Personalized image content generation and optimization method and system fusing generative AI

Abstract

The invention provides a personalized image content generation and optimization method and system fusing generative AI, and relates to the technical field of AI. The personalized image content generation and optimization method comprises the steps of analyzing a historical interaction track and a current intention expression of a target user based on a deep perception network, and constructing a multi-dimensional behavior portrait; user groups with similar generation preferences are identified by using a group feature recursive quantization technology, and a group wisdom feature map is constructed; performing multi-level mapping analysis on the user features and the group wisdom feature map, anchoring the group affiliation relationship of the target user and concretizing the guide features; constructing a multi-level guide vector according to current intention expression and guide features, performing accurate mapping on hidden space expression of an image generation model, quantizing deviation in real time and triggering intelligent calibration; and reconstructing a group distribution structure based on the evaluation data of the target user to realize dynamic evolution of the features. According to the invention, personalized demands of users can be accurately grasped, and the accuracy of image generation and the satisfaction degree of the users are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of AI technology, in particular to a personalized image content generation and optimization method and system fusing generative AI. BACKGROUND

[0002] With the rapid development of artificial intelligence technology, generative AI has become an important tool in the field of image content creation. Deep learning architectures such as diffusion models and generative adversarial networks enable machines to generate high-quality image content based on text descriptions or other inputs. However, existing image generation techniques mainly target general scenarios, lacking deep understanding and adaptation capabilities for individual user characteristics and preferences. When using image generation services, users often need to go through multiple rounds of repeated adjustments to obtain the desired results, which not only reduces the efficiency of creation, but also affects the user experience.

[0003] In order to improve the level of personalization of image generation, researchers have begun to explore the technical path of combining user behavior analysis with generative models. By collecting user's historical interaction data, including input prompt words, selected style parameters, feedback on generated results, etc., the user's creation preferences can be inferred to some extent. At the same time, recommendation algorithms such as collaborative filtering are also introduced into generative AI systems, trying to draw on the experience of user groups with similar preferences to optimize the generation effect of individual users. However, these exploratory work is still in its infancy, and there are many limitations in terms of technical implementation and application effect.

[0004] Existing methods are rough in characterizing user characteristics, usually only based on simple label classification or single-dimensional preference records, which cannot fully capture the complex intentions and diverse needs of users in different creation scenarios, resulting in insufficient accuracy and completeness of personalized modeling, making it difficult to provide effective support for subsequent generation optimization. SUMMARY

[0005] The embodiment of the present application provides a personalized image content generation and optimization method and system fusing generative AI, which can solve the problems in the prior art.

[0006] The first aspect of the embodiment of the present application provides a personalized image content generation and optimization method fusing generative AI, comprising: Based on the deep perception network, the historical interaction track and the current intention expression of the target user are analyzed, and a multi-dimensional behavior portrait is constructed in synchronization; based on the multi-dimensional behavior portrait, a user group with similar generation preferences is identified using group feature recursive quantization technology, and the distribution structure of the generation guide features of each group in high-dimensional space is described, and a group wisdom feature map is constructed; The user features contained in the historical interaction trajectory are multi-level mapped and analyzed with the wisdom of the crowd feature map, the group affiliation of the target user is anchored, and the guiding features corresponding to the affiliation group are visualized based on the deep feature perspective technology; According to the current intention expression and the guiding features, a multi-level guiding vector is constructed and the hidden space representation of the image generation model is accurately mapped, the deviation degree of the generation result and the guiding vector is quantified in real time, and when the deviation exceeds the preset deviation threshold, the vector intelligent calibration is triggered, and the generation result meeting the double constraints is output; The generation result is associated and projected with the wisdom of the crowd feature map, the distribution structure of the corresponding group is reconstructed based on the evaluation data of the target user, and the dynamic evolution of the features is realized.

[0007] Based on the multi-dimensional behavior portrait, the user groups with similar generation preferences are identified using group feature recursive quantization technology, and the distribution structure of the generation guiding features of each group in high-dimensional space is described, and a wisdom of the crowd feature map is constructed, including: Based on the multi-dimensional behavior portrait, the user groups with similar generation preferences are identified using group feature recursive quantization technology, the evolution chain of behavior trajectory is described, the resonance relationship and entropy distribution between trajectories are captured, the periodic commonality and turning point of user behavior mode in the time and space dimensions are dynamically analyzed, and the boundary profile and internal cohesion of the user group are determined according to the natural community formed by the resonance relationship in the feature space; Based on the boundary profile and the internal cohesion, the feature gravity and behavior spectrum of each user group in the multi-dimensional behavior portrait space are described, the dominant tendency and energy distribution proportion of each group in the generation guiding dimension are deconstructed, and the high-dimensional distribution structure of the user group generation guiding features is constructed. The high-dimensional distribution structure is mapped as a wisdom of the crowd feature map.

[0008] Based on the multi-dimensional behavior portrait, the user groups with similar generation preferences are identified using group feature recursive quantization technology, the evolution chain of behavior trajectory is described, the resonance relationship and entropy distribution between trajectories are captured, including: Based on the multi-dimensional behavior portrait, the user groups with similar generation preferences are identified using group feature recursive quantization technology, a recursive graph matrix of high-dimensional phase orbit is constructed, the recurrence mode and self-similar structure of the orbit in phase space are quantified through the recursive graph matrix, and the evolution chain and bifurcation features of user behavior trajectory are described according to the recurrence mode and the self-similar structure; Based on the evolution chain and the bifurcation features, the alignment degree of the diagonal line structure and the texture similarity between different user recursive graph matrices are calculated, and the resonance relationship between trajectories is captured through the combination analysis of the alignment degree of the diagonal line structure and the texture similarity; For the user group with significant resonance relationship, the Shannon entropy and permutation entropy of the recurrence plot matrix of the user group are calculated respectively, and the entropy value distribution of the user group is obtained by mapping the Shannon entropy and the permutation entropy in the feature space.

[0009] The user features contained in the historical interaction trajectory are mapped and analyzed at multiple levels with the collective wisdom feature map, the group affiliation of the target user is anchored, and the guiding features corresponding to the affiliation group are visualized based on the deep feature perspective technology, including: The user features contained in the historical interaction trajectory are mapped and analyzed at multiple levels with the collective wisdom feature map, the group affiliation of the target user is anchored, and the guiding features corresponding to the affiliation group are visualized based on the deep feature perspective technology, including: According to the group affiliation, the corresponding node in the collective wisdom feature map is located, the distribution parameters, evolution chain features and entropy labels contained in the corresponding node are deconstructed, the abstract guiding feature representation of the affiliation group is established by generating the deep perspective of the guiding feature space; Based on the abstract guiding feature representation, the generative preference of the group commonality is visualized as guiding parameters in the operation dimension, the dominant path deconstruction node and parameter strategy in the evolution chain feature are fused, and the visualized expression of the guiding feature is formed.

[0010] By constructing the multi-order projection link of the feature space, the semantic space and the topological space, the affinity relationship between the user feature vector and each group node is quantified, and the affiliation probability distribution is constructed based on the affinity relationship, including: By constructing the multi-order projection link of the feature space, the semantic space and the topological space, the affinity resonance degree of the user feature vector and the distribution parameters of each group node in the feature space is described to obtain the feature affinity, the semantic representation of the user feature vector and the cognitive synergy strength between the dominant guiding features of each group node in the semantic space are described to obtain the semantic affinity, and the coupling depth of the manifold distance of the user feature vector in the structure of the collective wisdom feature map and the topological centrality of each group node in the topological space is constructed to obtain the topological affinity. Based on the entropy labels of each group node, the expression tendency of the user feature vector in different cognitive levels is revealed, the synergy mechanism of the feature, semantic and topological triple affinity is constructed through the internal consistency of the entropy labels, the contribution proportion of each space affinity in different cognitive dimensions is guided according to the synergy mechanism, and a complete affinity relationship is obtained. The affinity relationship is condensed into the affiliation probability distribution in the probability space, so that the adjacent groups form a potential energy difference in the affiliation determination process, the fuzzy area of the group boundary is eliminated, and the affiliation probability distribution with clear dominant direction is obtained.

[0011] Constructing a multi-level guide vector and accurately mapping the hidden space representation of the image generation model, real-time quantifying the deviation degree of the generation result and the guide vector, triggering the vector intelligent calibration when the deviation breaks through the preset deviation threshold, and outputting the generation result meeting the double constraints including: Constructing a multi-level guide vector, constructing a deep resonance relationship based on the multi-level guide vector and the hidden space representation of the image generation model, and driving the image generation model to form an intermediate generation result; Condensing the hidden layer feature representation from the intermediate generation result, projecting the hidden layer feature representation to the multi-dimensional creative space constructed by the multi-level guide vector, and depicting the image projection of each level; According to the creative deviation degree of the intermediate generation result and the multi-level guide vector, constructing the level resonance strength based on the dependence association between levels, tuning the overall deviation degree based on the level resonance strength, and activating the vector calibration mechanism when the overall deviation degree breaks through the preset deviation threshold; Executing the vector calibration mechanism, locating the dominant deviation level and harmonizing the corresponding guide vector, constraining the harmonization amplitude based on the stability index, reconstructing the multi-level guide vector, establishing a new resonance relationship between the reconstructed multi-level guide vector and the hidden space representation, and repeatedly executing the deviation perception and calibration process until the overall deviation degree reaches the expected harmonious state, and outputting the generation result meeting the double constraints.

[0012] Constructing the level resonance strength based on the dependence association between levels, and tuning the overall deviation degree based on the level resonance strength, and activating the vector calibration mechanism when the overall deviation degree breaks through the preset deviation threshold, including: Constructing the level resonance strength based on the dependence association between level feature vectors, revealing the activity degree of information transmission between levels according to the level resonance strength through the information flow entropy ratio between level features, the activity degree reflecting the transmission depth and diffusion law of features between levels, depicting the guiding influence between levels based on the transmission depth and the diffusion law, and forming a dynamic evolution chain of level features according to the guiding influence; Based on the dynamic evolution chain, the level resonance strength is harmonically deconstructed, the stability of feature transmission is described by capturing the dominant frequency of resonance fluctuation, the deviation point in the creation process is identified based on the dominant frequency, and the vector calibration mechanism is activated when the overall deviation degree caused by the deviation point breaks through the preset deviation threshold.

