Geospatial data element circulation method and system based on smart contract

Abstract

The invention discloses a geographic space data element circulation method and system based on a smart contract. The geospatial data element circulation system comprises a perception and execution layer, an element and right layer, a calculation and discovery layer and an immunization and compliance layer. The perception and execution layer is used for determining a trust benchmark of the whole system through the distributed trusted bookkeeping network; the essential factor and right layer is used for endowing the data with rich connotation and unique identity through deep identity; the calculation and discovery layer is used for understanding the structure and evolution of data ecology through a dynamic pedigree map engine; the immunization and compliance layer is used for automatically identifying and clearing inferior and malicious data in the system through a data immunization system; and providing a special interface with analysis capability and a data visualization tool through a compliance audit and supervision interface. Trusted circulation, efficient calculation and value multiplication of geographic space data elements in the whole life cycle can be achieved, and the method can dynamically adapt to evolutionary data ecology and complex application requirements.

Description

Technical Field

[0001] This application relates to the field of geospatial data processing technology, and in particular to a method and system for the circulation of geospatial data elements based on smart contracts. Background Technology

[0002] Geospatial data, as a key production factor in the digital economy era, has yet to fully realize its potential value. Current technological systems face a series of deep-seated and interconnected bottlenecks in promoting the secure and efficient flow of data elements: 1. The Lack of a Basic Circulation System and a Trust Gap: The five core aspects of data—ownership verification, value assessment, fair trading, privacy protection, and compliance supervision—still lack unified and reliable technical solutions. This has led to a significant trust gap, trapping massive amounts of high-value data in the hands of individual institutions or individuals, forming "data silos" and preventing the formation of a large-scale value network. 2. Security Vulnerability and Functional Rigidity of Data Processing Endpoints: The final use and computation of data occur in the user's terminal environment. These common environments (such as personal computers or servers) are highly vulnerable to malware attacks, leading to the leakage of critical keys and sensitive data. Simultaneously, existing data processing tools are often functionally rigid. Whenever a new data format, cryptographic algorithm, or data model needs to be supported, independent and costly development, deployment, and security audits are required. There is a lack of a standardized computing endpoint that can provide both hardware-level security isolation and dynamically expandable functionality.

[0003] 3. Efficiency and Security Dilemmas in AI Collaboration: With the development of AI technology, joint modeling using multi-party data (such as federated learning) has become a trend. However, in this process, how to scientifically evaluate and incentivize high-quality data contributors, how to effectively resist malicious attacks such as "model poisoning" or "gradient leakage," and how to efficiently coordinate the computation process of distributed nodes while ensuring the data sovereignty and privacy of all parties have become core obstacles to achieving large-scale, trustworthy AI collaboration.

[0004] 4. Static Understanding of Deep Relationships Between Data: Current data catalogs or marketplaces mostly rely on metadata for superficial and static classification of data. This fails to reveal the inherent, dynamic structural relationships between different data assets (e.g., the spatiotemporal correlation between soil moisture data of a certain region and remote sensing images of crop growth in neighboring regions). When the scale of data assets grows rapidly, the lack of effective means to dynamically maintain and query this complex "data relationship map" hinders the discovery of deeper value.

[0005] 5. The Lax and Uncontrollable Nature of Data Usage Authorization: Traditional data authorization models are often "one-off" and "one-size-fits-all." Once a data user obtains access permission, their subsequent usage behavior falls outside the precise control of the data owner, posing a risk of using data beyond its scope or purpose. The system lacks a sophisticated control mechanism that can cryptographically bind data usage permissions to specific identities, specific computing tasks, and specific usage frequency, and be automatically enforced by the machine.

[0006] 6. The Challenge of Democratizing Cutting-Edge Model Applications: Advanced geospatial AI models, such as hyperspectral image analysis and geological change simulation, often have their powerful capabilities monopolized by a few institutions due to their high training costs and complex application environments, making them inaccessible to ordinary users. A key bottleneck in achieving the democratization of geospatial intelligent applications lies in enabling a broad range of users to utilize or fine-tune these powerful foundational models in a low-cost, efficient manner while protecting their data privacy. Summary of the Invention

[0007] The method and system for the circulation of geospatial data elements based on smart contracts provided in this application can realize the reliable circulation, efficient computing and value multiplication of geospatial data elements throughout their entire life cycle, and can dynamically adapt to the ever-evolving data ecosystem and complex application needs.

[0008] Firstly, this application provides a geospatial data element circulation system based on smart contracts. The geospatial data element circulation system includes: a perception and execution layer, used to establish a trust benchmark for the entire system through a distributed trusted ledger network, and to conduct multi-party secure collaboration through AI-assisted computing nodes, and to provide scalable and ultimate secure execution guarantees through hardware-level trusted computing endpoints; an elementization and rights layer, used to endow data with rich connotations and unique identities through deep identityization, enabling it to be deeply understood and recognized by machines; and to bind clear and rigid rights rules to data through asymmetric programmable contracts, enabling it to securely flow and interact in open networks; a computation and discovery layer, used to understand the structure and evolution of the data ecosystem through a dynamic genealogy graph engine; and to empower each participant with AI technology through a parameter-efficient Model-as-a-Service engine; and an immunity and compliance layer, used to automatically identify and remove inferior and malicious data within the system through a data immune system; and to provide dedicated interfaces with analytical capabilities and data visualization tools through compliance audit and regulatory interfaces.

[0009] The perception and execution layer includes: a distributed trusted ledger network, which uses consortium blockchain or access-controlled public blockchain technology to build a high-performance distributed ledger that records the highest level of trust in meta-state information, including element identity history, equity contracts and state machines, and key operation and transaction logs; AI collaborative computing nodes, which build a large model-driven federated intelligent collaboration framework, supporting homomorphic encryption, secure multi-party computation, and various privacy computing protocols on the hardware level, and acting as an intelligent commander on the software level through a large model that is centrally deployed or maintained by consensus among multiple nodes; and hardware-level trusted computing endpoints, which build a complete trusted application lifecycle management system, providing each participant with a standardized, secure vault running on their own devices.

[0010] The tasks of the large model acting as an intelligent commander include: intelligent contribution assessment and dynamic weighting; forward-looking adversarial training and model hardening; and adaptive privacy-efficiency balance scheduling.

[0011] The Trusted Application Lifecycle Management System includes the following tasks: In this operating system, a trusted application is a standardized, securely encapsulated software module that runs in a hardware-enclosed memory space; after a new trusted application is developed, it undergoes automated review and system re-signing; when a user needs to use a new function, the trusted computing endpoint requests the download of the double-signed trusted application from the distribution center; the downloaded feasible application is then sent to the trusted execution environment of the device hardware.

[0012] The element and rights layer includes: a geospatial data deep identity module, which performs multi-level feature fusion and deep representation on geospatial data to obtain deep genes, and performs structure-aware comparative hash encoding on the deep genes to obtain gene summaries; and an asymmetric programmable contract module for data rights, which provides contract casting and rights nomination functions, rights activation and verifiable execution functions, and rights non-transferability functions.

[0013] The computation and discovery layer includes: a data ecosystem dynamic genealogy graph engine, used to measure distances based on the deep genes generated for each data asset according to the element and equity layers, and to perform clustering using an incremental update mechanism when a new data asset is registered in the system; as well as to perform assisted intelligent pricing and proactive value discovery and recommendation for each piece of data; and a parameter-efficient model-as-a-service engine, used to provide a centralized base model and a lightweight LoRA adapter for users to download and train in a local trusted environment, and when inference is required, the user's trusted endpoint will dynamically combine the output of the centralized base model with the output of the locally trained LoRA adapter.

