An AI platform-based picture sharing method and device, electronic equipment and storage medium

Abstract

The application relates to the technical field of picture sharing, and discloses an AI platform-based picture sharing method and device, electronic equipment and a storage medium, the method comprising the following steps: acquiring a generated data stream of a picture uploaded by a first user, converting the generated data stream into a structured JSON file, and supporting multi-terminal sharing by built-in cross-platform compatible labels; based on a second user's creation preference portrait and key features, filtering non-key parameters through an AI semantic matching model, generating an adaptive priority adjustable component, and generating a difference heat map and a new picture after user adjustment; embedding enhanced copyright metadata containing an original author identifier, content fingerprints and the like into a JSON parameter stream; calculating a hierarchical hash value and block chain notarization every time the parameter is iterated, calculating a contribution degree based on multidimensional indexes, distributing a benefit, generating a corresponding NFT copyright certificate, and supporting on-chain reverse tracing. The application can solve the problem that a generated parameter lacks standardized structured encapsulation and dependency relationship annotation, resulting in inefficient secondary creation parameter adjustment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of picture sharing, and in particular to a picture sharing method and device based on an AI platform, an electronic device and a storage medium. BACKGROUND

[0002] The existing AI picture generation and sharing technology has been deeply integrated into many fields such as design, social interaction and marketing due to its efficient creation ability, and provides users with a convenient personalized picture generation and dissemination channel. However, in the current technology, picture sharing is mostly realized in the form of finished files, and the core parameters in the generation process are not standardized and structured, which not only leads to compatibility problems in multi-terminal adaptation, but also makes it impossible for subsequent users to carry out accurate secondary optimization based on the original core logic of creation, seriously limiting the continuity and expandability of creation.

[0003] This technical defect in the secondary creation scene derives the key problems: due to the lack of structured arrangement and dependency relationship annotation of the generation parameters, it is difficult for subsequent users to know the correlation between the parameters when adjusting the parameters, which is prone to parameter adjustment conflicts or invalid adjustments, resulting in low secondary creation efficiency; at the same time, there is a lack of unified parameter adaptation rules in multi-terminal adaptation, and the same parameters have different presentation effects on different terminals, further intensifying the fragmentation of the creation experience and hindering the large-scale promotion of AI picture collaborative creation.

[0004] From the above, how to solve the problem of low efficiency of secondary creation parameter adjustment caused by the lack of standardized and structured encapsulation and dependency relationship annotation of generation parameters still needs to be solved. SUMMARY

[0005] In order to solve the problem of low efficiency of secondary creation parameter adjustment caused by the lack of standardized and structured encapsulation and dependency relationship annotation of generation parameters, the present application provides a picture sharing method and device based on an AI platform, an electronic device and a storage medium.

[0006] In the first aspect, the present application provides a picture sharing method based on an AI platform, which adopts the following technical solution:

[0007] A picture sharing method based on an AI platform, comprising:

[0008] obtaining a first picture corresponding to the generation data stream uploaded by a first user, the generation data stream including a generation model unique identifier, a customized random seed, an iteration step number, a prompt word semantic analysis result and a model weight slice, converting the generation data stream into a structured JSON file containing a parameter dependency graph, and the JSON file is built-in cross-platform compatible tags to support multi-terminal sharing to at least one second user;

[0009] Based on the second user's creative preference profile and the input key features, the non-key parameters in the JSON file are filtered through an AI semantic matching model to generate self-adaptive priority-ordered adjustable components, each component is associated with a constraint threshold for parameter adjustment and an AI recommended adjustment range, and the default value is dynamically optimized according to the user's historical adjustment behavior; after the second user adjusts the parameters through the adjustable components, a difference heat map before and after the adjustment is generated in real time, and a new picture is generated based on the difference heat map and the adjustment parameters;

[0010] Embed enhanced copyright metadata in the JSON parameter stream, the enhanced copyright metadata including original author identification, creation timestamp, license type, AI-generated content fingerprint information, and contribution history record containing contributor identification, contribution timestamp, contribution type, contribution score, and adjustment parameter snapshot;

[0011] At each parameter iteration, the hierarchical hash value of the JSON parameter stream is calculated through the Merkle tree algorithm to generate a blockchain storage record including version number, timestamp, contributor information, hierarchical hash value, and parameter dependency graph abstract; based on multi-dimensional index calculation, the contribution degree is calculated, the multi-dimensional index including parameter adjustment dimension complexity, adjustment amplitude deviation, AI-quantified generation effect improvement value, and innovation score of adjustment behavior, wherein the effect improvement value includes aesthetic score and semantic consistency score, the copyright income proportion is dynamically allocated according to the contribution degree proportion, and the NFT copyright certificate bound with the hierarchical hash value is generated, the NFT copyright certificate supports reverse tracing of the corresponding parameter iteration version through on-chain query.

[0012] Optionally, in the process of generating self-adaptive priority-ordered adjustable components, the method further comprises:

[0013] The core semantic vector of the second user's key features is extracted through an AI semantic matching model, and the core semantic vector of the second user's key features is calculated with the functional semantic vector of each parameter in the JSON file for cosine similarity;

[0014] Parameters with a similarity lower than a preset similarity threshold are marked as non-key parameters, and the non-key parameters are prioritized in combination with the commonly used adjustment dimensions and adjustment frequency weights in the second user's creative preference profile;

[0015] For each non-key parameter after sorting, the associated parameters are identified based on the parameter dependency graph to generate a component group of main parameters and associated parameters, wherein the parameter linkage rules in the component group are labeled;

[0016] The second user's adjustment operation is monitored in real time, and if it is detected that the parameter adjustment exceeds the constraint threshold, multiple optimal adjustment schemes are pushed through the AI recommendation module for user selection or reference, wherein the optimal adjustment scheme includes an effect preview picture corresponding to the scheme.

[0017] Optionally, in the process of generating the difference heat map in real time before and after the adjustment, the method further comprises:

[0018] Pixel-level feature extraction is performed on the original picture before adjustment and the preview picture after adjustment to obtain feature data of color distribution, texture structure, and semantic region division;

[0019] The difference values of the original picture before adjustment and the preview picture after adjustment in each feature dimension are calculated by an AI difference quantization model, and the difference values are mapped to the color gradient of the heat map, wherein the higher the difference value, the brighter the color of the heat map;

[0020] Semantic labels are superimposed in the heat map to label the parameter adjustment items corresponding to the differences, and the user is supported to automatically locate to the corresponding adjustable component when clicking the semantic label;

[0021] If the second user continuously adjusts the same component group, the heat map is updated in real time and the cumulative difference value is calculated, and when the cumulative difference value exceeds a preset difference threshold, an effect solidification reminder is triggered and a parameter snapshot of the current adjustment version is saved.

[0022] Optionally, in the process of embedding the enhanced copyright metadata into the JSON parameter stream, the method further comprises:

[0023] The first picture and the generated data stream are processed by an AI content fingerprint generation model to extract dual content fingerprints based on visual features and parameter features;

[0024] An adjustment intention description field is added to the contribution history record, and the adjustment intention description field is automatically generated by a natural language processing model based on input text or adjustment behavior of the second user;

[0025] The enhanced copyright metadata is embedded into the specified field of the JSON parameter stream according to the hierarchical structure of original author information-core copyright information-contribution history record, and the copyright metadata is encrypted;

[0026] The contribution score in the contribution history record is updated synchronously each time the parameter is iterated, and the contribution score is dynamically calculated by an AI copyright evaluation model based on the adjustment effect improvement value and the innovation score.

