AIGC generation method and system based on multi-modal content copyright traceability

Abstract

The invention provides an AIGC generation method and system based on multi-modal content copyright traceability, and relates to the technical field of artificial intelligence, and the method comprises the steps: firstly constructing a copyright feature topology comprising an original creation material ownership core node, a creation evolution association edge and an authorization constraint surface, then anchoring the copyright feature topology to a parameter dimension of AI GC generation logic, and carrying out the copyright feature topology; a copyright generation logic anchoring relation is formed, a copyright generation logic symbiotic structure is generated, nodes, edges, surfaces and generated logic parameters form dynamic mapping and are adjusted in real time, the logic symbiotic structure is used for driving AIGC content generation, generated content feature dimensions carry mapping marks and are synchronously updated, finally, the mapping marks are analyzed, and the copyright generation logic symbiotic structure is generated. And reconstructing a dynamic mapping relation and inverting copyright feature topology to realize effective copyright traceability of the AIGC generation content.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to an AIGC generation method and system based on multimodal content copyright tracing. Background Technology

[0002] With the rapid development of AIGC (Artificial Intelligence Generated Content) technology, the efficiency and diversity of content creation and dissemination have been greatly improved. However, copyright issues surrounding AIGC-generated content are becoming increasingly prominent, posing a key obstacle to its healthy development.

[0003] Traditional copyright attribution methods rely primarily on manual registration and simple textual records, proving inadequate when faced with the massive, complex, and multimodal content generated by AIGC (AIGC-generated content). AIGC-generated content often integrates multiple original creative materials, which may originate from different creators and copyright holders, and their creation process has undergone multiple iterations and evolutions. Existing copyright attribution methods struggle to comprehensively and accurately record the ownership information, creative evolution process, and authorized usage scope of the materials.

[0004] For example, in the field of image generation, an image generated by AIGC may contain elements from multiple original image sources. These sources have different copyright ownership and have undergone complex operations such as style transfer and element fusion during the generation process. Traditional copyright attribution methods cannot clearly define the contribution of each source and the copyright ownership of the final generated content, easily leading to copyright disputes. Furthermore, similar problems exist in areas such as text generation and video generation. Summary of the Invention

[0005] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an AIGC generation method based on multimodal content copyright tracing, the method comprising:

[0006] Construct a copyright feature topology corresponding to AIGC generation. The copyright feature topology includes the ownership core node of the original creative material, the creative evolution association edge and the authorization constraint surface. The ownership core node records the ownership information of the material, the creative evolution association edge records the transformation relationship of the material from the initial form to the current form, and the authorization constraint surface defines the scope boundary of the material that can be used for AIGC generation.

[0007] The copyright feature topology is anchored to the parameter dimension of the AIGC generation logic, so that the ownership core node of the copyright feature topology corresponds to the subject identification parameters generated by AIGC, the creation evolution association edge corresponds to the feature iteration parameters generated by AIGC, and the authorization constraint face corresponds to the content boundary parameters generated by AIGC, thus forming a copyright generation logic anchoring relationship.

[0008] Based on the anchoring relationship of copyright generation logic, a symbiotic structure of copyright generation logic is generated. In this symbiotic structure, the nodes, edges, and faces of the copyright feature topology form a dynamic mapping with the parameters of the AIGC generation logic, and the mapping relationship is adjusted in real time with the AIGC generation process.

[0009] AIGC content generation is driven by a copyright generation logic symbiotic structure, so that each feature dimension of the generated content carries a mapping imprint of the copyright feature topology, and the mapping imprint is updated synchronously with the evolution of content features.

[0010] The mapping imprints in the generated content feature dimensions are analyzed, and the dynamic mapping relationship of the copyright generation logic symbiotic structure is reconstructed. Based on the dynamic mapping relationship, the ownership core node, creation evolution association edge and authorization constraint surface of the copyright feature topology are inverted.

[0011] In another aspect, embodiments of the present invention also provide an AIGC generation system based on multimodal content copyright tracing, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the machine-readable storage medium to implement the above-mentioned method.

[0012] Based on the above, this embodiment of the invention constructs a copyright feature topology containing core nodes of original creative material ownership, creative evolution association edges, and authorization constraint surfaces. This topology records the ownership information of the material, the transformation relationship from the initial form to the current form, and the scope boundaries that can be used for AIGC generation. The copyright feature topology is anchored to the parameter dimensions of the AIGC generation logic, forming a copyright generation logic anchoring relationship. This tightly integrates copyright features with the AIGC generation logic, ensuring the effective transmission of copyright information during the generation process. Based on this symbiotic structure of copyright generation logic, a dynamic mapping between the copyright feature topology and AIGC generation logic parameters is achieved. This mapping relationship is adjusted in real time during the generation process, flexibly responding to various changes during AIGC generation and ensuring the real-time nature and accuracy of copyright information. Driving AIGC content generation with this symbiotic structure ensures that each feature dimension of the generated content carries a mapping imprint of the copyright feature topology, and this mapping imprint is updated synchronously with the evolution of content features, achieving a deep integration of copyright information and generated content. Finally, by parsing the mapping imprint to reconstruct the dynamic mapping relationship and inverting the copyright feature topology, the copyright source of the generated content can be accurately traced, effectively solving the problem of copyright tracing for AIGC-generated content. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the execution flow of the AIGC generation method based on multimodal content copyright tracing provided in an embodiment of the present invention.

[0014] Figure 2 This is a schematic diagram of exemplary hardware and software components of the AIGC generation system based on multimodal content copyright tracing provided in an embodiment of the present invention. Detailed Implementation

[0015] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating an AIGC generation method based on multimodal content copyright tracing, provided by an embodiment of the present invention. The following is a detailed description of this AIGC generation method based on multimodal content copyright tracing.

[0016] Step S110: Construct the copyright feature topology corresponding to AIGC generation. The copyright feature topology includes the ownership core node of the original creative material, the creation evolution association edge, and the authorization constraint surface. The ownership core node records the ownership information of the material, the creation evolution association edge records the transformation relationship of the material from the initial form to the current form, and the authorization constraint surface defines the scope boundary of the material that can be used for AIGC generation.

[0017] In this embodiment, a copyright feature topology is constructed using a multimodal creative material (e.g., a comprehensive creative content containing multiple modalities such as text, images, and audio) as an example. First, the ownership information of the creative material is determined. For example, the creator of the creative material is a creative team, and its rights holding status includes the copyright, usage rights, etc. of the material, as well as changes in ownership, such as whether there has been any transfer or licensing to other entities.

[0018] Step S111: Extract the ownership information of the original creative material, and decompose the ownership information into ownership identification features, rights holding features, and ownership change features. Based on the ownership identification features, rights holding features, and ownership change features, construct the feature dimensions of the ownership core node. Each feature dimension corresponds to an attribute branch of the node.

[0019] For the aforementioned multimodal creative materials, their ownership information is extracted. Ownership identification features can include the name and team number of the creative team; rights holding features include the type of copyright the team holds for the creative materials and whether they possess exclusive rights; ownership change features include whether the creative materials have any authorization or transfer records. Then, based on these features, feature dimensions for core ownership nodes are constructed. For example, the owner identification feature corresponds to an attribute branch recording information such as the team's name; the rights holding feature corresponds to an attribute branch recording the specific content of the copyright; and the ownership change feature corresponds to an attribute branch recording information such as the time and object of the change.

[0020] Step S112: Track the morphological transformation process of the original creative material, record the morphological characteristics of the original creative material at different time stages, calculate the transformation similarity of morphological characteristics at adjacent time stages, and construct creative evolution association edges connecting different morphological characteristics with the transformation similarity as the weight. The length of the creative evolution association edges is negatively correlated with the transformation similarity.

[0021] For this multimodal creative material, its morphological transformation process is tracked. The initial form may be a textual creative concept, which is then visualized to form a concept map, and finally combined with audio elements to form multimedia presentation content. The morphological characteristics of different time stages are recorded: the theme and core elements in the creative concept stage; the visual style and composition in the concept map stage; and the audio style and the coordination between visuals and audio in the multimedia presentation stage. The transformation similarity of morphological characteristics between adjacent time stages is calculated. For example, from creative concept to concept map, the consistency of the theme and the presentation degree of core elements are analyzed as the basis for calculating transformation similarity. Using this transformation similarity as weight, creative evolution correlation edges are constructed. The length of the correlation edge is negatively correlated with the transformation similarity; that is, the higher the transformation similarity, the shorter the correlation edge.

[0022] Step S113: Collect the authorization rules for the original creative materials, transform the authorization rules into usage scenario features, dissemination medium features and derivative creation features, construct a three-dimensional authorization constraint surface with the usage scenario features, dissemination medium features and derivative creation features as coordinate axes, and the area within the three-dimensional authorization constraint surface is the legal range within which the material can be used for AIGC generation.

[0023] For this multimodal creative material, its licensing rules are collected, such as restrictions on usage scenarios (non-commercial art exhibitions), restrictions on dissemination media (specific art platforms), and restrictions on derivative creations (the core concept must not be altered). These licensing rules are then transformed into usage scenario characteristics (e.g., art exhibition scenarios, online exhibition scenarios), dissemination media characteristics (e.g., art websites, social media platforms), and derivative creation characteristics (e.g., whether secondary creation or adaptation is allowed). A three-dimensional licensing constraint surface is constructed using these three characteristics as coordinate axes. The area within this three-dimensional licensing constraint surface represents the legal scope within which this multimodal creative material can be used for AIGC generation.

[0024] Step S114: Input the ownership core node, creation evolution association edge, and authorization constraint surface into the topology construction module. Based on the association strength between the ownership core node and the creation evolution association edge, determine the projection position of the ownership core node in the authorization constraint surface, so that the distance between the projection position of the ownership core node and the authorization constraint surface reflects the degree of influence of material ownership on the authorization scope.

[0025] The ownership core node, creative evolution association edges, and authorization constraint surface constructed above are input into the topology construction module. The topology construction module analyzes the association strength between the ownership core node and the creative evolution association edges. For example, the association strength can be calculated by the subject's participation in the transformation of the material's form and the degree of influence of rights. Then, based on the association strength, the projection position of the ownership core node within the authorization constraint surface is determined. The distance between the projection position and the authorization constraint surface reflects the degree of influence of the material's ownership on the scope of authorization. For example, the stronger the ownership subject's control over the material's rights, the closer the projection position is to the authorization constraint surface.

[0026] Step S115: Calculate the intersection information of the creative evolution associated edge and the authorization constraint surface, and integrate the intersection information into the attributes of the creative evolution associated edge. The number and distribution density of the intersections in the intersection information reflect the changing trend of the authorization range during the material evolution process.

[0027] Step S1151: Establish the parametric equation of the creative evolution associated edge. The independent variable of the parametric equation is the time parameter of creative evolution, and the dependent variable is the spatial coordinate of the associated edge in the three-dimensional coordinate system of the authorization constraint surface. The range of the time parameter corresponds to the time span of material evolution.