[0014] Among them, the data ecosystem dynamic genealogy map engine is also used to assist in intelligent pricing for each piece of data by utilizing the dynamic pricing model of geospatial elements.

[0015] The immunity and compliance layer includes: a self-supervised distributed out-of-systems data immune system, used to perform immune checks on any new data asset attempting to register into the system; and a penetrating compliance audit and regulatory interface, used to assign different query and monitoring permissions to different regulators, and automatically pull and visualize the complete lifecycle of abnormal data transactions from the blockchain when regulators click on them.

[0016] Among them, the self-supervised out-of-distribution data immune system is equipped with a trained encoder-decoder model. The encoder-decoder model is used to reconstruct new data assets and calculate errors, and to perform graded anomaly response when the error is judged to be an anomaly. The graded anomalies include low-risk anomalies and high-risk anomalies.

[0017] Secondly, this application provides a method for the circulation of geospatial data elements based on smart contracts. This method is applied to the geospatial data element circulation system provided in the first aspect. The method includes: utilizing a perception and execution layer to establish a trust benchmark for the entire system through a distributed trusted ledger network; conducting multi-party secure collaboration through AI-assisted computing nodes; and providing scalable and ultimate secure execution guarantees through hardware-level trusted computing endpoints; utilizing an element-based and rights-based layer to endow data with rich connotations and unique identities through deep identity-based methods, enabling it to be deeply understood and recognized by machines; binding clear and rigid rights rules to data through asymmetric programmable contracts, enabling it to securely circulate and interact in open networks; utilizing a computation and discovery layer to understand the structure and evolution of the data ecosystem through a dynamic genealogical graph engine; and empowering each participant with AI technology through a parameter-efficient Model-as-a-Service engine; utilizing an immunity and compliance layer to automatically identify and remove inferior and malicious data within the system through a data immune system; and providing dedicated interfaces with analytical capabilities and data visualization tools through compliance audit and regulatory interfaces.

[0018] The beneficial effects of this application are as follows: Unlike existing technologies, the geospatial data element circulation method and system based on smart contracts provided in this application include: a perception and execution layer, used to establish a trust benchmark for the entire system through a distributed trusted ledger network, and to conduct multi-party secure collaboration through AI-assisted computing nodes, and to provide scalable and ultimate secure execution guarantees through hardware-level trusted computing endpoints; an elementization and rights layer, used to endow data with rich connotations and unique identities through deep identityization, enabling it to be deeply understood and recognized by machines; and to bind clear and rigid rights rules to data through asymmetric programmable contracts, enabling it to securely flow and interact in open networks; a computation and discovery layer, used to understand the structure and evolution of the data ecosystem through a dynamic genealogical graph engine; and to empower each participant with AI technology through a parameter-efficient Model-as-a-Service engine; and an immunity and compliance layer, used to automatically identify and remove inferior and malicious data within the system through a data immune system; and to provide dedicated interfaces with analytical capabilities and data visualization tools through compliance audit and regulatory interfaces. Through the above methods, a geospatial element operating system can be constructed that integrates hardware-level trusted roots, deep data identity, asymmetric programmable rights, intelligent AI collaboration, dynamic relationship hierarchy, and autonomous ecological evolution and immunity. By creatively encapsulating geospatial data, algorithms, computing power, and related rights into standardized element intelligent agents that can securely interact and combine in a trusted environment, the system can fundamentally achieve trusted circulation, efficient computing, and value multiplication of geospatial data elements throughout their entire lifecycle, and dynamically adapt to the ever-evolving data ecosystem and complex application needs. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a schematic diagram of the structure of an embodiment of the geospatial data element circulation system based on smart contracts provided in this application; Figure 2 This is a schematic diagram of the structure of an embodiment of the perception and execution layer provided in this application; Figure 3 This is a schematic diagram of the structure of an embodiment of the element and rights layer provided in this application; Figure 4 This is a schematic diagram of the structure of an embodiment of the computing and discovery layer provided in this application; Figure 5This is a schematic diagram of the structure of an embodiment of the immunity and compliance layer provided in this application; Figure 6 This is a flowchart illustrating an embodiment of the geospatial data element circulation method based on smart contracts provided in this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It is understood that the specific embodiments described herein are only for explaining this application and not for limiting it. Furthermore, it should be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all structures. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] See Figure 1 , Figure 1 This is a schematic diagram of an embodiment of the geospatial data element circulation system based on smart contracts provided in this application. The geospatial data element circulation system includes: a perception and execution layer 10, an elementization and rights layer 20, a computation and discovery layer 30, and an immunity and compliance layer 40.

[0023] In some embodiments, the geospatial data element circulation system is not an application platform, but a low-level operating system that provides standardized and reliable services for upper-level applications. Its architecture is divided into four layers that work together, such as the perception and execution layer 10, the elementization and rights layer 20, the computation and discovery layer 30, and the immunity and compliance layer 40.

[0024] The perception and execution layer 10 is used to establish a trust benchmark for the entire system through a distributed trusted ledger network, to conduct multi-party secure collaboration through AI-assisted computing nodes, and to provide scalable and ultimate secure execution assurance through hardware-level trusted computing endpoints.

[0025] In some embodiments, see Figure 2 The perception and execution layer 10 includes: a distributed trusted ledger network 11, an artificial intelligence collaborative computing node 12, and a hardware-level trusted computing endpoint 13.

[0026] The Distributed Trusted Ledger Network 11 is used to build a high-performance distributed ledger using consortium blockchain or access-controlled public blockchain technology, recording meta-state information with the highest level of trust. The meta-state information includes element identity history, equity contracts and state machines, and key operation and transaction logs.

[0027] The AI ​​collaborative computing node 12 is used to build a large model-driven federated intelligent collaboration framework. On the hardware level, it supports homomorphic encryption, secure multi-party computation, and various privacy computing protocols. On the software level, a large model, either centrally deployed or maintained by consensus among multiple nodes, plays the role of intelligent commander.

[0028] The hardware-level trusted computing endpoint 13 is used to build a complete trusted application lifecycle management system, providing each participant with a standardized, secure vault running on their own devices.

[0029] In some embodiments, the tasks of the large model acting as an intelligent commander include: intelligent contribution assessment and dynamic weighting; forward-looking adversarial training and model hardening; and adaptive privacy-efficiency balance scheduling.

[0030] In some embodiments, the work of the trusted application lifecycle management system includes: in this operating system, a trusted application is a standardized, securely encapsulated software module that runs in a hardware-enclosed memory space; after a new trusted application is developed, the new trusted application is automatically audited and re-signed by the system; when a user needs to use a new function, the trusted computing endpoint requests the download of the double-signed trusted application from the distribution center; the downloaded feasible application is sent to the trusted execution environment of the device hardware.

[0031] In some embodiments, the perception and execution layer 10 serves as a trusted foundation for the operating system.

[0032] The Perception and Execution Layer 10 is the lowest and most crucial infrastructure of the entire Trusted Geospatial Elements Operating System. Like the foundation and steel framework of the physical world, it provides a solid, reliable, and unshakeable security and operating environment for all upper layers' data elementization, rights management, intelligent computing, and value discovery. The core mission of this layer is to address three fundamental issues: the origin of trust, the isolation of computation, and the scalability of the ecosystem. It consists of three tightly coupled, collaborative core components: (a) Distributed Trusted Ledger Network 11: Trust Bus for the Entire System.

[0033] 1. Role and Position: This network serves as a "public fact book" for the operating system, its fundamental function being to provide a unified, immutable, and verifiable source of trust for all critical states and historical behaviors across the entire ecosystem. It solves the most fundamental challenges in the circulation of data elements: the determination of ownership, traceability, and oversight.