[0013] In a second aspect of the embodiments of the present application, a personalized image content generation and optimization system integrating generative AI is provided, including: The first unit is configured to analyze the historical interaction track and the current intention expression of a target user based on a deep perception network, and to construct a multi-dimensional behavior portrait synchronously; and to identify a user group with similar generation preferences based on a group feature recursive quantization technology based on the multi-dimensional behavior portrait, and to depict the distribution structure of generation guide features of each group in a high-dimensional space, and to construct a group wisdom feature map. The second unit is configured to perform multi-level mapping analysis on the user features contained in the historical interaction track and the group wisdom feature map, to anchor the group affiliation of the target user, and to visualize the guide features corresponding to the affiliation group based on a deep feature perspective technology. The third unit is configured to construct a multi-level guide vector based on the current intention expression and the guide features, and to perform accurate mapping on the hidden space representation of the image generation model, to quantize the deviation degree of the generation result and the guide vector in real time, and to trigger vector intelligent calibration when the deviation exceeds a preset deviation threshold, and to output a generation result meeting double constraints. The fourth unit is configured to associate and project the generation result with the group wisdom feature map, to reconstruct the distribution structure of the corresponding group based on the evaluation data of the target user, and to realize dynamic evolution of the features.

[0014] In a third aspect, an electronic device is provided, including: a processor; a memory for storing processor-executable instructions; The processor is configured to invoke the instructions stored in the memory to execute the method described above.

[0015] In a fourth aspect, a computer-readable storage medium is provided, which stores computer program instructions, and the computer program instructions are executed by a processor to implement the method described above.

[0016] The beneficial effects of the present application are as follows: The present application can comprehensively and accurately understand the personalized needs and preference features of the user by analyzing the historical interaction track and the current intention expression of the target user based on the deep perception network, and constructing a multi-dimensional behavior portrait synchronously, and can provide accurate user feature basis for subsequent image content generation, and effectively improves the pertinence and accuracy of personalized image generation.

[0017] The present application can fully utilize the collective experience and knowledge of the user group with similar preferences by combining individual user features with group wisdom through constructing a group wisdom feature map and performing multi-level mapping analysis, and simultaneously performing precise mapping and real-time deviation quantification on the hidden space representation of the image generation model through multi-level guide vectors, triggering an intelligent calibration mechanism when the deviation breaks through the preset threshold, ensuring that the generation result meets both the current intention of the user and the group preference constraints, significantly improving the quality of image generation and user satisfaction.

[0018] The present application establishes an association projection mechanism between the generation result and the group wisdom feature map, can dynamically reconstruct the distribution structure of the corresponding group based on the evaluation data of the target user, realize the continuous evolution and optimization of the features, make the system have the ability of self-learning and iterative improvement, and can continuously optimize the generation effect with user feedback, maintain the long-term effectiveness and adaptability of the technical scheme. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 The flowchart of the personalized image content generation and optimization method of the present application embodiment fused with generative AI is shown in the figure. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical scheme and advantages of the present application embodiment clearer, the technical scheme in the present application embodiment will be described clearly and completely below in combination with the drawings in the present application embodiment. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0021] The technical scheme of the present application will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in some embodiments.

[0022] Figure 1 The flowchart of the personalized image content generation and optimization method of the present application embodiment fused with generative AI is shown in the figure. Figure 1 As shown in the figure, the method comprises: Based on the deep perception network, the historical interaction track and the current intention expression of the target user are analyzed, and a multi-dimensional behavior portrait is constructed in synchronization; based on the multi-dimensional behavior portrait, the user group with similar generation preferences is identified by using group feature recursive quantization technology, and the distribution structure of the generation guide features of each group in the high-dimensional space is described, and a group wisdom feature map is constructed; The user features contained in the historical interaction trajectory are mapped and analyzed at multiple levels with the wisdom of the crowd feature map, the group affiliation of the target user is anchored, and the guiding features corresponding to the group affiliation are visualized based on the deep feature perspective technology; According to the current intention expression and the guiding features, a multi-level guiding vector is constructed and the hidden space representation of the image generation model is accurately mapped, the deviation degree of the generated result and the guiding vector is quantified in real time, and when the deviation exceeds the preset deviation threshold, the vector intelligent calibration is triggered, and the generated result meeting the double constraints is output; The generated result is associated and projected with the wisdom of the crowd feature map, the distribution structure of the corresponding group is reconstructed based on the evaluation data of the target user, and the dynamic evolution of the features is realized.

[0023] In an optional implementation, based on the multi-dimensional behavior portrait, a group of users with similar generation preferences is identified using a group feature recursive quantization technology, and the distribution structure of the generation guiding features of each group in high-dimensional space is depicted, and a wisdom of the crowd feature map is constructed, including: Based on the multi-dimensional behavior portrait, a group of users with similar generation preferences is identified using a group feature recursive quantization technology, the evolution chain of the behavior trajectory is depicted, the resonance relationship and entropy distribution between the trajectories are captured, the periodic commonality and turning point of the user behavior pattern in the time and space dimensions are dynamically analyzed, and the boundary profile and internal cohesion of the user group are determined according to the natural community formed by the resonance relationship in the feature space; Based on the boundary profile and the internal cohesion, the feature gravity and behavior spectrum of each user group in the multi-dimensional behavior portrait space are depicted, the dominant tendency and energy distribution proportion of each group in the generation guiding dimension are deconstructed, and the high-dimensional distribution structure of the user group generation guiding features is constructed. The high-dimensional distribution structure is mapped as a wisdom of the crowd feature map.

[0024] The process of identifying a group of users with similar generation preferences based on the multi-dimensional behavior portrait includes multiple detailed technical steps. The system first structures the multi-dimensional behavior portrait data of the user into a time sequence feature matrix, with the rows representing time steps and the columns representing behavior feature dimensions. The time step is set to 1 hour as the basic unit, and each user retains the behavior data of the last 30 days to form a 720x128 feature matrix. The behavior feature dimensions include image generation request frequency, modification operation times, style switching frequency, color adjustment amplitude, composition preference index, completion score, interaction dwell time, and other key indicators.

[0025] The implementation of recursive quantification analysis requires the reconstruction of the phase space of each user's behavior trajectory. The time-delay embedding method is used, with the embedding dimension m set to 5 and the delay time τ set to 3 time units. During the phase space reconstruction process, the original one-dimensional time series x(t) is converted into an m-dimensional vector sequence, and the i-th reconstructed vector is [x(i), x(i+τ), x(i+2τ), x(i+3τ), x(i+4τ)]. The reconstructed trajectory forms a complex geometric structure in the 5-dimensional phase space, reflecting the intrinsic dynamic characteristics of user behavior.

[0026] The construction of the recurrence matrix requires the calculation of the distance relationship between any two points in the phase space. The Euclidean distance is used to measure the distance between the i-th and j-th vector points in the reconstructed trajectory, and the distance value d(i,j) is calculated. The determination of the distance threshold ε uses the fixed recurrence rate method, and the threshold is set to maintain the overall recurrence rate at 5%. When d(i,j) is less than the threshold ε, the recurrence matrix R(i,j) takes the value of 1, otherwise it takes the value of 0. The dimension of the recurrence matrix is equal to the length of the reconstructed trajectory, usually a 715x715 square matrix.

[0027] The depiction of the behavior trajectory evolution chain is achieved through multi-scale time window analysis. Three different scale time windows are set, with a short-term window of 6 hours, a medium-term window of 3 days, and a long-term window of 10 days. The statistical quantities of user behavior characteristics are calculated within each window, including mean, standard deviation, maximum, minimum, and quartile descriptive statistics. The feature vector of the evolution chain node is composed of these statistical quantities, forming a 15-dimensional node feature representation. The nodes of adjacent time windows are connected by calculating the Mahalanobis distance between the feature vectors, and the nodes with a distance less than the adaptive threshold have an evolution connection.

[0028] The capture of inter-trajectory resonance relationships uses the cross-correlation function analysis method to calculate the correlation coefficient of any two user behavior sequences at different time lags, with the lag range set to -24 to +24 hours. The peak position of the cross-correlation function represents the time synchronization of two user behavior patterns, and the peak value represents the resonance strength. When the maximum value of the cross-correlation function exceeds 0.6 and the corresponding lag time is less than 6 hours, it is determined that there is a strong resonance relationship between the two users. The resonance network is constructed by connecting all user pairs of resonance relationships into a weighted undirected graph, with nodes representing users and edge weights representing resonance strength values.

[0029] The calculation of entropy value distribution adopts the multi-scale sample entropy method, and the user behavior sequence is coarsely granulated, and the scale factor is gradually increased from 1 to 20. For the scale factor s, the original sequence is divided into non-overlapping segments with a length of s, and the average value of the values in each segment is taken to form a coarsely granulated sequence. The calculation of sample entropy needs to set the pattern length m to 2 and the tolerance r to 0.15 times the standard deviation of the sequence. The system counts the number of patterns with lengths of m and m+1 in the sequence, calculates the negative logarithm of the conditional probability to obtain the sample entropy value. The multi-scale sample entropy curve reflects the complexity characteristics of the user behavior at different time scales.

[0030] The dynamic analysis of the temporal and spatial dimension periodicity commonality of the user behavior mode is realized by the wavelet packet decomposition technology, and the db4 wavelet in the Daubechies wavelet family is selected as the analysis wavelet, and the decomposition layer is set to 6 layers. The wavelet packet decomposition can decompose the behavior signal into different frequency components, and each frequency component corresponds to a specific periodic mode. The system calculates the energy proportion of each frequency component, and the component with an energy proportion exceeding 8% is identified as a significant periodic mode. The period length is calculated by the center frequency of the frequency component, and the period intensity is quantified by the energy proportion of the component.

[0031] The turning point recognition adopts the change point detection algorithm, and the cumulative sum test method is used to detect the statistical characteristic mutation point in the behavior sequence. The algorithm calculates the cumulative deviation statistic of the sequence, and when the statistic exceeds the critical value corresponding to the confidence level of 95%, it is marked as a turning point. The type of the turning point is classified by analyzing the changes of the mean and variance of the sequence in the time period before and after the turning point. The turning point with a significantly increased mean and a decreased variance is marked as an upward stable type, and the turning point with a significantly decreased mean and an increased variance is marked as a downward fluctuation type.

