of the visual communication design based on the original design data, and call the corresponding design specification knowledge base based on the design purpose information.

[0074] The design purpose information identification steps include: identifying the design purpose category by analyzing the text content and visual features in the original design data and using a keyword matching algorithm; specifically, extracting text elements from the design for keyword analysis, and simultaneously analyzing the dominant color features in the color distribution feature vector. When keywords such as "promotion," "discount," "special offer," and "offer" are identified and the HSV hue value of the dominant color is within the range of 0-15 degrees or 330-360 degrees of red, it is determined to be a promotional poster category; when keywords such as "brand," "trademark," "logo," and "LOGO" are identified and the color entropy value of the color distribution feature vector is less than 3.5, it is determined to be a brand image design category; when keywords such as "announcement," "notification," "public notice," and "announcement" are identified and the coefficient of variation of the font type encoding in the semantic content feature set is less than 0.2, it is determined to be a public information design category.

[0075] The steps for accessing the design specification knowledge base include: based on the identified design purpose category, retrieving the corresponding specification parameters from the preset design specification database; for the promotional poster category, retrieving the specification parameters that require a color contrast ratio of no less than 4.5:1 and a core information area ratio of no less than 30%; for the brand image design category, retrieving the specification parameters that allow a color difference range of ΔE≤5 for the brand standard color swatch and require a standard font usage rate of 100%; for the public information design category, retrieving the specification parameters that require a minimum font size of no less than 12pt and a line spacing of no less than 1.5 times the font height.

[0076] S3: Based on the spatial layout feature matrix, color distribution feature vector, semantic content feature set, and design specification knowledge base, perform simulated attention distribution analysis and design compliance analysis to generate simulated attention distribution features and design compliance features.

[0077] The simulated attention distribution analysis steps include: inputting the spatial layout feature matrix and color distribution feature vector into a pre-trained DeepGazeIIe deep visual saliency network; extracting multi-scale visual features through the network's convolutional layers; generating an initial attention heatmap through an attention mechanism layer; this heatmap represents the probability of each pixel region being noticed by the observer in the form of a probability distribution; determining the location of core information regions based on layout guidance criteria in the design specification knowledge base; generating the desired attention distribution using a Gaussian distribution model, where the distribution density of the core information regions is set to 3 times that of the surrounding regions; weighting and fusing the initial attention heatmap and the desired attention distribution, assigning a weight of 0.7 to the initial attention heatmap and a weight of 0.3 to the desired attention distribution; and generating the final attention distribution through pixel-level weighted summation; calculating the statistical characteristics of the final attention distribution, including the arithmetic mean as the average attention intensity, the square root of the mean of the sum of squares of the differences between all pixel values ​​and the average value as the standard deviation of the attention distribution, and calculating its information entropy as the attention distribution entropy value after normalizing the final attention distribution to a probability distribution. These three statistics are combined in the order of average attention intensity, standard deviation of attention distribution, and entropy of attention distribution to form a three-dimensional simulated attention distribution feature.

[0078] The design compliance analysis steps include: comparing the spatial layout feature matrix with the grid system in the design specification knowledge base to calculate element alignment, spacing uniformity, and the proportion of white space in the core area; specifically, calculating the alignment error between the element boundary box and the grid line, with an error less than 3 pixels considered aligned, and statistically analyzing the proportion of aligned elements to obtain an element alignment score; calculating the coefficient of variation of the spacing between adjacent elements, with a coefficient of variation less than 0.3 considered evenly spaced, and obtaining a spacing uniformity score; measuring the area ratio of the blank area around the core information area to obtain a white space ratio score; and weighting and summing these three scores with weights of 0.4, 0.3, and 0.3 to generate a layout specification compliance score. The primary color in the color distribution feature vector is compared with the standard brand color swatch in the design specification knowledge base, and the color difference ΔE is calculated in the CIELAB color space, with a color difference less than 5 considered compliant; simultaneously, the contrast value of the key color combination is calculated, with a contrast ratio of 4.5:1 or higher considered acceptable; and the color difference compliance and contrast compliance rate are weighted with weights of 0.6 and 0.4 respectively to generate a color system specification compliance score. The font type codes in the semantic content feature set are matched with the standard font list in the design specification knowledge base, and the proportion of text elements that match completely is calculated. Simultaneously, the font size hierarchy system is checked, requiring a minimum size difference of 2pt between different levels, and the proportion of text elements meeting the hierarchy requirement is calculated. These two proportions are averaged with weights of 0.5 each to generate a font usage specification compliance score. These three specification compliance scores are then weighted and merged with weights of 0.4 for layout, 0.35 for color system, and 0.25 for font usage to generate the final design compliance feature.

[0079] S4: Based on the simulated attention distribution characteristics and the design compliance characteristics, construct a communication effectiveness prediction model and generate a comprehensive communication effectiveness score and an initial design evaluation report.

[0080] The steps for constructing the communication effectiveness prediction model include: concatenating the three-dimensional simulated attention distribution features and scalar-valued design compliance features to form a four-dimensional combined feature vector. This combined feature vector is then input into a neural network model containing a feature cross-layer and a three-layer perceptron. The feature cross-layer uses a self-attention mechanism to calculate the interaction weights between features, where the dimensions of the query vector and key vector are set to 16, and the dimension of the value vector is set to 8. The correlation weights between features are calculated using scaled dot product attention. The multilayer perceptron contains three fully connected layers with 64, 32, and 16 neurons respectively, using the ReLU activation function. The output layer uses the Sigmoid activation function to map the output values ​​to the 0-1 range, yielding a comprehensive communication effectiveness score.