[0034] 2. Core Technology Mechanism: A high-performance distributed ledger is constructed using consortium blockchain or access-controlled public blockchain technology. Unlike traditional databases, each record in this ledger is linked to the previous record via a cryptographic hash pointer, forming an irreversible data chain. Any tiny modification to the historical record will cause the hash value of all subsequent blocks to change, thus being immediately detected and rejected by the system network.

[0035] 3. Specific applications in this system: This ledger is not used to record massive amounts of raw geospatial data, but focuses on recording high-value, top-level trusted "meta-state" information, including: Genesis of Element Identity: Whenever a new piece of geospatial data is "deeply identified," its unique "deep gene" summary, ownership, and creation timestamp are permanently recorded on the blockchain as a generative transaction.

[0036] Stake Contracts and State Machines: Asymmetric programmable smart contracts based on "nominated signatures" are deployed here. The contracts clearly define the data's users, purposes of use, validity periods, and other detailed rules. All requests for, authorizations of, and execution of state transitions related to this data are completed through interaction with the contract, and each state change is a publicly verifiable on-chain transaction.

[0037] Key Operations and Transaction Logs: All important data transactions, model calls, federated learning contribution evaluation results, abnormal data immunity records, and other key operations will have their summaries and results recorded on the blockchain, providing irrefutable evidence for post-event auditing, dispute resolution, and compliance supervision.

[0038] 4. Summary of Contributions: By establishing such a trust bus, this system fundamentally solves the problems of difficulty in confirming ownership, tracing, and regulating data transactions caused by the lack of a unified trust intermediary. It provides a fundamental guarantee for the "trustworthy" operation of all upper-layer functions.

[0039] (II) Artificial Intelligence Collaborative Computing Node 12: From "Privacy Computing" to "Intelligent Collaboration".

[0040] 1. Role and Position: These nodes are dedicated computing clusters used by the operating system for joint analysis of multi-party data and training of artificial intelligence models. Their goal extends beyond protecting privacy; it aims to improve the efficiency, fairness, and security of multi-party collaboration by introducing intelligent coordination mechanisms, thereby addressing deep-seated trust and incentive issues in AI collaboration.

[0041] 2. Core Technology Mechanisms and Integrated Innovation: This component constructs a "large model-driven federated intelligent collaboration framework." These nodes support various privacy computing protocols such as homomorphic encryption and secure multi-party computation in hardware, while in software, a large model that is centrally deployed (or maintained by consensus among multiple nodes) acts as the "intelligent commander."

[0042] 3. Detailed workflow of the large-scale commander: Intelligent contribution evaluation and dynamic weighting: In each round of federated learning, each participating node not only uploads encrypted model update parameters but also anonymized metadata, such as the loss reduction rate during local training, the model's accuracy on the local validation set, and the time taken to complete training. The large model commander comprehensively analyzes this metadata and scores the "quality of contribution" of each node. For example, a node that consistently provides high-quality, highly relevant data will receive a higher score. During global model aggregation, the system dynamically weights the model updates of each node based on this score, rather than simply averaging. This creates an effective incentive mechanism that encourages participants to provide authentic and valuable data.

[0043] Proactive adversarial training and model hardening: The large model commander can proactively and creatively generate "virtual adversarial samples" that best challenge the model's weaknesses, based on its understanding of the current global model. These samples are distributed to various nodes for the next round of local training. By allowing the model to "practice" locally how to resist these high-intensity attacks, the final aggregated global model will have stronger robustness and generalization ability against malicious behaviors such as model poisoning and data contamination that may occur in the real world.

[0044] Adaptive Privacy-Efficiency Balanced Scheduling: The large model commander also acts as the system's "privacy budgeter." It intelligently decides the strength of privacy protection protocols to activate based on the sensitivity of the task, the historical trustworthiness of participating nodes, and the current security posture of the network.

[0045] For scenarios involving extremely sensitive data or detecting abnormal node behavior, it will force the use of homomorphic encryption protocols, which are computationally expensive but offer the strongest security; while for routine, low-risk collaborative tasks, it will use differential privacy or gradient perturbation techniques, which have lower computational overhead.

[0046] This adaptive scheduling mechanism achieves a dynamic optimal balance between safety and efficiency.

[0047] 4. Summary of Contributions: This component creatively solves the efficiency and security dilemmas in AI collaboration. It goes beyond simple "privacy protection" and, by introducing intelligent coordination, establishes a truly trustworthy distributed intelligent collaboration system that integrates contribution measurement, fair incentives, proactive defense, and cost optimization.

[0048] (III) Hardware-level Trusted Computing Endpoint 13: From “Application Security” to “Ecosystem Scalability”.

[0049] 1. Role and Position: This is a standardized "security vault" provided by the operating system for every participant, running on the user's own device. It pushes the security boundary from the cloud to the edge closest to the data, fundamentally solving the problem of untrusted user endpoint environments. More importantly, it solves the fundamental problem of the entire ecosystem's functional rigidity and difficulty in evolution through an innovative modular and secure update mechanism.

[0050] 2. Core Technology Mechanisms and Integrated Innovation: The core idea of ​​this component is to combine the modular design of hardware wallets with the Trusted Execution Environment (TEE) of the operating system (such as TrustZone technology for ARM processors) and build a complete "Trusted Application (TA)" lifecycle management system.

[0051] 3. Trusted Application Lifecycle and Security Workflow: Definition and Encapsulation: In this operating system, a "trusted application" is a standardized, securely encapsulated software module that runs in a hardware-enclosed memory space. It comes in various forms, such as "data processing trusted applications" specifically designed for parsing and processing geospatial data in specific formats (e.g., hyperspectral data), and "algorithm trusted applications" that encapsulate specific analytical algorithms (e.g., landslide risk prediction models).

[0052] Development and Submission: Any third-party developer (such as research institutions or software companies) can develop new trusted applications by following the development kits provided by the operating system. After development is complete, the developer uses their private key to digitally sign the application and submits it to the operating system's "Module Authentication and Distribution Center".

[0053] Automated review and system re-signing: After receiving the submission, the distribution center will start an automated pipeline to conduct a rigorous security review of the trusted application, including static code vulnerability scanning and dynamic sandbox behavior analysis (detecting abnormal network requests or memory abuse).

[0054] Only applications that pass all reviews will be given a secondary authoritative signature by the operating system's root certificate. This dual-insurance mechanism of "developer signature + system signature" establishes a clear chain of responsibility and trust.

[0055] On-demand download and secure deployment: When a user needs to use a new feature (for example, a farmer wants to use the latest "wheat rust detection algorithm"), they simply select it on the operating system interface. The trusted computing endpoint on their device requests the download of this double-signed "trusted application of the algorithm" package from the distribution center.

[0056] Hardware-mandated signature verification and isolated loading: After downloading, the application package is sent to the device hardware's trusted execution environment. The firmware within this environment (whose own boot process is also protected by a secure boot chain) uses its system public key, embedded in the hardware, to perform a final signature verification of the application package. Only if the signature verification passes, proving its reliable origin and lack of tampering during transmission, is the application allowed to be decrypted and loaded into a strictly hardware-isolated memory region for execution. Any code without a valid signature cannot enter this "safe haven," thus completely preventing malware injection and supply chain attacks.

[0057] 4. Summary of Contributions: This component revolutionarily solves two major challenges: the security vulnerability of data processing endpoints and the rigidity of functionality. It not only provides the highest level of security for data and algorithms to date through hardware isolation, but more importantly, it endows the entire operating system with the ability to "evolve infinitely" through a modular and secure dynamic update mechanism. This transforms the system from a closed, functionally fixed platform into an open, community-driven, continuously growing and iterating, vibrant ecosystem.