[0027] Optionally, in the process of calculating the contribution degree based on multi-dimensional indicators, the method further comprises:

[0028] Initial weights of each dimension indicator are set, and the weights are dynamically adjusted according to the picture generation scene, wherein the picture generation scene includes commercial design, artistic creation, and daily sharing;

[0029] The specific values of each indicator are calculated, wherein the indicators include dimension complexity, adjustment amplitude deviation, effect improvement value, and innovation score;

[0030] Calculate the total contribution degree according to the adjusted index weight, and if there are multiple second users, calculate the contribution degree proportion of each user respectively to generate a contribution degree distribution matrix.

[0031] Optionally, in the process of generating the NFT copyright certificate bound with the hierarchical hash value, the method further comprises:

[0032] Based on the hierarchical hash value generated by the Merkle tree algorithm, the root hash value is extracted as the unique identifier of the NFT copyright certificate, and each layer hash value is used as the metadata attribute of the NFT;

[0033] Embed the parameter iterative traceability chain in the NFT copyright certificate, the parameter iterative traceability chain includes all iterative nodes from the original version to the current version, each node is associated with a corresponding hierarchical hash value, contributor information and parameter adjustment summary;

[0034] For secondary transactions of NFT copyright certificates, the copyright income distribution mechanism is automatically triggered during the transaction, the transaction income is split to each contributor account according to the current contribution degree distribution matrix, and a chain transaction record is generated;

[0035] If it is detected that the picture content corresponding to the NFT is infringed, the double content fingerprint and parameter dependence graph summary are extracted through on-chain query to serve as core evidence for infringement protection.

[0036] Optionally, in the process of embedding cross-platform compatible tags in the JSON file to support multi-terminal sharing, the method further comprises:

[0037] Analyze the picture rendering capabilities and interaction logic of the current mainstream terminals to generate a rendering rule library adapted to each terminal, wherein the current mainstream terminals include mobile terminals, PC terminals, tablet terminals and VR devices;

[0038] Embed terminal type identification tags and rendering parameter adaptation tags in the JSON file, wherein the rendering parameter adaptation tags include the picture resolution, color space and component interaction mode corresponding to each terminal;

[0039] When the second user receives the JSON file, the terminal automatically reads the identification tags, calls the corresponding adaptation rules from the rendering rule library, and dynamically adjusts the size, layout and operation mode of the adjustable components;

[0040] Lightweight processing is performed on the JSON file, redundant parameter data is removed based on the parameter dependence graph, the file size is compressed through the LZ77 compression algorithm, and a complete parameter recovery interface is retained, and the user can restore all parameter data when needed.

[0041] In a second aspect, the application provides a picture sharing device based on an AI platform, which adopts the following technical solution:

[0042] An AI platform-based picture sharing device, comprising:

[0043] A data stream structured cross-platform module acquires a generated data stream corresponding to a first picture uploaded by a first user, the generated data stream including a generated model unique identifier, a customized random seed, an iteration step number, a prompt word semantic analysis result, and a model weight slice, converts the generated data stream into a structured JSON file containing a parameter dependency graph, and the JSON file is built-in with a cross-platform compatible label to support multi-terminal sharing to at least one second user;

[0044] A parameter self-adaptive adjustment visualization module filters non-key parameters in the JSON file through an AI semantic matching model based on a second user's creation preference portrait and input key features, generates self-adaptive priority-ordered adjustable components, each component is associated with a constraint threshold of parameter adjustment and an AI recommended adjustment range, and a default value is dynamically optimized according to user historical adjustment behaviors; after the second user adjusts the parameters through the adjustable components, a difference heat map before and after adjustment is generated in real time, and a new picture is generated based on the difference heat map and the adjusted parameters;

[0045] A copyright metadata embedding update module embeds enhanced copyright metadata into the JSON parameter stream, the enhanced copyright metadata including an original author identifier, a creation timestamp, a license type, AI-generated content fingerprint information, and a contribution history record containing contributor identifier, contribution timestamp, contribution type, contribution score, and adjustment parameter snapshot;

[0046] A blockchain copyright tracing contribution allocation module calculates a hierarchical hash value of the JSON parameter stream through a Merkle tree algorithm at each parameter iteration, generates a blockchain notarization record including a version number, a timestamp, contributor information, a hierarchical hash value, and a parameter dependency graph abstract; calculates a contribution degree based on multi-dimensional indicators, the multi-dimensional indicators including a dimensional complexity of parameter adjustment, an adjustment amplitude deviation degree, an AI-quantified generation effect improvement value, and an innovation score of adjustment behavior, wherein the effect improvement value includes an aesthetic score and a semantic consistency score, dynamically allocates a copyright income proportion according to a contribution degree proportion, and generates an NFT copyright certificate bound to the hierarchical hash value, the NFT copyright certificate supporting reverse tracing of the corresponding parameter iteration version through on-chain query.

[0047] In a third aspect, the present application provides an electronic device, which adopts the following technical solution:

[0048] An electronic device, comprising a processor, wherein the processor runs a program of the AI platform-based picture sharing method of any one of the above.

[0049] In a fourth aspect, the present application provides a storage medium, which adopts the following technical solution:

[0050] A storage medium storing a program of the AI platform-based picture sharing method according to any one of the above.

[0051] In summary, the present application includes at least one of the following beneficial technical effects:

[0052] By converting the generated data stream into a structured JSON file containing a parameter dependency graph, the standardization of the generated parameters and the explicit labeling of the dependency relationship are realized, which fundamentally solves the problem of unclear parameter association in secondary creation. At the same time, based on the AI semantic matching model, non-key parameters are screened, and an adaptive priority-ordered component group is generated combined with user creation preferences, and parameter linkage rules are labeled, so that subsequent users can clearly understand the influence logic between parameters and avoid adjustment conflicts; combined with parameter adjustment constraint thresholds and AI recommendation schemes, invalid adjustments are greatly reduced, and the parameter adjustment efficiency of secondary creation is significantly improved.

[0053] Through the pixel-level feature extraction and the difference heat map generated by the AI difference quantization model, the influence of parameter adjustment on the picture is intuitively presented, and the semantic label and component positioning function are superimposed to further reduce the operation threshold of parameter adjustment; the cumulative difference value monitoring and solidification reminder during continuous adjustment ensure the continuity of secondary creation. In addition, the cross-platform compatible design and lightweight processing of the JSON file solve the problem of inconsistent presentation of multiple terminals, and the enhanced copyright metadata embedding and blockchain storage mechanism also realize the accurate definition and tracing of creation rights, which improves the efficiency of secondary creation while perfecting the ecological guarantee of AI picture sharing. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 is a flowchart of an AI platform-based picture sharing method according to an exemplary embodiment.

[0055] Figure 2 is a structural block diagram of an AI platform-based picture sharing device according to an exemplary embodiment. DETAILED DESCRIPTION

[0056] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0057] In the description of the specification, the description of the terms "certain embodiments", "one embodiment", "some embodiments", "illustrative embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the described embodiment or example are included in at least one embodiment or example of the present application. In the specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in one or more embodiments or examples.

[0058] The embodiment of the present application discloses a picture sharing method based on an AI platform, referring to Figure 1 , comprising:

[0059] S100, acquiring the generated data stream corresponding to the first picture uploaded by the first user, the generated data stream including the generated model unique identifier, the customized random seed, the iteration step number, the prompt word semantic analysis result and the model weight slice, converting the generated data stream into a structured JSON file containing a parameter dependency graph, and the JSON file is built-in cross-platform compatible label to support multi-terminal sharing to at least one second user.

[0060] The specific execution process of S100 is as follows:

[0061] Step 1, generated data stream acquisition:

[0062] First, the complete generated data stream corresponding to the first picture uploaded by the first user is acquired, which is not a picture finished product file, but a core "creation track data" in the picture generation process. Among them, the generated model unique identifier is used to accurately locate the original AI model for generating the picture (to avoid the effect deviation caused by model mismatch in subsequent secondary creation); the customized random seed is the initial random value of picture generation (to ensure that the original creation logic can be reproduced based on the seed); the iteration step number records the iteration number of the model when generating the picture (reflecting the fine degree of picture generation); the prompt word semantic analysis result is the structured disassembly of the user's original prompt word (such as disassembling "blue sea sunset" into "scene: sea, time: sunset, main color: blue" and other semantic units, which is convenient for subsequent parameter association); the model weight slice is the core parameter fragment of the original generation model (only the weight data directly related to the current picture generation is extracted, considering data integrity and lightweight).