[0081] The initial design evaluation report generation steps include: extracting the attention weight matrix from the feature cross-layer; sorting each feature according to its weight value from high to low; and selecting the top four features by weight value to generate a key design feature list. The key design feature list is then matched with the optimization case library in the design specification knowledge base, assigning a corresponding optimization suggestion case to each key design feature. Finally, the comprehensive communication effectiveness score, the key design feature list and its detailed analysis, and the optimization suggestion cases are integrated into a structured initial design evaluation report. The report is stored in JSON format and includes three components: a scoring module, a feature analysis module, and an optimization suggestion module.

[0082] S5: Based on the design purpose information and design specification knowledge base, perform joint analysis of interference effects on the spatial layout feature matrix and semantic content feature set to obtain a comprehensive quantitative evaluation result of interference effects.

[0083] The joint analysis steps for interference effects include:

[0084] Analyzing the spatial relationships between visual elements quantifies the intensity of their mutual interference, reflecting the rationality of element distribution in the design layout. A higher spatial interference intensity value indicates a greater degree of mutual interference and more pronounced visual conflict, facilitating accurate identification of problem areas in the layout and guiding subsequent layout optimization. The spatial interference intensity between any two visual elements is calculated by dividing the product of their area proportions by the squared Euclidean distance between their center points.

[0085] ;

[0086] In the formula and These represent the area proportions of the i-th and j-th visual elements in the design, respectively, used to describe the visual weight of the elements in the image; This represents the Euclidean distance between the center points of the i-th and j-th visual elements in the normalized coordinate system, used to describe the degree of spatial separation between elements; This represents the spatial interference intensity between the i-th and j-th visual elements, used to quantify the degree of mutual interference between elements due to their spatial positional relationship; It is a numerically stable term, which is much smaller than typical. The positive constant is used to prevent numerical singularities when the distance between elements approaches zero, while ensuring the continuity of the interference intensity assessment.

[0087] By analyzing the attribute differences between semantic elements, the degree of content conflict between them is quantified, reflecting the clarity of information hierarchy in the design. A higher semantic interference strength value indicates greater attribute differences between semantic elements and a more chaotic information hierarchy, facilitating the model's accurate identification of conflicts in semantic expression and guiding subsequent content hierarchy optimization. The semantic interference strength between any two semantic elements is calculated by weighting the difference in semantic importance, text length, and font size, and then multiplying this sum by an adjustment factor based on semantic content similarity.

[0088] ;

[0089] In the formula and These represent the semantic importance weights of the i-th and j-th semantic elements, respectively, used to describe the importance of the elements in the information hierarchy; and These represent the text lengths of the i-th and j-th semantic elements, respectively, used to describe the amount of text information contained in the elements; and represents the font size of the i-th and j-th semantic elements, respectively, used to describe the visual prominence of the elements; w1, w2, and w3 are weight coefficients of 0.5, 0.3, and 0.2, respectively, used to adjust the contribution ratio of different attributes in semantic interference; This represents a semantic content similarity adjustment factor calculated based on word vector similarity, used to adjust the interference intensity according to the relevance of semantic content; This represents the semantic interference strength between the i-th and j-th semantic elements, used to quantify the degree of mutual interference between semantic elements due to attribute differences.

[0090] The calculated spatial and semantic interference intensities are linearly combined with preset weights of 0.6 and 0.4 to form an n-order integrated interference matrix M, where n is the total number of visual elements. The spectral characteristics of the integrated interference matrix are analyzed to quantify the overall intensity and distribution characteristics of interference effects among design elements, reflecting the concentration and scope of interference effects in the design.

[0091] Performing eigenvalue decomposition on the matrix yields the set of eigenvalues:

[0092] ;

[0093] Sort the eigenvalues ​​in descending order, calculate the cumulative contribution rate of the first m eigenvalues, select the smallest k value that satisfies the condition that the cumulative contribution rate first reaches 85%, and use the sum of the first k eigenvalues ​​as the global interference index. At the same time, calculate the square root of the sum of the squares of all elements in the matrix as the average interference intensity.

[0094] ;

[0095] ;

[0096] in, represents the i-th eigenvalue after sorting from largest to smallest, used to describe the strength of the i-th interference mode; k represents the number of eigenvalues ​​selected, which is the number of eigenvalues ​​corresponding to the first time the cumulative contribution rate of the sum of all eigenvalues ​​of the matrix reaches 85%; GII represents the global interference index, used to quantify the cumulative strength of the main interference modes; The value of the element in the i-th row and j-th column of the integrated interference matrix is ​​represented by ; AIS represents the average interference intensity, which is used to quantify the average value of the overall interference level.

[0097] S6: Based on the quantitative evaluation results of the comprehensive interference effect, the initial design evaluation report is optimized and corrected, and the final design evaluation report is output.

[0098] The performance optimization and correction steps include: constructing a multi-objective optimization function with the objective functions being min (global interference index) and max (comprehensive communication performance score). Constraints include: a color contrast ratio of no less than 4.5:1, a core information area ratio of no less than 30%, a safety margin of no less than 5%, a font size no smaller than the minimum readable font size, brand color deviation within the allowable range, and key information visibility meeting standard requirements, as set by the design specification knowledge base. The constrained NSGA-II multi-objective optimization algorithm is used, with a population size of 50, 100 iterations, a crossover probability of 0.9, and a mutation probability of 0.1. Iterative optimization is performed under all constraints to obtain the Pareto optimal solution set. Design modification schemes are extracted from the optimal solution set, including suggestions for spatial layout adjustments such as element rearrangement and spacing optimization, and suggestions for visual attribute optimization such as improving color contrast and adjusting font size. Based on the optimized design parameters, the overall communication effectiveness score and the quantitative evaluation results of the overall interference effect are recalculated. The effectiveness improvement value and interference reduction value before and after optimization are calculated. The optimization effect comparison data and specific implementation guidance plan are integrated into the initial design evaluation report, and the final design evaluation report containing quantitative improvement data and executable suggestions is output.