[0028] In this embodiment, for the creation evolution association edges of the aforementioned multimodal creative materials, the form transformation time of the material is used as the time parameter. For example, from the initial creation time t0 of the material to the current time tn, the value range of the time parameter is [t0, tn]. n The three-dimensional coordinate system of the authorization constraint surface is set with the usage scenario feature, the dissemination medium feature, and the derivative creation feature as coordinate axes. For example, the value range of the usage scenario feature is [0, S], the value range of the dissemination medium feature is [0, M], and the value range of the derivative creation feature is [0, D], where S, M, and D are the maximum values ​​determined according to the authorization usage rules. The parametric equations of the creative evolution association edges are determined according to the position of the material's morphological features in the three-dimensional coordinate system at different times. For example, at time t, the usage scenario feature value corresponding to the material's morphological features is s(t), the dissemination medium feature value is m(t), and the derivative creation feature value is d(t). Then, the parametric equations of the creative evolution association edges can be expressed as: usage scenario feature coordinates = s(t), dissemination medium feature coordinates = m(t), derivative creation feature coordinates = d(t), where t∈[t0, t0, t1], t2, t3, t4, t5, t6, t7, t8, t9, t1, t2, t1, t2, t3, t1, t2, t3, t4, t5 ...2, t3, t4, t5, t6, t1, t2, t3, t4, t5, t1, t2, t3, t4, t5, t2, t3, t4, t5, t2, t3, t4, t5, t6 n ].

[0029] Step S1152: Combine the three-dimensional coordinate equation of the authorized constraint surface with the parametric equation of the creation evolution associated edge, and solve the solution of the combined equation. Each solution corresponds to the coordinates of an intersection point of the associated edge and the constraint surface. The coordinate values ​​contain the spatial coordinate dimension information in three-dimensional space.

[0030] The three-dimensional coordinate equation of the authorization constraint surface is determined according to the collected authorization usage rules. For example, if the authorization usage rules stipulate that the range of the usage scenario feature is [0, S0], the range of the dissemination medium feature is [0, M0], and the range of the derivative creation feature is [0, D0], then the three-dimensional coordinate equation of the authorization constraint surface is: 0 ≤ usage scenario feature coordinate ≤ S0, 0 ≤ dissemination medium feature coordinate ≤ M0, 0 ≤ derivative creation feature coordinate ≤ D0.联立 the parametric equations of the creation evolution association edge (usage scenario feature coordinate = s(t), dissemination medium feature coordinate = m(t), derivative creation feature coordinate = d(t)) with the three-dimensional coordinate equation of the authorization constraint surface to solve for the value of t that satisfies all equations, as well as the corresponding values of s(t), m(t), and d(t). These values are the intersection coordinates, which include the dimension information of the usage scenario feature coordinate, dissemination medium feature coordinate, and derivative creation feature coordinate in three-dimensional space. For example, when t = t1, s(t1) = 5, m(t1) = 8, d(t1) = 10, and 5 ≤ S0, 8 ≤ M0, 10 ≤ D0, and at the same time 5 ≥ 0, 8 ≥ 0, 10 ≥ 0, then (t1, 5, 8, 10) is the relevant information of an intersection point, where t1 is the time parameter value, and 5, 8, and 10 are the three-dimensional space coordinate values respectively.

[0031] Step S1153: Count the number of solutions of the simultaneous equations to obtain the number of intersection points between the creation evolution association edge and the authorization constraint surface. The change in the number corresponds to the change in the number of times of crossing the authorization range boundary during the material evolution process.

[0032] Count the number of solutions obtained by solving the simultaneous equations. For example, if a total of n solutions are obtained, then the number of intersection points is n. If in the subsequent process of material evolution, the number of solutions obtained by solving the simultaneous equations again becomes n + 1, it means that the material evolution process has crossed the authorization range boundary once again. The change in the number of intersection points reflects the change in the number of times of crossing the authorization range boundary.

[0033] Step S1154: Calculate the distribution distance of each intersection point coordinate within the authorization constraint surface, and calculate the distribution density of the intersection points based on the distribution distance. The distribution distance is the Euclidean distance between adjacent intersection point coordinates, and the distribution density is the number of intersection points distributed within the unit space range.

[0034] For the coordinates (s_i, m_i, d_i) and (s_j, m_j, d_j) (i < j) of each intersection point, calculate the Euclidean distance between them. The formula is: Euclidean distance = √[(s_i - s_j) 2 +(m_i - m_j) 2 +(d_i - d_j) 2Then, sum the Euclidean distances of all adjacent intersection points to obtain the total distribution distance. Next, calculate the distribution density of the intersection points based on the spatial extent of the authorized constraint surface (e.g., the volume of the spatial extent is V = S0 × M0 × D0), where distribution density = number of intersection points / V. For example, if the number of intersection points is 5 and the volume of the spatial extent is 1000, then the distribution density is 5 / 1000 = 0.005.

[0035] Step S1155: Analyze the time parameter values ​​corresponding to the intersection coordinates, determine the evolution time point of each intersection, arrange the evolution time points in chronological order, and generate an intersection time series. The interval of the intersection time series corresponds to the time interval between two crossings of the authorized boundary.

[0036] Based on the time parameter value t_i corresponding to each intersection point, determine the evolution time point at which each intersection point occurs, such as t1, t2, t3, etc., and then arrange these evolution time points in chronological order to generate the intersection point time series [t1, t2, t3, ...]. Calculate the difference between two adjacent time points, such as t2-t1, t3-t2, etc. These differences represent the time interval between two crossings of the authorization boundary.

[0037] Step S1156: Perform dimension normalization on the number of intersections, distribution density, and intersection time series to obtain unitless normalized number of intersections, normalized distribution density, and normalized intersection time series; based on the normalized number of intersections, normalized distribution density, and normalized intersection time series, construct an evaluation index for changes in the scope of authorization, wherein the change in the evaluation index of changes in the scope of authorization corresponds to the change in the significance of changes in the scope of authorization during the evolution of the material.

[0038] For the number of intersections, find the maximum value N_max of all possible intersections (e.g., determined based on the time span of material evolution and the extent of the authorized constraint surface), then the normalized number of intersections = number of intersections / N_max. For the distribution density, find the maximum value D_max of all possible distribution densities (e.g., the distribution density is N_max when the number of intersections is N_max and the spatial extent volume is 1), then the normalized distribution density = distribution density / D_max. For the intersection time series, calculate the relationship between each time interval and the total time span (t). n The ratio of -t0) is used to obtain the normalized time interval, thus generating a normalized intersection time series. Then, an evaluation index for changes in the authorized scope is constructed. For example, the evaluation index for changes in the authorized scope = a × number of normalized intersections + b × normalized distribution density + c × (1 - normalized average time interval), where a, b, and c are weighting coefficients, and a + b + c = 1. Changes in this evaluation index reflect the significant changes in the authorized scope during the evolution of the material; for example, a larger evaluation index value indicates a more significant change in the authorized scope.

[0039] Step S1157: Integrate the intersection coordinates, number of intersections, distribution density, intersection time series and authorization range change evaluation index into an intersection information set, and add the intersection information set as an attribute field to the corresponding creation evolution association edge.

[0040] The intersection coordinates (stored in list form, containing the time parameter value and 3D spatial coordinate value of each intersection), the number of intersections, the distribution density, the intersection time series (stored in list form, containing each time interval), and the authorization range change evaluation index obtained above are integrated into an intersection information set. For example, the intersection information set = {intersection coordinates: [(t1, s1, m1, d1), (t2, s2, m2, d2), ...], number of intersections: n, distribution density: d, intersection time series: [t2-t1, t3-t2, ...], authorization range change evaluation index: v}. Then, this intersection information set is added as an attribute field to the corresponding creative evolution association edge. For example, a field named "intersection information" is added to the attributes of the creative evolution association edge, and its value is this intersection information set.

[0041] Step S116: Based on the projection position, intersection information and feature dimensions of each component, generate a copyright feature topology that includes node attributes, edge weights, face coordinates and relationships between components. Each component in the copyright feature topology carries a unique identifier code, and the identifier code corresponds one-to-one with the features of the original creative material.

[0042] Based on the projection positions, intersection information, and feature dimensions of the core ownership nodes, creative evolution related edges, and authorization constraint surfaces obtained above, a copyright feature topology is generated. This copyright feature topology includes node attributes (i.e., the attribute branch information of each core ownership node), edge weights (i.e., the transformation similarity weight of creative evolution related edges, etc.), surface coordinates (i.e., the three-dimensional coordinate information of the authorization constraint surfaces), and relationships between components (such as the association strength between the core ownership nodes and creative evolution related edges, the intersection relationship between creative evolution related edges and authorization constraint surfaces, etc.). Furthermore, each component carries a unique identifier code, which corresponds one-to-one with the features of the original creative material. For example, the identifier code of the core ownership node corresponds to the relevant features of the ownership entity, and the identifier code of the creative evolution related edges corresponds to the relevant features of morphological transformation, etc.

[0043] Step S120: Anchor the copyright feature topology to the parameter dimension of the AIGC generation logic, so that the ownership core node of the copyright feature topology corresponds to the subject identification parameters generated by AIGC, the creation evolution association edge corresponds to the feature iteration parameters generated by AIGC, and the authorization constraint face corresponds to the content boundary parameters generated by AIGC, thus forming a copyright generation logic anchoring relationship.

[0044] In this embodiment, the copyright feature topology constructed above is anchored to the parameter dimension of the AIGC generation logic. First, the parameter system of the AIGC generation logic is analyzed. This parameter system includes subject identification parameters, feature iteration parameters, and content boundary parameters. The subject identification parameters control the degree to which AIGC incorporates the creator's information, the feature iteration parameters control the update frequency of AIGC content features, and the content boundary parameters control the scope limitation of the content generated by AIGC.

[0045] Step S121: Analyze the parameter system of the AIGC generation logic, and extract the subject recognition parameter, feature iteration parameter and content boundary parameter in the parameter system. The subject recognition parameter controls the degree to which AIGC integrates the information of the creator subject. The feature iteration parameter controls the update frequency of AIGC content features. The content boundary parameter controls the range limitation of AIGC generated content.

[0046] This paper provides a detailed analysis of the parameter system of the AIGC generation logic. For example, the subject recognition parameter can be a numerical range, with a larger value indicating more creative subject information incorporated into the AIGC-generated content. The feature iteration parameter can be a time interval or a limit on the number of iterations, controlling the frequency of content feature updates. The content boundary parameter can be a series of rules or numerical ranges, restricting the type, style, and other ranges of the AIGC-generated content. Then, the subject recognition parameter, feature iteration parameter, and content boundary parameter are extracted from this parameter system.

[0047] Step S122: Perform feature encoding on the core nodes of the copyright feature topology to generate a node encoding vector that is consistent with the dimension of the subject identification parameter. Each dimension value of the node encoding vector corresponds to an attribute branch information of the core node of the copyright feature.

[0048] The core nodes of the copyright feature topology described above are feature-encoded. For example, a certain encoding algorithm is used to encode the attribute branch information of the core nodes, such as the subject identification feature, the right holding feature, and the ownership change feature, to generate a node encoding vector. The dimension of the node encoding vector is consistent with the dimension of the subject identification parameter. Each dimension value corresponds to an attribute branch information. For example, the first dimension value corresponds to the encoding information of the subject identification feature, the second dimension value corresponds to the encoding information of the right holding feature, and so on.

