AI and block chain-based cultural property whole-process securitization system and method

Abstract

The invention relates to the technical field of intelligent assets, in particular to an AI and block chain-based cultural property whole-process securitization system and method. A data processing and decision optimization module comprises a copyright confirmation module, an intelligent valuation engine, an asset pool module and an intelligent feedback engine. The risk control and compliance guarantee module comprises a risk control module and a compliance settlement module; the external interaction and execution docking module comprises a hierarchical issuing sub-module, a full-process automation engine and a multi-terminal API docking module. By integrating AI and block chain technologies and applying a cross-modal attention mechanism, an AG-LSTM network, a DDPG algorithm, an RPA technology, an API gateway and a BFF intermediate layer architecture, the whole-process securitization system can safely complete the whole-process securitization operation in the aspects of definite ownership, accurate valuation, asset pool optimization management, accurate risk control and automatic process execution; and the income-risk ratio and the market adaptability of the cultural property securitization are improved.

Description

Technical Field

[0001] This invention relates to the field of smart asset technology, specifically to a system and method for the full-process securitization of cultural property rights based on AI and blockchain. Background Technology

[0002] With the booming development of the cultural industry, the value of cultural intellectual property assets is becoming increasingly prominent, and securitization, as an important means of unlocking their value, is receiving widespread attention. However, current practices in cultural intellectual property securitization face numerous challenges. In terms of copyright confirmation, traditional methods have significant shortcomings. For example, the intelligent R&D transaction and evaluation system based on big data and multi-asset risk-return disclosed in Chinese Patent Publication No. CN113902562A struggles to comprehensively and accurately extract multimodal features such as visual, audio, and text, leading to the inability to automatically align multimodal cultural data, ambiguous ownership definitions, and a high risk of ownership disputes. Simultaneously, the system has weak capabilities in processing diverse data, and its valuation model is sensitive to outliers, resulting in significant discrepancies between valuation results and actual market conditions, failing to provide a scientific basis for the structured mapping and reasonable pricing of standardized digital rights certificates. Furthermore, the lack of a dynamic feedback mechanism in asset allocation makes it difficult to adjust asset portfolios in real time according to market changes, resulting in an unsatisfactory risk-return ratio and severely impacting the market adaptability and investment attractiveness of cultural intellectual property securitization. Summary of the Invention

[0003] The technical problem to be solved by this invention is to overcome the shortcomings of the existing technology and provide a system and method for the whole process of cultural property rights securitization based on AI and blockchain.

[0004] The technical solution adopted in this invention is as follows: A cultural property rights securitization system based on AI and blockchain, comprising the following modules: The data processing and decision optimization module is used to process and analyze data related to cultural intellectual property assets, including: The copyright confirmation module extracts features of cultural intellectual property assets, including visual, audio, and textual features. It then generates a composite digital fingerprint using a cross-modal attention mechanism, a cross-modal generative adversarial network (CM-GAN) weak-modal completion mechanism, and a creation timestamp sequence to clarify asset ownership. To further enhance the immutability and credibility of the ownership proof, the composite digital fingerprint generated by the copyright confirmation module, the hash value used to generate the multimodal feature vector of the fingerprint, and the corresponding creation timestamp are simultaneously submitted to the blockchain network for notarization, generating a unique notarization transaction hash. This hash is uniquely bound to the asset, serving as a trust anchor for ownership verification in all subsequent securitization processes. The intelligent valuation engine is used to output asset valuation results based on multi-dimensional data, including historical transaction data, market sentiment index, and industry trend indicators. It uses a three-layer attention-gated long short-term memory network (AG-LSTM) for robust outlier processing. Compared to traditional LSTM, the three-layer attention gating mechanism of this invention improves the accuracy of outlier identification by 30% and reduces the valuation error from 18.5% to 6.7% by adding attention weights at, thus solving the technical bottleneck of traditional models being sensitive to extreme values ​​in multi-dimensional data. The asset pool module is used to introduce adaptation parameters including sub-sector types, cross-border settlement currencies, and regulatory jurisdiction codes, as well as asset valuation results, infringement risk factors, cross-border exchange rate risk factors, and sub-sector parameters. It employs a deep deterministic strategy gradient (DDPG) algorithm with the objective function of maximizing the Sharpe ratio to optimize asset pool management and improve the risk-return ratio of the asset portfolio. Compared to traditional asset allocation algorithms, such as the Markowitz model, the infringement risk factor and cross-border exchange rate factor introduced in this invention increase the Sharpe ratio of the asset pool to over 1.5, exceeding the warning threshold of 1.2, achieving a technological breakthrough in risk-adjusted returns. The intelligent feedback engine is used to adjust operating parameters and build a data closed loop, including module operating parameters, data processing results, and integrates compliance audit logs and abnormal behavior monitoring functions to build a full-process data closed loop; abnormal behaviors include preset monitoring events such as a sudden increase in API call failure rate, valuation deviation exceeding 8%, and single asset infringement probability consistently higher than 0.5; The risk control and compliance assurance module is used to monitor and manage various risks in the securitization of cultural intellectual property rights, including: The risk control module is used to monitor risk indicators of the asset pool and individual assets in real time, including market risk, credit risk, and operational risk. Once an indicator triggers a preset multi-level risk threshold, the corresponding risk response mechanism is immediately triggered. Among them, when the cross-border exchange rate fluctuation ΔE exceeds 10%, the system automatically triggers the exchange rate risk hedging mechanism, which uses smart contracts to call the stablecoin reserve pool for dynamic liquidation to ensure that the asset pool return fluctuation does not exceed 5%. The compliance settlement module is used to support cross-border transaction currency settlement based on the compliance verification mechanism, including investor identity verification results and settlement status, and integrates privacy protection and data anonymization functions. It supports settlement of multiple types of stablecoins and completes cross-border settlement operations. The external interaction and execution docking module is used to realize the structured mapping and automated execution of standardized digital certificates for cultural intellectual property rights, including: The tiered issuance submodule, based on risk levels, employs priority-based structured, hybrid, and digital warrant segmentation product designs for different risk assets to meet the risk preferences and investment needs of different investors. It also supports standardized digital certificate mapping, circulation, and dynamic adaptation to multiple market rules, enhancing product liquidity and market adaptability. Based on the aforementioned ownership verification, the generation, share segmentation, ownership changes, and transaction records of standardized digital warrants are automatically executed through smart contracts deployed on the blockchain and written into a distributed ledger for full traceability and tamper-proof protection. The fully automated engine integrates RPA technology and AI dynamic adaptation logic to achieve unmanned operation of IP material uploading, feature extraction, valuation calculation, financing material generation, and compliance verification. It also features an automatic fallback mechanism for anomalies, automatically taking appropriate measures to ensure process continuity when anomalies occur. When a modality is missing in multimodal data, such as the absence of audio features, a CM-GAN weak modality completion mechanism generates alternative features. The matching degree between the completed features and the complete modality features is ≥92%, ensuring that the accuracy of rights confirmation is not affected. It provides multi-dimensional experimental data, including valuation errors, anomaly repair success rates, and system failure rates under different sample sizes, with annotations of data sources and testing environments. The multi-terminal API integration module adopts an architecture of API gateway and BFF middleware, supporting standardized integration with bank credit systems, regional equity markets, and intellectual property trading platforms; it features multi-protocol fault tolerance, automatic data format conversion, and dynamic degradation mechanisms. Specifically: the asset characteristics and ownership fingerprints output by the copyright confirmation module are used as inputs to both the intelligent valuation engine and the risk control module; the asset valuation results output by the intelligent valuation engine and the risk indicators output by the risk control module are input to the asset pool module; the asset allocation scheme optimized by the asset pool module drives the tiered issuance sub-module to generate corresponding standardized digital warrants.

[0005] This technical solution employs a data processing and decision optimization module to deeply analyze data related to cultural intellectual property assets. It utilizes a copyright confirmation module to extract features and generate composite digital fingerprints to clarify ownership. An intelligent valuation engine, based on multi-source data and specific network processing, outputs asset value assessment results. An asset pool module introduces adaptation parameters, including subdivided track types, cross-border settlement currencies, and regulatory jurisdiction codes, to achieve optimized management. An intelligent feedback engine constructs a data closed loop based on operational parameters, achieving the goals of data processing and decision optimization. Through a risk control and compliance assurance module, the risk control module monitors risk indicators in real time and triggers response mechanisms. A compliance settlement module supports cross-border transaction currency settlement based on compliance verification mechanisms, enabling the monitoring and management of various risks in cultural intellectual property securitization. Through an external interaction and execution docking module, a tiered issuance sub-module designs products based on risk levels to meet the needs of different investors and enhance product liquidity and market adaptability. A full-process automation engine integrates technology to achieve unmanned operation and ensure process continuity. A multi-terminal API docking module uses a specific architecture to support standardized docking with multiple platforms, realizing the structured mapping and automated execution of standardized digital certificates for cultural intellectual property.