[0099] Please see Figure 2 , Figure 2 This is a structural schematic diagram of the detailed feature extraction process provided in this embodiment. This embodiment demonstrates the complete extraction process from the original design data to the three core features, including:

[0100] The initial design data input serves as the starting point for the entire feature extraction process, providing the system with the source of design materials to be processed. This input stage accepts design files in various formats, ensuring data consistency and compatibility in subsequent processing flows through a unified interface specification.

[0101] The spatial layout feature extraction channel establishes a standardized data processing environment by parsing and formatting the original design data. It then identifies all visual elements, including basic graphics, text blocks, and embedded images. Next, it calculates the geometric attributes of each visual element, including position coordinates, width, height, and pixel area. Finally, these geometric attributes are organized into feature vectors and combined row-wise to construct a spatial layout feature matrix. This process transforms the spatial composition of the design into a precise mathematical representation, providing a reliable quantitative basis for subsequent layout rationality analysis and spatial relationship evaluation.

[0102] The color distribution feature extraction process first converts the pixel data of the design draft from the RGB color space to the HSV color space, which better aligns with human visual perception. Then, it calculates the image's color statistical features, including the dominant hue extracted through clustering algorithms, the calculated color contrast, and the variance index of each color channel. Simultaneously, it constructs a color histogram by quantizing the color channel values ​​and calculates its information entropy value, obtaining the color entropy that characterizes color complexity and richness. Finally, all color statistical features and color entropy values ​​are concatenated in a preset order to form a complete color distribution feature vector. This process enables multi-dimensional quantitative analysis of the design's color attributes, providing comprehensive data support for evaluating the visual expressiveness and emotional communication effects of color combinations.

[0103] The semantic content feature extraction channel utilizes optical character recognition technology to accurately extract all text elements and their positional information within the design. Next, it calculates the text attributes of each text element, including character length, font size, and font type represented by a predefined encoding system. Then, a rule-based model comprehensively considers font size and thickness, spatial distribution, and matching degree with a pre-defined keyword library to scientifically evaluate the semantic importance weight of each text element. Finally, it constructs a complete feature tuple for each text element, containing character length, font size, font type encoding, and semantic importance weight, and integrates all feature tuples in a structured list format to form a standardized set of semantic content features. This processing flow not only accurately captures the visual presentation features of the text but also systematically quantifies its functional importance in the information transmission hierarchy, laying a solid foundation for in-depth analysis of the semantic structure and information transmission efficiency of the design.

[0104] The three main output features, serving as the final outcome of the feature extraction process, unify the output of the spatial layout feature matrix, color distribution feature vector, and semantic content feature set. This output stage provides high-quality feature input to subsequent modules such as attention analysis, compliance verification, performance prediction, and interference effect analysis through standardized data interfaces, ensuring the consistency and integrity of the entire evaluation system's data processing flow.

[0105] This feature extraction process establishes a technical framework for comprehensively analyzing design composition from multiple dimensions through the collaborative work of three parallel processing channels, transforming subjective visual design elements into an objective and computable feature representation system. This structured feature extraction architecture not only ensures the data completeness and analytical depth of design evaluation, but also provides a solid technical foundation for achieving accurate and reliable visual communication effectiveness evaluation.

[0106] Please see Figure 3 , Figure 3 This is a schematic diagram of the simulated attention distribution analysis process provided in this application embodiment. This embodiment demonstrates the complete computational process from multi-source input to generating simulated attention distribution features, including:

[0107] The spatial layout feature matrix and color distribution feature vector are inputs, serving as the foundational data source for visual attention simulation. These two types of features together describe the underlying visual attributes of the design. The spatial layout feature matrix encodes the geometric relationships and distribution of each visual element, while the color distribution feature vector characterizes the overall color composition and contrast of the image, providing comprehensive visual information support for subsequent visual saliency calculations.

[0108] The deep visual saliency network (DSN) processes the design image by receiving spatial layout and color feature inputs and then simulating visual attention using a pre-trained deep neural network model. Based on the cognitive mechanisms of the human visual system, this network can automatically extract multi-scale features from complex visual scenes and predict the distribution of unconscious gaze points, ultimately generating an initial attention heatmap that probabilistically represents the likelihood of attention for each pixel region. This step transforms the visual attributes of the design into attention predictions that conform to human visual cognition, providing a scientific basis for evaluating the visual guidance effect of the design.

[0109] The application of design specification knowledge base and layout guidance guidelines extracts corresponding layout guidance guidelines from the specification knowledge base based on the identified design purpose information. These guidelines clarify the professional layout requirements for different types of designs. The system uses these guidelines to determine the location and extent of core information areas and generates a desired attention distribution using a Gaussian distribution model, where the distribution density of core information areas is significantly higher than that of surrounding areas. This process transforms industry design specifications and experiential knowledge into quantified attention expectations, ensuring consistency between evaluation criteria and design goals.

[0110] Weighted fusion processing organically combines the initial attention heatmap generated by the deep visual saliency network with the desired attention distribution generated based on a canonical knowledge base. By assigning a first weight to the initial attention heatmap and a second weight to the desired attention distribution, and performing pixel-level weighted summation, a final attention distribution conforming to the design communication characteristics is formed. This fusion mechanism effectively balances the natural cognitive laws of human vision with the functional requirements of design, respecting the objective laws of visual appeal while emphasizing the degree to which the design intent is realized.

[0111] The final attention distribution output, as a direct result of the simulated attention distribution analysis, comprehensively reflects the attention distribution state considering both human visual cognitive characteristics and design specifications. This output provides an accurate visual attention representation for subsequent statistical feature calculations and communication effectiveness predictions, serving as a crucial bridge connecting underlying visual features with high-level design effectiveness evaluation.