[0049] Step S123: Perform numerical transformation on the weights and intersection information of the creative evolution related edges to generate an edge transformation matrix that matches the dimension of the feature iteration parameters. The row dimension of the edge transformation matrix corresponds to the number of creative evolution related edges, and the column dimension corresponds to the number of feature iteration parameters.

[0050] The weights (i.e., transformation similarity weights) and intersection information (such as the number of intersections, distribution density, etc.) of the creative evolution related edges are numerically transformed. For example, the numerical values ​​of the weights and intersection information are normalized to generate an edge transformation matrix. The row dimension of the edge transformation matrix corresponds to the number of creative evolution related edges, and the column dimension corresponds to the number of feature iteration parameters. The element values ​​in the edge transformation matrix represent the degree of correlation between the creative evolution related edges and the feature iteration parameters.

[0051] Step S124: Perform parameter mapping on the three-dimensional coordinates of the authorization constraint surface to generate a surface mapping function with the same format as the content boundary parameters. The input of the surface mapping function is the content feature value generated by AIGC, and the output is the determination result of whether the content feature value is within the authorization constraint surface.

[0052] The three-dimensional coordinates of the authorization constraint surface are parametrically mapped, for example, by converting the three-dimensional coordinates into a parameter form with the same format as the content boundary parameters, thus generating a surface mapping function. The input of this surface mapping function is the content feature value generated by AIGC, such as the style feature value and type feature value of the generated content. The output is the determination result of whether the content feature value is within the authorization constraint surface. That is, if the content feature value is within the three-dimensional coordinate range of the authorization constraint surface, it is determined to be valid; otherwise, it is determined to be invalid.

[0053] Step S125: Associate and bind the node encoding vector with the subject recognition parameters so that the value of the subject recognition parameters is dynamically adjusted according to the dimension value of the node encoding vector, wherein the adjustment range is positively correlated with the dimension value of the node encoding vector.

[0054] Associating and binding node encoding vectors with subject recognition parameters, for example, establishing a mapping relationship, allows the value of the subject recognition parameter to be dynamically adjusted as the dimension value of the node encoding vector changes. The adjustment range is positively correlated with the dimension value of the node encoding vector; that is, the larger the dimension value of the node encoding vector, the greater the adjustment range of the subject recognition parameter, thus incorporating more creative subject information into the AIGC-generated content.

[0055] Step S126: Associate and bind the edge transformation matrix with the feature iteration parameters so that the update frequency of the feature iteration parameters changes with the row dimension value of the edge transformation matrix, and the change trend is positively correlated with the column dimension value of the edge transformation matrix.

[0056] The edge transformation matrix is ​​associated with the feature iteration parameters. For example, the row dimension value of the edge transformation matrix represents the number of edges associated with creation and evolution, and the update frequency of the feature iteration parameters changes with this value. The column dimension value of the edge transformation matrix represents the degree of association with the feature iteration parameters, and the trend of the update frequency of the feature iteration parameters is positively correlated with the column dimension value. That is, the larger the column dimension value, the higher the update frequency of the feature iteration parameters.

[0057] Step S127: Associate and bind the surface mapping function with the content boundary parameters so that the range of the content boundary parameters is adjusted in real time according to the output of the surface mapping function, and the adjustment direction is consistent with the judgment result of the surface mapping function.

[0058] Step S1271: Extract the current range limit value of the content boundary parameter, the current range limit value includes an upper limit threshold and a lower limit threshold, and determine the adjustment step size of the current range limit value.

[0059] In this embodiment, the current range limits for the content boundary parameters are, for example, as follows: the upper limit threshold for the usage scenario feature is S_u, and the lower limit threshold is S_l; the upper limit threshold for the dissemination medium feature is M_u, and the lower limit threshold is M_l; and the upper limit threshold for the derivative creation feature is D_u, and the lower limit threshold is D_l. The adjustment step size can be determined based on the strictness of the licensing rules and the stability requirements of AIGC generation. For example, the adjustment step size for the usage scenario feature is ΔS, the adjustment step size for the dissemination medium feature is ΔM, and the adjustment step size for the derivative creation feature is ΔD, where ΔS, ΔM, and ΔD are preset fixed values, or dynamically adjusted according to the distance between the current content feature value and the boundary of the licensing constraint surface.

[0060] Step S1272: Set the determination period of the surface mapping function, collect the current content feature value generated by AIGC in each determination period, input the current content feature value into the surface mapping function, and obtain the determination result of whether the current content feature value is within the authorized constraint surface.

[0061] The decision period of the surface mapping function can be set according to the speed of AIGC generation and the frequency of content feature changes. For example, a decision can be made after each piece of content (such as a piece of text, an image, or an audio clip) is generated, or a decision can be made at regular intervals (such as 1 minute). Within each decision period, the feature values ​​of the current content generated by AIGC are collected. For example, the feature value of the current content's usage scenario is s_c, the feature value of the dissemination medium is m_c, and the feature value of derivative creation is d_c. The above current content feature values ​​are input into the surface mapping function. The surface mapping function determines whether (s_c, m_c, d_c) is within the authorization constraint surface based on the three-dimensional coordinate equation of the authorization constraint surface (0≤s≤S0, 0≤m≤M0, 0≤d≤D0). If s_c is in the range [0, S0], m_c is in the range [0, M0], and d_c is in the range [0, D0], the decision result is "within the authorization constraint surface"; otherwise, the decision result is "outside the authorization constraint surface", and further determination is made as to whether it exceeds the upper or lower boundary.

[0062] Step S1273: If the determination result is that the current content feature value is within the authorization constraint surface, then keep the current range limit value of the content boundary parameter unchanged, and record the feature value corresponding to this determination result and the determination time.

[0063] When the determination result is that the current content feature value is within the authorization constraint surface, for example, s_c ∈ [0, S0], m_c ∈ [0, M0], d_c ∈ [0, D0], then do not adjust the current range limit value of the content boundary parameter, that is, S_u, S_l, M_u, M_l, D_u, D_l remain unchanged. At the same time, record the feature value (s_c, m_c, d_c) corresponding to this determination result and the determination time t, for example, record it as {(s_c, m_c, d_c), t}.

[0064] Step S1274: If the determination result is that the current content feature value exceeds the upper boundary of the authorization constraint surface, then adjust the upper threshold of the content boundary parameter by the adjustment step size so that the adjusted upper threshold is within the reasonable interval corresponding to the current content feature value, and record the upper threshold before and after the adjustment and the reason for the adjustment.

[0065] When the determination result is that the current content feature value exceeds the upper boundary of the authorization constraint surface, for example, s_c > S0, m_c ≤ M0, d_c ≤ D0, it means that the upper threshold S0 of the usage scenario feature needs to be adjusted. Adjust the upper threshold S_u of the usage scenario feature by the adjustment step size ΔS, and the adjusted upper threshold S_u' = S_u + k × ΔS, where k is the number of adjustment times, such that S_u' ≥ s_c and S_u' - ΔS < s_c (that is, the adjusted upper threshold is within the reasonable interval corresponding to the current content feature value). For example, s_c = 12, S_u = 10, ΔS = 3, then k = 1, S_u' = 10 + 1 × 3 = 13, and at this time (13≥12且13 - 3 = 10 < 12), which meets the requirements. At the same time, record the upper threshold S_u = 10 before the adjustment, the upper threshold S_u' = 13 after the adjustment, and the reason for the adjustment is "the current content feature value s_c = 12 exceeds the upper boundary S0 = 10 of the usage scenario feature of the authorization constraint surface".

[0066] Step S1275: If the determination result is that the current content feature value exceeds the lower boundary of the authorization constraint surface, then adjust the lower threshold of the content boundary parameter by the adjustment step size so that the adjusted lower threshold is within the reasonable interval corresponding to the current content feature value, and record the lower threshold before and after the adjustment and the reason for the adjustment.

[0067] When the determination result indicates that the current content feature value exceeds the lower limit boundary of the authorization constraint surface, for example, s_c < 0, m_c ≥ 0, d_c ≥ 0, it means that the lower limit threshold S_l of the usage scenario feature needs to be adjusted. The lower limit threshold S_l of the usage scenario feature is adjusted by an adjustment step size ΔS. The adjusted lower limit threshold S_l' = S_l - k × ΔS, where k is the number of adjustments, such that S_l' ≤ s_c, and S_l' + ΔS > s_c (i.e., the adjusted lower limit threshold is within a reasonable range corresponding to the current content feature value). For example, if s_c = -2, S_l = 0, ΔS = 3, then k = 1, S_l' = 0 - 1 × 3 = -3. In this case, -3 ≤ -2 and -3 + 3 = 0 > -2, which meets the requirements. At the same time, the lower limit threshold S_l = 0 before adjustment and S_l' = -3 after adjustment are recorded. The reason for the adjustment is that "the current content feature value s_c = -2 exceeds the lower limit boundary of the usage scenario feature S_l = 0 of the authorization constraint surface".

[0068] Step S1276: Collect the judgment results, parameter adjustment status and current content feature values ​​for each judgment period, and construct a parameter adjustment log. Each record in the parameter adjustment log includes a timestamp, judgment result, parameter value and adjustment basis.

[0069] At the end of each judgment period, collect the judgment results ("within the authorization constraint plane" or "outside the authorization constraint plane" and the specific boundary exceeded), parameter adjustment status (if there is an adjustment, record the parameter values ​​before and after the adjustment; if there is no adjustment, record that the parameter values ​​remain unchanged), and the current content feature value, and construct a parameter adjustment log. The format of each record is as follows: timestamp: t; judgment result: "within the authorization constraint plane" / "outside the authorization constraint plane (upper limit / lower limit)"; parameter value: {S_u: value, S_l: value, M_u: value, M_l: value, D_u: value, D_l: value}; adjustment basis: "current content feature value is within the authorization constraint plane" / "current content feature value exceeds the XX boundary of the authorization constraint plane, adjust XX parameter".

[0070] Step S1277: Based on the parameter adjustment log, analyze the adjustment trend of the content boundary parameters, generate a parameter adjustment prediction model, the input of the parameter adjustment prediction model is the prediction result of future content feature values, and the output is the corresponding parameter adjustment suggestion.

[0071] The data in the parameter adjustment log is analyzed. For example, the number of adjustments, adjustment direction (increase or decrease), and adjustment step size of the upper and lower thresholds of usage scenario characteristics, dissemination media characteristics, and derivative creation characteristics are statistically analyzed to determine the adjustment trend of content boundary parameters. Then, machine learning algorithms (such as linear regression, decision trees, etc.) are used to generate a parameter adjustment prediction model. This model takes the predicted results of future content feature values ​​(e.g., the next content feature value predicted by the AIGC-generated model) as input and outputs corresponding parameter adjustment suggestions, such as the suggested parameters, adjustment direction, and adjustment step size.

[0072] Step S1278: Integrate the parameter adjustment prediction model with the surface mapping function and content boundary parameters to form a dynamic adjustment mechanism. This mechanism ensures that the range of content boundary parameters is updated in real time as the content feature values ​​generated by AIGC change and the judgment results of the surface mapping function change, so that the parameter adjustment remains consistent with the boundary requirements of the authorized constraint surface.