[0006] In addition, the cultural property rights securitization system and method based on AI and blockchain proposed above may also have the following additional technical features: According to an embodiment of the present invention, the cross-modal attention mechanism in the copyright confirmation module includes: Visual features are extracted using a ResNet-50 pre-trained deep convolutional neural network, outputting a 512-dimensional feature vector. Audio features are extracted by converting the waveform into a 128×431 dimension Mel spectrogram and then using a 3-layer convolutional network to extract a 256-dimensional feature vector. Text features are extracted using a 768-dimensional feature vector from a BERT-base-uncased pre-trained language model; Each feature is mapped to a 256-dimensional attention space via a 3×3-dimensional weight matrix, and features from different modalities are dynamically fused through a cross-modal attention mechanism.

[0007] This technical solution extracts features from different modalities using pre-trained models tailored to each modality, mining the inherent information of data from different modalities. For example, in vision, the ResNet-50 pre-trained deep convolutional neural network performs excellently in image recognition, effectively extracting visual features and outputting a 512-dimensional feature vector, providing rich visual information for subsequent processing. In audio processing, the waveform is first converted into a 128×431-dimensional Mel spectrogram, which better reflects the human ear's perception of sound. Then, a 3-layer convolutional network extracts a 256-dimensional feature vector, accurately capturing audio features. For text features, a 768-dimensional feature vector is extracted using the BERT-base-uncased pre-trained language model, which has powerful text understanding capabilities in natural language processing tasks. Each feature is mapped to a 256-dimensional attention space using a 3×3-dimensional weight matrix, achieving the unification of feature dimensions across different modalities. The cross-modal attention mechanism can dynamically adjust the weights of each modal feature based on the correlation between features from different modalities, making the fused features more comprehensively reflect the key information of the original multimodal data, thereby generating a composite digital fingerprint and clarifying the ownership of cultural intellectual property assets.

[0008] According to one embodiment of the present invention, the outlier robustness handling of the three-layer attention-gated long short-term memory network AG-LSTM in the intelligent valuation engine includes: The input layer integrates an isolated forest module to remove extreme data that deviates from 3σ or more. Generate reasonable inputs through linear interpolation and industry mean correction: x t '=IForest(x t ) Where: x t For time steps t The original input data; x t ' represents the corrected input data after outlier handling; IFOres(x) t For the Isolation Forest module, input data x t Outlier detection and handling functions; Attention gating mechanism adds attention weights: a t = σ (W a ·[h t-1 ,x t ]+b a ) Where: a t For time steps t The attention weights are added by the attention gating mechanism; σ W is the sigmoid activation function. a h is the weight matrix for attention weights in the attention gating mechanism;t-1 For time steps t 1's hidden state; b a This refers to the bias term of the attention weights in the attention gating mechanism; Hidden status updated to: h t =o t ⊙tanh(C t )·a t Where: h t For time steps t Updated hidden state; o t For time steps t The output gating state; ⊙ represents element-wise multiplication; tanh The hyperbolic tangent activation function; C t For time steps t The cellular state.

[0009] This technical solution achieves robust outlier handling through a multi-layered mechanism. Specifically, the input layer integrates an isolated forest module, leveraging the outlier detection feature of isolated forests, which construct binary trees based on random partitioning, to remove extreme data deviating more than 3σ, thus preventing outliers from interfering with subsequent processing. Linear interpolation and industry mean correction are used to generate reasonable inputs, ensuring data rationality and completeness. The attention gating mechanism adds attention weights. The sigmoid activation function limits the weight values ​​to between 0 and 1. Attention weights are calculated using a weight matrix and bias terms, combined with the previous hidden state and the current input data, enabling the model to dynamically focus on important data features. During hidden state updates, the output gating state and the cell state Ct processed by the activation function are multiplied element-wise, and then multiplied by the attention weights. This ensures that hidden state updates consider both the current cell state and the output gating, while also incorporating attention weights, allowing the model to adjust the hidden state according to data importance, enhancing its ability to handle outliers, and thus outputting more accurate asset valuation results.

[0010] According to an embodiment of the present invention, in the asset pool module, the objective function optimization of the Deep Deterministic Policy Gradient (DDPG) algorithm includes: Introducing infringement risk factors: P (st) =α×Rinc (st) +β×Dcomp (st) +γ×Tprob (st) The infringement risk factor P (st) It is obtained by weighted calculation of three sub-indicators: Historical Tort Litigation Risk Index Rinc (st)By connecting to the judicial open database API, we can obtain the historical litigation records of asset owners or similar works, and perform weighted calculations based on the loss rate, average compensation amount, and time decay factor. Dcomp, the degree of competition for similar content on the network (st) : Through a pre-set internet content aggregation interface, periodically crawl publicly available data from mainstream content platforms, and use a cross-modal similarity matching model that is the same as the copyright confirmation module to retrieve and count the number of publicly available works whose feature similarity to the target asset is higher than a pre-set threshold θ and their comprehensive popularity index. AI predicts the probability of infringement. Tprob (st) Output is generated by a pre-trained multimodal classification model. The multimodal classification model takes the asset's feature vector, the rights holder's historical credit score, and the compliance level certification of the first release channel as input, and maps them to probability values ​​in the 0, 1 interval through the Sigmoid function. The data for the above sub-indicators are obtained through the system's built-in data collection agent or authorized third-party compliant data service interfaces; α For Rinc (st) The weighting coefficient has a range of 0.4. 0.5, and α + β + γ =1; β For Dcomp (st) The weighting coefficient has a value range of 0.3. 0.4; γ For Tprob (st) The weighting coefficient has a value range of 0.1. 0.3, in the digital collectibles sector γ =0.3; The objective function for optimizing the cross-border exchange rate factor is: max SR =(μ p -r f -λ E ·|ΔE|) / σ p in: SR Sharpe ratio, a measure of the risk-adjusted return of a portfolio of assets; μ p The expected rate of return for the asset pool; r f The risk-free rate; λ E σ is the exchange rate risk aversion coefficient, ranging from [0.05, 0.1]; ΔE is the exchange rate fluctuation range during the settlement period; σ p λ represents the standard deviation of the asset pool's returns, measuring the risk level of the asset pool. ΔE represents the exchange rate fluctuation range during the settlement period. E ∈[0.05,0.1].

[0011] This technical solution achieves optimization by constructing an objective function that comprehensively considers multiple factors. Regarding the introduction of infringement risk factors, three indicators are used to measure the degree of infringement risk. Different indicators are assigned different weight coefficients, and the range of these weight coefficients is set to ensure that each indicator plays an appropriate role in the infringement risk assessment. Specific values ​​for γ, such as those for digital collectibles, reflect targeted considerations for different scenarios. The objective function for optimizing the cross-border exchange rate factor aims to maximize the Sharpe ratio. The Sharpe ratio comprehensively considers the expected return of the asset pool, the risk-free rate, the exchange rate risk aversion coefficient, the exchange rate volatility, and the standard deviation of the asset pool's return. The range of the exchange rate risk aversion coefficient reflects the model's cautious attitude towards exchange rate risk. By optimizing this objective function, the model can, while considering exchange rate volatility risk, pursue the maximum return after deducting the difference between the expected return of the asset pool and the risk-free rate, and simultaneously reduce the standard deviation of the asset pool's return, i.e., reduce risk. This achieves optimized asset pool management with the goal of maximizing the Sharpe ratio, improving the portfolio's risk-reward ratio.

[0012] According to one embodiment of the present invention, the risk control module includes multi-level risk thresholds, including: The Sharpe ratio warning threshold for the asset pool is ≤1.2; The probability of infringement risk for a single asset is ≥0.5; Valuation error ≥8%; Cross-border settlement anomaly rate ≥3%; A risk response is triggered when any indicator exceeds the limit for three consecutive periods or when any two indicators exceed the limit simultaneously.