[0032] The process of forming natural communities in the feature space through resonance relationship is realized by the community discovery algorithm, and the Louvain algorithm is used for community division of the resonance network. The algorithm finds the optimal community structure by maximizing the modularity function. The modularity calculation considers the difference between the internal connection density of the network and the expected connection density of the random network, and the modularity value ranges from -1 to 1, and the larger the value, the more obvious the community structure. In the iterative process of the algorithm, adjacent communities are gradually merged until the modularity no longer increases. Each detected community corresponds to a user group, and the users in the community have similar generation preference characteristics.

[0033] The determination of the boundary contour of the user group adopts the concave hull algorithm, calculates the set of coordinate points of the users in each group in the feature space, and then applies the alpha-shape algorithm to construct the group boundary. The selection of the alpha parameter is determined through cross-validation, so that the boundary can both contain the main distribution area of the group and not overfit the outliers. The boundary generated by the concave hull algorithm is a polygon contour, and the contour vertex coordinates are stored as an ordered point sequence. The complexity of the boundary contour is quantified by calculating the ratio of the contour perimeter to the perimeter of the equal-area circle.

[0034] Internal cohesion is calculated based on the spatial density of user feature points within the group. The group centroid is used as a reference point, and its coordinates are the arithmetic mean of the feature vectors of all users within the group. The Euclidean distance of each user to the group centroid reflects their deviation from the group center, and the standard deviation of the distances of all users to the centroid represents the internal dispersion of the group. The cohesion index is defined as the ratio of the maximum distance to the standard deviation of the actual distance, and the maximum distance is calculated from the maximum diameter of the group boundary profile.

[0035] Characterizing the distribution structure of the guiding features generated by each group in high-dimensional space requires extracting the dominant feature patterns of the groups and performing independent component analysis on the behavioral characteristics of users within each group to separate mutually independent feature components. The number of independent components is determined by the information criterion, typically retaining the top few components with a cumulative variance contribution rate of 85%. Each independent component corresponds to a dominant behavioral pattern of the group, and the weight coefficient of the component reflects the importance of that pattern in the group.

[0036] The construction of the swarm intelligence feature map maps the high-dimensional distribution structure to a two-dimensional visualization space. A t-SNE algorithm is used for nonlinear dimensionality reduction mapping, with the perplexity parameter set to one-third of the swarm size and the learning rate set to the square root of the swarm size. The mapping process optimizes the objective function through gradient descent, which measures the consistency between swarm similarity in the high-dimensional space and distance in the low-dimensional space. During iterative optimization, the system monitors the convergence of the objective function value, stopping optimization when the function value changes by less than 0.001 over 50 consecutive iterations.

[0037] In one optional implementation, based on the multi-dimensional behavioral profile, a group feature recursive quantization technique is used to identify user groups with similar generation preferences. By depicting the evolutionary chain of behavioral trajectories, the resonance relationship and entropy distribution between trajectories are captured, including: Based on the multidimensional behavioral profile, the recursive quantization technique of group features is used to identify user groups with similar generation preferences, construct a recursive graph matrix of high-dimensional phase orbits, and obtain the regression pattern and self-similar structure of the orbit in phase space through the quantization of the recursive graph matrix. The evolution chain and bifurcation features of the user behavior trajectory are jointly depicted according to the regression pattern and the self-similar structure. Based on the evolutionary chain and the bifurcation feature, the diagonal structure alignment and texture similarity between the recursive graph matrices of different users are calculated, and the resonance relationship between trajectories is captured by the combination analysis of the diagonal structure alignment and texture similarity. For user groups with significant resonance relationships, the Shannon entropy and permutation entropy of their recursive graph matrix are calculated respectively, and the entropy distribution of the group is obtained by mapping the Shannon entropy and permutation entropy in the feature space.

[0038] The implementation process of identifying user groups with similar generation preferences based on multi-dimensional behavioral profiling and recursive quantification of group features requires the establishment of a complete data processing pipeline. Raw user behavior data is stored in a distributed database, with each user record containing fields such as user identifier, timestamp, behavior type, and parameter values. The behavior type field enumerates 15 basic behavior types, including image generation requests, style switching operations, color adjustment actions, composition modification behaviors, and completion confirmation operations. The parameter value field stores specific behavior parameters in JSON format, such as the RGB three-channel values ​​of color adjustment amplitude, the nine-grid position coordinates of composition preferences, and the millisecond-level time value of dwell time, among other detailed information.

[0039] The multi-dimensional behavioral profile construction module first extracts behavioral sequences within a time window from the raw data. The time window adopts a sliding window mechanism with a window length of 24 hours and a sliding step size of 1 hour. Within each time window, the frequency distribution of user behavior is statistically analyzed, and statistical features such as the occurrence frequency, average duration, and parameter variation amplitude of each behavior type are calculated. The behavioral profile vector has 128 dimensions. The first 15 dimensions correspond to the frequency features of each behavior type, dimensions 16 to 30 correspond to the duration features, dimensions 31 to 45 correspond to the parameter variation amplitude features, and subsequent dimensions include higher-order statistical features such as Markov transition probabilities of the behavioral sequence, time interval distribution parameters, and periodicity intensity coefficients.

[0040] The high-dimensional phase orbit reconstruction module receives a 128-dimensional behavior profile vector sequence as input, with each user maintaining a vector sequence of 720 time points. Phase space reconstruction employs a specific implementation of Takens' embedding theorem. The embedding dimension is automatically determined using the pseudo-nearest neighbor method, with the calculation process gradually increasing from dimension 1. The proportion of pseudo-nearest neighbors is statistically analyzed, and the optimal embedding dimension is determined when this proportion drops below 5%. The delay time is determined by the first local minimum of the mutual information function. The mutual information between the behavior sequence and its delayed version is calculated under different delays, and the delay corresponding to the first local minimum of the mutual information is selected as the optimal delay time. The reconstructed phase orbit is represented as a multi-dimensional vector sequence, with each vector containing the values ​​of the original sequence at different delay times.

[0041] The recursive graph matrix construction module performs distance calculation and thresholding on the reconstructed phase orbits. Distance calculation uses weighted Euclidean distance, with the weight of each dimension determined by the reciprocal of the variance of that dimension's value. Dimensions with larger variances receive smaller weights, ensuring a relatively balanced contribution from each dimension to the distance calculation. The distance threshold is determined using a fixed recursion rate strategy, with a preset recursion rate of 5%. A binary search algorithm is used to find the threshold parameter that makes the actual recursion rate closest to the preset value. The recursive graph matrix is ​​stored in a sparse matrix format, recording only the positions of elements with a value of 1, significantly reducing memory usage. Texture feature extraction of the matrix uses a sliding window statistical method, with a window size of 20×20, statistically analyzing the geometric features such as the distribution density of 1-valued elements, the number of connected regions, and the area of ​​the largest connected region within the window.

[0042] The orbit regression pattern quantification module analyzes the structural features of the recursive graph matrix to identify dynamic behavior patterns. Deterministic structure detection is achieved through diagonal segment analysis, scanning 50 diagonals upwards and downwards from the main diagonal, and statistically analyzing the length distribution of consecutive 1-value segments on each diagonal. The deterministic index is calculated as the ratio of the total length of all segments with a length greater than or equal to 2 to the total number of non-zero elements in the matrix; this index reflects the predictability of the orbit. Laminar flow structure detection is achieved through vertical line analysis, scanning each column of the matrix, statistically analyzing the length distribution of consecutive 1-value segments in the vertical direction, and calculating the average vertical line length as the laminar flow index. Self-similar structure detection divides the recursive graph matrix into 50×50 sub-blocks, calculates the structural similarity between each sub-block and its eight neighboring sub-blocks, and uses the normalized cross-correlation coefficient as the similarity measure.

[0043] The user behavior trajectory evolution chain depiction module uses time segmentation analysis based on a recursive graph matrix. It divides the recursive graph matrix into 36 sub-matrices along the time dimension, each sub-matrix corresponding to a 20-hour behavior pattern time period. The sub-matrix feature vector contains five quantitative indicators for that time period: recursion rate, determinism, maximum diagonal length, average vertical line length, and laminar flow index. Transition characteristics between adjacent sub-matrices are measured by calculating the Euclidean distance between their feature vectors: a distance less than 0.3 indicates stationary evolution, a distance between 0.3 and 0.8 indicates gradual change, and a distance greater than 0.8 indicates abrupt evolution. Nodes in the evolution chain include attributes such as time period identifier, feature vector, transition type, and transition intensity; edges include attributes such as transition direction, transition probability, and duration.

[0044] The bifurcation feature identification module detects nonlinear dynamic phenomena in the evolutionary chain. Bifurcation points are identified by calculating the Lyapunov exponent of the evolutionary trajectory. Regions with a Lyapunov exponent greater than 0 indicate chaotic behavior, regions close to 0 indicate quasi-periodic behavior, and regions less than 0 indicate stable behavior. The bifurcation type is determined by analyzing the changes in the topological structure of the trajectory before and after the bifurcation point, including period-doubling bifurcation, Hopf bifurcation, and saddle-knot bifurcation. The bifurcation intensity is quantified by calculating the magnitude of the change in the attractor dimension of the trajectory before and after the bifurcation; a larger change in dimension indicates a stronger bifurcation.

[0045] The alignment calculation module for the diagonal structure of different user recursive graph matrices employs a dynamic time warping algorithm. It extracts a strip-shaped region composed of the main diagonal and the five diagonals above and below each recursive graph matrix, forming a feature sequence representing the temporal evolution pattern of user behavior. The dynamic time warping algorithm calculates the optimal alignment path between two user feature sequences, allowing for non-linear scaling and matching of sequences in the time dimension. The alignment index is defined as the normalized ratio of the cumulative distance of the optimal path to the sequence length; a smaller value indicates higher alignment. The algorithm's time complexity is controlled by setting a bandwidth constraint of 10% of the sequence length, significantly reducing computational overhead while maintaining alignment accuracy.

[0046] The texture similarity analysis module calculates two-dimensional texture features based on the recursive graph matrix. Treating the recursive graph matrix as a grayscale image, it extracts texture features using the gray-level co-occurrence matrix (GLCM) method. It calculates GLCMs with distances of 1 and 2 at four directions: 0°, 45°, 90°, and 135°, resulting in a total of eight GLCMs. From each GLCM, it extracts four Haralick texture features: contrast, correlation, energy, and uniformity, forming a 32-dimensional texture feature vector. Texture similarity is obtained by calculating the cosine similarity between the texture feature vectors of two users. The similarity value ranges from -1 to 1; the closer the value is to 1, the more similar the textures.