[0073] The parameter adjustment prediction model is integrated with the surface mapping function and content boundary parameters. For example, after the surface mapping function makes a determination, the range limit of the content boundary parameters is automatically adjusted based on the determination result and the output of the parameter adjustment prediction model. When AIGC generates new content feature values, the surface mapping function first makes a determination, and then, based on the determination result and the suggestion of the parameter adjustment prediction model, the range limit of the content boundary parameters is updated in real time to ensure that the parameter adjustment is consistent with the boundary requirements of the authorization constraint surface. That is, the range limit of the content boundary parameters can always reasonably restrict the content generated by AIGC to within the authorization range.

[0074] Step S128: Integrate the association and binding relationships between nodes, edges, faces and corresponding parameters to generate a copyright generation logic anchoring relationship that includes parameter adjustment rules, dimension matching relationships and dynamic association thresholds.

[0075] The associations and binding relationships between the above-mentioned node encoding vectors and subject identification parameters, edge transformation matrices and feature iteration parameters, and surface mapping functions and content boundary parameters are integrated to generate a copyright generation logic anchoring relationship. This copyright generation logic anchoring relationship includes parameter adjustment rules (such as the adjustment methods and conditions of subject identification parameters, feature iteration parameters, and content boundary parameters), dimensional matching relationships (such as the dimensional correspondence between node encoding vectors and subject identification parameters, and the dimensional correspondence between edge transformation matrices and feature iteration parameters), and dynamic association thresholds (such as the threshold for association strength and the threshold for adjustment magnitude).

[0076] Step S130: Generate a copyright generation logic symbiotic structure based on the copyright generation logic anchoring relationship. In this symbiotic structure, the nodes, edges, and faces of the copyright feature topology form a dynamic mapping with the parameters of the AIGC generation logic, and the mapping relationship is adjusted in real time as the AIGC generation process proceeds.

[0077] In this embodiment, a copyright generation logic symbiotic structure is generated based on the copyright generation logic anchoring relationship described above. First, based on the parameter adjustment rules in the copyright generation logic anchoring relationship, a mapping matrix between AIGC generation parameters and copyright feature topology components is constructed. The row dimension of this mapping matrix corresponds to the AIGC generation parameters, the column dimension corresponds to the copyright feature topology components, and the matrix element values ​​represent the mapping strength between the parameters and the components.

[0078] Step S131: Based on the parameter adjustment rules in the copyright generation logic anchoring relationship, construct a mapping matrix between AIGC generation parameters and copyright feature topology components. The row dimension of the mapping matrix corresponds to the AIGC generation parameters, the column dimension corresponds to the copyright feature topology components, and the matrix element values ​​represent the mapping strength between the parameters and the components.

[0079] Based on the parameter adjustment rules in the copyright generation logic anchoring relationship, such as the adjustment rules for subject identification parameters, feature iteration parameters, and content boundary parameters, a mapping matrix is ​​constructed between AIGC generation parameters and copyright feature topology components (ownership core nodes, creation evolution association edges, and authorization constraint surfaces). The row dimension of the matrix corresponds to the number of AIGC generation parameters, and the column dimension corresponds to the number of copyright feature topology components. The matrix element values ​​are determined by analyzing the degree of correlation between parameters and components, representing the mapping strength between parameters and components.

[0080] Step S132: Extract the initial parameter values ​​in the AIGC generation process, input the initial parameter values ​​into the mapping matrix, calculate the initial mapping strength corresponding to each copyright feature topology component, and determine the role weight of the copyright feature topology in the initial stage of AIGC generation based on the initial mapping strength.

[0081] Initial parameter values ​​are extracted during the AIGC generation process, such as initial values ​​for subject identification parameters, feature iteration parameters, and content boundary parameters. These initial parameter values ​​are input into the mapping matrix, and matrix operations are used to calculate the initial mapping strength corresponding to each copyright feature topology component, such as the initial mapping strength corresponding to the ownership core node, the initial mapping strength corresponding to the creation evolution association edge, and the initial mapping strength corresponding to the authorization constraint surface. Then, based on these initial mapping strengths, the role weight of the copyright feature topology in the initial stage of AIGC generation is determined. For example, the role weight can be obtained by normalizing the initial mapping strength; a larger weight indicates a greater influence of the component on AIGC generation in the initial stage.

[0082] Step S133: Set the mapping update cycle for the AIGC generation process. In each mapping update cycle, collect intermediate content features generated by AIGC. Based on the semantic association between intermediate content features and component features, analyze the matching degree between intermediate content features and copyright feature topology components. Adjust the element values ​​of the mapping matrix according to the matching degree so that the mapping strength between components with matching degree that meet the preset association requirements and their corresponding parameters is increased, and the mapping strength between components with matching degree that do not meet the preset association requirements and their corresponding parameters is reduced. The adjustment magnitude is positively correlated with the deviation value of the matching degree.

[0083] Step S1331: Set a standard threshold for matching degree, which is determined based on the ideal correlation between the copyright feature topology component and the AIGC generation parameters.

[0084] In this embodiment, the copyright feature topology components include ownership core nodes, creative evolution association edges, and authorization constraint surfaces. The AIGC generation parameters include subject identification parameters, feature iteration parameters, and content boundary parameters. For each copyright feature topology component and its corresponding AIGC generation parameters, a standard threshold for matching degree is set based on the ideal correlation between them. For example, the ideal correlation between the ownership core node and the subject identification parameters is high, so the standard threshold can be set to 0.8; the ideal correlation between the creative evolution association edges and the feature iteration parameters is moderate, so the standard threshold is set to 0.6; the ideal correlation between the authorization constraint surface and the content boundary parameters is high, so the standard threshold is set to 0.8. The above standard thresholds can be adjusted based on domain knowledge and experimental experience.

[0085] Step S1332: Calculate the deviation between the matching degree of each copyright feature topology component and the corresponding AIGC generation parameter and the standard threshold. The deviation is the difference between the matching degree and the standard threshold. A positive value indicates that the matching degree is higher than the standard threshold, and a negative value indicates that the matching degree is lower than the standard threshold.

[0086] First, the matching degree between each copyright feature topology component and the corresponding AIGC generated parameter is calculated. For example, for the ownership core node and the subject identification parameter, the matching degree is calculated by analyzing the semantic and numerical correlations between the features of the ownership core node (subject identification features, rights holding features, ownership change features) and the values ​​of the subject identification parameter (which control the degree to which AIGC integrates the author's information). Assuming the matching degree between the ownership core node and the subject identification parameter is 0.9 and the standard threshold is 0.8, the deviation value is 0.9 - 0.8 = 0.1 (a positive value, indicating that the matching degree is higher than the standard threshold); if the matching degree is 0.5 and the standard threshold is 0.6, the deviation value is 0.5 - 0.6 = -0.1 (a negative value, indicating that the matching degree is lower than the standard threshold).

[0087] Step S1333: Determine the adjustment coefficients for the elements of the mapping matrix. The values ​​of the adjustment coefficients are positively correlated with the absolute values ​​of the deviations. At the same time, the adjustment direction is determined according to the sign of the deviation value. Positive deviations correspond to positive adjustments, and negative deviations correspond to negative adjustments.

[0088] The adjustment coefficient can be determined using either a linear or nonlinear function. For example, using a linear function, the adjustment coefficient k = a × |deviation value| + b, where a and b are constants, and a > 0. This ensures that the value of the adjustment coefficient is positively correlated with the absolute value of the deviation. When the deviation value is positive, the adjustment direction is positive (i.e., increasing the element value of the mapping matrix); when the deviation value is negative, the adjustment direction is negative (i.e., decreasing the element value of the mapping matrix). For example, if the deviation value is 0.1, a = 2, and b = 0.1, then the adjustment coefficient k = 2 × 0.1 + 0.1 = 0.3; if the deviation value is -0.1, the adjustment coefficient k = 2 × 0.1 + 0.1 = 0.3 (the absolute values ​​are the same, so the adjustment coefficient is the same), but the adjustment direction is negative.

[0089] Step S1334: Based on the adjustment coefficient and adjustment direction, calculate the adjustment amount of the mapping matrix element values. The adjustment amount is the current mapping intensity multiplied by the adjustment coefficient, and the adjustment direction determines the positive or negative attribute of the adjustment amount.

[0090] Assuming the element value (mapping intensity) corresponding to the core node and subject identification parameter in the current mapping matrix is ​​m, the adjustment coefficient is k, and the adjustment direction is positive (positive deviation value), then the adjustment amount Δm = m × k (positive value); if the adjustment direction is negative (negative deviation value), then the adjustment amount Δm = -m × k (negative value). For example, if the current mapping intensity m = 0.5, the adjustment coefficient k = 0.3, and the adjustment direction is positive, then the adjustment amount Δm = 0.5 × 0.3 = 0.15; if the adjustment direction is negative, then the adjustment amount Δm = -0.5 × 0.3 = -0.15.

[0091] Step S1335: Add the adjustment amount to the current mapping matrix element value to obtain the adjusted mapping intensity. The adjusted mapping intensity must be kept within a preset reasonable range. If the adjusted intensity exceeds the preset reasonable range, the boundary value of the preset reasonable range shall be taken as the final mapping intensity.

[0092] The adjustment amount is added to the current mapping matrix element value to obtain the adjusted mapping intensity m' = m + Δm. A preset reasonable range is, for example, [0, 1]. If the adjusted mapping intensity m' > 1, then m' = 1; if m' < 0, then m' = 0. For example, if the current mapping intensity m = 0.9 and the adjustment amount Δm = 0.2, then m' = 0.9 + 0.2 = 1.1, which exceeds the reasonable range [0, 1], so the final mapping intensity is 1. If the current mapping intensity m = 0.1 and the adjustment amount Δm = -0.2, then m' = 0.1 - 0.2 = -0.1, which also exceeds the reasonable range, so the final mapping intensity is 0.

[0093] Step S1336: Repeat the above steps of calculating deviation value, determining adjustment coefficient, calculating adjustment amount, and updating element value until the element values ​​of the mapping matrix of all copyright feature topology components and corresponding AIGC generation parameters have been adjusted. Calculate the sum of row dimensions and column dimensions of the adjusted mapping matrix and compare it with the sum of row dimensions and column dimensions of the matrix before adjustment to ensure that the rate of change of the matrix sum before and after adjustment is within the preset stable threshold range. If it exceeds the preset stable threshold, readjust the adjustment coefficient until the rate of change meets the requirements.

[0094] For each copyright feature topology component and the corresponding AIGC generated parameter mapping matrix element values, the above steps are performed. After all adjustments are completed, the sum of row dimensions (the sum of all row element values) and the sum of column dimensions (the sum of all column element values) of the mapping matrix before adjustment are calculated, as well as the sum of row dimensions and the sum of column dimensions after adjustment. The rate of change is calculated, for example, the rate of change of the sum of row dimensions = (the sum of row dimensions after adjustment - the sum of row dimensions before adjustment) / the sum of row dimensions before adjustment × 100%, and the rate of change of the sum of column dimensions is calculated similarly. The preset stable threshold range is, for example, [-5%, 5%]. If the rate of change exceeds this range, the adjustment coefficients are readjusted (e.g., the values ​​of a and b are adjusted), and then the steps of deviation value calculation, adjustment coefficient determination, adjustment amount calculation, and element value update are performed again until the rate of change meets the requirements.