[0112] This simulated attention distribution analysis process establishes a complete computational framework from visual features to attention prediction by introducing a dual-drive approach of a deep visual saliency network and a design specification knowledge base. This analytical architecture considers both the objective cognitive laws of the human visual system and incorporates subjective experience and knowledge from the professional design field, making the assessment of design visual guidance capabilities both scientifically objective and professionally targeted, providing a reliable basis for accurately predicting the effectiveness of design communication based on attention distribution characteristics.

[0113] Please see Figure 4 , Figure 4 This is a schematic diagram of the design compliance analysis process provided in this application embodiment. This embodiment demonstrates the generation mechanism of design compliance features. The system performs professional evaluation of different dimensions of the design through three parallel compliance analysis channels and finally integrates them to form a unified design compliance feature.

[0114] The layout compliance analysis channel is responsible for verifying the spatial arrangement of the design against regulations. This channel precisely compares the visual element spatial layout feature matrix with the grid system and safety margin specifications in the design specification knowledge base. By calculating key indicators such as element alignment, spacing uniformity, and the proportion of white space in core areas, it generates a quantifiable layout compliance score. This analysis process transforms layout principles based on aesthetic experience into calculable objective standards, ensuring that the design meets professional typography requirements in terms of spatial organization.

[0115] The Color System Compliance Analysis Channel focuses on assessing the compliance of color applications in designs. This channel calculates the color similarity between the primary and secondary colors, represented by color distribution feature vectors, and the standard brand color palette in the design specification knowledge base. It also combines this with the contrast values ​​of key color combinations and their compliance with accessibility standard thresholds, performing a multi-dimensional weighted fusion to generate a color system compliance score. This processing mechanism not only ensures the consistency of brand visual identity but also fully considers the functionality and inclusivity of color use, making color assessment both artistic and scientific.

[0116] The font compliance analysis channel systematically reviews text elements in the design. This channel matches and verifies the font type codes in the semantic content feature set with the list of authorized standard fonts in the design specification knowledge base, while also checking the consistency of the font size hierarchy system. A font usage compliance score is generated by statistically analyzing the proportion of text elements that meet the specification requirements. This analysis process effectively ensures the legality of font usage and the clarity of information hierarchy, providing technical support for establishing a standardized visual reading order.

[0117] The weighted fusion processing stage receives compliance scores from three analysis channels and integrates these scores into a unified design compliance feature using a pre-defined weighted fusion algorithm. This fusion process comprehensively considers the relative importance of different compliance dimensions in the design evaluation, assigning different weight coefficients to layout, color scheme, and font, ultimately generating core features that comprehensively reflect the overall design compliance level. This weighted fusion mechanism retains the professional characteristics of each dimension's evaluation while achieving comprehensive integration of evaluation results, providing accurate and reliable compliance input for subsequent communication effectiveness prediction.

[0118] This design compliance analysis process establishes a multi-channel parallel analysis architecture to systematically verify the compliance of design works across three core dimensions: layout, color, and typography. This structured analysis approach not only ensures the comprehensiveness and professionalism of the evaluation process but, more importantly, transforms the traditional subjective experience-based design compliance check into an objective, quantitative, and automated assessment. This significantly improves the accuracy and efficiency of design compliance judgments, laying a solid foundation for building a standardized design quality assessment system.

[0119] Please see Figure 5 , Figure 5 This is a schematic diagram of the communication effectiveness prediction model structure provided in this application embodiment. This embodiment illustrates the model calculation architecture from feature input to comprehensive score output, including:

[0120] Simulated attention distribution characteristics and design compliance characteristics serve as dual inputs for predicting communication effectiveness. Simulated attention distribution characteristics describe the visual guidance properties of the design in three-dimensional vector form, including average attention intensity, standard deviation of attention distribution, and attention distribution entropy. Design compliance characteristics, on the other hand, integrate compliance with regulations across three dimensions—layout, color scheme, and font usage—using a weighted fusion algorithm. These two types of characteristics quantify design quality from two key perspectives: visual cognitive patterns and professional design standards, respectively, together forming the multi-dimensional input foundation of the predictive model.

[0121] Feature splicing processing combines the three-dimensional simulated attention distribution features with scalar-form design compliance features to form a unified combined feature vector. This process breaks down data barriers between different feature systems, establishes a data channel between visual cognitive attributes and professional standardization attributes, and enables subsequent models to simultaneously consider the natural appeal and human compliance of the design, creating the necessary data conditions for a comprehensive evaluation of communication effectiveness.

[0122] The combined feature vector, as the result of feature concatenation, integrates the originally scattered attention and compliance features into a unified mathematical representation. This combined feature not only retains the unique information of each dimension, but more importantly, establishes the basis for the correlation between features, enabling the model to learn the synergistic effect of different feature combinations on communication effectiveness, and providing a complete data carrier for subsequent deep feature analysis.

[0123] The feature cross-processing layer performs in-depth analysis of the combined feature vectors through a self-attention mechanism, automatically capturing the complex interaction relationships between the simulated attention distribution features and the various sub-features in the design compliance features, and calculating their relative weights. This processing layer can identify the impact patterns of different feature combinations on communication effectiveness. For example, it can discover that specific attention distribution patterns have a significant promoting effect on effectiveness improvement when certain regulatory conditions are met, thereby achieving intelligent learning and modeling of nonlinear relationships between features.

[0124] The multilayer perceptron processes the weighted feature representations received from the feature cross-layers, and performs nonlinear transformations and information compression through a deep neural network containing multiple fully connected layers. This network structure, through hierarchical feature abstraction and fusion, gradually extracts the most discriminative high-order feature representations for predicting communication effectiveness from complex feature interactions. Finally, the network output is mapped to the 0-1 range using the sigmoid activation function, forming a standardized performance score.