[0047] The trajectory resonance relationship capture module integrates diagonal structure alignment and texture similarity, employing an adaptive weight allocation mechanism to dynamically adjust the weights of these two similarity metrics based on the complexity of user behavior. Behavioral complexity is measured by the fractal dimension of the recursive graph matrix. Higher fractal dimensions indicate more complex user behavior patterns, leading to increased weight for texture similarity; conversely, lower fractal dimensions indicate simpler user behavior patterns, resulting in increased weight for diagonal structure alignment. The weight allocation formula is as follows: texture weight equals the ratio of the fractal dimension to the maximum fractal dimension, and diagonal weight equals 1 minus the texture weight. The overall similarity is calculated using a weighted average and serves as a quantitative indicator of the resonance intensity between users.

[0048] The module for identifying user groups with significant resonance relationships employs a density-based clustering algorithm to construct a user resonance network graph. Users are nodes, and pairs of users with a resonance intensity greater than 0.6 are connected by edges, with the edge weight being the resonance intensity value. The DBSCAN clustering algorithm is applied to identify high-density user groups, with a minimum number of contained points set to 8 users. The neighborhood radius is automatically determined using the elbow method of the K-distance graph. The algorithm first identifies core points, i.e., users whose neighborhood numbers reach the minimum number of contained points, and then groups users with attainable density into the same group. Boundary users are assigned to groups based on their distance to the core regions of each group, and the group with the closest distance accepts the boundary user.

[0049] The Shannon entropy calculation module for the recursive graph matrix quantifies the randomness of user behavior patterns within a group. It divides the recursive graph matrix into 25×25 grid regions based on row and column coordinates, and counts the number of elements with a value of 1 in each grid region. The number of elements in each grid region is normalized to a probability distribution, where the probability value equals the number of elements in that region divided by the total number of elements in the matrix. Shannon entropy is calculated as the negative sum of the product of the probability value of each grid region and its base-2 logarithm, with entropy values ​​ranging from 0 to 625 (base-2 logarithm). High entropy values ​​indicate that recursive points are uniformly and randomly distributed within the matrix, while low entropy values ​​indicate that recursive points are concentrated in specific regions.

[0050] The permutation entropy calculation module analyzes the permutation patterns of row vectors in a recursive graph matrix. It selects a permutation length of 3 and extracts permutation patterns consisting of three consecutive elements from each row of the recursive graph matrix. There are eight permutation patterns: 000, 001, 010, 011, 100, 101, 110, and 111. The frequency of each permutation pattern across all rows is counted. The permutation entropy is calculated as the negative sum of the relative frequencies of each permutation pattern and their base-2 logarithms, with entropy values ​​ranging from 0 to 3. Permutation entropy reflects the complexity of the row vector sequence in the recursive graph matrix; high permutation entropy indicates complex row vector pattern variations, while low permutation entropy indicates relatively simple and regular row vector patterns.

[0051] The group entropy distribution feature space mapping module constructs a group distribution map using Shannon entropy and permutation entropy as two-dimensional features. Standardization is employed to eliminate the dimensional differences between the two entropy values; the standardization formula is entropy value minus the mean, then divided by the standard deviation. A two-dimensional coordinate system is established, with the horizontal axis representing standardized Shannon entropy and the vertical axis representing standardized permutation entropy. Each user group corresponds to a coordinate point in the feature space, with the coordinate value being the arithmetic mean of the entropy values ​​of all users within the group. The group distribution is visualized using a scatter plot, where the size of the points is proportional to the number of users in the group, and the color depth reflects the similarity within the group; darker colors indicate more consistent user behavior patterns within the group.

[0052] In one optional implementation, the user characteristics contained in the historical interaction trajectory are mapped and analyzed at multiple levels with the collective intelligence feature map to anchor the group affiliation relationship of the target user, and the guiding features corresponding to the affiliation group are visualized based on deep feature perspective technology, including: The user characteristics contained in the historical interaction trajectory are mapped and analyzed in a multi-level manner with the collective intelligence feature map. By constructing multi-order projection links in feature space, semantic space and topological space, the affinity relationship between user feature vectors and each group node is quantified respectively. Based on the affinity relationship, the affiliation probability distribution is constructed. The group affiliation relationship of the target user is anchored according to the dominant peak of the affiliation probability distribution. Based on the group affiliation relationship, locate the corresponding node in the group intelligence feature map, deconstruct the distribution parameters, evolution chain features and entropy labels contained in the corresponding node, and establish an abstract guiding feature representation of the affiliation group by generating a deep perspective of the guiding feature space; Based on the abstract guidance feature representation, the common generation preferences of the group are visualized as guidance parameters in the operational dimension. The dominant path deconstruction control nodes and parameter strategies in the evolutionary chain features are integrated to form a concrete expression of the guidance feature.

[0053] In the process of achieving multi-level mapping and analysis between historical interaction trajectories and collective intelligence feature maps, basic feature vectors are extracted from users' historical interaction data. This historical interaction trajectory includes multi-dimensional data such as users' content generation records, editing operation sequences, tool call frequency, parameter adjustment magnitude, and session duration on the platform. The system converts this raw data into standardized numerical feature vectors, with each feature vector having 512 dimensions. The first 128 dimensions represent content type preferences, the middle 192 dimensions represent operational behavior patterns, and the last 192 dimensions represent temporal evolution characteristics. For a specific target user, their historical interaction trajectory shows that they completed 87 content generation tasks in the past 30 days, with image-based tasks accounting for 63%, text-based tasks for 25%, and mixed-type tasks for 12%. The user's average session duration was 18 minutes, and the parameter adjustment frequency was 7.3 times per session.

[0054] A projection mapping mechanism for the feature space is constructed, comparing and calculating the user feature vector with each group node in the swarm intelligence feature map. The swarm intelligence feature map contains 2048 group nodes, each representing a user cluster with similar behavioral characteristics. Projection into the feature space is achieved by calculating the distance metric between the user feature vector and the center vector of each group node. The distance metric uses the weighted Euclidean distance method, with the weight coefficients determined based on the inverse of the variance of the feature dimension. The system calculates that the distance between the target user and group node A1 is 0.23, the distance with group node B7 is 0.31, and the distance with group node C2 is 0.19. The smaller the distance value, the higher the affinity.

[0055] At the semantic space level, user interaction behaviors are semantically parsed, transforming operation sequences into semantic tag sequences. The target user's operation sequences show frequent use of style transfer functions, detail enhancement tools, and color adjustment modules. These operations are mapped to a set of semantic tags, including tags such as "visual optimization-oriented," "refined processing preference," and "tone-sensitive." The system matches these semantic tags with the group semantic features stored in the group nodes, using tag overlap as a measure of semantic affinity. The calculation results show that the target user's semantic tags have a 78% overlap with group node C2, a 65% overlap with group node A1, and a 52% overlap with group node B7.

[0056] The mapping of topological space focuses on the consistency between the temporal evolution path of user behavior and the evolutionary trajectory of the group. It extracts the operation transition patterns of target users in the time dimension and constructs a behavior transition topology map for each individual user. This topology map records each intermediate state of the user from the initial operation to the final result and its transition probability. For example, when a target user performs an image generation task, there is a 72% probability that they will first build the basic framework, then an 85% probability that they will enter the detail refinement stage, and finally a 91% probability that they will perform a global optimization operation. The system calculates the structural similarity between this user topology map and the standard topology patterns of each group node. The results show that the topological similarity with group node C2 is 0.83, and the topological similarity with group node A1 is 0.71.

[0057] Based on the affinity calculation results from the three levels of feature space, semantic space, and topological space, each group node is assigned an affiliation probability value. The affiliation probability value is obtained by weighted fusion of the affinity values ​​from the three spaces, with the feature space weight set to 0.3, the semantic space weight set to 0.35, and the topological space weight set to 0.35. For the target user, the overall affiliation probability of group node C2 is 0.76, that of group node A1 is 0.58, and that of group node B7 is 0.42; the affiliation probabilities of the remaining nodes are all below 0.35. The system identifies the dominant peak in the affiliation probability distribution as group node C2, determining the target user's group affiliation as group C2.

[0058] After locating the group node C2, the multi-dimensional group characteristic data stored in this node is deconstructed. The distribution parameters record the statistical distribution characteristics of C2 group members across various feature dimensions, including median, interquartile range, skewness coefficient, and kurtosis coefficient. Specifically, the data shows that the median for color saturation adjustment amplitude in the C2 group is 32 units, the interquartile range is 17 units, exhibiting a right-skewed distribution, and the skewness coefficient is 1.24. In the detail enhancement intensity dimension, the median for this group is 58 units, showing a relatively concentrated distribution, and the interquartile range is 9 units. These distribution parameters reveal the central tendency and dispersion of the C2 group in its choice of operating parameters.

[0059] The evolutionary chain characteristics record the evolutionary patterns of C2 group members' operational strategies over time. Data shows that C2 group members initially adjusted parameters an average of 4.2 times per session after joining the platform, increasing to 7.8 times after 30 days, and stabilizing at around 9.3 times after 90 days. The evolutionary chain also shows that group members gradually migrated from basic tools to advanced tools, with the usage rate of advanced tools increasing from 18% initially to 62% in the mature stage. The dominant path in the evolutionary chain identifies a typical three-stage evolutionary pattern for this group: the exploration stage focuses on mastering basic parameters, the growth stage focuses on the combined application of tools, and the mature stage focuses on stylistic customization.

[0060] Entropy labels quantify the degree of uncertainty in the behavioral choices of group C2. The calculation results show that the entropy value for content type selection is 1.86, indicating a certain degree of concentration in choices while still maintaining diversity. The entropy value for tool call sequence is 2.34, showing a high degree of diversity in operation paths. The entropy value for parameter value distribution is 1.52, indicating a clear preference pattern in parameter selection. These entropy labels provide a quantitative basis for understanding the randomness and determinism of group behavior.

[0061] Based on the deconstructed distribution parameters, evolutionary chain features, and entropy labels, a deep perspective model is constructed to establish an abstract representation of the C2 population in the guided feature space. This abstract representation employs a deep neural network architecture. The input layer receives a concatenated vector of distribution parameters, evolutionary chain features, and entropy labels, which undergoes three fully connected transformations and nonlinear activations. The output layer generates a 256-dimensional abstract guided feature vector. This vector encodes the core behavioral patterns and generation preferences of the C2 population in a high-dimensional space, specifically manifested as the correspondence between the activation patterns of the vector in a specific dimensional cluster and the population features.