[0095] Step S1337: Record the mapping matrix, deviation value, adjustment coefficient and adjustment amount before and after adjustment to form a matrix adjustment record.

[0096] Record the mapping matrix before adjustment, the mapping matrix after adjustment, the deviation value of each element, the adjustment coefficient, and the adjustment amount to form a matrix adjustment record. For example, the format of the matrix adjustment record is: Mapping matrix before adjustment: [[m 11 m 12 m 13 ], [m 21 m 22 m 23 ], [m 31 m32 m 33 Adjusted mapping matrix: [[m 11 ', m 12 ', m 13 '],[m 21 ', m 22 ', m 23 '],[m 31 ', m 32 ', m 33 Deviation value: [[d]]; 11 d 12 d 13 ],[d 21 d 22 d 23 ],[d 31 d 32 d 33 Adjustment coefficient: [[k] 11 k 12 k 13 ], [k 21 k 22 k 23 ], [k 31 k 32 k 33 Adjustment amount: [[Δm] 11 , Δm 12 , Δm 13 ], [Δm 21 , Δm 22 , Δm 23 ], [Δm 31 , Δm 32 , Δm 33 ]).

[0097] Step S134: Based on the adjusted mapping matrix, update the weight of the copyright feature topology component on the AIGC generation parameters, so that the weight is updated synchronously with the change of mapping intensity, so that the influence of the copyright feature topology on AIGC generation is consistent with the evolution of content features.

[0098] Based on the adjusted mapping matrix, the mapping strength of each copyright feature topology component to the AIGC generation parameters is recalculated, and then the role weight of the component is updated so that the role weight is updated synchronously with the change of mapping strength. For example, the role weight is proportional to the mapping strength. In this way, the influence of copyright feature topology on AIGC generation can be consistent with the evolution of content features. That is, as the content features change, the role weight of the copyright feature topology component is also adjusted accordingly, thereby better controlling the AIGC generation process.

[0099] Step S135: Collect the mapping matrix, impact weight and intermediate content features for each mapping update cycle, and construct a dynamic mapping dataset containing time series information. Each record in the dynamic mapping dataset corresponds to the mapping relationship state within an update cycle.

[0100] After each mapping update cycle ends, the mapping matrix, the weights, and the intermediate content features within that cycle are collected. The above data, along with time information (such as the update cycle number and timestamp), are used to construct a dynamic mapping dataset. Each record corresponds to the mapping relationship status within an update cycle, thus recording the dynamic changes in the mapping relationship during the AIGC generation process.

[0101] Step S136: Based on the dynamic mapping dataset, generate a dynamic association model between copyright feature topology components and AIGC generation parameters. The output of the dynamic association model is the mapping relationship prediction result at different generation stages. Integrate the dynamic association model with the AIGC generation logic to form a symbiotic structure of copyright generation logic.

[0102] Based on the aforementioned dynamic mapping dataset, machine learning algorithms (such as LSTM, GRU, and other time series models) are used to generate a dynamic association model between copyright feature topology components and AIGC generation parameters. This dynamic association model can predict the mapping relationship results at different generation stages based on historical mapping relationship data. Then, this dynamic association model is integrated with the AIGC generation logic to form a symbiotic structure for copyright generation logic. This symbiotic structure can adjust the parameters in the AIGC generation process in real time based on the prediction results of the dynamic association model, thereby achieving a dynamic symbiosis between copyright feature topology and AIGC generation logic.

[0103] Step S140: Drive AIGC content generation based on the copyright generation logic symbiotic structure, so that each feature dimension of the generated content carries the mapping imprint of the copyright feature topology, and the mapping imprint is updated synchronously with the evolution of content features.

[0104] In this embodiment, AIGC content generation is driven by the copyright generation logic symbiotic structure generated above. First, the dynamic association model in the copyright generation logic symbiotic structure is analyzed, and the prediction results of the mapping relationship between different generation stages output by the model are extracted to determine the copyright feature topology components and their weights corresponding to each generation stage.

[0105] Step S141: Analyze the dynamic association model in the copyright generation logic symbiotic structure, extract the prediction results of the mapping relationship of different generation stages output by the dynamic association model, and determine the copyright feature topology components and their weights corresponding to each generation stage.

[0106] This paper provides a detailed analysis of the dynamic association model in the symbiotic structure of copyright generation logic, including its structure and parameter settings. Then, it extracts the mapping relationship prediction results of different generation stages (such as the initial generation stage, intermediate generation stage, and final generation stage) from the model output. Based on these prediction results, it determines the copyright feature topology components (ownership core node, creative evolution association edge, and authorization constraint surface) corresponding to each generation stage and the role weight of each component. For example, in the initial generation stage, the ownership core node has a higher role weight, and in the intermediate generation stage, the creative evolution association edge has a higher role weight.

[0107] Step S142: Input the mapping relationship prediction result of the initial generation stage into the initial parameter configuration module generated by AIGC. The initial parameter configuration module adjusts the initial values ​​of the subject recognition parameter, feature iteration parameter and content boundary parameter based on the component role weight in the mapping relationship prediction result.

[0108] The mapping relationship prediction results from the initial generation stage are input into the initial parameter configuration module generated by AIGC. This module adjusts the initial values ​​of the subject identification parameters, feature iteration parameters, and content boundary parameters based on the component role weights in the mapping relationship prediction results. For example, if the role weight of the core node is high, the initial value of the subject identification parameters is increased to increase the creative subject information incorporated into the AIGC-generated content; if the role weight of the creative evolution association edge is high, the initial value of the feature iteration parameters is adjusted to control the update frequency of content features, etc.

[0109] Step S143: Start the AIGC content generation process. In the first generation stage, generate the first stage intermediate content features based on the adjusted initial parameters, and record the mapping relationship between the copyright feature topology components and content features in this generation stage to form the first stage mapping imprint.

[0110] The AIGC content generation process is initiated. In the first generation stage (i.e., the initial generation stage), the AIGC generation model generates intermediate content features based on adjusted initial parameters (subject identification parameters, feature iteration parameters, and content boundary parameters). This could include generating a piece of text, an image, or an audio clip. Simultaneously, the mapping relationship between the copyright feature topology components (ownership core node, creation evolution association edges, and authorization constraint surfaces) and the content features is recorded. For example, it records which attribute branches of the ownership core node influence the generation of content features, and which weight factors of the creation evolution association edges participate in updating the content features. This mapping relationship forms the first-stage mapping imprint.

[0111] Step S144: Enter the next generation stage, call the dynamic association model to output the mapping relationship prediction result of this generation stage, and adjust the current parameter value of the AIGC generation parameter based on the mapping relationship prediction result so that the current parameter value matches the component action weight of this generation stage.

[0112] Entering the next generation stage (such as the intermediate generation stage), the dynamic association model in the copyright generation logic symbiotic structure is invoked, and the mapping relationship prediction result of this generation stage is output. Based on the prediction result, the current parameter values ​​of the AIGC generation parameters are adjusted. For example, if the role weight of the creation evolution association edge increases in this stage, the current value of the feature iteration parameter is adjusted so that the update frequency of the feature iteration parameter matches the role weight of the creation evolution association edge, so as to better control the evolution of content features.

[0113] Step S145: Generate intermediate content features for this generation stage based on the adjusted parameters, compare the differences between the intermediate content features of the current generation stage and the intermediate content features of the previous generation stage, and update the feature information of the mapping imprint according to the degree of difference, so that the mapping imprint reflects the changes in copyright association during the evolution of content features.

[0114] For example, step S1451: Extract the feature dimension data of the intermediate content features of the current generation stage and the intermediate content features of the previous generation stage. Each feature dimension data includes feature value and feature weight. The feature weight indicates the importance of the feature dimension in the generated content.

[0115] In this embodiment, the feature dimensions of the intermediate content features in the current generation stage and the intermediate content features in the previous generation stage include, for example, theme features, style features, and type features. For each feature dimension, its feature value and feature weight are extracted. For example, the theme feature value in the current generation stage is t_c, and the feature weight is w_tc; the theme feature value in the previous generation stage is t_p, and the feature weight is w_tp; the style feature value in the current generation stage is s_c, and the feature weight is w_sc; the style feature value in the previous generation stage is s_p, and the feature weight is w_sp, etc.

[0116] Step S1452: Calculate the feature value difference rate of the corresponding feature dimension between the two generation stages. The difference rate is the ratio of the absolute value of the difference between the feature values ​​of the two generation stages to the feature value of the previous generation stage. The change in the ratio corresponds to the change in the significance of the feature change in the corresponding feature dimension.

[0117] For each feature dimension, calculate the feature value difference rate. For example, the feature value difference rate for the topic feature is |t_c - t_p| / t_p (when t_p ≠ 0). If t_p = 0, then the difference rate is |t_c - t_p| / 1 (to avoid dividing by zero). The feature value difference rate for the style feature is |s_c - s_p| / s_p (similarly). The larger the difference rate, the more significant the feature change in that feature dimension.

[0118] Step S1453: Calculate the feature weight difference of the corresponding feature dimensions between the two generation stages. The weight difference is the difference between the feature weight of the current generation stage and the feature weight of the previous generation stage. A positive value indicates that the weight of the current generation stage is increased compared to the previous stage, and a negative value indicates that the weight of the current generation stage is decreased compared to the previous generation stage.

[0119] For each feature dimension, calculate the feature weight difference. For example, the feature weight difference for topic features = w_tc - w_tp, and the feature weight difference for style features = w_sc - w_sp, etc. If the weight difference is positive, it indicates that the importance of that feature dimension has increased in the current generation stage; if it is negative, it indicates that the importance has decreased.

[0120] Step S1454: Based on the feature value difference rate and feature weight difference, construct a content feature difference evaluation value, wherein the content feature difference evaluation value is a weighted sum of the feature value difference rate and the feature weight difference.

[0121] Weights are assigned to the eigenvalue difference rate and feature weight difference for each feature dimension. For example, the weight of the eigenvalue difference rate is α, and the weight of the feature weight difference is β, where α + β = 1. Then, the content feature difference assessment value = α × eigenvalue difference rate + β × feature weight difference. For example, the content feature difference assessment value for the topic feature = 0.6 × |t_c - t_p| / t_p + 0.4 × (w_tc - w_tp), and the content feature difference assessment value for the style feature = 0.6 × |s_c - s_p| / s_p + 0.4 × (w_sc - w_sp), etc. Then, the content feature difference assessment values ​​for all feature dimensions are summed to obtain the total content feature difference assessment value.

[0122] Step S1455: Set the update threshold for the difference evaluation value. If the difference evaluation value of the content features between the current stage and the previous stage exceeds the update threshold, it is determined that the mapping imprint needs to be updated. If it does not exceed the threshold, the mapping imprint feature information of the previous stage remains unchanged.