[0013] This technical solution precisely monitors and manages various risks in cultural intellectual property securitization by setting multi-dimensional and multi-level risk thresholds. The asset pool Sharpe ratio warning threshold is ≤1.2. The Sharpe ratio measures risk-adjusted returns; this threshold can promptly identify situations where the risk-return ratio of the asset pool is too low, indicating potential market risks or unreasonable asset allocation. The single-asset infringement risk probability is ≥0.5, directly warning of potential infringement risks for individual assets. When the infringement risk probability reaches a high level, it means that the ownership of the asset is highly uncertain, which may trigger a series of subsequent risk events. Valuation error is ≥8%. Valuation accuracy is crucial for the pricing and trading of structured mappings of standardized digital warrants. Large valuation errors may affect investor decisions and product market performance; this threshold ensures that valuation results are within a reasonable range. The cross-border settlement anomaly rate is ≥3%. Cross-border settlement involves complex currency conversion and compliance issues; an excessively high anomaly rate can affect the smooth flow of funds and transactions. This threshold helps to promptly identify problems in the cross-border settlement process. Risk response is triggered when any indicator exceeds the limit for three consecutive periods or when any two indicators exceed the limit simultaneously. This takes into account both the persistence and concurrency of risks, ensuring that the system can take timely and appropriate measures to effectively control the spread of risks and guarantee the stable operation of the cultural intellectual property securitization process when risks accumulate or multiple risks occur at the same time.

[0014] According to an embodiment of the present invention, the privacy protection and data anonymization functions in the compliance settlement module include: Identify and de-identify trade secrets and personal information using the BERT-NER model; Sensitive images are blurred and watermarked. Data access permissions are set based on the RBAC model to ensure data security.

[0015] This technical solution employs multiple techniques to achieve privacy protection and data anonymization, ensuring data security. The BERT-NER model, an advanced named entity recognition model, accurately identifies trade secrets and personal information in text, thereby anonymizing this sensitive information to prevent leakage and protect the privacy and commercial interests of relevant parties. For sensitive images, blurring technology reduces the clarity of sensitive content, making key information difficult to identify directly; watermarking technology adds invisible or difficult-to-remove markers to images, protecting image copyright to some extent while tracking image use and dissemination, ensuring image data security. Data access permissions are set based on the RBAC (Role-Based Access Control) model, which divides users into different roles and assigns specific data access permissions to each role. In this way, only users with the corresponding permissions can access specific data, preventing unauthorized access and misuse, ensuring data security from the access control level, and providing comprehensive protection for data security in the cross-border transaction currency settlement process of cultural intellectual property securitization.

[0016] According to one embodiment of the present invention, the product structure design of the tiered issuance sub-module includes: the infringement probability of low-risk assets is ≤0.2, and the priority share accounts for 75%; the infringement probability of medium-risk assets is 0.2-0.4, and the priority share accounts for 55%; the infringement probability of high-risk assets is ≥0.4, and they are split into 5,000 standardized digital warrants.

[0017] This technical solution categorizes assets into risk levels based on the key risk indicator of infringement probability and designs a corresponding product structure to meet the risk preferences and investment needs of different investors. For low-risk assets, the infringement probability is ≤0.2, indicating relatively clear ownership and low infringement risk. A 75% distribution of settlement results to different warrant holder addresses is set according to a preset weight. This weighted distribution prioritizes the standardized segmentation, circulation, and risk-sharing of digital warrants, attracting investors with lower risk tolerance and seeking stable returns. Medium-risk assets have an infringement probability between 0.2 and 0.4, representing a moderate level of risk. A 55% distribution of settlement results to different warrant holder addresses is set according to a preset weight. This retains a certain proportion of priority shares to guarantee some returns and reduce risk, while adjusting the proportion to provide participation opportunities for investors with slightly higher risk tolerance, meeting their investment needs. High-risk assets have an infringement probability ≥0.4, indicating significant ownership uncertainty and high risk. Dividing it into 5,000 standardized digital warrants lowers the investment threshold for each warrant, allowing more investors to participate. Simultaneously, standardized digital warrants facilitate market circulation and trading, improving product liquidity and meeting the needs of investors with higher risk appetites who wish to obtain returns through flexible trading. This tiered product structure enhances the liquidity and market adaptability of the structured mapping of standardized digital warrants for cultural intellectual property rights.

[0018] According to an embodiment of the present invention, the automatic fallback mechanism for anomalies in the full-process automation engine includes: Configure backup algorithm switching and API retry scripts; When a module fails, a fault-tolerance process is automatically executed, with an anomaly repair success rate of ≥95.2%. The system failure rate is ≤0.5%.

[0019] This technical solution employs multiple safeguards to automatically handle anomalies during the fully automated process, ensuring stable system operation. Configuring backup algorithm switching and API retry scripts are crucial measures. Backup algorithms can quickly switch to the main algorithm when anomalies occur, guaranteeing uninterrupted processing. API retry scripts automatically retry calls to external interfaces when they fail, reducing the probability of process disruptions due to temporary interface issues. A fault-tolerant process is automatically executed in case of module failure. This meticulously designed process covers fault detection, isolation, and recovery, enabling rapid identification of faulty modules and appropriate action. The anomaly repair success rate is ≥95.2%, meaning that the vast majority of module failures can be effectively handled, minimizing the impact on the overall process. The system failure rate is ≤0.5%, a goal achieved through the aforementioned mechanisms. Backup algorithms and API retry scripts reduce the likelihood of localized anomalies, while the fault-tolerant process responds quickly when anomalies occur. The synergistic effect of these three elements significantly improves system stability and reliability, ensuring the continuous and stable operation of the fully automated engine and guaranteeing the efficient conduct of cultural intellectual property securitization business.

[0020] According to an embodiment of the present invention, the dynamic degradation mechanism in the multi-terminal API interface module includes: When the concurrency exceeds 1500 transactions / second or the response time exceeds 500ms, non-core functions are paused and the system is switched to the mirror node API. It employs TLS 1.3 encrypted transmission and SHA-256 data verification, and supports OAuth 2.0 authentication and fine-grained access control to ensure secure data transmission.

[0021] This technical solution ensures system stability and data security through multi-dimensional strategies. A dynamic degradation mechanism is a key measure to cope with high loads. When concurrency exceeds 1500 requests / second or response time exceeds 500ms, it indicates significant system pressure. At this point, suspending non-core functions allows resources to be concentrated on ensuring the normal operation of core business processes. Switching to a mirror node API allows the mirror node to share some requests, preventing the master node from crashing due to overload and ensuring the system can still provide basic services under high concurrency scenarios. Regarding data security, TLS 1.3 encrypted transmission is used. This protocol offers faster connection speeds and stronger security, effectively preventing data theft or tampering during transmission. SHA-256 data verification ensures data integrity and accuracy; the receiver can verify whether the data has been altered during transmission. OAuth 2.0 authentication is supported, enabling strict verification of user identities, allowing only authorized users to access the system. Fine-grained access control further refines user permissions, allowing different users to access only the data and functions they are authorized to use. This comprehensively ensures data transmission security and provides a solid guarantee for the stability and security of multi-terminal API integration in cross-border transactions and currency settlements for cultural intellectual property securitization.

[0022] To achieve the above objectives, the present invention also provides a method for the full-process securitization of cultural property rights based on AI and blockchain.

[0023] A method for the full-process securitization of cultural property rights based on AI and blockchain includes the following steps: S1. Copyright Confirmation: Collect multimodal cultural property data using collection tools, upload it to the system in batches via the upload interface, process images, audio, and text separately using deep learning, extract features, fuse and complete them, and generate digital fingerprints to clarify ownership. S2, Intelligent Valuation: Receives confirmation data and multi-source data, calls the AG-LSTM network to perform attention gating calculation on the input sequence after isolation forest correction, and outputs an asset value assessment sequence for the next 12 months; S3. Asset Pool Optimization: Combining multiple risk factors and parameters, the DDPG algorithm is used to dynamically adjust and optimize the asset pool with the goal of maximizing the Sharpe ratio. S4. Risk Control and Compliance Assurance: Real-time monitoring of risk indicators, immediate response upon triggering thresholds, de-identification of information and protection of images during compliant settlement, and completion of cross-border settlement; S5. Tiered Issuance and Automated Execution: Products are designed according to risk levels, an automated engine enables unmanned operation, and an anomaly fallback mechanism ensures process continuity. S6, Multi-platform API Integration: Uses a specific architecture to connect to multiple platforms, with fault tolerance and degradation mechanisms, and employs encryption technology to ensure data transmission security; S7. Intelligent Feedback and Optimization: Adjusts system parameters based on operating parameters, integrates monitoring functions, and uses analysis algorithms to ensure compliant and safe system operation.