[0062] When transforming the abstract guiding feature vector into executable operational-level guiding parameters, the mapping relationship between the activation intensity of each dimension in the analytical vector and the specific operational parameters is determined. For the color saturation parameter, the average activation intensity of the clusters in dimensions 23 to 35 of the abstract vector is 0.68, which maps to a recommended saturation increase range of 28 to 36 units. For the detail enhancement parameter, the activation intensity of the clusters in dimensions 78 to 94 is 0.82, mapping to a recommended enhancement intensity range of 54 to 62 units. For the contrast parameter, the activation intensity of the clusters in dimensions 156 to 171 is 0.54, mapping to a recommended contrast adjustment range of 12 to 18 units. These concrete guiding parameters constitute the initial recommended configuration scheme for the target user.

[0063] By combining the dominant paths identified in the evolutionary chain characteristics, the key regulatory nodes of each evolutionary stage are deconstructed. During the exploration phase, regulatory nodes focus on setting basic parameter thresholds, with the system defining recommended ranges for basic parameters for target users to prevent parameter values ​​from exceeding the beginner's adaptation range. During the growth phase, regulatory nodes shift to tool combination mode guidance, with the system recommending advanced tools highly compatible with the target user's currently proficient tools, prioritizing recommendations based on the synergistic effect coefficient between tools. During the maturity phase, regulatory nodes focus on stylistic path branches, with the system recommending personalized parameter combination templates based on the target user's historical style preferences. Each regulatory node is configured with a corresponding parameter strategy, defining the type, intensity, and timing of guidance information the system should provide at that node, ultimately forming a dynamic guidance feature expression system covering the entire user usage cycle.

[0064] In one optional implementation, by constructing multi-order projection links in the feature space, semantic space, and topological space, the affinity between user feature vectors and nodes of each group is quantified respectively, and the affiliation probability distribution is constructed based on the affinity, including: By constructing a multi-order projection link of feature space, semantic space and topological space, feature affinity is obtained by depicting the affinity and resonance degree between user feature vectors and the distribution parameters of each group node in the feature space. Semantic affinity is obtained by characterizing the semantic representation of user feature vectors and the cognitive synergy strength between the dominant and guiding features of each group node in the semantic space. Topological affinity is obtained by constructing the manifold distance of user feature vectors in the collective intelligence feature graph structure and the coupling depth of the topological centrality of each group node in the topological space. The entropy labels of each group node reveal the expression tendencies of user feature vectors at different cognitive levels. The inherent consistency of the entropy labels is used to construct a collaborative mechanism of feature, semantic and topological affinity. Based on the collaborative mechanism, the contribution ratio of each spatial affinity in different cognitive dimensions is guided to obtain a complete affinity relationship picture. The affinity relationship picture is condensed into an affiliation probability distribution in the probability space, so that the potential energy difference between adjacent groups is formed in the affiliation determination process, eliminating the fuzzy area of ​​the group boundary and obtaining an affiliation probability distribution with a clear dominant orientation.

[0065] The multi-level projection link construction module establishes data projection channels in the feature space, semantic space, and topological space. The feature space projection module receives the user's 128-dimensional behavioral profile vector as input and uses principal component analysis (PCA) to project the high-dimensional features into a 32-dimensional feature subspace. The PCA selection strategy retains the top 32 principal components with a cumulative variance contribution rate of 92%. The projection matrix is ​​a 128×32 real matrix with element values ​​ranging from -1 to 1. The semantic space projection uses a pre-trained text embedding model to convert the semantic description of user behavior into a 256-dimensional semantic vector. The semantic description includes text tags such as the user's preferred art style, creative theme, and technical complexity. The topological space projection uses a graph embedding algorithm to map the user's positional relationship in the collective intelligence feature graph into a 64-dimensional topological coordinate vector. The Node2Vec algorithm is used to train the graph embedding representation, with a random walk length of 40 steps and 20 random walk paths generated for each node.

[0066] The feature affinity calculation module quantifies the resonance degree between the user feature vector and the distribution parameters of each group node in the feature space. The group node distribution parameters include statistical descriptors such as the mean, covariance matrix, skewness coefficient, and kurtosis coefficient of the group feature vector. The affinity resonance degree is measured using the reciprocal of the Mahalanobis distance, which requires the difference between the user feature vector and the group mean vector, as well as the inverse of the group covariance matrix. The covariance matrix inversion uses singular value decomposition to handle approximately singular cases; components with singular values ​​less than 0.001 are set to 0.001 to ensure numerical stability. The feature affinity value range is normalized to between 0 and 1 using the Sigmoid function, with the function parameters set to a slope factor of 2.5 and a center offset of 0.5.

[0067] The semantic affinity calculation module characterizes the cognitive synergy strength between the semantic representation of user feature vectors and the dominant guiding features of each group node. The dominant guiding features of group nodes are obtained through weighted aggregation of user semantic vectors within the group. The weights are determined based on the user's representativeness within the group, which is quantified by the inverse of the distance between the user and the group center. Cognitive synergy strength is calculated using cosine similarity, dividing the inner product of the user's semantic vector and the dominant semantic vector of the group by the product of the magnitudes of the two vectors. Preprocessing of the semantic vectors includes L2 norm normalization and centering operations to ensure the geometric meaning of the similarity calculation is clear. The dynamic adjustment mechanism of semantic affinity considers the confidence level of the user's semantic representation; the semantic affinity weight of users with a confidence level below 0.6 is halved.

[0068] The topological affinity calculation module constructs the coupling depth between the manifold distance of user feature vectors in the swarm intelligence feature graph structure and the topological centrality of each swarm node. The manifold distance is calculated using geodesic distance, and the Dijkstra algorithm is used to find the shortest paths from user nodes to each swarm center node in the swarm intelligence feature graph. The path weights are set as a weighted combination of Euclidean distance and connectivity strength between nodes, with Euclidean distance weighting at 0.7 and connectivity strength weighting at 0.3. The topological centrality of swarm nodes is measured using eigenvector centrality, and the principal eigenvectors of the graph adjacency matrix are calculated using a power iteration method, with an iteration convergence threshold set to 0.0001. The coupling depth is calculated by multiplying the negative exponential function of the manifold distance by the topological centrality, with a decay coefficient of 0.15 for the exponential function.

[0069] The entropy labeling module reveals the expression tendencies of user feature vectors at different cognitive levels based on the entropy distribution characteristics of each group node. Group node entropy labels include statistical descriptions of various entropy indicators such as Shannon entropy, permutation entropy, approximate entropy, and sample entropy. Expression tendency quantification is achieved by calculating the probability density function values ​​of individual user entropy and group entropy distributions. The probability density function is constructed using the kernel density estimation method, with a Gaussian kernel selected as the kernel function, and the bandwidth parameter automatically selected through cross-validation. The cognitive level is divided into three dimensions: innovativeness, regularity, and complexity. Shannon entropy reflects innovativeness, permutation entropy reflects regularity, and approximate entropy reflects complexity.

[0070] The triple affinity coordination mechanism module guides the synergistic effect of each spatial affinity through the inherent consistency of entropy labels. Inherent consistency is measured by the Pearson correlation coefficient between entropy label vectors; an absolute value greater than 0.6 indicates high consistency, 0.3 to 0.6 indicates moderate consistency, and less than 0.3 indicates low consistency. The coordination mechanism dynamically adjusts the weight allocation of each spatial affinity according to the degree of consistency. In the case of high consistency, the weights are: feature affinity 0.5, semantic affinity 0.3, and topological affinity 0.2; in the case of moderate consistency, the weights are adjusted to 0.4, 0.35, and 0.25; and in the case of low consistency, the weights are adjusted to 0.35, 0.4, and 0.25.

[0071] The cognitive dimension contribution weighting module determines the contribution allocation of each spatial affinity across the dimensions of innovativeness, regularity, and complexity based on a collaborative mechanism. The innovativeness dimension primarily relies on semantic affinity, with a base weight of 0.6, fine-tuned according to the matching degree between the user's innovativeness entropy value and the group distribution; the weight increases by up to 0.15 when the matching degree is high. The regularity dimension primarily relies on feature affinity, with a base weight of 0.55, adjusted according to the stability of user behavior patterns, which is quantified by the peak value of the autocorrelation function of the behavior sequence. The complexity dimension primarily relies on topological affinity, with a base weight of 0.5, corrected based on the user's local clustering coefficient in the graph; a high clustering coefficient indicates dense connections around the user, enhancing the complexity contribution.

[0072] The affinity mapping integration module combines the weighted combinations of the three affinity levels across various cognitive dimensions to form a complete affinity description. The integration process employs tensor multiplication, performing an outer product operation between the three affinity vectors and the three cognitive dimension weight vectors to obtain a 3×3 affinity tensor. Tensor elements represent the contribution strength of a specific affinity level on a specific cognitive dimension, the trace of the tensor represents the overall affinity strength, and the determinant of the tensor represents the stability of the affinity relationship. The affinity mapping is visualized using a radar chart, with the three axes representing feature, semantic, and topological affinity, respectively. Different colored regions represent the contributions of different cognitive dimensions.

[0073] The probability space condensation module transforms the affinity relationship map into an attribution probability distribution. The basic form of this distribution uses the Softmax function to convert the overall affinity value of each group into a probability value, ensuring that the sum of the attribution probabilities of all groups equals 1. The temperature parameter of the Softmax function is set to 1.5; a higher temperature value makes the probability distribution smoother and avoids over-concentration on a single group. The probability adjustment mechanism considers the uncertainty of user characteristics. This uncertainty is quantified by the mean of the variance of each dimension of the feature vector; users with higher uncertainty have a more uniform probability distribution.

[0074] The potential energy difference mechanism module eliminates the boundary ambiguity between adjacent groups during the attribution determination process. Adjacent groups are defined as pairs of groups directly connected in the collective intelligence feature map. The connection strength is calculated by a weighted average of user interaction frequency and feature similarity between groups. The potential energy difference is implemented by introducing a competition suppression mechanism between adjacent groups. When the probability difference between a user's attribution to two adjacent groups is less than 0.15, the potential energy difference adjustment is activated, increasing the probability value of the higher-probability group by 0.1 and decreasing the probability value of the lower-probability group by 0.1. After adjustment, probability normalization is performed again to ensure the rationality of the probability distribution.