[0123] Based on the sensitivity requirements of changes in content features generated by AIGC, an update threshold for the difference evaluation value is set, for example, an update threshold of 0.3. If the total content feature difference evaluation value exceeds 0.3, it is determined that the mapping imprint needs to be updated; if it does not exceed 0.3, the mapping imprint feature information of the previous stage remains unchanged.

[0124] Step S1456: When it is determined that an update is needed, extract the association information between the copyright feature topology components and content features in the current stage, including component identifier, role weight and association time, and add it to the feature information of the mapping imprint. At the same time, delete the content in the previous generation stage that conflicts with the association information in the current stage.

[0125] When an update is deemed necessary, the association information between the current stage's copyright feature topology components (ownership core node, creative evolution association edge, and authorization constraint surface) and content features is extracted. For example, the ownership core node's component identifier is N1, its weight is w_N1, and its association time is t_c; the creative evolution association edge's component identifier is E1, its weight is w_E1, and its association time is t_c; the authorization constraint surface's component identifier is F1, its weight is w_F1, and its association time is t_c. This association information is added to the feature information of the mapping imprint. Simultaneously, the association information from the previous generation stage is checked, and content conflicting with the current stage is deleted. For instance, if the weight of the ownership core node in the association information from the previous generation stage is inconsistent with the weight in the current stage, and the current stage's association information is more accurate based on changes in content features, then the association information of the ownership core node from the previous generation stage is deleted.

[0126] Step S1457: Update the feature evolution identifier in the mapping imprint. The evolution identifier records the type of change of the content feature from the previous stage to the current stage. The change type includes enhancement, weakening, addition and deletion. The evolution identifier corresponds to the change of the content feature difference evaluation value. The change of the content feature difference evaluation value corresponds to the change of the degree of change of the evolution identifier.

[0127] Based on changes in the content feature difference assessment value and feature dimensions, the feature evolution identifier in the mapping imprint is updated. For example, if the feature value difference rate of the theme feature is large and the feature weight difference is positive, it indicates that the theme feature is enhanced, and the evolution identifier is recorded as "enhanced"; if the feature value difference rate of the style feature is small and the feature weight difference is negative, it indicates that the style feature is weakened, and the evolution identifier is recorded as "weakened"; if a new feature dimension (such as sentiment feature) is added, the evolution identifier is recorded as "added"; if a feature dimension (such as scene feature) is deleted, the evolution identifier is recorded as "deleted". The degree of change in the evolution identifier corresponds to the change in the content feature difference assessment value. For example, the larger the content feature difference assessment value, the more significant the change in the evolution identifier (e.g., from "slightly enhanced" to "significantly enhanced").

[0128] Step S1458: Integrate the updated association information and evolution identifiers to form the mapping imprint feature information for the current stage.

[0129] The updated copyright feature topology components are integrated with the association information (component identifier, role weight, association time) and feature evolution identifier to form the mapping imprint feature information for the current stage. For example, mapping imprint feature information = {association information: [(N1, w_N1, t_c), (E1, w_E1, t_c), (F1, w_F1, t_c)], evolution identifier: {theme feature: "enhanced", style feature: "weakened", sentiment feature: "added", scene feature: "deleted"}}.

[0130] Step S146: Repeat the parameter adjustment, content generation, and imprint update steps of the above generation stage until the content evolution of all generation stages is completed. The mapping imprint of each generation stage contains the copyright feature topology component identifier, role weight, and content feature association information of that generation stage.

[0131] Repeat steps S144 to S145 to enter the next generation stage, adjusting parameters, generating content, and updating imprints until the content evolution of all generation stages is completed, such as the initial generation stage, intermediate generation stage, and final generation stage. The mapping imprint of each generation stage includes the copyright feature topology component identifier (such as the identifier of the core ownership node, the identifier of the creation evolution related edge, etc.), the role weight (such as the role weight value of each component in this stage), and the content feature association information (such as the mapping relationship between components and content features, etc.).

[0132] Step S147: In the final content generation stage, integrate the mapping imprints from all generation stages to form a mapping imprint chain that runs through the entire process of content feature evolution, so that each feature dimension of the generated content corresponds to an imprint fragment in the mapping imprint chain.

[0133] In the final content generation stage, the mapping imprints of all generation stages (initial generation stage, intermediate generation stage, final generation stage, etc.) are integrated and connected in chronological order to form a mapping imprint chain that runs through the entire process of content feature evolution. Each feature dimension of the generated content (such as the theme feature of text, the color feature of image, the melody feature of audio, etc.) corresponds to an imprint fragment in the mapping imprint chain. Through this imprint fragment, the topological relationship between that feature dimension and the copyright feature during the generation process can be traced.

[0134] Step S150: Analyze the mapping imprints in the generated content feature dimension, reconstruct the dynamic mapping relationship of the copyright generation logic symbiotic structure, and invert the core ownership node, creation evolution association edge, and authorization constraint surface of the copyright feature topology based on the dynamic mapping relationship.

[0135] In this embodiment, for the generated AIGC content, the mapping imprints in its feature dimensions are analyzed to reconstruct the dynamic mapping relationship of the copyright generation logic symbiotic structure and to invert the relevant components of the copyright feature topology. First, the mapping imprint fragments corresponding to each feature dimension of the generated content are extracted. Based on the generation stage identifier in the mapping imprint fragment, the imprint fragments are spliced ​​together in chronological order to restore the complete mapping imprint chain.

[0136] Step S151: Extract the mapping imprint fragments corresponding to each feature dimension of the generated content, and based on the generation stage identifier in the mapping imprint fragments, splice the imprint fragments in chronological order to restore the complete mapping imprint chain.

[0137] Extract the mapping imprint fragments corresponding to each feature dimension of the generated content (such as each theme feature of the text, each color feature of the image, each melody feature of the audio, etc.), and then splice the above imprint fragments in chronological order according to the generation stage identifiers in the mapping imprint fragments (such as the initial stage, intermediate stage, final stage, etc.) to restore the complete mapping imprint chain. This mapping imprint chain records the topological relationship between the content feature evolution process and the copyright feature.

[0138] Step S152: Analyze the copyright feature topology component identifier, role weight, and content feature association information of each imprint fragment in the mapping imprint chain, establish the correspondence between the imprint fragment and the generation stage, and form an imprint stage association table.

[0139] The copyright feature topology component identifiers (such as the identifiers of the core ownership node, the identifiers of the creation evolution association edges, the identifiers of the authorization constraint surfaces, etc.), the role weights (such as the role weights of each component in the generation stage corresponding to the imprint fragment), and the content feature association information (such as the mapping relationship between components and content features) of each imprint fragment in the mapping imprint chain are analyzed. Then, the correspondence between imprint fragments and generation stages is established, such as which imprint fragment corresponds to the initial generation stage, which corresponds to the intermediate generation stage, etc. The above correspondence is formed into an imprint stage association table.

[0140] Step S153: Based on the component identifiers in the imprint stage association table, count the frequency of occurrence and the trend of change of the role weight of each copyright feature topology component in different generation stages, and generate a component role heatmap. The horizontal axis of the component role heatmap is the generation stage, and the vertical axis is the role weight.

[0141] Based on the component identifiers in the imprint stage association table, the frequency of occurrence of each copyright feature topology component (ownership core node, creation evolution association edge, authorization constraint surface) in different generation stages (initial stage, intermediate stage, final stage, etc.) is statistically analyzed. For example, the number of times the ownership core node appears in the initial stage and the number of times it appears in the intermediate stage. Simultaneously, the trend of the change in the influence weight of each component in different generation stages is analyzed, such as whether the influence weight gradually increases or decreases. Then, a component influence heatmap is generated based on these statistical results. The horizontal axis of the heatmap represents the generation stage, and the vertical axis represents the influence weight. The influence weight of the component in different generation stages is represented by the color intensity or numerical value of the heatmap.

[0142] Step S154: Based on the component action heatmap, determine the target action stage of each copyright feature topology component in the AIGC generation process, and extract the imprint fragment information of the target action stage. The target action stage is a continuous generation stage in which the component action weight exceeds a preset threshold.

[0143] Based on the component effect heatmap, the changes in the effect weight of each copyright feature topology component are analyzed to determine the target effect stage of each component in the AIGC generation process, i.e., the continuous generation stage where the component effect weight exceeds a preset threshold. For example, if the effect weight of the ownership core node exceeds the preset threshold in the initial stage and the first half of the intermediate stage, then these two stages are the target effect stages of the ownership core node. Then, imprint fragment information of these target effect stages is extracted, including component identifier, effect weight, content feature association information, etc.

[0144] Step S155: Input the imprint fragment information of the target action stage into the dynamic mapping reconstruction module. Based on the content feature association information in the imprint fragment, calculate the mapping strength between the copyright feature topology component and the AIGC generation parameters, and reconstruct the mapping matrix of each target action stage.

[0145] The imprint fragment information of the target action stage is input into the dynamic mapping reconstruction module. This module analyzes the correlation between the copyright feature topology components and the AIGC generation parameters (subject identification parameters, feature iteration parameters, and content boundary parameters) based on the content feature association information in the imprint fragments, and calculates the mapping strength. Then, based on the calculated mapping strength, the mapping matrix for each target action stage is reconstructed. The row dimension of this mapping matrix corresponds to the AIGC generation parameters, the column dimension corresponds to the copyright feature topology components, and the matrix element values ​​represent the mapping strength between the parameters and the components.

[0146] Step S156: Integrate the mapping matrices of each target action stage to generate a dynamic mapping relationship sequence of the copyright generation logic symbiotic structure, wherein each element in the dynamic mapping relationship sequence corresponds to a mapping relationship state of a target action stage.

[0147] The mapping matrices of each target action stage are integrated and arranged in chronological order to generate a dynamic mapping relationship sequence of the copyright generation logic symbiotic structure. Each element in this dynamic mapping relationship sequence corresponds to the mapping relationship state of a target action stage. Through this dynamic mapping relationship sequence, the changes in the mapping relationship between the copyright feature topology components and AIGC generation parameters at different target action stages can be clearly understood.

[0148] Step S157: Based on the dynamic mapping relationship sequence, reverse the original features of the copyright feature topology components. The features of the core ownership node are reversed through the correlation between the subject identification parameters and the mapping strength. The features of the creation evolution related edges are reversed through the correlation between the feature iteration parameters and the mapping strength. The features of the authorization constraint surface are reversed through the correlation between the content boundary parameters and the mapping strength.

[0149] Based on the dynamic mapping relationship sequence, this study analyzes the changes in the mapping intensity between copyright feature topology components and AIGC generation parameters to infer the original features of the copyright feature topology components. For core ownership nodes, their features are inferred by the correlation between subject identification parameters and mapping intensity. For example, changes in subject identification parameters reflect the influence of subject identification features and rights holding features of the core ownership nodes. For creative evolution association edges, their features are inferred by the correlation between feature iteration parameters and mapping intensity. For example, changes in feature iteration parameters reflect the influence of transformation similarity weights and intersection information of creative evolution association edges. For authorization constraint surfaces, their features are inferred by the correlation between content boundary parameters and mapping intensity. For example, changes in content boundary parameters reflect the influence of usage scenario features and dissemination media features of the authorization constraint surfaces.