[0125] The comprehensive communication effectiveness score, as the final result of the entire prediction process, comprehensively reflects the overall performance of the design in both visual guidance and compliance dimensions through a single scalar value. This quantitative indicator not only provides an intuitive evaluation standard for design quality, but more importantly, it establishes a calculable correlation between design attributes and communication effectiveness. This enables design optimization to shift from experience-driven to data-driven approaches, significantly improving the objectivity of design evaluation and the efficiency of decision-making.

[0126] This communication effectiveness prediction model constructs a complete mapping relationship from multi-source design features to comprehensive effectiveness scores through a progressive architecture of feature concatenation, cross-computation, and deep neural network processing. This model design considers the uniqueness of different feature systems while fully exploring the synergistic effects between features. This ensures that the final effectiveness score comprehensively reflects the overall quality of the design and accurately reveals key influencing factors, providing core computational capabilities for achieving accurate and reliable design communication effectiveness assessment.

[0127] Please see Figure 6 , Figure 6 This is a schematic diagram of a data analysis-based visual communication design evaluation system provided in this application embodiment. This embodiment provides a data analysis-based visual communication design evaluation system, including:

[0128] The data acquisition and feature extraction module 210 is used to acquire the original design data of the visual communication design to be evaluated, and to perform parsing and format unification processing in order to extract the spatial layout feature matrix of visual elements, calculate the color distribution feature vector, and extract the semantic content feature set.

[0129] The attention and compliance analysis module 220 is used to simulate visual attention distribution based on spatial layout feature matrix and color distribution feature vector to generate simulated attention distribution features, and to perform design compliance analysis based on design specification knowledge base to generate design compliance features.

[0130] The communication effectiveness prediction module 230 is used to construct a communication effectiveness prediction model based on the simulated attention distribution characteristics and design compliance characteristics, and generate a comprehensive communication effectiveness score.

[0131] The interference effect analysis module 240 is used to calculate the spatial interference intensity and semantic interference intensity between elements based on the spatial layout feature matrix and the semantic content feature set, and to generate a quantitative evaluation result of the comprehensive interference effect by forming a comprehensive interference matrix and performing spectral analysis.

[0132] The report generation and optimization module 250 is used to generate an initial design evaluation report based on the comprehensive communication effectiveness score, and to optimize and revise the initial design evaluation report based on the comprehensive interference effect quantitative evaluation results, and output a final design evaluation report containing quantitative improvement data and actionable suggestions.

[0133] In this embodiment, the data acquisition and feature extraction module 210 is used to acquire the original design data of the visual communication design to be evaluated, and to perform parsing and format unification processing to extract the spatial layout feature matrix of visual elements, calculate the color distribution feature vector, and extract the semantic content feature set, specifically including:

[0134] The original design data is parsed and format-unified to identify all visual elements and calculate their geometric attributes, thus constructing a spatial layout feature matrix. The original design data comes from structured design files or raster images. The structured design files extract visual element objects by parsing their internal format specifications, and the raster images identify visual element regions using a deep learning-based instance segmentation model. The visual elements include basic graphics, text blocks, embedded images, and icons. The spatial layout feature matrix is ​​constructed by calculating the position coordinates, width, height, area percentage, and aspect ratio of each visual element in a standardized coordinate system, and arranging these geometric attributes of all visual elements in sequence.

[0135] The original design data is preprocessed and converted to the HSV color space. The color statistical features and color entropy values ​​of the image are calculated to form a color distribution feature vector. The preprocessing includes converting the design draft pixel data from the RGB color space to the HSV color space. The color statistical features include the dominant hue extracted using a clustering algorithm, the calculated color contrast, and the color channel variance. The color entropy value is obtained by quantizing the color channel values, constructing a color histogram, and calculating its information entropy. The color distribution feature vector is a seven-dimensional vector formed by concatenating the average value of the hue channel, the standard deviation of the hue channel, the average value of the saturation channel, the standard deviation of the saturation channel, the average value of the lightness channel, the standard deviation of the lightness channel, and the color entropy value in this order.

[0136] The original design data is processed through text recognition to extract all text elements and calculate their text attributes and semantic importance weights. A feature tuple containing character length, font size, font type encoding, and semantic importance weights is constructed for each text element. All feature tuples are then structured and integrated in a list format to obtain a semantic content feature set. The text recognition utilizes optical character recognition technology to extract text content and its location information. The text attributes include character length, font size, and font type represented by a predefined encoding. The semantic importance weights are obtained through a rule-based model that comprehensively considers font size, font style quantization, vertical position coordinate quantization, and the degree of matching with a pre-defined keyword library.

[0137] In this embodiment, the attention and compliance analysis module 220 is used to generate simulated attention distribution features based on the spatial layout feature matrix and color distribution feature vector, and to perform design compliance analysis based on the design specification knowledge base to generate design compliance features, specifically including:

[0138] The spatial layout feature matrix and color distribution feature vector are processed through a pre-trained deep visual saliency network to simulate visual attention to the input design image and generate an initial attention heatmap.

[0139] The design specification knowledge base is extracted using layout guidance principles, and the core information area is determined based on the design purpose information to generate the desired attention distribution.

[0140] The initial attention heatmap and the desired attention distribution are processed by weighting and pixel-level weighted summation, where the initial attention heatmap is assigned a first weight of 0.7 and the desired attention distribution is assigned a second weight of 0.3, forming the final attention distribution that conforms to the design communication characteristics;

[0141] The final attention distribution is calculated using statistical features to obtain the average attention intensity, the standard deviation of the attention distribution, and the entropy of the attention distribution. The average attention intensity is obtained by calculating the arithmetic mean of the final attention distribution. The standard deviation of the attention distribution is obtained by calculating the square root of the sum of the squares of the differences between all values ​​in the final attention distribution and its average value. The entropy of the attention distribution is obtained by first normalizing the final attention distribution to a probability distribution and then calculating its information entropy.