[0063] Step 2, structured JSON file conversion:

[0064] The key to converting the generated data stream into a structured JSON file is to embed the parameter dependency graph. Specifically, by analyzing the correlation logic and influence weight between parameters, for example, adjusting the "main color" parameter will affect the "light intensity" and "color saturation" parameters, and adjusting the "iteration step" will affect the "detail richness" and "rendering speed", these correlation relationships are written into the JSON file in the form of a graph (such as a node-edge structure), so that the originally scattered parameters form a logical whole, realizing the standardization and structured packaging of parameters.

[0065] Step 3, cross-platform compatible label embedding:

[0066] Two types of core labels are built into the JSON file: terminal type identification label (used to identify the type of receiving terminal device) and rendering parameter adaptation label (contains adaptation rules for mainstream terminals such as mobile terminals, PC terminals, tablet terminals, and VR devices, such as mobile terminal adaptation to low resolution, simplified component layout, PC terminal support for high resolution, batch parameter operation, etc.). At the same time, based on the parameter dependency graph, redundant data is removed (such as deleting model weight fragments unrelated to the current picture generation), the file size is compressed through the LZ77 compression algorithm, and the "complete parameter recovery interface" is retained (to ensure that users can restore all original parameters when needed).

[0067] Step 4, multi-terminal sharing adaptation:

[0068] After completing the structure and label embedding, the JSON file can support cross-terminal transmission to at least one second user, and the receiving terminal will automatically read the identification label in the file and call the corresponding adaptation rules, without the need for manual adjustment by the user, to achieve parameter and terminal adaptation compatibility.

[0069] Through the standardized and structured packaging of generated data stream, parameter dependency graph construction, and cross-platform label embedding, the problems of "scattered and irregular generation parameters, unclear dependency relationships, and poor multi-terminal adaptation" in traditional technology are solved from the root: on the one hand, the structured JSON file and parameter dependency graph provide a basis for data for precise optimization when adjusting parameters in subsequent secondary creation; on the other hand, cross-platform compatible design and lightweight processing ensure the convenience of sharing, realizing consistent presentation of parameters on different terminals.

[0070] S200, based on the second user's creative preference profile and the input key features, filtering the non-key parameters in the JSON file through the AI semantic matching model, generating self-adaptive priority-ordered adjustable components, each component associated with the constraint threshold of parameter adjustment and the AI recommended adjustment range, and the default value dynamically optimized according to the user's historical adjustment behavior; After the second user adjusts the parameters through the adjustable components, real-time difference heat map is generated before and after adjustment, and new picture is generated based on the difference heat map and the adjustment parameters.

[0071] The specific execution process of S200 is as follows:

[0072] Step 1, user feature and key data call:

[0073] Firstly, the second user's creative preference profile (which is based on user historical adjustment records, favorite content, creative tags, etc. and contains common adjustment dimensions such as "color optimization" and "detail enhancement", and adjustment frequency weights such as high-frequency adjustments of "brightness" and "contrast") is called, and the second user's input key features (such as "preserve the seaside scene and optimize the sky color" and "enhance the facial details of the characters") are obtained.

[0074] Step 2, AI semantic matching and non-key parameter filtering:

[0075] The text information of the second user's input key features is converted into core semantic vectors, and cosine similarity calculation is performed between the AI semantic matching model and the function semantic vectors of each parameter in the JSON file (such as "color adjustment" semantic corresponding to "main color" and "detail richness" semantic corresponding to "iteration step number"). Set a preset similarity threshold (such as 0.6), and mark the parameters with a similarity lower than the threshold as non-key parameters, i.e. parameters that the user does not explicitly focus on and can be flexibly adjusted. The key parameters (such as the parameters corresponding to the user's mentioned "sky color") with a similarity higher than the threshold are locked and cannot be adjusted arbitrarily to ensure that the core requirements are not damaged.

[0076] Step 3, self-adaptive adjustable component generation:

[0077] The non-critical parameters selected are prioritized by combining the "common adjustment dimensions" and "adjustment frequency weights" in the second user's creative preference profile (e.g., if the user frequently adjusts "saturation," it is prioritized). An adaptive priority-adjustable component is generated. Each component is associated with two core pieces of information: one is the constraint threshold for parameter adjustment (based on the parameter dependency graph, to avoid adjusting beyond a reasonable range and causing picture distortion, such as a "brightness" adjustment threshold limited to -30%~+30%); the second is the AI-recommended adjustment range (based on similar picture optimization cases and user preferences, such as a recommended "saturation" adjustment range of +5%~+15%). At the same time, the default value of the component is not a fixed value, but is dynamically optimized based on the user's historical adjustment behavior (e.g., if the user has a habit of increasing the saturation of "warm-toned pictures" by 10%, the default value for this scenario is automatically set to +10%).

[0078] Step 4, real-time generation and interaction of difference heat map:

[0079] After the second user adjusts the parameters through the adjustable component, the system immediately starts a double-image comparison analysis: first, pixel-level feature extraction is performed on the original picture before adjustment and the preview picture after adjustment, obtaining feature data such as color distribution, texture structure, and semantic region division (e.g., "sky," "person," "background"); then, an AI difference quantification model is used to calculate the difference values of the two images in each feature dimension, and the difference values are mapped to the color gradient of the heat map (the higher the difference value, the brighter the color, e.g., a high difference value in "sky color" adjustment presents a deep red color); subsequently, semantic labels (e.g., "saturation +12%" and "brightness +8%") are superimposed on the heat map, and interactive functions are supported, allowing the user to click on the semantic labels to automatically locate the corresponding adjustable component for precise correction.

[0080] Step 5, new picture generation:

[0081] After the second user confirms the parameter adjustment, the system generates a new picture that meets the user's needs based on the final adjustment parameters and effective optimization data in the difference heat map (excluding invalid adjustment difference values), calls the appropriate generation model (matched through the unique identifier of the generation model in S100), and retains the parameter snapshot of this adjustment for future tracing.

[0082] Through the whole process design of "user feature-driven parameter intelligent screening, adaptive component generation, and adjustment effect visualization," the problems of "blind parameter adjustment, unclear dependency, and multiple inefficient adjustments" in traditional secondary creation are accurately solved.

[0083] AI semantic matching and priority sorting allow users to focus only on non-critical parameters, avoid core requirements from being broken, and reduce adjustment errors by setting threshold values and AI recommendation ranges. In addition, a difference heat map visually presents adjustment effects and associated parameters, reducing the operation threshold, and dynamically optimized default values further adapt to user habits.

[0084] S300, enhanced copyright metadata is embedded in the JSON parameter stream, including original author identification, creation timestamp, license type, AI-generated content fingerprint information, and contribution history records containing contributor identification, contribution timestamp, contribution type, contribution score, and adjustment parameter snapshot.

[0085] The specific execution process of S300 is as follows:

[0086] Step 1, enhanced copyright metadata collection and generation:

[0087] First, collect basic copyright core data, including original author identification (the unique account ID or identity of the first user), creation timestamp (the exact time the first picture was generated, accurate to milliseconds), and license type (such as commercial use license, non-commercial use license, secondary creation authorization license, etc., preset by the first user). Then generate enhanced core data: joint process the first picture (visual features) and the generated data stream (parameter features) in S100 through the AI content fingerprint generation model to extract dual content fingerprints (to avoid copyright invalidation due to single feature tampering); initialize the contribution history record, including basic fields (contributor identification, contribution timestamp, contribution type such as "parameter adjustment" "demand optimization", contribution score, and adjustment parameter snapshot), and add an adjustment intent description field, automatically generate structured descriptions by analyzing the second user's input text (such as "optimize sky color") or adjustment behavior (such as continuous adjustment of "saturation" "color temperature" parameters) through a natural language processing model, ensuring that the adjustment behavior is traceable.