[0075] The clear dominant orientation generation module determines the user's final group affiliation based on the probability distribution adjusted for potential energy difference. The dominant orientation is determined using the maximum probability principle, with the group with the highest probability becoming the user's dominant affiliation. Confidence assessment is calculated using the ratio of the highest probability to the second-highest probability; a ratio greater than 2.0 indicates high confidence, 1.5 to 2.0 indicates medium confidence, and less than 1.5 indicates low confidence. In cases of low confidence, a multi-round iterative mechanism is initiated, recalculating affinity and adjusting the probability distribution, with a maximum of three iterations.

[0076] The data case demonstrates the effectiveness of this technique on a dataset containing 950 users and 8 groups. User feature vectors have a dimension of 128 and are projected onto a 32-dimensional feature space using principal component analysis. The semantic space uses 768-dimensional pre-trained embedding vectors, compressed to 256 dimensions. The topological space embedding dimension is set to 64. The calculated average feature affinity is 0.67, the average semantic affinity is 0.58, and the average topological affinity is 0.72. Collaborative mechanism analysis shows that 74% of users have highly consistent entropy labels, 21% have moderate consistency, and 5% have low consistency.

[0077] In one optional implementation, a multi-level guiding vector is constructed and accurately mapped to the latent space representation of the image generation model. The deviation between the generated result and the guiding vector is quantified in real time. When the deviation exceeds a preset deviation threshold, intelligent vector calibration is triggered, and the generated result that satisfies the dual constraints is output, including: A multi-level guiding vector is constructed, and a deep resonance relationship is built based on the multi-level guiding vector and the latent space representation of the image generation model to drive the image generation model to form an intermediate generation result; The hidden layer feature representation is condensed from the intermediate generated results, and the hidden layer feature representation is projected onto the multi-dimensional creative space constructed by the multi-level guiding vector to depict the image projection of each level. Based on the degree of creative deviation between the intermediate generated results and the multi-level guiding vectors, a hierarchical resonance intensity is constructed based on the inter-level dependency relationship. The hierarchical resonance intensity is harmonized to obtain the overall deviation. When the overall deviation exceeds the preset deviation threshold, the vector calibration mechanism is activated. The vector calibration mechanism is executed to locate the dominant deviation level and harmonize the corresponding guiding vector. Based on the stability index constraint, the harmonization amplitude is constrained, and the multi-level guiding vector is reconstructed. The reconstructed multi-level guiding vector is then used to establish a new resonance relationship with the latent space representation. The deviation perception and calibration process is executed cyclically until the overall deviation reaches the expected harmonious state, and the generation result that satisfies the dual constraints is output.

[0078] In the latent space operations of the image generation model, the construction of multi-level guide vectors adopts a hierarchical semantic coding architecture. The input text description is converted into an initial vector representation through a word embedding layer. This initial vector has a dimension of 512, with each dimension corresponding to a specific semantic component. In the semantic decomposition stage, the initial vector is decomposed along different levels of abstraction, forming three levels: high-level conceptual vectors, mid-level attribute vectors, and low-level detail vectors. The high-level conceptual vector captures overall style and theme information, with a dimension of 128, and each component's value ranges from -1 to +1. The mid-level attribute vector encodes attribute features such as color distribution and spatial layout, with a dimension of 192. The low-level detail vector is responsible for expressing fine-grained features such as texture and edges, with a dimension of 192. The three levels of vectors are then concatenated to form a complete 512-dimensional guide vector.

[0079] To establish a deep resonance relationship between multi-level guiding vectors and the latent space representation of the image generation model, a feature alignment module is used for spatial mapping. This module receives the guiding vector as input and transforms it into a representation that matches the latent space of the generation model through a three-layer fully connected network. The first fully connected layer maps the 512-dimensional guiding vector to a 1024-dimensional intermediate representation, using a leak-corrected linear unit as the activation function and setting the negative slope parameter to 0.2. The second layer maintains the 1024-dimensional representation and introduces a residual connection structure to maintain information flow. The third layer compresses the representation to the 768-dimensional latent space vector required by the generation model. The latent space vector is embedded into each decoding layer of the generation model through an injection mechanism. Guiding information of the corresponding level is injected into layers 3, 6, and 9, respectively, forming a hierarchical feature modulation effect. The generation model performs forward propagation based on the injected latent space representation and outputs an intermediate generation result with a resolution of 512 x 512 pixels.

[0080] The process of extracting hidden layer features from the intermediate generated results employs a multi-scale feature acquisition strategy. A feature extraction network is applied to the generated image data, comprising four downsampling stages, each halving the spatial resolution and doubling the number of channels. The initial input is a 512x512x3 RGB image. After the first stage of convolution, a 256x256x64 feature map is obtained, with a kernel size of 3x3, a stride of 2, and padding of 1. The second stage outputs a 128x128x128 feature map. The third stage generates a 64x64x256 feature map. The fourth stage produces a 32x32x512 deep feature representation. Global average pooling is performed on the feature maps of each of the four stages to compress the spatial dimensions into a single value, resulting in stage feature vectors with dimensions of 64, 128, 256, and 512, respectively. The four vectors are then concatenated to form a 960-dimensional comprehensive hidden layer feature representation.

[0081] Projecting hidden layer features onto a multi-dimensional creative space constructed from multi-level guided vectors requires establishing feature mapping relationships. The creative space is defined based on three levels of the multi-level guided vectors, each corresponding to a subspace. The high-level subspace has a dimension of 128, the middle-level subspace has a dimension of 192, and the low-level subspace has a dimension of 192. The 960-dimensional hidden layer features are mapped to these three subspaces through three independent projection networks. The high-level projection network consists of two fully connected layers: the first layer compresses the 960 dimensions to 256 dimensions, and the second layer outputs a 128-dimensional high-level image projection. The middle-level projection network also uses a two-layer structure: the middle layer has a dimension of 384, and the output layer has a dimension of 192. The low-level projection network has a middle layer dimension of 384 and an output dimension of 192. The three image projection vectors reflect the feature distribution of the intermediate generated results at different levels of abstraction.

[0082] When calculating the creative deviation between the intermediate generated results and the multi-level guiding vectors, a difference index is measured for each level. The high-level deviation index is obtained by calculating the Euclidean distance between the high-level image projection and the high-level concept vector. Specifically, this is achieved by squaring the difference between corresponding positions of two 128-dimensional vectors, summing the squares of all positions, and taking the square root to obtain the scalar distance value. When the value of a certain dimension of the high-level concept vector is 0.8, and the corresponding image projection dimension value is 0.6, the difference contributed by that dimension is 0.04. Summing the differences of all 128 dimensions and taking the square root, assuming the sum is 2.56, the high-level deviation index is 1.6. The mid-level deviation index uses the same calculation method as the low-level deviation index, measuring the distance between the mid-level attribute vector and the mid-level image projection, and the low-level detail vector and the low-level image projection, respectively, in a 192-dimensional space. Assuming the mid-level deviation index is calculated to be 2.3 and the low-level deviation index to be 1.9.

[0083] Constructing hierarchical resonance intensity requires introducing inter-level dependency weights. These weights reflect the degree of influence of different levels on the final generated effect and are pre-set according to the characteristics of the generation task. For generation tasks that emphasize overall style consistency, the high-level dependency weight is set to 0.5, the mid-level dependency weight to 0.3, and the low-level dependency weight to 0.2. Multiplying the deviation index of each level by its corresponding dependency weight, the weighted deviation for the high level is 1.6 multiplied by 0.5 equals 0.8, for the mid-level it is 2.3 multiplied by 0.3 equals 0.69, and for the low-level it is 1.9 multiplied by 0.2 equals 0.38. Summing the three weighted deviation values ​​yields the overall deviation, which in this example is 1.87. The preset deviation threshold is determined based on the generation quality requirements; when set to 1.5, the current overall deviation of 1.87 exceeds the threshold, triggering the vector calibration mechanism.

[0084] When executing the vector calibration mechanism, the dominant deviation level is located by comparing the weighted deviation values ​​of each level. In the current example, the weighted deviation of the high-level concept vector is 0.8, which is the maximum value, thus identifying the high-level concept vector as the dominant deviation level. A harmonic operation is performed on the high-level concept vector, with the harmonic direction determined by the difference vector between the high-level image projection and the high-level concept vector. Each dimension of the difference vector represents the direction and relative magnitude of adjustment required for that dimension; a positive dimension value indicates an increase in the corresponding dimension of the original vector, while a negative value indicates a decrease. A stability constraint coefficient is introduced to control the harmonic magnitude. This coefficient is dynamically adjusted according to the current iteration round, initially set to 0.3, and gradually decaying to 0.1 as iterations progress. The difference vector is multiplied by the stability constraint coefficient to obtain the adjustment vector, which is then added to the original high-level concept vector to complete the vector harmonic adjustment. For example, if the difference value of a certain dimension is -0.2 and the stability constraint coefficient is 0.3, then the adjustment amount for that dimension is -0.06, and the original value of 0.8 becomes 0.74 after adjustment.

[0085] When reconstructing the multi-level guide vector, the original high-level vector is replaced by the harmonized high-level concept vector, while the mid-level attribute vector and low-level detail vector remain unchanged. A new 512-dimensional complete guide vector is formed through concatenation. The reconstructed guide vector is then re-input into the feature alignment module, performing the same mapping transformation as the initial process to generate a new 768-dimensional latent space representation. This new latent space representation is then injected into the image generation model, triggering a new round of forward propagation and producing updated intermediate generation results. The complete process of feature extraction, projection mapping, and deviation calculation is repeated on the updated results to obtain a new overall deviation value. Assuming that the overall deviation decreases to 1.62 after one calibration, but still exceeds the threshold of 1.5, the next round of calibration is performed. In the second round of calibration, the mid-level weighted deviation becomes the dominant factor, and a harmonization operation is performed on the mid-level attribute vector. This iterative process continues until the overall deviation decreases to 1.48, satisfying the harmony condition below the threshold, at which point the calibration loop terminates.