[0058] This section can be summarized as follows: The perception and execution layer 10 establishes a trust benchmark for the entire system through a distributed trusted ledger network 11, achieves intelligent multi-party secure collaboration through artificial intelligence collaborative computing nodes 12, and provides scalable and ultimate secure execution assurance through hardware-level trusted computing endpoints 13. These three components are interconnected, jointly constructing a solid foundation for the entire "Trusted Geospatial Element Operating System," making it both rock-solid and full of vitality.

[0059] In some embodiments, the elementization and rights layer 20 is used to endow data with rich connotations and unique identities through deep identityization, enabling it to be deeply understood and recognized by machines; and to bind clear and rigid rights rules to data through asymmetric programmable contracts, enabling it to securely flow and interact in open networks.

[0060] In some embodiments, see Figure 3 The element and rights layer 20 includes: a geospatial data deep identity module 21 and an asymmetric programmable contract module 22 for data rights.

[0061] The geospatial data deep identity module 21 is used to perform multi-level feature fusion and deep representation on geospatial data to obtain deep genes, and to perform structure-aware comparative hash encoding on the deep genes to obtain gene summaries.

[0062] The asymmetric programmable contract module 22 for data rights is used to provide contract casting and rights nomination functions, rights activation and verifiable execution functions, and rights non-transferability functions.

[0063] In some embodiments, elementization and rights layer 20: identity and contract of data.

[0064] Above the robust perception and execution layer 10, the elementization and rights layer 20 undertakes the core task of transforming raw, chaotic geospatial data into standardized, intelligent digital assets. This layer's mission is to assign a rich and unique "identity" to each piece of data and bind it to a programmable, automatically executable "contract." Through this process, data is upgraded from a passive, waiting-to-be-called "file" to an "intelligent element agent" with self-description capabilities and endogenous rules. This layer consists of two innovative components, addressing the challenges of deep data understanding and precise rights control.

[0065] (I) Deep Identity of Geospatial Data: The Evolution from “Fingerprint” to “Gene”.

[0066] 1. Role and Position: Traditional data management relies on manually defined, sparse metadata tags (such as "Land use map of Haidian District, Beijing, summer 2024"). This approach has low information content and fails to capture the complex patterns inherent in the data. The goal of this component is to automatically generate a comprehensive, machine-understandable, and unique "deep identity" for each data asset, which we call the "data deep gene." This "gene" not only contains the semantic information of the data but also its rich structural and textural features, forming the foundation for content-level data retrieval, intelligent pricing, and anomaly detection.

[0067] 2. Core Technology Mechanisms and Integrated Innovation: The core idea of ​​this component originates from the technology used in medical image retrieval for deep feature integration and structured analysis, and is creatively applied to the field of geospatial data. It constructs the "deep genes" of the data through a two-stage process.

[0068] Phase 1: Multi-level feature fusion and deep representation.

[0069] A hybrid deep learning encoder was designed, combining the advantages of convolutional neural networks (excel at capturing local textures) and transformer networks (excel at capturing global correlations). When geospatial data (such as a remote sensing image) is input into this encoder, it does not simply take the output of the last layer as the final feature, as traditional models do, but instead: 1. Shallow Feature Extraction: In the initial convolutional layers of the network, low-level visual features of the data are extracted, such as the edges, textures, and color details of ground features. These features are like "bases" in a gene sequence, forming the basis of complex patterns.

[0070] 2. Deep Feature Extraction: At the final transformer layer of the network, high-level semantic features of the data are extracted, such as identifying global concepts in the image like "urban building complexes," "rivers," and "farmland." These features are like "functional proteins" in gene expression, representing the core content of the data.

[0071] 3. Adaptive Fusion: The most crucial step is that the system uses an attention mechanism to adaptively and weight-wise fuse shallow "base" information with deep "functional protein" information. This means that the system can dynamically determine whether detailed texture or global semantics is more important based on the specific content of the data, thereby generating a highly condensed high-dimensional feature vector that contains both macroscopic semantics and retains key microscopic details. This vector is the unique "deep gene" of this data.

[0072] Second stage: Structure-aware contrastive hash encoding.

[0073] While comprehensive, the high-dimensional nature of "deep genes" hinders rapid retrieval. Therefore, it needs to be compressed into low-dimensional binary hash codes (i.e., "gene summaries"). Traditional hashing methods often misclassify samples that are "similar but visually different" (e.g., one is a neatly planned modern farmland, the other is terraced fields, both belonging to the "farmland" category) due to subtle differences. To address this "over-centralization" problem, we introduce a structure-aware mechanism: 1) Generating a macroscopic structural fingerprint: Before hash encoding, the system downsamples the original image to generate a low-resolution thumbnail. This image ignores details but retains the core spatial layout and structural information.

[0074] 2) Structured Pair Learning: During the training of the hash network, the system generates three types of sample pairs: Positive sample pairs: Two datasets belong to the same category and their "macrostructural fingerprints" are also highly similar (e.g., both are images of regular farmland).

[0075] Negative sample pair: Two data points belonging to different categories.

[0076] Neutral sample pairs: Two data sets belong to the same category, but their "macrostructural fingerprints" differ greatly (e.g., a regular farmland and a terraced field).

[0077] 3) Differentiated Loss Function: During learning, the network strongly narrows the distance between "positive sample pairs" and widens the distance between "negative sample pairs," while treating "neutral sample pairs" with a more lenient constraint. This allows the hash network to learn to "find commonalities while respecting differences": it can identify the commonalities of similar data while tolerating reasonable differences in the macroscopic structure of similar data.

[0078] 3. Summary of Contributions: By constructing a "data depth gene," this component fundamentally changes the way the system perceives data. It enables data retrieval to move beyond simple text tags and instead achieve a true understanding of the content. Simultaneously, the introduction of a structure-aware mechanism significantly improves the robustness and accuracy of data classification and retrieval in complex geospatial scenarios, addressing the core pain point of static and superficial understanding of data relationships.

[0079] (ii) Asymmetric programmable contracts for data rights: an upgrade from “license” to “code as law”.

[0080] 1. Role and Position: Traditional data authorization models are like issuing an "access card"—once issued, the cardholder's behavior after entering is difficult to control in detail. The goal of this component is to create a cryptographically-based, tamper-proof, and automatically executable "intelligent legal document" for each data right (such as usage rights and profit rights). This "document" precisely defines the boundaries and execution conditions of the rights in code form, fundamentally solving the problems of crude data authorization and uncontrollable processes.

[0081] 2. Core Technology Mechanisms and Integrated Innovation: The core idea of ​​this component stems from combining "nominated signatures" in cryptography with smart contract technology in blockchain to create an "asymmetric programmable equity contract." Here, "asymmetric" does not refer to the encryption algorithm, but rather to the unequal distribution of power and the one-way flow of power.

[0082] 3. Lifecycle and Workflow of Equity Contracts: Contract Casting and Stake Nomination: When a data owner (Party A) wants to authorize a user (Party B) a specific stake (e.g., allowing Party B to use their land cover data for a month-long drought analysis), they don't simply send a permission. Instead, they initiate a transaction on the operating system's blockchain to deploy a smart contract. The core of this contract is a "nomination" act: 1. Signer: Party A, i.e., the data owner.

[0083] 2. Nominee: The user of the data. Here, the nominee's identity is the public key address of its trusted computing endpoint or a specific trusted application, not a username that can be easily impersonated.

[0084] 3. Subject matter of the contract: A "deep gene" summary of the authorized data.