[0088] Step 2, metadata hierarchical arrangement and encryption:

[0089] Arrange all enhanced copyright metadata according to the hierarchical structure of "original author information-core copyright information-contribution history record" to avoid data confusion. Among them, the original author information is the top layer of data, the core copyright information (including dual content fingerprints and license type) is the middle layer of core, and the contribution history record is the bottom layer of dynamic expansion data. Then encrypt the entire copyright metadata: the encryption key uses a hash combination of "original author identification + creation timestamp" (to ensure uniqueness and strong association with the copyright subject), and the encrypted data is embedded in the specified hidden field of the JSON parameter stream to prevent unauthorized tampering or theft.

[0090] Step 3, metadata embedding JSON parameter stream and synchronous update:

[0091] The hierarchical encrypted enhanced copyright metadata is embedded in the JSON parameter stream generated by S100, and is associated with the parameter dependency graph. Each parameter iteration version corresponds to a complete copyright metadata record. At the same time, a synchronous update mechanism is set up: each time the second user completes parameter adjustment (i.e. when S200 generates a new picture), the system automatically adds a new record to the contribution history record, supplements the current contributor identification (second user ID), contribution timestamp (adjustment completion time), contribution type (determined according to adjustment operation), adjustment parameter snapshot (parameter name, original value, adjusted value of this adjustment), and through the AI copyright evaluation model based on the adjustment effect improvement value and the innovation score in S200, dynamically calculates and updates the "contribution score", ensuring that the copyright metadata is synchronized with the creation process in real time.

[0092] Through double content fingerprint and encryption processing, the uniqueness and security of the copyright are ensured, and the infringement behavior is effectively prevented; in addition, the detailed contribution history record (including adjustment intention, parameter snapshot, contribution score) completely retains the full-link track of multi-user collaborative creation, at the same time, the rights boundary of the original author and the contributor is clear and traceable, greatly improving the copyright protection strength in AI picture sharing and secondary creation, and stimulating the enthusiasm of multi-user collaborative creation.

[0093] S400, at each parameter iteration, the hierarchical hash value of the JSON parameter stream is calculated by the Merkle tree algorithm, and the block chain storage record including version number, timestamp, contributor information, hierarchical hash value and parameter dependency graph abstract is generated; based on multi-dimensional index, the contribution degree is calculated, the multi-dimensional index includes the dimensional complexity of parameter adjustment, adjustment amplitude deviation, AI quantitative generation effect improvement value, and innovation score of adjustment behavior, according to the contribution degree proportion, the copyright income proportion is dynamically allocated, and the NFT copyright certificate bound with hierarchical hash value is generated, the NFT copyright certificate supports reverse tracing corresponding parameter iteration version through on-chain query.

[0094] Among them, S400 is realized through blockchain technology to realize traceability of creation process, accurately calculate contribution degree based on multi-dimensional index, and generate transferable NFT copyright certificate, the specific execution process is as follows:

[0095] Step 1, hierarchical hash value calculation:

[0096] Each time the second user completes parameter adjustment (i.e. once parameter iteration), the system calls the Merkle tree algorithm to process the current JSON parameter stream. First, the JSON parameter stream is split into multiple data blocks according to "core parameters - associated parameters - copyright metadata", and the hash value of each data block (constituting the leaf nodes of the Merkle tree) is calculated; then the parent node hash value is calculated layer by layer, and finally the root hash value and each layer branch hash value (i.e. hierarchical hash value) are generated, ensuring that any minor tampering of the parameter stream will result in a change in the hash value, achieving data integrity verification.

[0097] Step 2, blockchain record generation:

[0098] Based on the hierarchical hash value, a structured blockchain record is generated, containing five core fields: version number (increasing with the number of parameter iterations, such as V1.0 for the initial version and V1.1 for the first adjustment), timestamp (the exact time of parameter iteration completion), contributor information (the unique identifier of the second user for this adjustment), hierarchical hash value (including root hash value and key branch hash value), and parameter dependency graph summary (a compressed representation of the graph in S100, used for quick association of original parameter logic). The generated record is synchronized to the blockchain network, taking advantage of the decentralized and tamper-proof nature of the blockchain to achieve full-link evidence of the creation process.

[0099] Step 3, multi-dimensional contribution calculation:

[0100] First, set the initial weights of each dimension index (dimension complexity 0.2, adjustment amplitude deviation 0.15, effect improvement value 0.4, and innovation score 0.25), and dynamically adjust according to the picture generation scenario (commercial design, artistic creation, daily sharing) (such as increasing the effect improvement value weight to 0.5 in the commercial design scenario); then calculate the specific values of each index: dimension complexity is calculated according to "adjustment dimension number x 1 + associated parameter number x 0.5", adjustment amplitude deviation is the deviation ratio of actual adjustment value to the average of AI recommendation range, effect improvement value is the weighted average of AI aesthetic score (weight 0.6) and semantic consistency score (weight 0.4), and innovation score is determined by comparing historical adjustment records on the platform (repetition rate less than 5% for full score); finally, calculate the total contribution according to the adjusted weight, and if there are multiple second users, generate a contribution ratio matrix for each user.

[0101] Step 4, dynamic distribution of copyright benefits:

[0102] According to the contribution degree distribution matrix, the system automatically splits the copyright income proportion, the original author obtains the income fixedly according to the basic proportion (such as 30%), and the remaining income is distributed according to the contribution degree proportion of each second user (such as a user with a contribution degree proportion of 20% obtains 20% x 70% = 14% of the income). The income distribution rules and the distribution matrix are written into the blockchain storage record synchronously, ensuring that the distribution process is transparent and traceable, and supporting subsequent income settlement and reconciliation.

[0103] Step 5, NFT copyright certificate generation and application:

[0104] The Merkle tree root hash value is extracted as the unique identifier of the NFT copyright certificate, and each layer hash value, parameter iteration traceability chain (including all version node hash values, contributor information, and adjustment summary) is used as the NFT metadata attribute; NFT supports secondary transactions, and the income distribution mechanism is automatically triggered during transaction, and the transaction income is split to each contributor account according to the current contribution degree matrix, and the on-chain transaction record is generated; if image infringement is detected, the double content fingerprint (S300 generation) and parameter dependence graph summary are extracted through on-chain query as the core evidence for infringement right protection.

[0105] Through blockchain storage, multi-dimensional contribution quantification and NFT copyright certificate generation, the problems of "difficulty in copyright tracing, difficulty in quantifying contribution degree, and unfair income distribution" in traditional multi-user collaborative creation are completely solved: on the one hand, hierarchical hash values and blockchain storage ensure the non-tamperability and full traceability of the creation process, providing solid evidence for copyright ownership; on the other hand, multi-dimensional indicators accurately quantify the contribution value of each user, realize fair and equitable income distribution, and stimulate the enthusiasm of original authors and secondary creators; at the same time, NFT copyright certificate endows digital copyright with the attributes of transferability and tradability, perfects the commercial ecological closed loop of AI picture creation, and provides core rights protection for the large-scale promotion of multi-user collaborative creation.

[0106] Based on the above scheme, the following is an example combined with a case, assuming that the first user A generates a "blue seaside sunset" theme picture through an AI platform, and shares it with the second user B through the scheme, and the core demand of user B is "to retain the seaside and sunset core scene, optimize the color level of the sky, and enhance the details of the sea waves". In traditional technology, user B can only get the picture product and cannot know the original generation parameters and associated logic, and needs to blindly try and error when adjusting, which is extremely inefficient, while the present scheme accurately solves this problem through full-process design.