[0150] Step S158: Verify the consistency between the component features obtained by reverse engineering and the original copyright feature topology component features. Adjust the reverse engineering parameters based on the consistency verification results until the association error between the reverse engineering features and the original component features is within the preset range.

[0151] The copyright feature topology component features obtained through reverse engineering are compared with the original copyright feature topology component features to verify their consistency of association. For example, the similarity or error between the reverse-engineered features and the original features is calculated. If the association error exceeds a preset range, the reverse engineering parameters are adjusted, such as adjusting the calculation method of the mapping strength and the analysis model of the degree of association. Then, the reverse engineering is performed again until the association error between the reverse-engineered features and the original component features is within the preset range.

[0152] Step S159: Integrate the core ownership nodes, creative evolution related edges, and authorization constraint surface features obtained by reverse inference to restore the complete copyright feature topology. Based on the restored copyright feature topology, output the copyright tracing result of AIGC generated content. The copyright tracing result includes ownership subject information, creative evolution path, and authorization scope definition.

[0153] The features of the core ownership nodes, creative evolution related edges, and authorization constraint surfaces obtained through reverse engineering are integrated to reconstruct the complete copyright feature topology. Then, based on the reconstructed copyright feature topology, the copyright tracing results of AIGC-generated content are output. These results include information on the ownership entity (such as the name of the creative team and the status of rights holding), the creative evolution path (such as the transformation process of the material from its initial form to its current form), and the scope of authorization (such as the legal scope within which the material can be used for AIGC generation).

[0154] Figure 2 The illustration shows exemplary hardware and software components of an AIGC generation system 100 based on multimodal content copyright tracing, which can implement the ideas of this application, according to some embodiments of this application. For example, processor 120 can be used in the AIGC generation system 100 based on multimodal content copyright tracing and to perform the functions in this application.

[0155] The AIGC generation system 100 based on multimodal content copyright tracing can be a general-purpose server or a special-purpose server; both can be used to implement the AIGC generation method based on multimodal content copyright tracing of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0156] For example, the AIGC generation system 100 based on multimodal content copyright tracing may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the AIGC generation system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The AIGC generation system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0157] For ease of explanation, only one processor is described in the AIGC generation system 100 based on multimodal content copyright tracing. However, it should be noted that the AIGC generation system 100 based on multimodal content copyright tracing in this application may also include multiple processors. Therefore, the steps performed by one processor as described in this application may also be performed jointly or individually by multiple processors. For example, if the processor of the AIGC generation system 100 based on multimodal content copyright tracing performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0158] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the AIGC generation method based on multimodal content copyright tracing as described above is implemented.

[0159] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. An AIGC generation method based on multimodal content copyright tracing, characterized in that, The method includes: Construct a copyright feature topology corresponding to AIGC generation. The copyright feature topology includes the ownership core node of the original creative material, the creative evolution association edge and the authorization constraint surface. The ownership core node records the ownership information of the material, the creative evolution association edge records the transformation relationship of the material from the initial form to the current form, and the authorization constraint surface defines the scope boundary of the material that can be used for AIGC generation. The copyright feature topology is anchored to the parameter dimension of the AIGC generation logic, so that the ownership core node of the copyright feature topology corresponds to the subject identification parameters generated by AIGC, the creation evolution association edge corresponds to the feature iteration parameters generated by AIGC, and the authorization constraint face corresponds to the content boundary parameters generated by AIGC, thus forming a copyright generation logic anchoring relationship. Based on the anchoring relationship of copyright generation logic, a symbiotic structure of copyright generation logic is generated. In this symbiotic structure, the nodes, edges, and faces of the copyright feature topology form a dynamic mapping with the parameters of the AIGC generation logic, and the mapping relationship is adjusted in real time with the AIGC generation process. AIGC content generation is driven by a copyright generation logic symbiotic structure, so that each feature dimension of the generated content carries a mapping imprint of the copyright feature topology, and the mapping imprint is updated synchronously with the evolution of content features. The mapping imprints in the generated content feature dimensions are analyzed, and the dynamic mapping relationship of the copyright generation logic symbiotic structure is reconstructed. Based on the dynamic mapping relationship, the ownership core node, creation evolution association edge and authorization constraint surface of the copyright feature topology are inverted.

2. The AIGC generation method based on multimodal content copyright tracing according to claim 1, characterized in that, The construction of the copyright feature topology corresponding to AIGC generation includes: Extract the ownership information of the original creative material, and decompose the ownership information into ownership identification features, rights holding features and ownership change features. Based on the ownership identification features, rights holding features and ownership change features, construct the feature dimensions of the core ownership node, and each feature dimension corresponds to an attribute branch of the node. The process of morphological transformation of the original creative material is tracked, the morphological characteristics of the original creative material at different time stages are recorded, the transformation similarity of morphological characteristics at adjacent time stages is calculated, and the creative evolution association edges connecting different morphological characteristics are constructed with the transformation similarity as the weight. The length of the creative evolution association edge is negatively correlated with the transformation similarity. Collect the authorization rules for the use of original creative materials, transform the authorization rules into usage scenario features, dissemination medium features and derivative creation features, construct a three-dimensional authorization constraint surface with the usage scenario features, dissemination medium features and derivative creation features as coordinate axes, and the area within the three-dimensional authorization constraint surface is the legal range within which the material can be used for AIGC generation; Input the ownership core node, creative evolution related edge, and authorization constraint surface into the topology construction module. Based on the association strength between the ownership core node and the creative evolution related edge, determine the projection position of the ownership core node in the authorization constraint surface, so that the distance between the projection position of the ownership core node and the authorization constraint surface reflects the degree of influence of material ownership on the scope of authorization. Calculate the intersection information of the creative evolution associated edge and the authorization constraint surface, and integrate the intersection information into the attributes of the creative evolution associated edge. The number and distribution density of the intersection points in the intersection information reflect the changing trend of the authorization range during the material evolution process. Based on the projection position, intersection information, and feature dimensions of each component, a copyright feature topology is generated that includes node attributes, edge weights, face coordinates, and relationships between components. Each component in the copyright feature topology carries a unique identifier code, which corresponds one-to-one with the features of the original creative material.

3. The AIGC generation method based on multimodal content copyright tracing according to claim 1, characterized in that, The process of anchoring copyright feature topology to the parameter dimensions of AIGC generation logic, such that the core ownership nodes of the copyright feature topology correspond to the subject identification parameters generated by AIGC, the creation evolution association edges correspond to the feature iteration parameters generated by AIGC, and the authorization constraint surfaces correspond to the content boundary parameters generated by AIGC, forms a copyright generation logic anchoring relationship, including: The parameter system of AIGC generation logic is analyzed, and the subject recognition parameter, feature iteration parameter and content boundary parameter are extracted from the parameter system. The subject recognition parameter controls the degree to which AIGC integrates the information of the creator subject, the feature iteration parameter controls the update frequency of AIGC content features, and the content boundary parameter controls the scope limitation of AIGC generated content. Feature encoding is performed on the ownership core node of the copyright feature topology to generate a node encoding vector consistent with the dimension of the subject identification parameter. Each dimension value of the node encoding vector corresponds to an attribute branch information of the ownership core node. The weights and intersection information of the creative evolution related edges are numerically transformed to generate an edge transformation matrix that matches the dimension of the feature iteration parameters. The row dimension of the edge transformation matrix corresponds to the number of creative evolution related edges, and the column dimension corresponds to the number of feature iteration parameters. The three-dimensional coordinates of the authorization constraint surface are parameterized to generate a surface mapping function with the same format as the content boundary parameters. The input of the surface mapping function is the content feature value generated by AIGC, and the output is the determination result of whether the content feature value is within the authorization constraint surface. The node encoding vector is associated and bound with the subject recognition parameters, so that the value of the subject recognition parameters is dynamically adjusted according to the dimension value of the node encoding vector, wherein the adjustment magnitude is positively correlated with the dimension value of the node encoding vector; The edge transformation matrix is ​​associated with the feature iteration parameters, so that the update frequency of the feature iteration parameters changes with the row dimension value of the edge transformation matrix, and the trend of change is positively correlated with the column dimension value of the edge transformation matrix. By associating and binding the surface mapping function with the content boundary parameter, the range of the content boundary parameter is adjusted in real time according to the output of the surface mapping function, and the adjustment direction is consistent with the judgment result of the surface mapping function. Integrate the association and binding relationships between nodes, edges, faces and corresponding parameters to generate a copyright generation logic anchoring relationship that includes parameter adjustment rules, dimension matching relationships and dynamic association thresholds.

4. The AIGC generation method based on multimodal content copyright tracing according to claim 1, characterized in that, The copyright generation logic symbiotic structure generated based on the copyright generation logic anchoring relationship includes: Based on the parameter adjustment rules in the copyright generation logic anchoring relationship, a mapping matrix between AIGC generation parameters and copyright feature topology components is constructed. The row dimension of the mapping matrix corresponds to the AIGC generation parameters, the column dimension corresponds to the copyright feature topology components, and the matrix element values ​​represent the mapping strength between the parameters and the components. Extract the initial parameter values ​​during the AIGC generation process, input the initial parameter values ​​into the mapping matrix, calculate the initial mapping strength corresponding to each copyright feature topology component, and determine the role weight of the copyright feature topology in the initial stage of AIGC generation based on the initial mapping strength. Set the mapping update cycle for the AIGC generation process. In each mapping update cycle, collect intermediate content features generated by AIGC. Based on the semantic association between intermediate content features and component features, analyze the matching degree between intermediate content features and copyright feature topology components. Adjust the element values ​​of the mapping matrix according to the matching degree so that the mapping strength between components with matching degree that meet the preset association requirements and their corresponding parameters is increased, and the mapping strength between components with matching degree that do not meet the preset association requirements and their corresponding parameters is reduced. The adjustment magnitude is positively correlated with the deviation value of the matching degree. Based on the adjusted mapping matrix, the weight of the copyright feature topology component on the AIGC generation parameters is updated so that the weight is updated synchronously with the change of mapping intensity, so that the influence of the copyright feature topology on AIGC generation is consistent with the evolution of content features. Collect the mapping matrix, impact weight, and intermediate content features for each mapping update cycle to construct a dynamic mapping dataset containing time series information. Each record in the dynamic mapping dataset corresponds to the mapping relationship status within an update cycle. Based on the dynamic mapping dataset, a dynamic association model is generated between copyright feature topology components and AIGC generation parameters. The output of the dynamic association model is the mapping relationship prediction result at different generation stages. The dynamic association model is integrated with the AIGC generation logic to form a symbiotic structure of copyright generation logic.