[0085] 4. Rights and Benefits: Precisely defined through contract code, such as: "Limited to executing the 'Drought Analysis' algorithm (specified by the hash value of its trusted application)," "Valid until December 31, 2025," and "A total of 100 calls allowed," etc. Both parties, through an interactive process, jointly complete the signing confirmation of this contract, generating a "nominated signature," which is permanently locked in the contract. This process ensures the mutual consent and non-repudiation of the authorization.

[0086] Rights activation and verifiable execution: When Party B needs to exercise its granted rights, the workflow is as follows: 1. Identity Verification and Request: The B party's trusted application initiates a call to the smart contract and uses its own private key to sign, proving that it is indeed the "nominated party" that was originally nominated.

[0087] 2. Contract Verification and Token Generation: The smart contract automatically verifies the client's signature on the blockchain. Upon successful verification, the contract generates a one-time, time-limited "execution token" bound to the specific parameters of the request (such as the range of data blocks invoked), based on pre-defined rights and obligations. Simultaneously, the contract's internal counter is updated (e.g., the remaining number of calls is decremented by one).

[0088] 3. Forced execution at the system kernel level: When Party B initiates an actual data access request to the operating system's data storage or computing module, it must attach this newly generated valid "execution token." The operating system's underlying kernel will verify the token; only after successful verification will the corresponding data reading or computing operation be executed.

[0089] The key to this mechanism is that the logic for generating execution tokens is completely embedded in an immutable smart contract, while the verification of tokens is enforced by the underlying kernel of the operating system. Together, they form a closed loop that cannot be bypassed.

[0090] Non-transferability of rights: Since the generation of the "execution token" requires the signature of the nominated Party B, it means that Party B cannot transfer this right to any third party (Party C). Party C does not possess Party B's private key and therefore can never obtain a valid execution token from the smart contract. This cryptographically guarantees the strict non-transferability of data rights.

[0091] 3. Summary of Contributions: Through "asymmetric programmable rights contracts," this component advances the granularity of data authorization from "whether access is allowed" to fine-grained and automated control over "under what conditions, by whom, in what way, and what operations are performed." It replaces legal documents that require post-audit with immutable code, realizing a new paradigm of data governance that prioritizes "prevention" over "punishment," and completely solving the core pain points of extensive data use authorization and uncontrollable processes.

[0092] This section can be summarized as follows: The element-based and rights-based layer 20 endows data with rich connotations and unique identities through deep identity-based processing, enabling it to be deeply understood and recognized by machines. Through asymmetric programmable contracts, it binds data with clear and rigid rights rules, giving it the ability to flow and interact securely in open networks. The combination of these two elements completes the crucial transformation of data from raw resources to standardized "intelligent element agents," laying a solid foundation for building a truly trustworthy, efficient, and intelligent data element circulation ecosystem.

[0093] In some embodiments, the computation and discovery layer 30 is used to understand the structure and evolution of the data ecosystem through a dynamic phylogenetic graph engine; and to empower each participant with AI technology through a parameter-efficient model-as-a-service engine.

[0094] In some embodiments, see Figure 4 The computation and discovery layer 30 includes: a data ecosystem dynamic genealogy graph engine 31 and a parameter-efficient model-as-a-service engine 32.

[0095] The data ecosystem dynamic genealogy engine 31 is used to measure the distance of each data asset based on the deep genes generated by the element and equity layers 20, and to perform clustering using an incremental update mechanism when a new data asset is registered in the system; as well as to provide assisted intelligent pricing for each piece of data and to perform proactive value discovery and recommendation.

[0096] The parameter-efficient Model-as-a-Service engine 32 provides a centralized base model and a lightweight LoRA adapter for users to download and train in a local trusted environment. When inference is required, the user's trusted endpoint dynamically combines the output of the centralized base model with the output of the locally trained LoRA adapter.

[0097] Among them, the Data Ecosystem Dynamic Genealogy Map Engine 31 is also used to assist in intelligent pricing for each piece of data by utilizing the dynamic pricing model of geospatial elements.

[0098] In some embodiments, the computing and discovery layer 30: the intelligence and insights of the system.

[0099] If the Perception and Execution Layer 10 is the skeleton of the operating system, and the Elementization and Rights Layer 20 is the blood, then the Computation and Discovery Layer 30 is the "brain" and "central nervous system" of the entire system. Its core mission is to endow the system with the ability to perform advanced intelligent computing and proactively discover potential value from massive amounts of data. This layer not only executes passive computational tasks but also endows the system with the ability to "think" and "insight" through dynamic analysis and intelligent modeling. It mainly consists of two core engines, responsible for macroscopic structural insight and microscopic high-efficiency computing, respectively.

[0100] (I) Data Ecosystem Dynamic Genealogy Map Engine 31: From “Data Catalog” to “Economic Census”.

[0101] 1. Role and Position: Traditional data markets resemble static catalogs, merely listing available data without revealing the complex and dynamic relationships between these "products." This engine aims to construct and maintain a dynamic genealogical map of the entire data ecosystem in real time. This map is not a simple classification tree, but a complex network reflecting the similarities, correlations, complementarities, and evolutionary trends among all data elements. It functions like a real-time "economic census" of the entire data element market, providing macro-level, structured decision-making support for value assessment and intelligent recommendations.

[0102] 2. Core Technology Mechanisms and Integrated Innovation: The core idea of ​​this component is to fundamentally transform the "single-link clustering" algorithm in graph theory by introducing efficient parallelization and dynamicization techniques, enabling it to handle ultra-large-scale and constantly changing datasets.

[0103] 3. Construction and workflow of the "Dynamic Genealogy Map": Distance metric based on "deep genes": The engine's basic input is the "deep genes" (i.e., high-dimensional feature vectors) generated for each data asset by "Elementization and Equity Layer 20". The engine quantifies the intrinsic similarity between any two data assets by calculating the distance between these "genes" (such as Euclidean distance or cosine similarity). The closer the distance, the more similar their content.

[0104] Efficient incremental clustering: When a new data asset is registered in the system, traditional clustering algorithms require recalculating the entire dataset, which is impractical in massive data environments. Our dynamic engine employs an incremental update mechanism: 1. Local calculation: The engine only needs to calculate the distance between the new data node and the existing nodes in the graph, without affecting the distance relationships between existing nodes.

[0105] 2. Dynamic Maintenance of the Tree Structure: The system maintains a data structure called a "single-linked tree graph" in the background. When a new node is added, the engine uses a parallel batch update algorithm specifically designed for tree graphs to perform efficient "pruning" and "grafting" operations only on the affected local areas of the tree, thereby integrating the new node into the existing hierarchical structure with minimal computational cost. The theoretical time complexity of the entire update process is independent of the total amount of data and only depends on the height of the tree, achieving near real-time response.

[0106] Output and application of structured information: The "dynamic phylogenetic map" continuously output by the engine has rich application value: 1. Assisted Intelligent Pricing: When evaluating the value of data (e.g., a groundwater distribution map of a region), the system considers not only its intrinsic quality but also its phylogenetic relationships. If the map shows that highly correlated data such as soil composition and crop growth are scarce, the potential application value of this hydrological data (i.e., its ability to create value when combined with other data) is assessed more highly. This solves the problem of data pricing difficulties, moving it from isolated assessment to ecological evaluation. The dynamic pricing model for geospatial elements driven by ecological value is as follows: Core Idea: Traditional data pricing models are often static, primarily focusing on the intrinsic attributes of the data itself (such as accuracy and timeliness). The core innovation of this model lies in its view that the true value (price) of a data element does not exist in isolation, but is determined by its network position, collaborative potential, scarcity, and level of trust within the entire data ecosystem. Therefore, it is a dynamic, multidimensional, and non-linear value assessment function.

[0107] The dynamic pricing formula is as follows: Define the fair market price (i, t) of data element i at time t as the product of its basic quality value and a dynamic ecological value multiplier.