[0107] Firstly, S100 lays the foundation of structured data for secondary creation: after user A uploads the picture, the system generates a data stream (including core data such as unique model identifier, customized random seed, and iteration step number) and converts it into a JSON file containing a parameter dependency graph. The file clearly marks the association logic of "main tone" and "light intensity" "color saturation", and the influence weight of "iteration step number" and "sea wave detail richness", completely solving the problem of scattered and irregular parameters in traditional methods. At the same time, the JSON file has built-in cross-platform tags, so user B can receive the structured parameters directly on the mobile terminal without manual adaptation, which eliminates data obstacles for efficient adjustment.

[0108] Secondly, S200 reduces invalid adjustments through intelligent filtering and adaptive components: the system calls user B's creation preference profile (frequently adjusted "color saturation" and "detail enhancement" dimensions), combines the input key features, and filters non-key parameters through an AI semantic matching model. The "sky color" and "sea wave detail" related parameters are locked as key parameters (not adjustable), and the "contrast" and "sharpening degree" non-key parameters are sorted according to user preference to generate adaptive adjustable components. Each component is associated with a constraint threshold (such as "sharpening degree" limited to -20% to +25%) and an AI recommendation range (such as "contrast" recommended +8% to +15%), and the default value is set to "contrast +10%" based on user B's historical adjustment habits, avoiding the inefficient problem of "ambiguous parameter range and blind trial and error" in traditional adjustment, allowing user B to focus only on core optimization dimensions.

[0109] Finally, the difference heat map visualization of S200 further improves the accuracy of adjustment: after user B adjusts "contrast" to +12% and "sharpening degree" to +20% through the component, the system generates a difference heat map in real time. The sky area presents deep red (high difference value) due to color level optimization, and the sea wave area presents light red (medium difference value) due to detail enhancement, and the "contrast +12%" and "sharpening degree +20%" semantic labels are superimposed. User B clicks the sky area label, and the system automatically locates to the "color saturation" component, without the need to search one by one to make accurate corrections. After continuous adjustment, the system also calculates the cumulative difference value and triggers a solidification reminder to ensure that the adjustment effect is traceable. The whole process allows user B to change from "blind trial and error" to "precise optimization", greatly improving the efficiency of secondary creation.

[0110] In the embodiments of the present application, in the process of generating the adaptive priority-ordered adjustable components, the method further comprises:

[0111] Step 1: Semantic vector extraction and similarity quantization

[0112] Firstly, the key features input by the second user (such as "optimizing sky color") are converted into computer-recognizable core semantic vectors through the AI semantic matching model; at the same time, a preset functional semantic vector is set for each parameter in the JSON file (such as "saturation" "color temperature" "sharpening degree"), such as "saturation" corresponding to "color density adjustment" semantics, and the matching degree of the two groups of vectors is calculated by the cosine similarity algorithm to quantify the correlation strength of the parameters and the user's core needs.

[0113] Step 2, non-key parameter screening and priority sorting:

[0114] Set a preset similarity threshold (such as 0.6), mark the parameters with a matching degree lower than the threshold and a weak correlation with the user's core needs as non-key parameters (such as "contrast" "shadow intensity" not mentioned by the user); then retrieve the second user's creation preference profile, extract the "frequently adjusted dimensions" (such as the user frequently adjusts "color parameters") and "adjustment frequency weights" (such as adjusting "saturation" 15 times a month, with a higher weight than adjusting "shadow intensity" 3 times a month), and prioritize the non-key parameters, with high-frequency and frequently used parameters at the top.

[0115] Step 3, component group generation and linkage rule labeling:

[0116] Based on the parameter dependency graph constructed by S100, identify the strongly correlated parameters for each non-key parameter in the sorted list (such as "main parameter - saturation" is associated with "color temperature" "brightness"), and generate a component group of "main parameter - associated parameter" (such as "saturation + color temperature + brightness" component group); clearly label the parameter linkage rules within the component group, such as "when the main parameter'saturation' is increased by more than 10%, the associated parameter 'color temperature' is automatically decreased by 3% to avoid color distortion", so that the user can clearly understand the chain effect of parameter adjustment.

[0117] Step 4, adjustment monitoring and optimal solution pushing:

[0118] During the second user's operation of the adjustable component, the system monitors in real time whether the adjustment value exceeds the preset constraint threshold of the component (such as "saturation" constraint threshold -30% to +30%); if it is detected that the adjustment value exceeds the threshold (such as saturation increased by 35%), the AI recommendation module is immediately started, based on the platform's similar picture optimization cases, user creation preferences and parameter dependency logic, to generate 3-5 optimal adjustment schemes (such as "saturation +25% and color temperature -5%" "saturation +28% and brightness +3%"), each scheme is accompanied by a corresponding picture effect preview, for the user to intuitively select or reference adjustment.

[0119] By defining parameter linkage rules, providing real-time early warning for threshold adjustment, and providing visual optimal solutions, the problems of "parameter correlation leading to adjustment conflicts" and "no guidance for out-of-range adjustments" in traditional secondary creation are solved, further improving the accuracy, safety, and convenience of parameter adjustment, making secondary creation more efficient and more user-friendly.

[0120] In the embodiments of the present application, in the process of generating the difference heat map before and after real-time adjustment, the method further comprises:

[0121] Step 1: Double-image pixel-level feature depth extraction:

[0122] After the second user adjusts the parameters, the system synchronously acquires the original picture before adjustment and the preview picture after adjustment, and performs pixel-level traversal on the two pictures through computer vision algorithms to extract three categories of core feature data: color distribution (RGB channel value, hue proportion), texture structure (edge intensity, detail texture density), and semantic region division (independent semantic regions such as "sky", "sea waves", and "person" are labeled through image segmentation technology), providing accurate data support for difference calculation.

[0123] Step 2: Multi-dimensional difference value calculation and heat map mapping:

[0124] An AI difference quantization model is called to calculate the difference values of the two pictures in the three feature dimensions of color, texture, and semantic region (such as color difference value = adjusted RGB mean - unadjusted RGB mean), and map the comprehensive difference values of each region to the color gradient of the heat map (preset gradient rule: difference value 0-30 is green, 31-60 is yellow, and 61 and above is red, the higher the difference value, the brighter the color), directly presenting the adjustment impact range and degree.

[0125] Step 3: Semantic label superposition and interactive positioning: In the generated heat map, superimpose semantic labels for each high-difference region, and the label content directly labels the corresponding parameter adjustment item (such as "saturation +12%" and "sharpness +18%"); At the same time, build an association mapping between the label and the adjustable component, when the user clicks on a semantic label, the system automatically jumps and positions to the corresponding adjustable component, without the need for the user to manually search, realizing "what you see is what you adjust".

[0126] Step 4: Continuous adjustment tracking and solidification reminder triggering:

[0127] The system monitors the operation behavior of the second user in real time. If it is detected that the user continuously adjusts the same component group (such as the “saturation + color temperature + brightness” component group), the heat map is dynamically updated, and the cumulative difference value after multiple adjustments is calculated synchronously (cumulative difference value = weighted sum of each adjustment difference value, and the weight decreases in the order of adjustment). When the cumulative difference value exceeds the threshold (such as 80), the system triggers the “effect solidification reminder” through a pop-up window and sound effects (such as “the current adjustment has significantly optimized the picture effect. Do you want to solidify this version?”), and automatically saves the parameter snapshot of the current adjustment version (including the original value, adjusted value, and adjustment timestamp of all parameters in the component group), supporting subsequent backtracking or recovery.