[0086] When outputting a generated result that satisfies dual constraints—deviation and stability—the deviation constraint requires the overall deviation to be below a preset threshold, ensuring consistency between the generated content and the guiding intent. The stability constraint prevents generation collapse caused by excessive vector correction by limiting the harmonic amplitude. The final generated result retains the 512x512 pixel image data generated in the last iteration. This image maintains a high degree of semantic alignment with the original guiding vector, achieving the output standard of the generative model in terms of visual quality. The execution time of the entire process depends on the initial deviation and the calibration convergence speed; typically, 3 to 5 iterations are required to complete the calibration process.

[0087] In one optional implementation, a hierarchical resonance intensity is constructed based on the inter-level dependency relationship, and the hierarchical resonance intensity is tuned to obtain an overall deviation. When the overall deviation exceeds a preset deviation threshold, a vector calibration mechanism is activated, including: The hierarchical resonance intensity is constructed based on the dependency relationship between feature vectors at each level. The information transmission activity is revealed by the information flow entropy ratio between hierarchical features according to the hierarchical resonance intensity. The activity reflects the transmission depth and diffusion law of features between levels. The guiding influence between levels is described based on the transmission depth and diffusion law. The dynamic evolution chain of hierarchical features is constructed based on the guiding influence. Based on the dynamic evolution chain, the hierarchical resonance intensity is harmonic deconstructed. By capturing the dominant frequency of the resonance fluctuation, the stability of the characteristic transmission is characterized. Based on the dominant frequency, deviation points in the creation process are identified. When the overall deviation caused by the deviation point exceeds the preset deviation threshold, the vector calibration mechanism is activated.

[0088] In the specific implementation process, the dependency relationships between feature vectors at each level are quantitatively analyzed, extracting the following feature vectors: Level 1: A1 = [0.82, 0.35, 0.67, 0.91], Level 2: A2 = [0.74, 0.42, 0.59, 0.88], Level 3: A3 = [0.69, 0.51, 0.63, 0.85]. For feature vectors between adjacent levels, the system calculates the degree of difference in values ​​at each dimension. The difference between the first and second levels in the first dimension is calculated as 0.08, in the second dimension as 0.07, in the third dimension as 0.08, and in the fourth dimension as 0.03. The system comprehensively evaluates all difference values, summing the squares of the difference values ​​in each dimension and taking the square root, resulting in a dependency relationship degree of 0.135 between levels. By traversing all adjacent level pairs, the system constructs a complete dependency relationship matrix, which describes the tightness of the connection between features at different levels.

[0089] A hierarchical resonance strength index is constructed based on the dependency correlation matrix. For a creative network structure containing five levels, the dependency correlation degree between the first and second levels is denoted as 0.135, between the second and third levels as 0.142, between the third and fourth levels as 0.168, and between the fourth and fifth levels as 0.193. The system progressively accumulates the dependency correlation values ​​between adjacent levels, assigning a positional weight coefficient to each correlation degree, with higher weights for correlation degrees closer to the output layer. In the specific calculation, the correlation degree between the first and second levels is multiplied by a weight of 0.8, between the second and third levels by a weight of 0.9, between the third and fourth levels by a weight of 1.0, and between the fourth and fifth levels by a weight of 1.2. The system sums the weighted correlation degree values ​​and divides them by the total weights to obtain a comprehensive hierarchical resonance strength of 0.162.

[0090] The information flow entropy ratio is calculated based on the hierarchical resonance intensity. The number of activated feature nodes in each level is counted: 128 nodes in the first level, 96 in the second, 84 in the third, 72 in the fourth, and 58 in the fifth. The system calculates the ratio of activated nodes between adjacent levels: the node retention rate from the first to the second level is 0.75, from the second to the third level is 0.875, from the third to the fourth level is 0.857, and from the fourth to the fifth level is 0.806. The system normalizes the logarithm of each level's retention rate, yielding information flow entropy values ​​of 0.32, 0.15, 0.17, and 0.24 for each level. The system also calculates the ratio of information flow entropy values ​​between adjacent levels, resulting in entropy ratio sequences of 2.13, 0.88, and 1.41. This entropy ratio sequence reveals the activity level of information transmission in the hierarchical process. The larger the value, the more active the information transmission between the levels and the wider the feature diffusion range.

[0091] Based on the entropy ratio of information flow, the transmission depth and diffusion law are determined. Layers with an entropy ratio greater than 1.5 are marked as high-activity transmission regions, those with an entropy ratio between 0.8 and 1.5 are marked as medium-activity transmission regions, and those with an entropy ratio less than 0.8 are marked as low-activity transmission regions. In this case, the first to second layer is a high-activity transmission region with a transmission depth score of 9.2, indicating that the feature information can be deeply transmitted to subsequent layers; the second to third layer is a low-activity transmission region with a transmission depth score of 4.6, indicating that feature transmission is attenuated; and the third to fourth layer is a medium-activity transmission region with a transmission depth score of 6.8. The system constructs a diffusion law model, dividing the feature activation distribution of each layer according to spatial regions and statistically analyzing the concentration of feature activation in each spatial region. A concentration below 0.4 indicates a divergent diffusion law, a concentration between 0.4 and 0.7 indicates a balanced diffusion law, and a concentration above 0.7 indicates a convergent diffusion law.

[0092] The system describes the guiding influence between levels based on transmission depth and diffusion patterns. A cross-analysis is performed on the transmission depth score and the concentration of diffusion patterns. When the transmission depth score is higher than 7.5 and the diffusion pattern is divergent, the guiding influence of that level on subsequent levels is rated as strong, with an influence coefficient set at 1.35. When the transmission depth score is between 5.0 and 7.5 and the diffusion pattern is balanced, the guiding influence of that level on subsequent levels is rated as moderate, with an influence coefficient set at 1.0. When the transmission depth score is lower than 5.0 and the diffusion pattern is convergent, the guiding influence of that level on subsequent levels is rated as weak, with an influence coefficient set at 0.65. In this case, the system determines that the first level is strongly guiding, the second level is weakly guiding, the third level is moderately guiding, and the fourth level is moderately guiding. The system connects the guiding influence coefficients of each level in chronological order to form a dynamic evolution chain [1.35, 0.65, 1.0, 1.0]. This chain reflects the trajectory of the change in the transmission intensity of features between different levels.

[0093] Harmonic deconstruction of hierarchical resonance intensity is performed based on a dynamic evolution chain. The influence coefficient sequence in the dynamic evolution chain is treated as a discrete signal, and frequency domain decomposition is performed on this signal. The system constructs a sampling window of length 8 and periodically expands the influence coefficient sequence [1.35, 0.65, 1.0, 1.0] to obtain the expanded sequence [1.35, 0.65, 1.0, 1.0, 1.35, 0.65, 1.0, 1.0]. The system performs frequency component extraction on the expanded sequence and identifies three main frequency components: the first frequency component has an amplitude of 0.87 and a period of 2 levels; the second frequency component has an amplitude of 0.43 and a period of 4 levels; and the third frequency component has an amplitude of 0.21 and a period of 8 levels. The system determines the frequency component with the largest amplitude as the dominant frequency. In this case, the dominant frequency has a period of 2 levels and an amplitude of 0.87. Based on the periodic characteristics of the dominant frequency, the system performs periodic alignment analysis on the influence coefficients of each level. When the influence coefficient of a certain level deviates from the theoretical value of the dominant frequency by more than 0.35, the level is marked as a potential deviation point.

[0094] The stability of feature transmission is characterized by the dominant frequency. The ratio of the dominant frequency's amplitude to the sum of the amplitudes of all frequency components is calculated, yielding a dominant frequency proportion of 0.576. This proportion reflects the stability of the dominant mode during feature transmission. When the dominant frequency proportion is higher than 0.65, the system determines that the feature transmission has high stability, and the creative process exhibits strong regularity. When the dominant frequency proportion is between 0.45 and 0.65, the system determines that the feature transmission has moderate stability, and the creative process exhibits some fluctuations. When the dominant frequency proportion is lower than 0.45, the system determines that the feature transmission has low stability, and the creative process exhibits irregularity. In this case, the stability is rated as moderate, with a stability score of 58.3. The system further analyzes the distribution characteristics of non-dominant frequency components. When the number of non-dominant frequency components exceeds 5 and the amplitude distribution dispersion is greater than 0.6, the system determines that there is multi-mode competition and lowers the stability score by 15 percentage points.

[0095] Based on the dominant frequency identification of deviation points in the creative process, the actual influence coefficient of each level is compared with the theoretical influence coefficient predicted by the dominant frequency. The actual influence coefficient of the second level is 0.65, while the theoretical coefficient predicted according to the periodicity of the dominant frequency should be 1.15. The difference is 0.50, exceeding the deviation threshold of 0.35. Therefore, the second level is identified as a deviation point. The system records the location information, deviation degree value, and deviation direction type of the deviation point. In this case, the deviation point is located in the second level, the deviation degree is a negative deviation of 0.50, and the deviation direction is a suppressive deviation. The system constructs a deviation point influence range assessment model to analyze the propagation depth of the deviation point's influence on subsequent levels. When the influence of the deviation point decays to less than 30% of the initial deviation degree within two levels, it is judged as a local deviation; when the influence of the deviation point remains above 30% of the initial deviation degree in three or more levels, it is judged as a diffuse deviation.

[0096] The system calculates the overall deviation index by weighting and summing the deviation values ​​of all identified deviation points. Deviation points at the higher levels are assigned higher weight coefficients, with the weight coefficient for second-level deviation points set at 1.25. The system multiplies the deviation degree of this deviation point (0.50) by the weight coefficient 1.25, resulting in a weighted deviation of 0.625. The system also considers the moderating effect of the stability score on the overall deviation. Subtracting the stability score (58.3) from 100 yields an instability index of 41.7, which is then converted into a moderating factor of 1.417. The system multiplies the weighted deviation of 0.625 by the moderating factor 1.417, resulting in a final overall deviation of 0.886. The system's preset deviation threshold is 0.75. When the overall deviation value exceeds this threshold, the system determines that a significant deviation has occurred in the creative process, requiring intervention and adjustment via a vector calibration mechanism to ensure that feature transfer returns to the expected trajectory.