[0142] The average attention intensity, the standard deviation of attention distribution, and the entropy of attention distribution are combined in a preset order through feature combination processing to form a three-dimensional simulated attention distribution feature.

[0143] By comparing and analyzing the spatial layout feature matrix of visual elements with the grid system and safety margin specifications in the design specification knowledge base, the element alignment, spacing uniformity, and white space ratio of the core area are calculated to generate a standard compliance score for the layout.

[0144] The primary and secondary colors represented by the color distribution feature vectors are weighted and fused with the standard brand color palettes in the design specification knowledge base through similarity calculation and contrast analysis, combined with accessibility standard threshold verification, to generate a color system specification compliance score.

[0145] The font type encoding in the semantic content feature set is matched and verified with the list of authorized standard fonts in the design specification knowledge base. At the same time, the consistency of the font size hierarchy system is checked, and a font usage specification compliance score is generated.

[0146] The scores for compliance with layout specifications, color system specifications, and font usage specifications are integrated using a weighted fusion algorithm to generate design compliance features.

[0147] In this embodiment, the communication effectiveness prediction module 230 is used to construct a communication effectiveness prediction model based on the simulated attention distribution characteristics and design compliance characteristics, and generate a comprehensive communication effectiveness score, specifically including:

[0148] The simulated attention distribution characteristics and design compliance characteristics are combined into a combined feature vector through feature concatenation.

[0149] The combined feature vectors are processed by a pre-trained neural network model, which includes a feature crossover layer and a multilayer perceptron. The feature crossover layer captures the interaction between the simulated attention distribution features and the sub-features in the design compliance features through a self-attention mechanism and calculates their weights.

[0150] A comprehensive communication effectiveness score is obtained from the output of the neural network model.

[0151] In this embodiment, the interference effect analysis module 240 is used to calculate the spatial interference intensity and semantic interference intensity between elements based on the spatial layout feature matrix and the semantic content feature set, and to quantify and generate a comprehensive interference effect quantitative evaluation result by forming a comprehensive interference matrix and performing spectral analysis, specifically including:

[0152] The spatial positional relationship between visual elements is analyzed to quantify the intensity of mutual interference between elements. The spatial interference intensity between any two visual elements is calculated, and the spatial interference intensity value is obtained by calculating the ratio of the product of the area ratio of the two elements to the square of the Euclidean distance between the center points of the two elements.

[0153] The degree of content conflict between semantic elements is quantified by analyzing the attribute differences between semantic elements. The semantic interference strength between any two semantic elements is calculated by calculating the weighted sum of the difference in semantic importance weight, text length difference, and font size difference, and then multiplying it by an adjustment factor based on semantic content similarity to obtain the semantic interference strength value.

[0154] The calculated spatial interference intensity and semantic interference intensity are linearly combined with preset weights of 0.6 and 0.4 to form an n-order comprehensive interference matrix M, where n is the total number of visual elements;

[0155] By analyzing the spectral characteristics of the integrated interference matrix, the overall strength and distribution characteristics of the interference effect between design elements are quantified. The integrated interference matrix is ​​decomposed into eigenvalues ​​to obtain a set of eigenvalues. The eigenvalues ​​are sorted from largest to smallest, and the cumulative contribution rate of the first m eigenvalues ​​is calculated. The value of k is an integer from 1 to the matrix order n. The smallest k value that satisfies the requirement that the cumulative contribution rate first reaches 85% is selected. The sum of the first k eigenvalues ​​is used as the global interference index. At the same time, the square root of the sum of squares of all elements in the matrix is ​​calculated as the average interference intensity.

[0156] The global interference index and the average interference intensity are combined into a two-dimensional feature vector to obtain a quantitative evaluation result of the comprehensive interference effect.

[0157] In this embodiment, the report generation and optimization module 250 is used to generate an initial design evaluation report based on the comprehensive communication effectiveness score, and to optimize and correct the initial design evaluation report based on the comprehensive interference effect quantitative evaluation results, outputting a final design evaluation report containing quantitative improvement data and actionable suggestions, specifically including:

[0158] By analyzing the feature importance of the self-attention weight matrix in the feature cross layer, the top N features with the highest correlation weight with the comprehensive communication effectiveness score are selected from the sub-features included in the simulated attention distribution features and design compliance features, respectively, to generate a list of key design features; where N is a positive integer determined according to preset rules, and its value ranges from 3 to 6.

[0159] The report generation module integrates the comprehensive communication effectiveness score, the list of key design features and their analysis with optimization cases from the design specification knowledge base to generate an initial design evaluation report.

[0160] Construct a multi-objective optimization function with the goal of minimizing the global interference index and maximizing the comprehensive communication effectiveness score;

[0161] An optimization algorithm with constraints is used to iteratively find the optimal solution set under the constraints set by the design specification knowledge base.

[0162] Design modification schemes are extracted from the optimized solution set, including suggestions for adjusting the spatial layout and suggestions for optimizing the visual attributes of color and font;

[0163] Based on the optimized design parameters, the overall communication effectiveness score and the quantitative evaluation results of the overall interference effect are recalculated. The effectiveness improvement value and interference reduction value before and after optimization are used as the basis for correction. An optimization effect comparison chapter and a specific implementation guidance module are added to the initial design evaluation report. The final design evaluation report containing quantitative improvement data and actionable suggestions is output.