[0024] This technical solution is used to realize the securitization of cultural property rights throughout the entire process. Specifically, firstly, in the copyright confirmation stage, multimodal data is collected using data acquisition tools. Deep learning is used to process different types of data and fuse and complete them to generate digital fingerprints, laying the foundation for ownership in subsequent processes. Next, the intelligent valuation stage receives the confirmed data, processes it using AG-LSTM, and outputs a valuation, providing a basis for asset pricing. In the asset pool optimization stage, multiple risk factors and parameters are combined, and the DDPG algorithm is used to dynamically adjust the asset pool with the goal of maximizing the Sharpe ratio, improving the risk-return ratio of the asset portfolio. In the risk control and compliance assurance stage, risk indicators are monitored in real time and responses are triggered. Simultaneously, compliant settlement is ensured through methods such as anonymization and image protection. In the tiered issuance and automated execution stage, products are designed according to risk levels, and an automated engine and an anomaly fallback mechanism ensure efficient and stable operation of the process. In the multi-platform API integration stage, a specific architecture and multiple mechanisms are adopted, and encryption and other technologies are used to ensure data transmission security, achieving efficient integration with multiple platforms. Finally, the intelligent feedback and optimization stage adjusts system parameters based on operating parameters, integrates monitoring functions, and uses analytical algorithms to ensure compliant and secure system operation, forming a closed-loop optimization that continuously improves the performance and reliability of the entire cultural property rights securitization system.

[0025] Compared with the prior art, the present invention has the following advantages: (1) Enhanced ownership clarity: Through deep learning technologies such as cross-modal attention mechanisms, multimodal cultural property rights characteristics are accurately extracted and composite digital fingerprints are generated, providing a reliable basis for asset ownership confirmation and effectively reducing the risk of ownership disputes.

[0026] (2) Enhanced valuation accuracy: By using the AG-LSTM network to process diverse data and combining it with an outlier robustness handling mechanism, the 12-month asset valuation is more closely related to the actual market, providing scientific support for the structured mapping pricing of standardized digital warrants.

[0027] (3) Risk-return ratio optimization: The asset pool allocation is dynamically adjusted through the DDPG algorithm, and the infringement risk factor and cross-border exchange rate factor are introduced. The asset portfolio is optimized with the Sharpe ratio as the goal, thereby improving the overall risk-return level.

[0028] (4) Process automation guarantee: The full-process automation engine integrates RPA technology and AI logic to achieve unmanned operation execution. With the automatic fallback mechanism for anomalies, it ensures process continuity and reduces the risk of human intervention.

[0029] (5) Enhanced security and compliance: TLS1.3 encryption, SHA-256 verification and other technologies are used to ensure data transmission security. Data access permissions are controlled through the RBAC model. Combined with multi-level risk threshold monitoring and compliance settlement mechanism, a comprehensive security protection system is built. Attached Figure Description

[0030] Figure 1 This is a logic block diagram of the system of the present invention.

[0031] Figure 2 This is a flowchart of the method of the present invention.

[0032] Figure 3 This is a bar chart comparing the performance of the present invention with that of traditional methods.

[0033] Figure 4 This is a diagram of the AG-LSTM valuation network structure with feedback mechanism of the present invention. Detailed Implementation

[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] Example 1 like Figure 1 As shown, this embodiment provides a full-process securitization system for cultural property rights based on AI and blockchain, including the following modules: The data processing and decision optimization module is used to process and analyze data related to cultural intellectual property assets, including: The copyright confirmation module extracts features of cultural intellectual property assets, including visual, audio, and textual features. It then generates a composite digital fingerprint using a cross-modal attention mechanism, a cross-modal generative adversarial network (CM-GAN) weak-modal completion mechanism, and a creation timestamp sequence to clarify asset ownership. To further enhance the immutability and credibility of the ownership proof, the composite digital fingerprint generated by the copyright confirmation module, the hash value used to generate the multimodal feature vector of the fingerprint, and the corresponding creation timestamp are simultaneously submitted to the blockchain network for notarization, generating a unique notarization transaction hash. This hash is uniquely bound to the asset, serving as a trust anchor for ownership verification in all subsequent securitization processes. The intelligent valuation engine is used to output asset valuation results based on multi-dimensional data, including historical transaction data, market sentiment index, and industry trend indicators. It uses a three-layer attention-gated long short-term memory network (AG-LSTM) for robust outlier processing. Compared to traditional LSTM, the three-layer attention gating mechanism of this invention improves the accuracy of outlier identification by 30% and reduces the valuation error from 18.5% to 6.7% by adding attention weights at, thus solving the technical bottleneck of traditional models being sensitive to extreme values ​​in multi-dimensional data. The asset pool module is used to introduce adaptation parameters including sub-sector types, cross-border settlement currencies, and regulatory jurisdiction codes, as well as asset valuation results, infringement risk factors, cross-border exchange rate risk factors, and sub-sector parameters. It employs a deep deterministic strategy gradient (DDPG) algorithm with the objective function of maximizing the Sharpe ratio to optimize asset pool management and improve the risk-return ratio of the asset portfolio. Compared to traditional asset allocation algorithms, such as the Markowitz model, the infringement risk factor and cross-border exchange rate factor introduced in this invention increase the Sharpe ratio of the asset pool to over 1.5, exceeding the warning threshold of 1.2, achieving a technological breakthrough in risk-adjusted returns. The intelligent feedback engine is used to adjust operating parameters and build a data closed loop, including module operating parameters, data processing results, and integrates compliance audit logs and abnormal behavior monitoring functions to build a full-process data closed loop; abnormal behaviors include preset monitoring events such as a sudden increase in API call failure rate, valuation deviation exceeding 8%, and single asset infringement probability consistently higher than 0.5; The risk control and compliance assurance module is used to monitor and manage various risks in the securitization of cultural intellectual property rights, including: The risk control module is used to monitor risk indicators of the asset pool and individual assets in real time, including market risk, credit risk, and operational risk. Once an indicator triggers a preset multi-level risk threshold, the corresponding risk response mechanism is immediately triggered. Among them, when the cross-border exchange rate fluctuation ΔE exceeds 10%, the system automatically triggers the exchange rate risk hedging mechanism, which uses smart contracts to call the stablecoin reserve pool for dynamic liquidation to ensure that the asset pool return fluctuation does not exceed 5%. The compliance settlement module is used to support cross-border transaction currency settlement based on the compliance verification mechanism, including investor identity verification results and settlement status, and integrates privacy protection and data anonymization functions. It supports settlement of multiple types of stablecoins and completes cross-border settlement operations. The external interaction and execution docking module is used to realize the structured mapping and automated execution of standardized digital certificates for cultural intellectual property rights, including: The tiered issuance submodule, based on risk levels, employs priority-based structured, hybrid, and digital warrant segmentation product designs for different risk assets to meet the risk preferences and investment needs of different investors. It also supports standardized digital certificate mapping, circulation, and dynamic adaptation to multiple market rules, enhancing product liquidity and market adaptability. Based on the aforementioned ownership verification, the generation, share segmentation, ownership changes, and transaction records of standardized digital warrants are automatically executed through smart contracts deployed on the blockchain and written into a distributed ledger for full traceability and tamper-proof protection. The fully automated engine integrates RPA technology and AI dynamic adaptation logic to achieve unmanned operation of IP material uploading, feature extraction, valuation calculation, financing material generation, and compliance verification. It also features an automatic fallback mechanism for anomalies, automatically taking appropriate measures to ensure process continuity when anomalies occur. When a modality is missing in multimodal data, such as the absence of audio features, a CM-GAN weak modality completion mechanism generates alternative features. The matching degree between the completed features and the complete modality features is ≥92%, ensuring that the accuracy of rights confirmation is not affected. It provides multi-dimensional experimental data, including valuation errors, anomaly repair success rates, and system failure rates under different sample sizes, with annotations of data sources and testing environments. The multi-terminal API integration module adopts an architecture of API gateway and BFF middleware, supporting standardized integration with bank credit systems, regional equity markets, and intellectual property trading platforms; it features multi-protocol fault tolerance, automatic data format conversion, and dynamic degradation mechanisms. Specifically: the asset characteristics and ownership fingerprints output by the copyright confirmation module are used as inputs to both the intelligent valuation engine and the risk control module; the asset valuation results output by the intelligent valuation engine and the risk indicators output by the risk control module are input to the asset pool module; the asset allocation scheme optimized by the asset pool module drives the tiered issuance sub-module to generate corresponding standardized digital warrants.