5. The AIGC generation method based on multimodal content copyright tracing according to claim 1, characterized in that, The AIGC content generation driven by the copyright generation logic symbiotic structure ensures that each feature dimension of the generated content carries a mapping imprint of the copyright feature topology. This mapping imprint is updated synchronously with the evolution of the content features, including: The dynamic association model in the symbiotic structure of copyright generation logic is analyzed, and the prediction results of the mapping relationship of different generation stages output by the dynamic association model are extracted. The copyright feature topology components and their weights corresponding to each generation stage are determined. The mapping relationship prediction results from the initial generation stage are input into the initial parameter configuration module generated by AIGC. The initial parameter configuration module adjusts the initial values ​​of the subject identification parameters, feature iteration parameters, and content boundary parameters based on the component role weights in the mapping relationship prediction results. Initiate the AIGC content generation process. In the first generation stage, generate intermediate content features for the first stage based on the adjusted initial parameters. At the same time, record the mapping relationship between the copyright feature topology components and content features in this generation stage to form the first stage mapping imprint. Entering the next generation stage, the dynamic association model is called to output the mapping relationship prediction result of this generation stage. Based on the mapping relationship prediction result, the current parameter value of the AIGC generation parameter is adjusted so that the current parameter value matches the component role weight of this generation stage. Based on the adjusted parameters, the intermediate content features of this generation stage are generated. The differences between the intermediate content features of the current generation stage and the intermediate content features of the previous generation stage are compared. The feature information of the mapping imprint is updated according to the degree of difference, so that the mapping imprint reflects the changes in copyright association during the evolution of content features. Repeat the parameter adjustment, content generation, and imprint update steps of the above generation stage until the content evolution of all generation stages is completed. The mapping imprint of each generation stage contains the copyright feature topology component identifier, role weight, and content feature association information of that generation stage. In the final content generation stage, the mapping imprints from all generation stages are integrated to form a mapping imprint chain that runs through the entire process of content feature evolution, so that each feature dimension of the generated content corresponds to an imprint fragment in the mapping imprint chain.

6. The AIGC generation method based on multimodal content copyright tracing according to claim 1, characterized in that, The parsing generates mapping imprints in the content feature dimension, reconstructs the dynamic mapping relationship of the copyright generation logic symbiotic structure, and inverts the core ownership nodes, creation evolution association edges, and authorization constraint surfaces of the copyright feature topology based on the dynamic mapping relationship, including: Extract the mapping imprint fragments corresponding to each feature dimension of the generated content, and based on the generation stage identifier in the mapping imprint fragments, splice the imprint fragments in chronological order to restore the complete mapping imprint chain. The copyright feature topology component identifier, role weight, and content feature association information of each imprint fragment in the mapping imprint chain are analyzed to establish the correspondence between the imprint fragment and the generation stage, forming an imprint stage association table; Based on the component identifiers in the imprint stage association table, the frequency of occurrence and the trend of change of the role weight of each copyright feature topology component in different generation stages are statistically analyzed to generate a component role heatmap. The horizontal axis of the component role heatmap is the generation stage, and the vertical axis is the role weight. Based on the component action heatmap, the target action stage of each copyright feature topology component in the AIGC generation process is determined, and the imprint fragment information of the target action stage is extracted. The target action stage is a continuous generation stage in which the component action weight exceeds a preset threshold. The imprint fragment information of the target action stage is input into the dynamic mapping reconstruction module. Based on the content feature association information in the imprint fragment, the mapping strength between the copyright feature topology component and the AIGC generation parameter is calculated, and the mapping matrix of each target action stage is reconstructed. By integrating the mapping matrices of each target action stage, a dynamic mapping relationship sequence of the copyright generation logic symbiotic structure is generated, where each element in the dynamic mapping relationship sequence corresponds to a mapping relationship state of a target action stage. Based on the dynamic mapping relationship sequence, the original features of the copyright feature topology components are inferred. The features of the core ownership node are inferred through the correlation between the subject identification parameters and the mapping strength. The features of the creation evolution related edges are inferred through the correlation between the feature iteration parameters and the mapping strength. The features of the authorization constraint surface are inferred through the correlation between the content boundary parameters and the mapping strength. Verify the consistency between the component features obtained by reverse engineering and the original copyright feature topology component features. Adjust the reverse engineering parameters based on the consistency verification results until the association error between the reverse engineering features and the original component features is within the preset range. By integrating the core ownership nodes, creative evolution related edges, and authorization constraint surface features obtained through reverse inference, a complete copyright feature topology is restored. Based on the restored copyright feature topology, the copyright tracing results of AIGC-generated content are output. The copyright tracing results include ownership subject information, creative evolution path, and authorization scope definition.

7. The AIGC generation method based on multimodal content copyright tracing according to claim 2, characterized in that, The intersection information of the creative evolution association edge and the authorization constraint surface is calculated, and the intersection information is integrated into the attributes of the creative evolution association edge, including: Establish a parametric equation for the creative evolution associated edge. The independent variable of the parametric equation is the time parameter of creative evolution, and the dependent variable is the spatial coordinate of the associated edge in the three-dimensional coordinate system of the authorization constraint surface. The range of the time parameter corresponds to the time span of material evolution. The three-dimensional coordinate equations of the authorized constraint surface and the parametric equations of the creation evolution associated edge are combined and the solution of the combined equations is obtained. Each solution corresponds to the coordinates of an intersection point of the associated edge and the constraint surface. The coordinate values ​​contain spatial coordinate dimension information in three-dimensional space. By counting the number of solutions to the simultaneous equations, we can obtain the number of intersections between the creative evolution-related edges and the authorization constraint surfaces. The change in the number corresponds to the change in the number of times the material crosses the boundary of the authorization range during the evolution process. Calculate the distribution distance of the coordinates of each intersection point within the authorized constraint surface, and calculate the distribution density of the intersection points based on the distribution distance. The distribution distance is the Euclidean distance between the coordinates of adjacent intersection points, and the distribution density is the number of intersection points distributed within a unit space. Analyze the time parameter values ​​corresponding to the intersection coordinates, determine the evolution time point of each intersection, arrange the evolution time points in chronological order, and generate an intersection time series. The interval of the intersection time series corresponds to the time interval between two crossings of the authorization boundary. The number of intersection points, distribution density, and intersection point time series are normalized to obtain unitless normalized number of intersection points, normalized distribution density, and normalized intersection point time series. Based on the normalized number of intersection points, normalized distribution density, and normalized intersection point time series, an evaluation index for changes in the scope of authorization is constructed. The change in the evaluation index of changes in the scope of authorization corresponds to the change in the significance of changes in the scope of authorization during the evolution of the material. The intersection coordinates, number of intersections, distribution density, intersection time series, and evaluation indicators of changes in authorized scope are integrated into an intersection information set, and this intersection information set is added as an attribute field to the corresponding creative evolution association edge.

8. The AIGC generation method based on multimodal content copyright tracing according to claim 3, characterized in that, The step of associating and binding the surface mapping function with the content boundary parameters, so that the range of the content boundary parameters is adjusted in real time according to the output of the surface mapping function, and the adjustment direction is consistent with the determination result of the surface mapping function, includes: Extract the current range limit value of the content boundary parameter, which includes an upper limit threshold and a lower limit threshold, and determine the adjustment step size of the current range limit value; Set the determination period of the surface mapping function, collect the current content feature value generated by AIGC in each determination period, input the current content feature value into the surface mapping function, and obtain the determination result of whether the current content feature value is within the authorized constraint surface; If the determination result is that the current content feature value is within the authorization constraint plane, then keep the current range limit value of the content boundary parameter unchanged, and record the feature value and determination time corresponding to the determination result. If the determination result is that the current content feature value exceeds the upper limit boundary of the authorization constraint surface, the upper limit threshold of the content boundary parameter is adjusted by the adjustment step size so that the adjusted upper limit threshold is in the reasonable range corresponding to the current content feature value. At the same time, the upper limit threshold before and after the adjustment and the reason for the adjustment are recorded. If the determination result is that the current content feature value exceeds the lower limit boundary of the authorization constraint surface, then the lower limit threshold of the content boundary parameter is adjusted by the adjustment step size so that the adjusted lower limit threshold is in the reasonable range corresponding to the current content feature value. At the same time, the lower limit threshold before and after the adjustment and the reason for the adjustment are recorded. Collect the judgment results, parameter adjustment status, and current content feature values ​​for each judgment period, and construct a parameter adjustment log. Each record in the parameter adjustment log includes a timestamp, judgment result, parameter value, and adjustment basis. Based on the parameter adjustment log, the adjustment trend of content boundary parameters is analyzed, and a parameter adjustment prediction model is generated. The parameter adjustment prediction model takes the prediction result of future content feature values ​​as input and outputs the corresponding parameter adjustment suggestions. By integrating the parameter adjustment prediction model with the surface mapping function and content boundary parameters, a dynamic adjustment mechanism is formed. This mechanism ensures that the range of content boundary parameters is updated in real time as the content feature values ​​generated by AIGC change and the judgment results of the surface mapping function change, so that the parameter adjustment remains consistent with the boundary requirements of the authorized constraint surface.

9. The AIGC generation method based on multimodal content copyright tracing according to claim 4, characterized in that, The step of adjusting the element values ​​of the mapping matrix according to the matching degree increases the mapping strength between components whose matching degree meets the preset association requirements and their corresponding parameters, and decreases the mapping strength between components whose matching degree does not meet the preset association requirements and their corresponding parameters. The adjustment magnitude is positively correlated with the deviation value of the matching degree, including: A standard threshold for matching degree is set, which is determined based on the ideal correlation between the copyright feature topology component and the AIGC generation parameter; Calculate the deviation between the matching degree of each copyright feature topology component and the corresponding AIGC generation parameter and the standard threshold. The deviation is the difference between the matching degree and the standard threshold. A positive value indicates that the matching degree is higher than the standard threshold, and a negative value indicates that the matching degree is lower than the standard threshold. Determine the adjustment coefficients for the elements of the mapping matrix. The values ​​of the adjustment coefficients are positively correlated with the absolute value of the deviation. At the same time, the direction of adjustment is determined according to the sign of the deviation value: a positive deviation corresponds to a positive adjustment, and a negative deviation corresponds to a negative adjustment. Based on the adjustment coefficient and adjustment direction, the adjustment amount of the mapping matrix element values ​​is calculated. The adjustment amount is the current mapping intensity multiplied by the adjustment coefficient, and the adjustment direction determines the positive or negative attribute of the adjustment amount. The adjustment amount is superimposed on the current mapping matrix element value to obtain the adjusted mapping strength. The adjusted mapping strength must be kept within a preset reasonable range. If the adjusted strength exceeds the preset reasonable range, the boundary value of the preset reasonable range is taken as the final mapping strength. Repeat the above steps of calculating deviation value, determining adjustment coefficient, calculating adjustment amount, and updating element value until the element values ​​of the mapping matrix of all copyright feature topology components and corresponding AIGC generated parameters have been adjusted. Then, calculate the sum of row dimensions and the sum of column dimensions of the adjusted mapping matrix and compare it with the sum of row dimensions and the sum of column dimensions of the matrix before adjustment. This ensures that the rate of change of the sum of the matrix before and after adjustment is within the preset stable threshold range. If it exceeds the preset stable threshold, readjust the adjustment coefficient until the rate of change meets the requirements. Record the mapping matrix, deviation value, adjustment coefficient, and adjustment amount before and after adjustment to form a matrix adjustment record.

10. An AIGC generation system based on multimodal content copyright tracing, characterized in that, The AIGC generation system based on multimodal content copyright tracing includes a processor and a memory, the memory and the processor being connected. The memory is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the memory to implement the AIGC generation method based on multimodal content copyright tracing as described in any one of claims 1-9.

Images