[0108] Price(i, t) = Value_Base(i) × (1 + Multiplier_Ecology(i, t)).

[0109] The detailed explanations of each component of the formula are as follows: 1. Basic quality value: Value_Basic(i).

[0110] This is the static intrinsic value of a data element, determined by its inherent attributes. It can be seen as the "factory price" of this data.

[0111] Value_Base(i) = Σ_{k=1}^{N} (weight_k × rating_k(i)).

[0112] in: N is the total number of basic quality evaluation dimensions.

[0113] Weight_k is the importance weight of the k-th dimension, which is dynamically adjusted by the system based on market feedback, satisfying Σweight_k = 1.

[0114] Rating_k(i) is the standardized rating (between 0 and 1) of data element i on the k-th dimension. These dimensions include, but are not limited to: Data accuracy: The degree of conformity with the actual ground values.

[0115] Spatial resolution: The actual geographic area represented by each pixel.

[0116] Time resolution: the frequency of data updates.

[0117] Timeliness: How close the data collection time is to the current time t.

[0118] Completeness: The absence of data coverage areas.

[0119] Standardization level: The degree to which the data format conforms to system specifications.

[0120] 2. Ecological value multiplier: Multiplier_ecology(i, t).

[0121] This is used to quantify the added value of data element i at time t due to its unique position and relationship within the ecosystem. It consists of three main nonlinear terms: network structure value, collaborative creation value, and dynamic trust discounting.

[0122] Multiplier_Ecosystem(i, t) = Value_Network(i, t) + Value_Collaboration(i, t) - Discounted_Trust(i, t) 2.1. Network structure value: value_network(i, t).

[0123] This term measures the value of a data element i due to its structural position in a "dynamic hierarchical graph." Its core idea is that the value of a piece of data depends not only on itself, but also on its association with its "valuable" neighbors and its own scarcity. It borrows the concept of PageRank from graph theory, but with adaptive modifications.

[0124] Value_network(i, t) = α × ∫_{j ∈ Neighbor(i)} [ (Similarity(i, j) / Σ_{k ∈ Neighbor(i)} Similarity(i, k) ) × Value_network(j, t) ] dμ(j) + (1 - α) × Scarcity(Category(i), t).

[0125] This is a single-step expression of the weighted PageRank iterative formula in continuous form: α is a damping coefficient (usually taken as 0.85), representing the proportion of value transmitted in the network.

[0126] Neighbor(i) is the set of neighbor nodes that are directly connected to data i in the "Dynamic Genealogy".

[0127] ∫...dμ(j) represents the integration or summation of the value contributions of all neighboring nodes.

[0128] Similarity (i, j) is the similarity score between nodes i and j calculated by the "data depth gene".

[0129] The denominator Σ_{k ∈ Neighbor(i)} similarity(i, k) is a normalization term that ensures that the sum of the weights of all outgoing chains is 1.

[0130] The recursively defined value_network(j, t) represents the network value of its neighbor node j. In actual computation, this is a value that requires multiple iterations to converge.

[0131] Creative introduction of scarcity (category(i), t): This part represents the value injection of "exogeneity". It measures the scarcity of the category to which data i belongs (determined by the clustering results of the phylogenetic map) within the entire ecosystem. It is defined as follows: Scarcity (C, t) = -log ( N_C / N_total).

[0132] Where N_C is the number of data elements in category C, and N_ is always the total number of data elements in the ecosystem. This logarithmic form makes extremely rare categories have very high fundamental value.

[0133] The explanation for this is as follows: The network value of a data element is comprised of α proportions from the weighted contribution of the values ​​of all its "neighbors" (the more similar the neighbors, the higher their contribution weight), and an additional (1-α) proportion from the global scarcity of its category. This means that even data of average quality can have its network value significantly increased if it happens to connect with multiple high-value, highly scarce data points.

[0134] 2.2. Collaborative Value Creation: Value_Collaboration(i, t).

[0135] This measure assesses the “collaborative ability” and “value creation ability” of data element i in past participation in collaborative AI computing (such as federated learning and model fine-tuning).

[0136] Value_Collaboration(i, t) = β × tanh( Σ_{m ∈ Task} [ ( Contribution(i, m) × e^(-λ(t - t_m)) ) / Cost(i, m) ] ).

[0137] β is a coefficient that modulates the influence of synergistic value.

[0138] tanh(...) is the hyperbolic tangent function, which smoothly maps the accumulated collaborative scores to the (-β,β) interval, avoiding infinite value growth, while providing a large marginal effect when the scores are low.

[0139] Σ_{m ∈ Task} represents the summation of all historical tasks m in which data i has participated.

[0140] The contribution score (i, m) is given by the large model coordinator of "AI Collaborative Computing Node 12", which is the contribution score of data i in task m (e.g., the percentage contribution it makes to the improvement of global model accuracy).

[0141] We creatively introduce the term e^(-λ(t - t_m)): This is a time decay factor. t_m is the time when task m occurs, and λ is the decay rate. This factor implies that the contribution of recent tasks has a higher weight than that of distant tasks, reflecting the "timeliness" of data value.

[0142] Cost (i, m) represents the computational and communication resources consumed by data i when participating in task m. Dividing the contribution by the cost yields the "contribution efficiency," which encourages low-cost, high-return collaboration.

[0143] The explanation for this is as follows: The collaborative value of a data element is the cumulative sum of all its historically "efficient" collaborative contributions, which decays over time, and is smoothed by a saturation function. This rewards data that continuously and efficiently participates in value co-creation.

[0144] 2.3. Dynamic Trust Discounting: Discount_Trust(i, t).

[0145] This is a penalty item used to discount the value of data element i based on its historical "bad records".

[0146] Discount_trust(i, t) = γ × ( 1 - e^(-Σ_{e ∈ event} severity(e) × e^(-δ(t- t_e))) ).

[0147] γ is the maximum penalty coefficient that modulates the influence of trust discounting.

[0148] 1 - e^(-x) is a function with a shape similar to tanh that maps the accumulated “bad record” score x to the interval [0, γ), and has the characteristic of diminishing marginal effect (i.e., the first offense is punished the most severely).

[0149] Σ_{e ∈ event} represents the summation of all historical negative events e related to data i.

[0150] Severity (e) is the system's rating of the severity of a negative event e. For example: Determined as mild OOD by the "immune system": Severity = 0.2.

[0151] Contractual disputes arising in transactions: Severity = 0.5.

[0152] Data identified as malicious poisoning: Severity = 1.0.

[0153] e^(-δ(t - t_e)) is also a time decay factor, but its decay rate δ may differ from λ. This means that the impact of "bad records" will gradually fade over time, giving the data owner a chance to "turn over a new leaf".

[0154] The explanation for this is as follows: The trust discount of a data element is the cumulative sum of the severity of all its historical negative events, decaying over time, mapped non-linearly through a penalty function. This ensures that data elements with a "criminal record" will have their trading prices suppressed accordingly, thereby maintaining market fairness and order.

[0155] In summary: The innovation of this dynamic pricing formula lies in its expression of the price (i, t) of a data element as a complex system driven by both static intrinsic quality and dynamic ecological value. Specifically, the ecological value multiplier creatively integrates: Structural value analysis (value network) from graph theory and network science.

[0156] A collaborative contribution metric (value_collaboration) derived from federated learning and game theory.

[0157] Time decay and trust penalty (discounted trust) from risk management and behavioral economics.

[0158] Through this formula, the operating system's computation and discovery layer 30 can provide a highly intelligent and fair reference price for each data element that reflects both its own strengths and weaknesses and its unique role in the entire dynamic ecosystem. This provides a core mathematical engine for building a transparent, efficient, and fair data element market.