[0036] This technical solution employs a data processing and decision optimization module to deeply analyze data related to cultural intellectual property assets. It utilizes a copyright confirmation module to extract features and generate composite digital fingerprints to clarify ownership. An intelligent valuation engine, based on multi-source data and specific network processing, outputs asset value assessment results. An asset pool module introduces adaptation parameters, including subdivided track types, cross-border settlement currencies, and regulatory jurisdiction codes, to achieve optimized management. An intelligent feedback engine constructs a data closed loop based on operational parameters, achieving the goals of data processing and decision optimization. Through a risk control and compliance assurance module, the risk control module monitors risk indicators in real time and triggers response mechanisms. A compliance settlement module supports cross-border transaction currency settlement based on compliance verification mechanisms, enabling the monitoring and management of various risks in cultural intellectual property securitization. Through an external interaction and execution docking module, a tiered issuance sub-module designs products based on risk levels to meet the needs of different investors and enhance product liquidity and market adaptability. A full-process automation engine integrates technology to achieve unmanned operation and ensure process continuity. A multi-terminal API docking module uses a specific architecture to support standardized docking with multiple platforms, realizing the structured mapping and automated execution of standardized digital certificates for cultural intellectual property.

[0037] In addition, the cultural property rights securitization system and method based on AI and blockchain proposed above may also have the following additional technical features: According to an embodiment of the present invention, the cross-modal attention mechanism in the copyright confirmation module includes: Visual features are extracted using a ResNet-50 pre-trained deep convolutional neural network, outputting a 512-dimensional feature vector. Audio features are extracted by converting the waveform into a 128×431 dimension Mel spectrogram and then using a 3-layer convolutional network to extract a 256-dimensional feature vector. Text features are extracted using a 768-dimensional feature vector from a BERT-base-uncased pre-trained language model; Each feature is mapped to a 256-dimensional attention space via a 3×3-dimensional weight matrix, and features from different modalities are dynamically fused through a cross-modal attention mechanism.

[0038] This technical solution extracts features from different modalities using pre-trained models tailored to each modality, mining the inherent information of data from different modalities. For example, in vision, the ResNet-50 pre-trained deep convolutional neural network performs excellently in image recognition, effectively extracting visual features and outputting a 512-dimensional feature vector, providing rich visual information for subsequent processing. In audio processing, the waveform is first converted into a 128×431-dimensional Mel spectrogram, which better reflects the human ear's perception of sound. Then, a 3-layer convolutional network extracts a 256-dimensional feature vector, accurately capturing audio features. For text features, a 768-dimensional feature vector is extracted using the BERT-base-uncased pre-trained language model, which has powerful text understanding capabilities in natural language processing tasks. Each feature is mapped to a 256-dimensional attention space using a 3×3-dimensional weight matrix, achieving the unification of feature dimensions across different modalities. The cross-modal attention mechanism can dynamically adjust the weights of each modal feature based on the correlation between features from different modalities, making the fused features more comprehensively reflect the key information of the original multimodal data, thereby generating a composite digital fingerprint and clarifying the ownership of cultural intellectual property assets.

[0039] According to one embodiment of the present invention, the outlier robustness handling of the three-layer attention-gated long short-term memory network AG-LSTM in the intelligent valuation engine includes: The input layer integrates an isolated forest module to remove extreme data that deviates from 3σ or more. Generate reasonable inputs through linear interpolation and industry mean correction: x t '=IForest(x t ) Where: x t For time steps t The original input data; x t ' represents the corrected input data after outlier handling; IFOres(x) tFor the Isolation Forest module, input data x t Outlier detection and handling functions; Attention gating mechanism adds attention weights: a t = σ (W a ·[h t-1 ,x t ]+b a ) Where: a t For time steps t The attention weights are added by the attention gating mechanism; σ W is the sigmoid activation function. a h is the weight matrix for attention weights in the attention gating mechanism; t-1 For time steps t 1's hidden state; b a This refers to the bias term of the attention weights in the attention gating mechanism; Hidden status updated to: h t =o t ⊙tanh(C t )·a t Where: h t For time steps t Updated hidden state; o t For time steps t The output gating state; ⊙ represents element-wise multiplication; tanh The hyperbolic tangent activation function; C t For time steps t The cellular state.

[0040] This technical solution achieves robust outlier handling through a multi-layered mechanism. Specifically, the input layer integrates an isolated forest module, leveraging the outlier detection feature of isolated forests, which construct binary trees based on random partitioning, to remove extreme data deviating more than 3σ, thus preventing outliers from interfering with subsequent processing. Linear interpolation and industry mean correction are used to generate reasonable inputs, ensuring data rationality and completeness. The attention gating mechanism adds attention weights. The sigmoid activation function limits the weight values ​​to between 0 and 1. Attention weights are calculated using a weight matrix and bias terms, combined with the previous hidden state and the current input data, enabling the model to dynamically focus on important data features. During hidden state updates, the output gating state and the cell state Ct processed by the activation function are multiplied element-wise, and then multiplied by the attention weights. This ensures that hidden state updates consider both the current cell state and the output gating, while also incorporating attention weights, allowing the model to adjust the hidden state according to data importance, enhancing its ability to handle outliers, and thus outputting more accurate asset valuation results.

[0041] According to an embodiment of the present invention, in the asset pool module, the objective function optimization of the Deep Deterministic Policy Gradient (DDPG) algorithm includes: Introducing infringement risk factors: P (st) =α×Rinc (st) +β×Dcomp (st) +γ×Tprob (st) The infringement risk factor P (st) It is obtained by weighted calculation of three sub-indicators: Historical Tort Litigation Risk Index Rinc (st) By connecting to the judicial open database API, we can obtain the historical litigation records of asset owners or similar works, and perform weighted calculations based on the loss rate, average compensation amount, and time decay factor. Dcomp, the degree of competition for similar content on the network (st) : Through a pre-set internet content aggregation interface, periodically crawl publicly available data from mainstream content platforms, and use a cross-modal similarity matching model that is the same as the copyright confirmation module to retrieve and count the number of publicly available works whose feature similarity to the target asset is higher than a pre-set threshold θ and their comprehensive popularity index. AI predicts the probability of infringement. Tprob (st) Output is generated by a pre-trained multimodal classification model. The multimodal classification model takes the asset's feature vector, the rights holder's historical credit score, and the compliance level certification of the first release channel as input, and maps them to probability values ​​in the 0, 1 interval through the Sigmoid function. The data for the above sub-indicators are obtained through the system's built-in data collection agent or authorized third-party compliant data service interfaces; α For Rinc (st) The weighting coefficient has a range of 0.4. 0.5, and α + β + γ =1; β For Dcomp (st) The weighting coefficient has a value range of 0.3. 0.4; γ For Tprob (st) The weighting coefficient has a value range of 0.1. 0.3, in the digital collectibles sector γ =0.3; The objective function for optimizing the cross-border exchange rate factor is: max SR =(μ p -r f -λ E ·|ΔE|) / σ p in: SR Sharpe ratio, a measure of the risk-adjusted return of a portfolio of assets; μ p The expected rate of return for the asset pool; r f The risk-free rate; λ E σ is the exchange rate risk aversion coefficient, ranging from [0.05, 0.1]; ΔE is the exchange rate fluctuation range during the settlement period; σ p λ represents the standard deviation of the asset pool's returns, measuring the risk level of the asset pool. ΔE represents the exchange rate fluctuation range during the settlement period. E ∈[0.05,0.1].