[0164] The steps for implementing the corresponding functions of each parameter and unit module in the data analysis-based visual communication design evaluation system of this application can be referred to the parameters and steps in the above-described embodiment of the data analysis-based visual communication design evaluation method, and will not be repeated here.

[0165] Please see Figure 7 Embodiments of the present invention also provide an electronic device 300, including a memory 310, a processor 320, and a communication bus 330; the memory 310 and the processor 320 are connected via the communication bus 330. The memory 310 stores a visual communication design evaluation method based on data analysis, which can be loaded and executed by the processor 320 as provided in the above embodiments.

[0166] The memory 310 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 310 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function, and instructions for implementing a data analysis-based visual communication design evaluation method provided in the above embodiments; the data storage area may store data involved in the data analysis-based visual communication design evaluation method provided in the above embodiments.

[0167] Processor 320 may include one or more processing cores. Processor 320 executes instructions, programs, code sets, or instruction sets stored in memory 310, and calls data stored in memory 310 to perform various functions and process data as described in this application. Processor 320 may be at least one of the following: Application-Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), Central Processing Unit (CPU), controller, microcontroller, and microprocessor. It is understood that, for different devices, the electronic devices used to implement the functions of processor 320 may also be other types, and this application embodiment does not specifically limit the specific implementation.

[0168] The communication bus 330 may include a path for transmitting information between the aforementioned components. The communication bus 330 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 330 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 The symbol is represented by a single double arrow, but this does not mean that there is only one bus or one type of bus.

[0169] This application provides a computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described in the above embodiments, a visual communication design evaluation method based on data analysis.

[0170] In this embodiment, a computer-readable storage medium can be a tangible device that holds and stores instructions used by an instruction execution device. A computer-readable storage medium can be, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof. Specifically, a computer-readable storage medium can be a portable computer disk, a hard disk, a USB flash drive, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), spoofing random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory stick, floppy disk, optical disk, magnetic disk, mechanical encoding device, or any combination thereof.

[0171] 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 process, method, article, or apparatus.

[0172] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing application concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions claimed in this application.

Claims

1. A data-driven visual communication design evaluation method, characterized in that, Includes the following steps: S1: Obtain the original design data of the visual communication design to be evaluated, perform feature analysis, and obtain the spatial layout feature matrix, color distribution feature vector, and semantic content feature set; S2: Identify the design purpose information of the visual communication design based on the original design data, and call the corresponding design specification knowledge base based on the design purpose information; S3: Based on the spatial layout feature matrix, color distribution feature vector, semantic content feature set, and design specification knowledge base, perform simulated attention distribution analysis and design compliance analysis to generate simulated attention distribution features and design compliance features; Based on the spatial layout feature matrix, color distribution feature vector, semantic content feature set, and design specification knowledge base, simulated attention distribution analysis is performed to generate simulated attention distribution features, including the following steps: S310: Based on the spatial layout feature matrix and color distribution feature vector, a pre-trained deep visual saliency network is applied to simulate visual attention to the input design image to generate an initial attention heatmap; S320: Based on the design specification knowledge base, obtain layout guidance criteria corresponding to the design purpose; determine the core information area according to the layout guidance criteria, and generate the desired attention distribution; S330: The initial attention heatmap and the desired attention distribution are weighted and summed at the pixel level by assigning a first weight to the initial attention heatmap and a second weight to the desired attention distribution to form a final attention distribution that conforms to the design communication characteristics. S340: The average attention intensity is obtained by calculating the arithmetic mean of the final attention distribution; the standard deviation of the attention distribution is obtained by calculating the square root of the mean of the sum of squares of the differences between all values ​​in the final attention distribution and their average value; the attention distribution entropy value is obtained by first normalizing the final attention distribution to a probability distribution and then calculating its information entropy. S350: The average attention intensity, the standard deviation of the attention distribution, and the entropy value of the attention distribution are combined as sub-features in a preset order to form a three-dimensional simulated attention distribution feature; S4: Based on the simulated attention distribution characteristics and the design compliance characteristics, construct a communication effectiveness prediction model and generate a comprehensive communication effectiveness score and an initial design evaluation report; Based on the simulated attention distribution characteristics and the design compliance characteristics, a communication effectiveness prediction model is constructed to generate a comprehensive communication effectiveness score, including: S410: The simulated attention distribution features and the design compliance features are concatenated to form a combined feature vector; S420: Input the combined feature vector into a pre-trained neural network model; the neural network model includes a feature cross layer and a multilayer perceptron, wherein the feature cross layer captures the interaction relationship between the simulated attention distribution features and the sub-features in the design compliance features through a self-attention mechanism and calculates their weights; S430: Outputs a comprehensive communication effectiveness score from the neural network model; S5: Based on the design purpose information and design specification knowledge base, perform joint analysis of interference effects on the spatial layout feature matrix and semantic content feature set to obtain a comprehensive quantitative evaluation result of interference effects; S6: Based on the quantitative evaluation results of the comprehensive interference effect, the initial design evaluation report is optimized and corrected, and a final design evaluation report containing quantitative improvement data and executable suggestions is output.

2. The visual communication design evaluation method based on data analysis according to claim 1, characterized in that, Obtain the original design data of the visual communication design to be evaluated and perform feature analysis. The steps include: S110: By parsing and unifying the format of the original design data, all visual elements are identified and their geometric properties are calculated to construct a spatial layout feature matrix; S120: By preprocessing the original design data and converting it to the HSV color space, the color statistical characteristics and color entropy values ​​of the image are calculated to form a color distribution feature vector; S130: By performing text recognition on the original design data, extract all text elements and calculate their text attributes and semantic importance weights. Construct feature tuples containing character length, font size, font type encoding and semantic importance weights for each text element, and integrate all feature tuples in a structured list to obtain a semantic content feature set.