[0159] 2. Proactive Value Discovery and Recommendation: The engine can proactively discover "value gaps" in the spectrum. For example, it might find that "urban heat island effect data" and "regional electricity consumption data" form a tight cluster in the graph, even though they traditionally belong to different fields. The system can push this insight to relevant researchers, suggesting that there may be undiscovered strong correlations between these two types of data, thereby fostering new interdisciplinary research and applications.

[0160] 4. Summary of Contributions: This engine solves the problem of static understanding of deep relationships between data through dynamic and parallel technologies. It enables the operating system to gain macroscopic, real-time, and structured insights into its own ecosystem, elevating data management from a passive "catalog-based" approach to a proactive "census and analysis" level.

[0161] (II) A parameter-efficient Model-as-a-Service engine 32: From “model-private” to “capability-sharing”.

[0162] 1. Role and Position: The geospatial domain possesses a wealth of extremely powerful but costly foundational AI models (such as general hyperspectral image analysis models). These models, like "heavy industrial mother machines," are powerful tools driving industry progress, but their high acquisition and usage costs deter most users. The goal of this engine is to build a "model capability sharing platform," allowing a broad range of users to securely "rent" the capabilities of these top-tier models at extremely low cost and with extremely high data efficiency, and adapt them to their specific, niche task requirements.

[0163] 2. Core Technology Mechanisms and Integrated Innovation: The core idea of ​​this component stems from the deep integration of efficient parameter fine-tuning techniques, especially Low-Rank Adaptation (LoRA), with the hardware-level trusted computing endpoint 13 of the "Perception and Execution Layer 10".

[0164] 3. Workflow and security of "Model as a Service": Basic Model and LoRA Adapter: 1. Centralized Base Model: The operating system operator is responsible for training and maintaining one or more large geospatial base models in the cloud. All parameters of these models remain "frozen" during fine-tuning and are not changed in any way.

[0165] 2. Lightweight LoRA Adapter: For each key module in the model (such as the attention layer in the transformer network), the system injects two additional, extremely small, trainable low-rank matrices (A and B). The combination of these two small matrices constitutes the so-called "LoRA adapter." During fine-tuning, the user actually only trains the parameters of these adapters, which may account for only 0.1% of the total parameters of the entire base model.

[0166] A safe and efficient personalized fine-tuning process: 1. Task Definition and Adapter Download: A user (e.g., a tea plantation owner) wants to use a hyperspectral analysis model to identify a local tea disease. He first defines his task through the operating system interface and prepares a small amount of self-labeled sample data. Then, he downloads the untrained LoRA adapter, which is compatible with the basic hyperspectral analysis model, from the system's model library to his hardware-level trusted computing endpoint 13.

[0167] 2. Training in a Local Trusted Environment: The fine-tuning process takes place entirely within the user's trusted endpoint. A small amount of local data provided by the user is used to train the LoRA adapter's parameters. Due to the extremely small number of parameters being trained, the entire process requires very little computing power and can likely be completed within minutes on a typical edge device. More importantly, the user's private data never leaves their local secure environment, perfectly protecting data privacy and sovereignty.

[0168] 3. Dynamic Synthesis of Model Capabilities: When inference is required, the user's trusted endpoint dynamically combines the output of the "frozen base model" with the output of the locally trained "LoRA adapter." This is like adding a "turbocharger" tuned for a specific track to a general-purpose "engine," enabling it to unleash powerful performance on a particular task.

[0169] The value transfer of fine-tuning results: The LoRA adapter trained by this tea plantation owner specifically for identifying tea diseases is itself a valuable, small, and sophisticated "algorithmic asset." He can choose to share or sell it to other tea plantation owners with similar needs through the operating system's trading mechanism. This creates a vibrant secondary market for "model capabilities."

[0170] 4. Summary of Contributions: This engine perfectly solves the challenge of democratizing cutting-edge model applications through an innovative combination of efficient parameter fine-tuning and trusted computing endpoints. It pioneers a new model of "centralizing basic capabilities and marginalizing personalized innovation," which greatly lowers the application threshold of advanced artificial intelligence technologies while ensuring data security and privacy, making it possible for every ordinary user to become a participant and contributor to AI innovation.

[0171] This section can be summarized as follows: The computation and discovery layer 30 is the core embodiment of the operating system's intelligence. The dynamic genealogy graph engine endows the system with macroscopic and global insight capabilities, enabling it to "understand" the structure and evolution of the data ecosystem; while the parameter-efficient model-as-a-service engine 32 provides microscopic and inclusive computing capabilities, enabling it to empower every participant with cutting-edge AI technologies. These two engines support each other and work together to form the powerful "brain" of the system for advanced value discovery and intelligent computing.

[0172] In some embodiments, the immune and compliance layer 40 is used to automatically identify and remove substandard and malicious data within the system through a data immune system; and to provide a dedicated interface with analytical capabilities and data visualization tools through a compliance audit and regulatory interface.

[0173] In some embodiments, see Figure 5The immunity and compliance layer 40 includes: a self-supervised off-distribution data immune system 41 and a penetrating compliance audit and regulatory interface 42.

[0174] The self-supervised out-of-distribution data immune system 41 is used to perform immune checks on any new data asset when it attempts to register into the system.

[0175] The transparent compliance audit and regulatory interface 42 is used to assign different query and monitoring permissions to different regulators, and automatically pulls and visualizes the complete lifecycle of abnormal data transactions from the blockchain when regulators click on them.

[0176] Among them, the self-supervised out-of-distribution data immune system 41 is equipped with a trained encoder-decoder model. The encoder-decoder model is used to reconstruct new data assets and calculate errors, and to perform graded anomaly response when the error is judged to be an anomaly. The graded anomalies include low-risk anomalies and high-risk anomalies.

[0177] In some embodiments, the immune and compliance layer 40: the health and order of the ecosystem.

[0178] A truly mature operating system not only needs powerful functions and high performance, but also the ability to maintain its long-term health, resist external threats, and comply with external rules. The Immune and Compliance Layer 40 plays precisely this role; it is the "defense system" and "legal advisor" of the entire Trusted Geospatial Elements Operating System. Its core mission is to ensure the purity of the data ecosystem, the security of the system, and the compliance of its operations, providing a predictable and sustainable environment for all participants. It mainly consists of two key modules, responsible for internal "proactive immunity" and external "compliance auditing," respectively.

[0179] (a) Self-supervised out-of-distribution data immune system 41: From “passive defense” to “active purification”.

[0180] 1. Role and Position: In any open ecosystem, low-quality, anomalous, or even malicious participants or data are inevitably mixed in. Traditional security systems often rely on rule-based "blacklists" or "firewalls" for passive defense, which is insufficient to deal with unknown and emerging threats. The goal of this component is to build a proactive, self-learning "data immune system" that, like a biological immune system, automatically identifies and eliminates "abnormal pathogens" in the ecosystem—that is, "out-of-distribution" data that differs significantly in distribution from normal data.

[0181] 2. Core Technology Mechanisms and Integrated Innovation: The core idea of ​​this component originates from the innovative application of the "self-supervised reconstruction" technique used in medical image analysis to detect abnormal lesions, to the quality monitoring and security auditing of geospatial data elements. A deep learning model trained on a massive amount of normal samples will exhibit a strong "understanding" and reconstruction ability for normal samples that it is "familiar" with; however, it will appear "helpless" with abnormal samples that it is "unfamiliar" with, and its reconstruction results will inevitably be of low quality.

[0182] 3. Detailed workflow of the "Data Immune System": Training the immune "memory": Building a self-supervised decoder.