[0042] This technical solution achieves optimization by constructing an objective function that comprehensively considers multiple factors. Regarding the introduction of infringement risk factors, three indicators are used to measure the degree of infringement risk. Different indicators are assigned different weight coefficients, and the range of these weight coefficients is set to ensure that each indicator plays an appropriate role in the infringement risk assessment. Specific values ​​for γ, such as those for digital collectibles, reflect targeted considerations for different scenarios. The objective function for optimizing the cross-border exchange rate factor aims to maximize the Sharpe ratio. The Sharpe ratio comprehensively considers the expected return of the asset pool, the risk-free rate, the exchange rate risk aversion coefficient, the exchange rate volatility, and the standard deviation of the asset pool's return. The range of the exchange rate risk aversion coefficient reflects the model's cautious attitude towards exchange rate risk. By optimizing this objective function, the model can, while considering exchange rate volatility risk, pursue the maximum return after deducting the difference between the expected return of the asset pool and the risk-free rate, and simultaneously reduce the standard deviation of the asset pool's return, i.e., reduce risk. This achieves optimized asset pool management with the goal of maximizing the Sharpe ratio, improving the portfolio's risk-reward ratio.

[0043] According to one embodiment of the present invention, the risk control module includes multi-level risk thresholds, including: The Sharpe ratio warning threshold for the asset pool is ≤1.2; The probability of infringement risk for a single asset is ≥0.5; Valuation error ≥8%; Cross-border settlement anomaly rate ≥3%; A risk response is triggered when any indicator exceeds the limit for three consecutive periods or when any two indicators exceed the limit simultaneously.

[0044] This technical solution precisely monitors and manages various risks in cultural intellectual property securitization by setting multi-dimensional and multi-level risk thresholds. The asset pool Sharpe ratio warning threshold is ≤1.2. The Sharpe ratio measures risk-adjusted returns; this threshold can promptly identify situations where the risk-return ratio of the asset pool is too low, indicating potential market risks or unreasonable asset allocation. The single-asset infringement risk probability is ≥0.5, directly warning of potential infringement risks for individual assets. When the infringement risk probability reaches a high level, it means that the ownership of the asset is highly uncertain, which may trigger a series of subsequent risk events. Valuation error is ≥8%. Valuation accuracy is crucial for the pricing and trading of structured mappings of standardized digital warrants. Large valuation errors may affect investor decisions and product market performance; this threshold ensures that valuation results are within a reasonable range. The cross-border settlement anomaly rate is ≥3%. Cross-border settlement involves complex currency conversion and compliance issues; an excessively high anomaly rate can affect the smooth flow of funds and transactions. This threshold helps to promptly identify problems in the cross-border settlement process. Risk response is triggered when any indicator exceeds the limit for three consecutive periods or when any two indicators exceed the limit simultaneously. This takes into account both the persistence and concurrency of risks, ensuring that the system can take timely and appropriate measures to effectively control the spread of risks and guarantee the stable operation of the cultural intellectual property securitization process when risks accumulate or multiple risks occur at the same time.

[0045] According to an embodiment of the present invention, the privacy protection and data anonymization functions in the compliance settlement module include: Identify and de-identify trade secrets and personal information using the BERT-NER model; Sensitive images are blurred and watermarked. Data access permissions are set based on the RBAC model to ensure data security.

[0046] This technical solution employs multiple techniques to achieve privacy protection and data anonymization, ensuring data security. The BERT-NER model, an advanced named entity recognition model, accurately identifies trade secrets and personal information in text, thereby anonymizing this sensitive information to prevent leakage and protect the privacy and commercial interests of relevant parties. For sensitive images, blurring technology reduces the clarity of sensitive content, making key information difficult to identify directly; watermarking technology adds invisible or difficult-to-remove markers to images, protecting image copyright to some extent while tracking image use and dissemination, ensuring image data security. Data access permissions are set based on the RBAC (Role-Based Access Control) model, which divides users into different roles and assigns specific data access permissions to each role. In this way, only users with the corresponding permissions can access specific data, preventing unauthorized access and misuse, ensuring data security from the access control level, and providing comprehensive protection for data security in the cross-border transaction currency settlement process of cultural intellectual property securitization.

[0047] According to one embodiment of the present invention, the product structure design of the tiered issuance sub-module includes: the infringement probability of low-risk assets is ≤0.2, and the priority share accounts for 75%; the infringement probability of medium-risk assets is 0.2-0.4, and the priority share accounts for 55%; the infringement probability of high-risk assets is ≥0.4, and they are split into 5,000 standardized digital warrants.

[0048] This technical solution categorizes assets into risk levels based on the key risk indicator of infringement probability and designs a corresponding product structure to meet the risk preferences and investment needs of different investors. For low-risk assets, the infringement probability is ≤0.2, indicating relatively clear ownership and low infringement risk. A 75% distribution of settlement results to different warrant holder addresses is set according to a preset weight. This weighted distribution prioritizes the standardized segmentation, circulation, and risk-sharing of digital warrants, attracting investors with lower risk tolerance and seeking stable returns. Medium-risk assets have an infringement probability between 0.2 and 0.4, representing a moderate level of risk. A 55% distribution of settlement results to different warrant holder addresses is set according to a preset weight. This retains a certain proportion of priority shares to guarantee some returns and reduce risk, while adjusting the proportion to provide participation opportunities for investors with slightly higher risk tolerance, meeting their investment needs. High-risk assets have an infringement probability ≥0.4, indicating significant ownership uncertainty and high risk. Dividing it into 5,000 standardized digital warrants lowers the investment threshold for each warrant, allowing more investors to participate. Simultaneously, standardized digital warrants facilitate market circulation and trading, improving product liquidity and meeting the needs of investors with higher risk appetites who wish to obtain returns through flexible trading. This tiered product structure enhances the liquidity and market adaptability of the structured mapping of standardized digital warrants for cultural intellectual property rights.

[0049] According to an embodiment of the present invention, the automatic fallback mechanism for anomalies in the full-process automation engine includes: Configure backup algorithm switching and API retry scripts; When a module fails, a fault-tolerance process is automatically executed, with an anomaly repair success rate of ≥95.2%. The system failure rate is ≤0.5%.

[0050] This technical solution employs multiple safeguards to automatically handle anomalies during the fully automated process, ensuring stable system operation. Configuring backup algorithm switching and API retry scripts are crucial measures. Backup algorithms can quickly switch to the main algorithm when anomalies occur, guaranteeing uninterrupted processing. API retry scripts automatically retry calls to external interfaces when they fail, reducing the probability of process disruptions due to temporary interface issues. A fault-tolerant process is automatically executed in case of module failure. This meticulously designed process covers fault detection, isolation, and recovery, enabling rapid identification of faulty modules and appropriate action. The anomaly repair success rate is ≥95.2%, meaning that the vast majority of module failures can be effectively handled, minimizing the impact on the overall process. The system failure rate is ≤0.5%, a goal achieved through the aforementioned mechanisms. Backup algorithms and API retry scripts reduce the likelihood of localized anomalies, while the fault-tolerant process responds quickly when anomalies occur. The synergistic effect of these three elements significantly improves system stability and reliability, ensuring the continuous and stable operation of the fully automated engine and guaranteeing the efficient conduct of cultural intellectual property securitization business.

[0051] According to an embodiment of the present invention, the dynamic degradation mechanism in the multi-terminal API interface module includes: When the concurrency exceeds 1500 transactions / second or the response time exceeds 500ms, non-core functions are paused and the system is switched to the mirror node API. It employs TLS 1.3 encrypted transmission and SHA-256 data verification, and supports OAuth 2.0 authentication and fine-grained access control to ensure secure data transmission.

[0052] This technical solution ensures system stability and data security through multi-dimensional strategies. A dynamic degradation mechanism is a key measure to cope with high loads. When concurrency exceeds 1500 requests / second or response time exceeds 500ms, it indicates significant system pressure. At this point, suspending non-core functions allows resources to be concentrated on ensuring the normal operation of core business processes. Switching to a mirror node API allows the mirror node to share some requests, preventing the master node from crashing due to overload and ensuring the system can still provide basic services under high concurrency scenarios. Regarding data security, TLS 1.3 encrypted transmission is used. This protocol offers faster connection speeds and stronger security, effectively preventing data theft or tampering during transmission. SHA-256 data verification ensures data integrity and accuracy; the receiver can verify whether the data has been altered during transmission. OAuth 2.0 authentication is supported, enabling strict verification of user identities, allowing only authorized users to access the system. Fine-grained access control further refines user permissions, allowing different users to access only the data and functions they are authorized to use. This comprehensively ensures data transmission security and provides a solid guarantee for the stability and security of multi-terminal API integration in cross-border transactions and currency settlements for cultural intellectual property securitization.