[0097] A second aspect of the present invention provides a personalized image content generation and optimization system that integrates generative AI, comprising: The first unit is used to analyze the historical interaction trajectory and current intent expression of target users based on deep perception networks, and simultaneously construct a multi-dimensional behavioral profile; based on the multi-dimensional behavioral profile, the group feature recursive quantization technology is used to identify user groups with similar generation preferences, and the distribution structure of the generation guidance features of each group in high-dimensional space is depicted to construct a group intelligence feature map. The second unit is used to perform multi-level mapping and analysis of the user characteristics contained in the historical interaction trajectory and the collective intelligence feature map, anchor the group affiliation relationship of the target user, and visualize the guiding features corresponding to the affiliation group based on deep feature perspective technology. The third unit is used to construct a multi-level guidance vector based on the current intention expression and the guidance features, and to accurately map the latent space representation of the image generation model. It quantifies the degree of deviation between the generated result and the guidance vector in real time. When the deviation exceeds the preset deviation threshold, it triggers intelligent vector calibration and outputs the generated result that satisfies the dual constraints. The fourth unit is used to establish a correlation projection between the generated result and the collective intelligence feature map, and to reconstruct the distribution structure of the corresponding group based on the evaluation data of the target user, so as to realize the dynamic evolution of the features.

[0098] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0099] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0100] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A personalized image content generation and optimization method with fusion generative AI, characterized in that, Comprise: Based on the deep perception network analysis of the historical interaction trajectory of the target user and the current intention expression, a multi-dimensional behavior portrait is constructed synchronously; Based on the multi-dimensional behavior portrait, a user group with similar generation preference is identified using group feature recursive quantization technology, and the distribution structure of the generation guide features of each group in high-dimensional space is depicted to construct a group wisdom feature map; The user features contained in the historical interaction trajectory are mapped and analyzed with the group wisdom feature map at multiple levels to anchor the group affiliation of the target user, and the guide features corresponding to the affiliation group are visualized based on deep feature perspective technology; According to the current intention expression and the guide features, a multi-level guide vector is constructed and the hidden space representation of the image generation model is accurately mapped, the deviation degree of the generation result and the guide vector is quantified in real time, and when the deviation exceeds the preset deviation threshold, the vector intelligent calibration is triggered, and the generation result meeting the double constraints is output; The generation result is associated and projected with the group wisdom feature map, and the distribution structure of the corresponding group is reconstructed based on the evaluation data of the target user to realize the dynamic evolution of the features.

2. The method of claim 1, wherein, Based on the multi-dimensional behavior portrait, a user group with similar generation preference is identified using group feature recursive quantization technology, and the distribution structure of the generation guide features of each group in high-dimensional space is depicted to construct a group wisdom feature map, comprising: Based on the multi-dimensional behavior portrait, a user group with similar generation preference is identified using group feature recursive quantization technology, the evolution chain of behavior trajectory is depicted, the resonance relationship and entropy distribution between trajectories are captured, the periodic commonness and turning point of user behavior mode in time and space dimensions are dynamically analyzed, and the boundary profile and internal cohesion of the user group are determined according to the natural community formed by the resonance relationship in the feature space; Based on the boundary profile and the internal cohesion, the feature gravity and behavior spectrum of each user group in the multi-dimensional behavior portrait space are depicted, the dominant tendency and energy distribution proportion of each group in the generation guide dimension are deconstructed, and the high-dimensional distribution structure of the user group generation guide features is constructed, and the high-dimensional distribution structure is mapped as a group wisdom feature map.

3. The method of claim 2, wherein, Based on the multi-dimensional behavior portrait, a user group with similar generation preference is identified using group feature recursive quantization technology, the evolution chain of behavior trajectory is depicted, the resonance relationship and entropy distribution between trajectories are captured, the periodic commonness and turning point of user behavior mode in time and space dimensions are dynamically analyzed, and the boundary profile and internal cohesion of the user group are determined according to the natural community formed by the resonance relationship in the feature space; Based on the evolution chain and the bifurcation features, the alignment degree of the diagonal line structure and the texture similarity between different user recursive graph matrices are calculated, and the resonance relationship between trajectories is captured through the combined analysis of the alignment degree of the diagonal line structure and the texture similarity; ​ For the user group with significant resonance relationship, the Shannon entropy and permutation entropy of the recurrence plot matrix of the user group are calculated respectively, and the entropy value distribution of the user group is obtained by mapping the Shannon entropy and the permutation entropy in the feature space.

4. The method of claim 1, wherein, The user features contained in the historical interaction trajectory are mapped and analyzed at multiple levels with the collective wisdom feature map, the group affiliation of the target user is anchored, and the guiding features corresponding to the affiliation group are visualized based on the deep feature perspective technology, including: The user features contained in the historical interaction trajectory are mapped and analyzed at multiple levels with the collective wisdom feature map, the group affiliation of the target user is anchored, and the guiding features corresponding to the affiliation group are visualized based on the deep feature perspective technology, including: The user features contained in the historical interaction trajectory are mapped and analyzed at multiple levels with the collective wisdom feature map, the group affiliation of the target user is anchored, and the guiding features corresponding to the affiliation group are visualized based on the deep feature perspective technology, including: According to the group affiliation relationship, the corresponding node in the collective wisdom feature map is located, the distribution parameters, evolution chain features and entropy labels contained in the corresponding node are deconstructed, and the abstract guiding feature representation of the affiliation group is established by generating a deep perspective of the guiding feature space.

5. The method of claim 4, wherein, Based on the abstract guiding feature representation, the generative preference of the group commonality is embodied as the guiding parameters in the operation dimension, and the dominant path deconstruction node and parameter strategy in the evolution chain feature are fused to form the embodiment expression of the guiding feature. By constructing the multi-order projection link of the feature space, semantic space and topological space, the affinity relationship between the user feature vector and each group node is quantified, and the affinity probability distribution is constructed based on the affinity relationship, including: By constructing the multi-order projection link of the feature space, semantic space and topological space, the affinity resonance degree of the user feature vector and the distribution parameters of each group node in the feature space is depicted to obtain the feature affinity, the semantic representation of the user feature vector and the cognitive synergy strength between the dominant guiding features of each group node in the semantic space are depicted to obtain the semantic affinity, and the manifold distance of the user feature vector in the structure of the collective wisdom feature map and the coupling depth of the topological centrality of each group node in the topological space are constructed to obtain the topological affinity. Based on the entropy labels of each group node, the expression tendency of the user feature vector in different cognitive levels is revealed, the synergy mechanism of the feature, semantic and topological triple affinity is constructed through the internal consistency of the entropy labels, the contribution proportion of each space affinity in different cognitive dimensions is guided according to the synergy mechanism, and a complete affinity relationship is obtained.

6. The method of claim 1, wherein, The affinity relationship is condensed into the affiliation probability distribution in the probability space, so that the adjacent groups form a potential energy difference in the affiliation determination process, the fuzzy area of the group boundary is eliminated, and the affiliation probability distribution with clear dominant direction is obtained. A multi-level guiding vector is constructed and the hidden space representation of the image generation model is accurately mapped, the deviation degree of the generation result and the guiding vector is quantified in real time, the vector intelligent calibration is triggered when the deviation exceeds the preset deviation threshold, and the generation result meeting the double constraints is output, including: Constructing a multi-level guidance vector, constructing a deep resonance relationship based on the multi-level guidance vector and the hidden space representation of the image generation model, driving the image generation model to form an intermediate generation result; Condensing the hidden layer feature representation from the intermediate generation result, projecting the hidden layer feature representation to the multi-dimensional creative space constructed by the multi-level guidance vector, and depicting the image projection of each level; According to the creative deviation degree of the intermediate generation result and the multi-level guidance vector, constructing the level resonance strength based on the dependence association between levels, harmonizing the level resonance strength to obtain the overall deviation degree, and activating the vector calibration mechanism when the overall deviation degree breaks through the preset deviation threshold; Executing the vector calibration mechanism, locating the dominant deviation level and harmonizing the corresponding guidance vector, constraining the harmonization amplitude based on the stability index, reconstructing the multi-level guidance vector, establishing a new resonance relationship between the reconstructed multi-level guidance vector and the hidden space representation, and repeating the deviation perception and calibration process until the overall deviation degree reaches the expected harmonious state, and outputting the generation result satisfying the double constraints.

7. The method of claim 6, wherein, Based on the dependence association between levels, the level resonance strength is constructed, and the overall deviation degree is obtained by harmonizing the level resonance strength. When the overall deviation degree breaks through the preset deviation threshold, the vector calibration mechanism is activated, which includes: Based on the dependence association between levels, the level resonance strength is constructed, and the active degree of information transmission is revealed by the information flow entropy ratio between level features according to the level resonance strength. The active degree reflects the transmission depth and diffusion law of features between levels, and the guidance influence between levels is depicted based on the transmission depth and the diffusion law. According to the guidance influence, a dynamic evolution chain of level features is formed; Based on the dynamic evolution chain, the level resonance strength is deconstructed by harmonic, the stability of feature transmission is described by capturing the dominant frequency of resonance fluctuation, and the deviation point in the creation process is identified based on the dominant frequency. When the overall deviation degree caused by the deviation point breaks through the preset deviation threshold, the vector calibration mechanism is activated.

8. A personalized image content generation and optimization system fusing generative AI for implementing the method of any of the preceding claims 1-7, characterized in that, Including: The first unit is used for constructing a multi-dimensional behavior portrait based on a deep perception network analyzing the historical interaction trajectory and current intention expression of a target user; Based on the multi-dimensional behavior portrait, a group feature recursive quantization technology is used to identify a user group with similar generation preferences, and the distribution structure of the generation guidance features of each group in high-dimensional space is depicted to construct a group wisdom feature map; The second unit is used for multi-level mapping and analysis of the user features contained in the historical interaction trajectory and the group wisdom feature map, anchoring the group attribution relationship of the target user, and visualizing the guidance features corresponding to the attribution group based on a deep feature perspective technology; The third unit is used for constructing a multi-level guidance vector and accurately mapping the hidden space representation of the image generation model according to the current intention expression and the guidance features, real-time quantifying the deviation degree of the generation result and the guidance vector, triggering vector intelligent calibration when the deviation breaks through the preset deviation threshold, and outputting the generation result satisfying the double constraints. A fourth unit is configured to associate the generation result with the wisdom-of-the-crowd feature map, reconstruct the distribution structure of the corresponding crowd based on the evaluation data of the target user, and realize dynamic evolution of the features.

9. An electronic device, comprising: The computer program instructions are executed by the processor to implement the method in any one of claims 1 to 7. The computer program instructions are executed by the processor to implement the method in any one of claims 1 to 7. The computer program instructions are executed by the processor to implement the method in any one of claims 1 to 7. ​ 10. A computer-readable storage medium having stored thereon computer program instructions, wherein, ​

Images