[0128] Through continuous adjustment tracking, cumulative difference value calculation, and intermediate version solidification, the problems of “uncontrollable continuous adjustment effect and easy loss of intermediate optimized version” in traditional secondary creation are solved, which not only enables users to master the adjustment superposition effect in real time, but also retains key optimization nodes through parameter snapshots, further improving the continuity and traceability of secondary creation.

[0129] In the process of embedding the enhanced copyright metadata into the JSON parameter stream in the embodiment of the application, the method further comprises:

[0130] Step 1, dual content fingerprint extraction:

[0131] First, the AI content fingerprint generation model is called to perform feature extraction on the first picture (visual dimension) and the generated data stream (parameter dimension) collected by S100, respectively. The visual dimension extracts core visual features such as color distribution, texture features, and semantic region contour of the picture, and the parameter dimension extracts core parameter features such as unique identification of the generation model, customized random seed, and key adjustment parameters. Then, the two types of features are fused to generate a unique dual content fingerprint (stored in Base64 encoding format to ensure tamper resistance and uniqueness).

[0132] Step 2, adjustment intention description field generation: In the basic field of the contribution history record, a new “adjustment intention description” subfield is added. The input text of the second user (such as “enhance the sky level of detail”) is parsed through a natural language processing (NLP) model to directly convert it into a structured description. If the user does not input text, the adjustment behavior (such as continuously adjusting “saturation”, “color temperature”, and “sharpness degree”) is parsed, and the adjustment intention (such as “optimizing color richness and picture sharpness”) is automatically inferred based on the parameter function semantics, ensuring that the intention of each contribution behavior is traceable.

[0133] Step 3, hierarchical embedding and encryption processing: In the JSON parameter stream, a dedicated "copyright metadata field" (field name such as "copyright_metadata") is preset, and the enhanced copyright metadata is written into the corresponding subfield according to the hierarchical structure of "original author information (top layer) - core copyright information (middle layer) - contribution history record (bottom layer)". The top layer stores the original author identification and creation timestamp; the middle layer stores the license type and dual content fingerprint; the bottom layer stores the complete contribution history record containing "adjustment intention description"; then the hash combination of "original author identification + creation timestamp" is used as the encryption key to symmetrically encrypt the entire "copyright metadata field", preventing unauthorized tampering or illegal reading.

[0134] Step 4, dynamic update of contribution score: Each time the second user completes parameter adjustment (i.e., once parameter iteration), the system automatically triggers the AI copyright evaluation model, inputs the "adjustment effect improvement value" (aesthetic score + semantic consistency score weighted value) and "innovation score" (repetition rate determination value compared to platform historical adjustment records) calculated in S200, and the model calculates the score of the current contribution according to the preset algorithm (such as effect improvement value x 0.6 + innovation score x 0.4), and synchronously updates the corresponding entry in the contribution history record, ensuring that the contribution value is quantified and synchronized with the creation process in real time.

[0135] Through the dual content fingerprint to strengthen the uniqueness of copyright, hierarchical encryption to ensure data security, adjustment intention description to perfect the contribution traceability, and dynamic score to quantify the contribution value, the integrity and reliability of the copyright metadata are further improved.

[0136] In the embodiments of the present application, in the process of calculating the contribution degree based on multiple dimensions, the method further comprises:

[0137] Step 1, dual content fingerprint extraction:

[0138] First, call the AI content fingerprint generation model to extract features from the first picture (visual dimension) and the generated data stream (parameter dimension) collected in S100, respectively. The visual dimension extracts core visual features such as color distribution, texture features, and semantic region outline of the picture, and the parameter dimension extracts core parameter features such as unique identification of the generation model, customized random seed, and key adjustment parameters. Then, the two types of features are fused to generate a unique dual content fingerprint (stored in Base64 encoding format to ensure tamper resistance and uniqueness).

[0139] Step 2, adjustment intention description field generation:

[0140] In the basic field of the contribution history record, a new subfield "adjustment intention description" is added. The input text of the second user (such as "enhance the sky level") is directly converted into a structured description through natural language processing (NLP) model. If the user does not input text, the adjustment behavior (such as continuous adjustment of "saturation", "color temperature" and "sharpening degree") is analyzed, and the adjustment intention (such as "optimize color richness and picture sharpness") is automatically inferred by combining parameter function semantics, ensuring that the intention of each contribution behavior is traceable.

[0141] Step 3, hierarchical embedding and encryption processing: In the JSON parameter stream, a dedicated "copyright metadata field" (field name such as "copyright_metadata") is preset, and the enhanced copyright metadata is written into the corresponding subfield according to the hierarchical structure of "original author information (top layer) - core copyright information (middle layer) - contribution history record (bottom layer)". The top layer stores the original author identification and creation timestamp. The middle layer stores the license type and double content fingerprint. The bottom layer stores the complete contribution history record containing "adjustment intention description". Then, the hash combination of "original author identification + creation timestamp" is used as the encryption key to symmetrically encrypt the entire "copyright metadata field", preventing unauthorized tampering or illegal reading.

[0142] Step 4, dynamic update of contribution score: Each time the second user completes parameter adjustment (i.e. one parameter iteration), the system automatically triggers the AI copyright evaluation model, inputs the "adjustment effect improvement value" (aesthetic score + semantic consistency score weighted value) and "innovation score" (repetition rate determination value compared with platform historical adjustment records) calculated in S200, and the model calculates the score of the current contribution according to the preset algorithm (such as effect improvement value x 0.6 + innovation score x 0.4). The score is updated to the corresponding entry in the contribution history record, ensuring that the contribution value is quantified and the creation process is synchronized in real time.

[0143] Through double content fingerprint, hierarchical encryption ensures data security, adjustment intention description perfects contribution traceability, and dynamic score quantifies contribution value.

[0144] In the embodiments of the present application, in the process of generating the NFT copyright certificate bound with the layered hash value, the method further includes:

[0145] Step 1, hash value association and NFT identification establishment:

[0146] First, from the hierarchical hash value generated in step 1 of S400, the root hash value of the Merkle tree (globally unique hash value) is extracted as the unique identification (TokenID) of the NFT copyright certificate, ensuring that each creative version of NFT is unique; At the same time, the branch hash value of each layer (such as the core parameter layer, copyright metadata layer hash value) is stored as the metadata attribute of NFT in the format of "layer-hash" key-value pair, providing data support for subsequent on-chain tracing.

[0147] Step 2, parameter iteration tracing chain embedding:

[0148] Construct a parameter iteration tracing chain from "original version to all iteration versions", each iteration node in the chain corresponds to one parameter adjustment; Each node contains three core information: corresponding hierarchical hash value (associated with on-chain storage record), contributor information (second user unique identification), parameter adjustment summary (such as "saturation +12%, color temperature -3%", extracted from parameter snapshot of S200); Embed the tracing chain in the metadata extension field of NFT in JSON format, realize the reverse tracing from the current version to the original version of the full link creative track.

[0149] Step 3, automatic distribution of secondary transaction income:

[0150] For NFT copyright certificate, preset transaction trigger mechanism, when NFT occurs secondary transaction in transaction platform, the system automatically calls the contribution degree distribution matrix stored on the blockchain (generated in step 3 of S400); According to the proportion of each contributor in the matrix (such as original author 30%, secondary creator A 20%, secondary creator B 15%), the transaction income (such as NFT selling amount) is split into the corresponding contributor's on-chain account; At the same time, generate on-chain transaction record, including transaction amount, income splitting details, transaction timestamp, to ensure the transparency and traceability of the distribution process.

[0151] Step 4, infringement evidence extraction:

[0152] Through the real-time monitoring of NFT corresponding picture propagation by blockchain node, if suspected infringement content (such as unauthorized copy, commercial use) is detected, the system starts the evidence extraction process: through the unique identification (root hash value) of NFT on the chain, the double content fingerprint (visual + parameter feature) generated by S300 and the parameter dependence graph summary of S100 are called; Pack these data into a standardized evidence package (including on-chain storage record hash, evidence generation timestamp), as the core basis of infringement lawsuit or platform complaint, to ensure efficient and legally effective rights protection.