3. The visual communication design evaluation method based on data analysis according to claim 2, characterized in that, Based on the aforementioned spatial layout feature matrix, color distribution feature vector, semantic content feature set, and design specification knowledge base, compliance analysis is performed to generate design compliance features, including the following steps: S360: By comparing the spatial layout feature matrix of visual elements with the grid system and safety margin specifications in the design specification knowledge base, the alignment of elements, spacing uniformity and white space ratio of core areas are calculated to generate a standard compliance score for the layout. S370: By calculating the color similarity between the primary and secondary colors represented by the color distribution feature vector and the standard brand color palette in the design specification knowledge base, and combining the contrast value of the key color combination composed of the primary and secondary colors with the degree of conformity to the accessibility standard threshold, a weighted fusion is performed to generate a color system specification conformity score. S380: By matching and verifying the font type encoding in the semantic content feature set with the list of authorized standard fonts in the design specification knowledge base, and at the same time checking the consistency of the font size hierarchy system, a font usage specification compliance score is generated. S390: By using a weighted fusion algorithm, the compliance scores of the layout, the color system, and the font usage are integrated as sub-features to generate design compliance features.

4. The visual communication design evaluation method based on data analysis according to claim 3, characterized in that, Based on the comprehensive communication effectiveness score, an initial design evaluation report is generated, including: S440: By reading the self-attention weight matrix in the feature cross layer, select the top N features with the highest correlation weight with the comprehensive communication effectiveness score from each sub-feature contained in the simulated attention distribution feature and the design compliance feature, respectively, and generate a key design feature list; where N is a positive integer determined according to preset rules, and its value ranges from 3 to 6; S450: Combine optimization cases from the design specification knowledge base to generate an initial design evaluation report that includes a comprehensive communication effectiveness score, a list of key design features and their analysis.

5. The visual communication design evaluation method based on data analysis according to claim 4, characterized in that, A joint analysis of the interference effects is performed on the spatial layout feature matrix and semantic content feature set to quantitatively evaluate the comprehensive interference effect of design elements on the transmission of core information, including: S510: Spatial interference intensity is obtained by calculating the ratio of the product of the area proportions of any two visual elements in the normalized coordinate system to the square of the Euclidean distance between the center points of the two elements. S520: The semantic interference intensity is obtained by calculating the weighted sum of the absolute differences between any two semantic elements in three dimensions: semantic importance weight, text length, and font size, and multiplying the weighted sum by an adjustment factor based on the semantic content similarity between the two elements. S530: The calculated spatial interference intensity and semantic interference intensity are linearly combined according to preset weights to form a comprehensive interference matrix; S540: Perform spectral analysis on the comprehensive interference matrix, extract the sum of the first k largest eigenvalues ​​of the matrix to obtain the global interference index, where k is the number of eigenvalues ​​corresponding to the first time the cumulative contribution rate of the sum of all eigenvalues ​​in the matrix reaches a preset threshold, and the minimum value of k is 1, and the maximum value is the order of the matrix; and calculate the square root of the sum of squares of all elements of the matrix to obtain the average interference intensity. S550: Generate a two-dimensional feature vector containing the global interference index and the average interference intensity to obtain the quantitative evaluation result of the comprehensive interference effect.

6. The visual communication design evaluation method based on data analysis according to claim 5, characterized in that, Based on the quantitative evaluation results of the comprehensive interference effect, the initial design evaluation report is optimized and revised to output the final design evaluation report, including: S610: Construct a multi-objective optimization function with the objectives of minimizing the global interference index and maximizing the comprehensive communication effectiveness score; S620: An optimization algorithm with constraints is used to iteratively find the optimal solution set under the constraints set by the design specification knowledge base. S630: Extract design modification schemes from the optimized solution set, the schemes including suggestions for adjusting the spatial layout and suggestions for optimizing the visual attributes of color and font; S640: Based on the optimized design parameters, recalculate the overall communication effectiveness score and the quantitative evaluation results of the overall interference effect. Use the performance improvement value and interference reduction value before and after optimization as the basis for correction. Add a section on the comparison of optimization effects and a specific implementation guidance module to the initial design evaluation report. Output a final design evaluation report containing quantitative improvement data and executable suggestions.

7. A data analysis-based visual communication design evaluation system, applied to the data analysis-based visual communication design evaluation method according to any one of claims 1-6, characterized in that, The system includes: The data acquisition and feature extraction module is used to acquire the original design data of the visual communication design to be evaluated, and to perform parsing and format unification processing in order to extract the spatial layout feature matrix of visual elements, calculate the color distribution feature vector, and extract the semantic content feature set. The attention and compliance analysis module is used to simulate visual attention distribution based on spatial layout feature matrix and color distribution feature vector to generate simulated attention distribution features, and to perform design compliance analysis based on design specification knowledge base to generate design compliance features. The communication effectiveness prediction module is used to construct a communication effectiveness prediction model based on the simulated attention distribution characteristics and design compliance characteristics, and generate a comprehensive communication effectiveness score. The interference effect analysis module is used to calculate the spatial interference intensity and semantic interference intensity between elements based on the spatial layout feature matrix and the semantic content feature set, and to generate a quantitative evaluation result of the comprehensive interference effect by forming a comprehensive interference matrix and performing spectral analysis. The report generation and optimization module is used to generate an initial design evaluation report based on the comprehensive communication effectiveness score, and to optimize and revise the initial design evaluation report based on the comprehensive interference effect quantitative evaluation results, and output a final design evaluation report containing quantitative improvement data and actionable suggestions.

8. An electronic device, comprising: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; characterized in that the processor executes a data analysis-based visual communication design evaluation method as described in any one of claims 1-6 by calling the computer program stored in the memory.

Images