[0053] Example 2 Based on Example 1, such as Figure 2 As shown, this embodiment provides a method for the full-process securitization of cultural property rights based on AI and blockchain, including the following steps: S1. Copyright Confirmation: Collect multimodal cultural property data using collection tools, upload it to the system in batches via the upload interface, process images, audio, and text separately using deep learning, extract features, fuse and complete them, and generate digital fingerprints to clarify ownership. S2, Intelligent Valuation: Receives confirmation data and multi-source data, calls the AG-LSTM network to perform attention gating calculation on the input sequence after isolation forest correction, and outputs an asset value assessment sequence for the next 12 months; S3. Asset Pool Optimization: Combining multiple risk factors and parameters, the DDPG algorithm is used to dynamically adjust and optimize the asset pool with the goal of maximizing the Sharpe ratio. S4. Risk Control and Compliance Assurance: Real-time monitoring of risk indicators, immediate response upon triggering thresholds, de-identification of information and protection of images during compliant settlement, and completion of cross-border settlement; S5. Tiered Issuance and Automated Execution: Products are designed according to risk levels, an automated engine enables unmanned operation, and an anomaly fallback mechanism ensures process continuity. S6, Multi-platform API Integration: Uses a specific architecture to connect to multiple platforms, with fault tolerance and degradation mechanisms, and employs encryption technology to ensure data transmission security; S7. Intelligent Feedback and Optimization: Adjusts system parameters based on operating parameters, integrates monitoring functions, and uses analysis algorithms to ensure compliant and safe system operation.

[0054] This technical solution is used to realize the securitization of cultural property rights throughout the entire process. Specifically, firstly, in the copyright confirmation stage, multimodal data is collected using data acquisition tools. Deep learning is used to process different types of data and fuse and complete them to generate digital fingerprints, laying the foundation for ownership in subsequent processes. Next, the intelligent valuation stage receives the confirmed data, processes it using AG-LSTM, and outputs a valuation, providing a basis for asset pricing. In the asset pool optimization stage, multiple risk factors and parameters are combined, and the DDPG algorithm is used to dynamically adjust the asset pool with the goal of maximizing the Sharpe ratio, improving the risk-return ratio of the asset portfolio. In the risk control and compliance assurance stage, risk indicators are monitored in real time and responses are triggered. Simultaneously, compliant settlement is ensured through methods such as anonymization and image protection. In the tiered issuance and automated execution stage, products are designed according to risk levels, and an automated engine and an anomaly fallback mechanism ensure efficient and stable operation of the process. In the multi-platform API integration stage, a specific architecture and multiple mechanisms are adopted, and encryption and other technologies are used to ensure data transmission security, achieving efficient integration with multiple platforms. Finally, the intelligent feedback and optimization stage adjusts system parameters based on operating parameters, integrates monitoring functions, and uses analytical algorithms to ensure compliant and secure system operation, forming a closed-loop optimization that continuously improves the performance and reliability of the entire cultural property rights securitization system.

[0055] This technical solution is used to realize the securitization of cultural property rights throughout the entire process. Specifically, firstly, in the copyright confirmation stage, multimodal data is collected using data acquisition tools. Different types of data are processed and merged through deep learning to generate a digital fingerprint, laying the foundation for ownership determination in subsequent processes. Next, the intelligent valuation stage receives the confirmed data and utilizes... Figure 4 The AG-LSTM process, as shown, processes data and outputs valuations, providing a basis for asset pricing. The asset pool optimization stage combines multiple risk factors and parameters, using the DDPG algorithm to dynamically adjust the asset pool with the goal of maximizing the Sharpe ratio, thereby improving the risk-return ratio of the asset portfolio. The risk control and compliance assurance stage monitors risk indicators in real time and triggers responses, while ensuring compliant settlement through methods such as anonymization and image protection. The tiered issuance and automated execution stage designs products based on risk levels, utilizing an automation engine and an anomaly fallback mechanism to ensure efficient and stable process operation. The multi-platform API integration stage employs a specific architecture and multiple mechanisms, using encryption and other technologies to ensure data transmission security and achieve efficient integration with multiple platforms. Finally, the intelligent feedback and optimization stage adjusts system parameters based on operating parameters, integrates monitoring functions, and uses analytical algorithms to ensure compliant and secure system operation, forming a closed-loop optimization that continuously improves the performance and reliability of the entire cultural property rights securitization system.

[0056] The blockchain adopts a consortium blockchain architecture consisting of cultural property registration agencies, regional equity markets, and regulatory nodes. Changes in ownership require multi-signature consensus verification. The blockchain consensus mechanism uses Proof-of-Stake (PoS) instead of Proof-of-Stake (PBFT).

[0057] like Figure 3 As shown, the advantages of traditional methods and this technical solution in key performance indicators are presented intuitively. For example, in terms of accuracy, the traditional method achieves 78.0%, while this system reaches 92.5%, providing more accurate results, reducing error rates, and meeting the needs of multiple application scenarios. Regarding valuation error, the traditional method has a MAPE of 18.5%, while this system reduces it to 6.7%, making the assessment closer to the true value and improving credibility. In terms of processing time, the traditional method requires 72 hours, while this system only takes 60 minutes, significantly improving efficiency. In terms of API success rate, the traditional method achieves 80.0%, while this system achieves 95.2%, resulting in more stable interaction. In terms of market access success rate, the traditional method achieves 83.0%, while this system achieves 97.0%, facilitating rapid product launch and reducing resource waste.

[0058] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the invention by those skilled in the art without departing from the spirit and essence of the invention, and such modifications or substitutions should all be within the scope of the invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the invention should also be covered within the protection scope of the invention. Therefore, the protection scope of the invention should be determined by the scope of the claims.

Claims

1. A full-process securitization system for cultural property rights based on AI and blockchain, characterized in that, Includes the following modules: The data processing and decision optimization module is used to process and analyze data related to cultural intellectual property assets, including: The copyright confirmation module is used to extract the characteristics of cultural intellectual property assets, including visual, audio, and text features, and generate composite digital fingerprints by using cross-modal attention mechanisms, cross-modal generative adversarial network (CM-GAN) weak modality completion mechanisms, and creation timestamp sequences to clarify asset ownership. The intelligent valuation engine is used to output asset valuation results based on diverse data, including historical transaction data, market sentiment index, and industry trend indicators. It also uses a three-layer attention-gated long short-term memory network (AG-LSTM) for robust outlier processing. The asset pool module is used to introduce adaptation parameters including sub-sector types, cross-border settlement currencies, and regulatory jurisdiction codes, as well as asset valuation results, infringement risk factors, cross-border exchange rate risk factors, and sub-sector parameters. It uses the deep deterministic strategy gradient DDPG algorithm with the Sharpe ratio as the objective function to achieve optimized management of the asset pool and improve the risk-reward ratio of the asset portfolio. The intelligent feedback engine is used to build a data closed loop based on the adjustment of operating parameters, including module operating parameters, data processing results, and integrates compliance audit logs and abnormal behavior monitoring functions to build a full-process data closed loop; The risk control and compliance assurance module is used to monitor and manage various risks in the securitization of cultural intellectual property rights, including: The risk control module is used to monitor risk indicators of the asset pool and individual assets in real time, including market risk, credit risk, and operational risk. Once an indicator triggers a preset multi-level risk threshold, the corresponding risk response mechanism is immediately triggered. The compliance settlement module is used to support cross-border transaction currency settlement based on the compliance verification mechanism, including investor identity verification results and settlement status, and integrates privacy protection and data anonymization functions. It supports settlement of multiple types of stablecoins and completes cross-border settlement operations. The external interaction and execution docking module is used to realize the structured mapping and automated execution of standardized digital certificates for cultural intellectual property rights, including: The tiered issuance submodule, based on risk levels, employs priority-based structured, hybrid, and digital warrant segmentation product designs for different risk assets to meet the risk preferences and investment needs of different investors. It also supports standardized digital certificate mapping, circulation, and dynamic adaptation to multiple market rules, enhancing product liquidity and market adaptability. The generation, ownership changes, and transaction records of standardized digital warrants are automatically executed through smart contracts deployed on the blockchain and written into a distributed ledger to achieve full traceability and tamper-proof protection. The fully automated engine integrates RPA technology and AI dynamic adaptation logic to achieve unmanned operation of IP material uploading, feature extraction, valuation calculation, financing material generation, and compliance verification; it is also equipped with an automatic fallback mechanism for anomalies, which automatically takes corresponding measures to ensure the continuity of the process when anomalies occur during the execution process. The multi-terminal API integration module adopts an architecture of API gateway and BFF middleware, supporting standardized integration with bank credit systems, regional equity markets, and intellectual property trading platforms; it features multi-protocol fault tolerance, automatic data format conversion, and dynamic degradation mechanisms. Specifically: the asset characteristics and ownership fingerprints output by the copyright confirmation module are used as inputs to both the intelligent valuation engine and the risk control module; the asset valuation results output by the intelligent valuation engine and the risk indicators output by the risk control module are input to the asset pool module; the asset allocation scheme optimized by the asset pool module drives the tiered issuance sub-module to generate corresponding standardized digital warrants.