[0153] By means of strong binding by hash value, embedding of full-link traceability chain, automatic splitting of transaction income, and solidification of right protection evidence, the NFT copyright certificate not only symbolizes digital copyright, but also becomes a core carrier of "traceable creation track, fair income distribution, and reliable right protection", solving the problems of "non-unique identification, incoherent traceability, difficult income distribution, and no ironclad evidence for right protection" in the circulation of digital copyright, and perfecting the whole life cycle guarantee of AI creation copyright.

[0154] In the process of embedding cross-platform compatible tags in the JSON file to support multi-terminal sharing in the embodiments of the present application, the method further comprises:

[0155] Step 1, main terminal adaptation rule library construction:

[0156] First, the technical characteristics of the current mainstream terminals (mobile terminal, PC terminal, tablet terminal, VR device) are investigated, and the picture rendering capabilities (such as the maximum resolution supported by the mobile terminal, the 3D rendering protocol of the VR device) and the interaction logic (such as the mobile terminal mainly supporting touch screen operation, the PC terminal supporting precise operation of keyboard and mouse) of each terminal are analyzed; based on the research results, a rendering rule library adapted to each terminal is generated, which contains the "rendering parameter threshold" and "component interaction specification" (such as mobile terminal components supporting finger touch size, PC terminal components supporting batch import and export) corresponding to each terminal.

[0157] Step 2, double-tag embedding and parameter association: embed two types of functional tags in the JSON file: one is the terminal type identification tag; the other is the rendering parameter adaptation tag (stored in the field "render_adapt_tag", including the picture resolution (such as 1080P for mobile terminal, 4K for PC terminal), color space (such as sRGB for mobile terminal, DCI-P3 for VR device), and component interaction mode (such as vertical layout for mobile terminal, horizontal layout for PC terminal) corresponding to each terminal), and the tag is associated with the parameter dependency graph of S100, to ensure that the adaptation parameters are consistent with the core generation parameter logic.

[0158] Step 3, dynamic adaptation after terminal reception:

[0159] After the terminal of the second user receives the JSON file, it automatically reads the "device_type_tag" to identify the terminal type (such as identifying as an iOS mobile terminal); the adaptation rule corresponding to the terminal is called from the rendering rule library to dynamically adjust the presentation form of the adjustable components, and the component size is adapted to the screen size (the mobile terminal component is enlarged by 20%), the layout is optimized according to the interaction logic (the mobile terminal components are arranged vertically), and the operation mode is adapted to the input device (the mobile terminal supports sliding adjustment, and the PC terminal supports numerical input), so that "plug and play" can be achieved without manual setting.

[0160] Step 4, file lightweight processing and parameter recovery:

[0161] After embedding the label, the core data is screened based on the parameter dependency graph, and redundant parameters unrelated to the generation of the current picture (such as weight fragments of other models and invalid parameter default values) are removed; the JSON file is compressed through the LZ77 compression algorithm to reduce the transmission bandwidth occupation and storage volume; at the same time, a "complete parameter recovery interface" (stored in the field "full_param_recover") is retained, so that the user can restore the complete data through the interface if necessary in the future, and the lightness and data integrity are taken into account.

[0162] By constructing a terminal-specific rendering rule library, double-label precise adaptation, dynamically adjusting the component form, and the design of lightness and complete parameters, the problems of "parameter presentation inconsistency, operation experience fragmentation, and inefficient file transmission" in traditional cross-platform sharing are completely solved, and the convenience of parameter adjustment, the consistency of effects, and the integrity of data under multiple terminals are realized.

[0163] The embodiment of the application discloses a picture sharing device based on an AI platform, referring to Figure 2 , comprising:

[0164] The data flow structured cross-platform module 001 obtains the generated data flow corresponding to the first picture uploaded by the first user, the generated data flow including the generated model unique identifier, the customized random seed, the iteration step number, the prompt word semantic analysis result and the model weight slice, converts the generated data flow into a structured JSON file containing a parameter dependency graph, and the JSON file is built-in cross-platform compatible label to support multi-terminal sharing to at least one second user;

[0165] The parameter self-adaptive adjustment visualization module 002 filters the non-key parameters in the JSON file based on the creative preference portrait of the second user and the input key features through an AI semantic matching model, generates self-adaptive priority ordered adjustable components, each component is associated with the constraint threshold and the AI recommended adjustment range of parameter adjustment, and the default value is dynamically optimized according to the user's historical adjustment behavior; after the second user adjusts the parameters through the adjustable component, a difference heat map before and after adjustment is generated in real time, and a new picture is generated based on the difference heat map and the adjusted parameters;

[0166] The copyright metadata embedding update module 003 embeds enhanced copyright metadata into the JSON parameter stream, and the enhanced copyright metadata includes the original author identifier, the creation timestamp, the license type, the AI generated content fingerprint information, and the contribution history record containing the contributor identifier, the contribution timestamp, the contribution type, the contribution score and the adjustment parameter snapshot;

[0167] The blockchain copyright tracing contribution allocation module 004 calculates the hierarchical hash value of the JSON parameter stream through the Merkle tree algorithm at each parameter iteration, generates a blockchain storage record including the version number, timestamp, contributor information, hierarchical hash value and parameter dependency graph abstract, calculates the contribution degree based on multi-dimensional indexes, the multi-dimensional indexes including the dimensional complexity of parameter adjustment, adjustment amplitude deviation, AI quantitative generation effect improvement value, innovation score of adjustment behavior, wherein the effect improvement value includes aesthetic score and semantic consistency score, dynamically allocates the copyright income proportion according to the contribution degree proportion, and generates the NFT copyright certificate bound with the hierarchical hash value, the NFT copyright certificate supporting reverse tracing of the corresponding parameter iteration version through on-chain query.

[0168] The embodiment of the present application further discloses an electronic device comprising a processor, wherein the processor runs a program of the picture sharing method based on the AI platform according to any one of the above.

[0169] The embodiment of the present application further discloses a storage medium storing a program of the picture sharing method based on the AI platform according to any one of the above.

[0170] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. An AI platform-based picture sharing method, characterized in that, The method comprises the following steps: obtaining a first picture uploaded by a first user, and generating a data stream corresponding to the first picture, wherein the data stream comprises a unique identification of a generation model, a customized random seed, a number of iteration steps, a semantic analysis result of a prompt word, and a model weight slice; converting the data stream into a structured JSON file containing a parameter dependency graph; the JSON file is embedded with cross-platform compatible tags to support multi-terminal sharing to at least one second user; based on the creative preference profile and the input key features of the second user, filtering non-key parameters in the JSON file through an AI semantic matching model to generate an adjustable component with adaptive priority ranking, wherein each component is associated with a constraint threshold of parameter adjustment and an AI recommended adjustment range, and the default value is dynamically optimized according to the user's historical adjustment behavior; after the second user adjusts the parameters through the adjustable component, a difference heat map before and after the adjustment is generated in real time, and a new picture is generated based on the difference heat map and the adjusted parameters; embedding enhanced copyright metadata into the JSON parameter stream, wherein the enhanced copyright metadata comprises an original author identification, a creation timestamp, a license type, AI-generated content fingerprint information, and a contribution history record containing contributor identification, contribution timestamp, contribution type, contribution score, and adjustment parameter snapshot; at each parameter iteration, calculating a hierarchical hash value of the JSON parameter stream through a Merkle tree algorithm to generate a blockchain storage record comprising a version number, a timestamp, contributor information, a hierarchical hash value, and a parameter dependency graph summary; calculating the contribution degree based on multi-dimensional indicators, including the dimensional complexity of parameter adjustment, the adjustment amplitude deviation, the AI-quantified generation effect improvement value, and the innovation score of adjustment behavior, wherein the effect improvement value includes an aesthetic score and a semantic consistency score; dynamically allocating the copyright income proportion according to the contribution degree proportion, and generating an NFT copyright certificate bound to the hierarchical hash value, wherein the NFT copyright certificate supports reverse tracing of the corresponding parameter iteration version through on-chain query.