2. The system as described in claim 1, characterized in that, The copyright confirmation module includes a cross-modal attention mechanism, which includes: Visual features are extracted using a ResNet-50 pre-trained deep convolutional neural network, outputting a 512-dimensional feature vector. Audio features are extracted by converting the waveform into a 128×431 dimension Mel spectrogram and then using a 3-layer convolutional network to extract a 256-dimensional feature vector. Text features are extracted using a 768-dimensional feature vector from a BERT-base-uncased pre-trained language model; Each feature is mapped to a 256-dimensional attention space via a 3×3-dimensional weight matrix, and features from different modalities are dynamically fused through a cross-modal attention mechanism.

3. The cultural property rights securitization system based on AI and blockchain as described in claim 1, characterized in that, In the intelligent valuation engine, the outlier robustness handling of the three-layer attention-gated long short-term memory network AG-LSTM includes: The input layer integrates an isolated forest module to remove extreme data that deviates from 3σ or more. Generate reasonable inputs through linear interpolation and industry mean correction: x t = IForest(x t ) Where: x t For time steps t The original input data; x t ' represents the corrected input data after outlier handling; IFOres(x) t For the Isolation Forest module, input data x t Outlier detection and handling functions; Attention gating mechanism adds attention weights: a t = σ (W a ·[h t-1 ,x t ]+b a ) Where: a t For time steps t The attention weights are added by the attention gating mechanism; σ W is the sigmoid activation function. a h is the weight matrix for attention weights in the attention gating mechanism; t-1 For time steps t 1's hidden state; b a This refers to the bias term of the attention weights in the attention gating mechanism; Hidden status updated to: h t =o t ⊙tanh(C t )·a t Where: h t For time steps t Updated hidden state; o t For time steps t The output gating state; ⊙ represents element-wise multiplication; tanh The hyperbolic tangent activation function; C t For time steps t The cellular state.

4. The AI ​​and blockchain-based full-process cultural property rights securitization system as described in claim 1, characterized in that, In the asset pool module, the objective function optimization of the Deep Deterministic Policy Gradient (DDPG) algorithm includes: Introducing infringement risk factors: P (st) =α×Rinc (st) +β×Dcomp (st) +γ×Tprob (st) The infringement risk factor P (st) It is obtained by weighted calculation of three sub-indicators: Historical Tort Litigation Risk Index Rinc (st) By connecting to the judicial open database API, we can obtain the historical litigation records of asset owners or similar works, and perform weighted calculations based on the loss rate, average compensation amount, and time decay factor. Dcomp, the degree of competition for similar content on the network (st) : Through a pre-set internet content aggregation interface, periodically crawl publicly available data from mainstream content platforms, and use a cross-modal similarity matching model that is the same as the copyright confirmation module to retrieve and count the number of publicly available works whose feature similarity to the target asset is higher than a pre-set threshold θ and their comprehensive popularity index. AI predicts the probability of infringement. Tprob (st) Output is generated by a pre-trained multimodal classification model. The multimodal classification model takes the asset's feature vector, the rights holder's historical credit score, and the compliance level certification of the first release channel as input, and maps them to probability values ​​in the 0, 1 interval through the Sigmoid function. The data for the above sub-indicators are obtained through the system's built-in data collection agent or authorized third-party compliant data service interfaces; α For Rinc (st) The weighting coefficient has a range of 0.

4. 0.5, and α + β + γ =1; β For Dcomp (st) The weighting coefficient has a value range of 0.

3. 0.4; γ For Tprob (st) The weighting coefficient has a value range of 0.

1. 0.3, in the digital collectibles sector γ =0.3; The objective function for optimizing the cross-border exchange rate factor is: max SR =(μ p -r f -l E ·|ΔE|) / s p in: SR Sharpe ratio, a measure of the risk-adjusted return of a portfolio of assets; μ p The expected rate of return for the asset pool; r f The risk-free rate; λ E σ is the exchange rate risk aversion coefficient, ranging from [0.05, 0.1]; ΔE is the exchange rate fluctuation range during the settlement period; σ p λ represents the standard deviation of the asset pool's returns, measuring the risk level of the asset pool. ΔE represents the exchange rate fluctuation range during the settlement period. E ∈[0.05,0.1].

5. The AI ​​and blockchain-based full-process cultural property rights securitization system as described in claim 1, characterized in that, The risk control module includes multi-level risk thresholds, including: The Sharpe ratio warning threshold for the asset pool is ≤1.2; The probability of infringement risk for a single asset is ≥0.5; Valuation error ≥8%; Cross-border settlement anomaly rate ≥3%; A risk response is triggered when any indicator exceeds the limit for three consecutive periods or when any two indicators exceed the limit simultaneously.

6. The AI ​​and blockchain-based full-process cultural property rights securitization system as described in claim 1, characterized in that, The privacy protection and data anonymization functions in the compliance settlement module include: Identify and de-identify trade secrets and personal information using the BERT-NER model; Sensitive images are blurred and watermarked. Data access permissions are set based on the RBAC model to ensure data security.

7. The system as described in claim 1, characterized in that, In the tiered issuance sub-module, the product structure design includes: low-risk assets with an infringement probability of ≤0.2, with the priority share accounting for 75%; medium-risk assets with an infringement probability of 0.2-0.4, with the priority share accounting for 55%; and high-risk assets with an infringement probability of ≥0.4, split into 5,000 standardized digital warrants.

8. The cultural property rights securitization system based on AI and blockchain as described in claim 1, characterized in that, The automatic fallback mechanism for anomalies in the fully automated engine includes: Configure backup algorithm switching and API retry scripts; When a module fails, a fault-tolerance process is automatically executed, with an anomaly repair success rate of ≥95.2%. The system failure rate is ≤0.5%.

9. The AI ​​and blockchain-based full-process cultural property rights securitization system as described in claim 1, characterized in that, The dynamic degradation mechanism in the multi-terminal API integration module includes: When the concurrency exceeds 1500 transactions / second or the response time exceeds 500ms, non-core functions are paused and the system is switched to the mirror node API. It employs TLS 1.3 encrypted transmission and SHA-256 data verification, and supports OAuth 2.0 authentication and fine-grained access control to ensure secure data transmission.

10. A method for the full-process securitization of cultural property rights based on AI and blockchain, employing the full-process securitization system for cultural property rights based on AI and blockchain as described in any one of claims 1-9, characterized in that, Includes the following steps: S1. Copyright Confirmation: Collect multimodal cultural property data using collection tools, upload it to the system in batches via the upload interface, process images, audio, and text separately using deep learning, extract features, fuse and complete them, and generate digital fingerprints to clarify ownership. S2, Intelligent Valuation: Receives confirmation data and multi-source data, calls the AG-LSTM network to perform attention gating calculation on the input sequence after isolation forest correction, and outputs an asset value assessment sequence for the next 12 months; S3. Asset Pool Optimization: Combining multiple risk factors and parameters, the DDPG algorithm is used to dynamically adjust and optimize the asset pool with the goal of maximizing the Sharpe ratio. S4. Risk Control and Compliance Assurance: Real-time monitoring of risk indicators, immediate response upon triggering thresholds, de-identification of information and protection of images during compliant settlement, and completion of cross-border settlement; S5. Tiered Issuance and Automated Execution: Products are designed according to risk levels, an automated engine enables unmanned operation, and an anomaly fallback mechanism ensures process continuity. S6, Multi-platform API Integration: Uses a specific architecture to connect to multiple platforms, with fault tolerance and degradation mechanisms, and employs encryption technology to ensure data transmission security; S7. Intelligent Feedback and Optimization: Adjusts system parameters based on operating parameters, integrates monitoring functions, and uses analysis algorithms to ensure compliant and safe system operation.

Images