2. The AI platform-based picture sharing method of claim 1, wherein, In the process of generating the adaptive priority ranking adjustable component, the method further comprises: extracting the core semantic vector of the second user's key features through an AI semantic matching model, and performing cosine similarity calculation on the core semantic vector of the second user's key features and the functional semantic vector of each parameter in the JSON file; marking parameters with a similarity lower than a preset similarity threshold as non-key parameters, and prioritizing the non-key parameters in combination with the commonly used adjustment dimensions and adjustment frequency weights in the second user's creative preference profile; for each non-key parameter after sorting, identifying associated parameters based on the parameter dependency graph to generate a component group of main parameters and associated parameters, wherein parameter linkage rules are labeled in the component group; real-time monitoring of the second user's adjustment operation, if the parameter adjustment exceeds the constraint threshold, pushing multiple optimal adjustment schemes through the AI recommendation module for user selection or reference, wherein the optimal adjustment scheme includes an effect preview graph corresponding to the scheme. 3.The AI platform-based picture sharing method of claim 1, wherein, In the process of generating the difference heat map before and after the adjustment in real time, the method further comprises: Pixel-level feature extraction is performed on the original picture before adjustment and the preview picture after adjustment to obtain feature data of color distribution, texture structure, and semantic region division; The difference values of the original picture before adjustment and the preview picture after adjustment in each feature dimension are calculated by an AI difference quantization model, and the difference values are mapped to the color gradient of the heat map. The higher the difference value, the brighter the color of the heat map. Semantic labels are superimposed in the heat map to mark the parameter adjustment items corresponding to the differences, and the user is automatically positioned to the corresponding adjustable component when clicking the semantic label. If the second user continuously adjusts the same component group, the heat map is updated in real time and the cumulative difference value is calculated. When the cumulative difference value exceeds the preset difference threshold, an effect solidification reminder is triggered and a parameter snapshot of the current adjustment version is saved.

4. The AI platform-based picture sharing method of claim 1, wherein, In the process of embedding the enhanced copyright metadata into the JSON parameter stream, the method further includes: The first picture and the generated data stream are processed by an AI content fingerprint generation model to extract dual content fingerprints based on visual features and parameter features; An adjustment intention description field is added to the contribution history record, which is automatically generated by a natural language processing model based on the input text or adjustment behavior of the second user; The enhanced copyright metadata is embedded into the specified field of the JSON parameter stream according to the hierarchical structure of the original author information-core copyright information-contribution history record, and the copyright metadata is encrypted; The contribution score in the contribution history record is updated synchronously each time the parameter is iterated. The contribution score is dynamically calculated by an AI copyright evaluation model based on the adjustment effect improvement value and the innovation score.

5. The AI platform-based picture sharing method of claim 1, wherein, In the process of calculating the contribution degree based on multi-dimensional indicators, the method further includes: Setting the initial weight of each dimension indicator, and dynamically adjusting the weight according to the picture generation scene, wherein the picture generation scene includes commercial design, artistic creation, and daily sharing; Calculating the specific value of each indicator, wherein the indicators include dimension complexity, adjustment amplitude deviation, effect improvement value, and innovation score; According to the adjusted indicator weight, the total contribution degree is calculated. If there are multiple second users, the contribution degree proportion of each user is calculated respectively to generate a contribution degree distribution matrix.

6. The AI platform-based picture sharing method of claim 1, wherein, In the process of generating the NFT copyright certificate bound with the hierarchical hash value, the method further includes: Based on the hierarchical hash value generated by the Merkle tree algorithm, the root hash value is extracted as the unique identifier of the NFT copyright certificate, and each layer hash value is used as the metadata attribute of the NFT; Embedding a parameter iteration trace chain in the NFT copyright certificate, which includes all iteration nodes from the original version to the current version, each node is associated with the corresponding hierarchical hash value, contributor information, and parameter adjustment summary; For secondary transactions of NFT copyright certificates, the copyright income distribution mechanism is automatically triggered during the transaction, and the transaction income is split into the accounts of each contributor according to the current contribution degree distribution matrix, and a chain transaction record is generated; If it is detected that the picture content corresponding to the NFT is infringed, the dual content fingerprints and parameter dependency graph summaries are extracted through on-chain query to serve as core evidence for infringement and protection.

7. The AI platform-based picture sharing method of claim 1, wherein, In the process of embedding cross-platform compatible tags in the JSON file to support multi-terminal sharing, the method further comprises: analyzing the picture rendering capabilities and interaction logic of the current mainstream terminals, including mobile terminals, PC terminals, tablet terminals, and VR devices, to generate a rendering rule library adapted to each terminal; embedding terminal type identification tags and rendering parameter adaptation tags in the JSON file, wherein the rendering parameter adaptation tags include the picture resolution, color space, and component interaction mode corresponding to each terminal; When the second user receives the JSON file, the terminal automatically reads the identification tags, calls the corresponding adaptation rules from the rendering rule library, and dynamically adjusts the size, layout, and operation mode of the adjustable components; Lightweight processing of the JSON file, based on the parameter dependency graph, to eliminate redundant parameter data, compress the file size through the LZ77 compression algorithm, and retain complete parameter recovery interfaces, while supporting users to restore all parameter data when needed.

8. An AI platform-based picture sharing device, characterized by, It includes: A data flow structured cross-platform module that obtains a generated data flow corresponding to a first picture uploaded by a first user, the generated data flow including a unique model identifier, a customized random seed, an iteration step number, a prompt word semantic analysis result, and a model weight slice, converts the generated data flow into a structured JSON file containing a parameter dependency graph, and embeds cross-platform compatible tags in the JSON file to support multi-terminal sharing to at least one second user; A parameter self-adaptive adjustment visualization module that filters non-key parameters in the JSON file based on the second user's creative preference profile and input key features, generates self-adaptive priority-ordered adjustable components, each component associated with a constraint threshold for parameter adjustment and an AI-recommended adjustment range, and the default value dynamically optimized according to the user's historical adjustment behavior; After the second user adjusts the parameters through the adjustable components, a difference heat map before and after the adjustment is generated in real time, and a new picture is generated based on the difference heat map and the adjusted parameters; A copyright metadata embedding and updating module that embeds enhanced copyright metadata into the JSON parameter stream, including original author identification, creation timestamp, license type, AI-generated content fingerprint information, and contribution history records containing contributor identification, contribution timestamp, contribution type, contribution score, and adjustment parameter snapshot; A blockchain copyright tracing and contribution allocation module that calculates the hierarchical hash value of the JSON parameter stream through the Merkle tree algorithm at each parameter iteration, generates a blockchain record of evidence including version number, timestamp, contributor information, hierarchical hash value, and parameter dependency graph summary; Based on multi-dimensional indicators, including the dimension complexity of parameter adjustment, adjustment amplitude deviation, AI-quantified generation effect improvement value, and innovation score of adjustment behavior, the contribution degree is calculated, and the copyright income proportion is dynamically allocated according to the contribution degree proportion, and an NFT copyright certificate bound to the hierarchical hash value is generated, which supports reverse tracing of the corresponding parameter iteration version through on-chain query.

9. An electronic device, comprising: A processor, wherein a program of the AI platform-based picture sharing method according to any one of claims 1-7 is run in the processor.

10. A storage medium, characterized by A storage, wherein a program of the AI platform-based picture sharing method according to any one of claims 1-7 is stored.

Images