[0183] a. Encoder-Decoder Architecture: The system utilizes a deep feature encoder (i.e., a model used to generate "data depth genes") that has already been trained in the featureization and equity layer 20. Based on this, the system additionally constructs a corresponding lightweight decoder.

[0184] The encoder is responsible for compressing the raw input image into deep features, while the decoder is responsible for using these deep features to try to restore the image to its original state.

[0185] b. Self-supervised learning based on normal samples: The system uses a massive amount of filtered, verified "normal" geospatial data samples (e.g., millions of clear, unobstructed satellite images) to train the decoder. The training goal is very simple: to make the reconstructed image output by the decoder as similar as possible to the original input image. Through this process, the decoder gradually "learns" the inherent statistical regularities, texture patterns, and structural features of normal geospatial data, forming a deep "memory" of "normal".

[0186] c. Mandatory Information Bottleneck: To enhance the sensitivity of the immune system, we specifically removed direct "skip connections" between the encoder and decoder in the design. This means that when reconstructing an image, the only information the decoder can rely on is the deep features extracted by the encoder; it cannot "take a shortcut" by directly copying the pixels of the original image. This design forces the decoder to truly "understand" the essence of the data, rather than simply copying pixels.

[0187] Performing an immune "inspection": a reconstruction test of new data.

[0188] a. Mandatory "physical examination": Any new data asset attempting to register with this operating system must undergo a mandatory "physical examination" by this immune system.

[0189] b. Reconstruction and Error Calculation: New data is fed into a pre-trained encoder-decoder model. The system calculates the "reconstruction error" between the original image and the image reconstructed by the model. This error can be measured using various metrics; for example, for ordinary images, "Multi-Scale Structural Similarity (MS-SSIM)" can be used; for more specialized data, pixel-wise norm distance can be used.

[0190] c. Anomaly Detection and Thresholds: The system maintains a dynamic "normal error threshold," which is derived through statistical analysis of the reconstruction errors of a large number of normal samples (e.g., taking the mean plus three standard deviations). If the reconstruction error of a new dataset is significantly higher than this threshold, the immune system will classify it as "out-of-distribution" anomalous data.

[0191] Implementing an immune “response”: risk isolation and labeling.

[0192] Data deemed abnormal will be automatically subject to tiered response measures by the system: Low-risk anomalies (which may simply be slightly poor data quality, such as cloud cover in images): The system will label them as "low-quality," provide risk warnings in transactions, or automatically reduce their contribution weight when participating in collaborative computing such as federated learning.

[0193] High-risk anomalies (which may be forged data, adversarial attack samples, or completely irrelevant data): The system will directly reject their registration or place them in an isolated "observation zone" for manual auditing.

[0194] 4. Summary of Contributions: This component introduces a self-supervised refactoring mechanism, creating a proactive, self-learning, and rule-free "data immune system" for the operating system. It elevates the concept of security protection from addressing known "blacklist" attacks to identifying unknown "abnormal patterns," fundamentally ensuring the purity, high quality, and inherent security of the entire data ecosystem and effectively solving the performance issues of weakly distributed out-of-systems data.

[0195] (II) Penetrating Compliance Audit and Regulatory Interface 42: From “post-event traceability” to “real-time penetration”.

[0196] 1. Role and Position: A successful, large-scale data element operating system must operate transparently and proactively embrace regulation. This component aims to provide national or industry regulatory agencies with a powerful, controllable, and transparent "regulatory cockpit," ensuring that all actions within the system comply with laws, regulations, and industry standards, thereby addressing the long-standing issues of compliance difficulties and high regulatory costs in the data element market.

[0197] 2. Core Technology Mechanisms and Integrated Innovation: The technological foundation of this component is the distributed trusted ledger network 11 of the perception and execution layer 10. All critical operation records are on-chain and immutable, providing "ironclad evidence" for auditing. The innovation of this component lies in the design of a dedicated interface and data visualization tool with "penetrating" analytical capabilities, tailored to regulatory needs, based on a blockchain explorer.

[0198] 3. Functions and workflow of the "Supervisory Cockpit": Role-based access control: Regulatory agencies' access is not unrestricted. The system employs a strict role-based access control (RBAC) mechanism to assign different query and monitoring permissions to different regulatory personnel. For example, financial regulators may only be able to see records related to the value and transactions of data assets, while national security agencies may be granted higher permissions to query the flow of data in specific sensitive areas. All the granting and modification of permissions are themselves on-chain activities that need to be recorded.

[0199] End-to-end traceability: Traditional regulation is often fragmented and piecemeal. This component provides tools that enable "one-click traceability" of a data element or a single transaction. For example, regulators can click on an abnormal data transaction, and the system will automatically retrieve and visualize its entire lifecycle from the blockchain. a. Source tracing: Tracing back to the original ownership record of the data asset to identify its original owner.

[0200] b. Transfer Path: Clearly show the history of all rights and interests in this data asset, including which users nominated and used it through which "asymmetric rights contracts".

[0201] c. Value Relevance: By linking with the "Dynamic Genealogy Engine", other data associated with the data asset can be displayed, helping regulators understand its potential combined application scenarios and risk transmission paths.

[0202] Real-time risk warnings and macro-situation awareness: The regulatory dashboard not only allows for post-event queries but also real-time monitoring and risk warnings. Regulatory agencies can preset sensitive rules, such as "automatically triggering an alarm when the frequency of data transactions within a certain geographical coordinate range suddenly increases by 500% within one hour." Furthermore, through macro-analysis of transaction data across the entire network, the system can also provide regulatory agencies with a macro-situation map of the entire data element market, such as the trading activity of different types of data and average price fluctuations, providing data support for the formulation of sound industrial policies.

[0203] 4. Summary of Contributions: This component, by building a regulatory-oriented, penetrating analysis tool on top of an immutable blockchain ledger, transforms data regulation from a reactive, fragmented, and costly "evidence collection model" to a real-time, end-to-end, and intelligent "patrol model." While ensuring high system transparency and compliant operation, it also significantly reduces societal regulatory costs, providing an indispensable institutional guarantee for building a healthy and orderly data element market.

[0204] This section can be summarized as follows: The immune and compliance layers are the "ballast" that ensures the long-term stability of the operating system. The data immune system ensures the health and purity of the ecosystem, protecting it from the erosion of inferior and malicious data; while the compliance audit and regulatory interface ensures the system's adherence to and transparency to external social order, enabling it to develop healthily within the framework of laws and regulations. These two elements, one internal and one external, together constitute the operating system's powerful "risk control" capability, which is crucial for its eventual widespread trust and acceptance by society.

[0205] See Figure 6 , Figure 6 This is a flowchart illustrating an embodiment of the geospatial data element circulation method based on smart contracts provided in this application. The method is applied to the geospatial data element circulation system provided in the above embodiment, and includes: Step 61: Utilize the perception and execution layers to establish a trust benchmark for the entire system through a distributed trusted ledger network, conduct multi-party secure collaboration through AI-assisted computing nodes, and provide scalable and ultimate secure execution assurance through hardware-level trusted computing endpoints.

[0206] Step 62: By leveraging element-based and rights-based layers, data is endowed with rich connotations and unique identities through deep identity-based methods, enabling it to be deeply understood and recognized by machines; and by binding clear and rigid rights rules to data through asymmetric programmable contracts, enabling it to securely flow and interact in open networks.

[0207] Step 63: Utilize the computation and discovery layer to understand the structure and evolution of the data ecosystem through a dynamic genealogical graph engine; and empower every participant with AI technology through a parameter-efficient Model-as-a-Service engine.

[0208] Step 64: Utilize the immunity and compliance layer to automatically identify and remove substandard and malicious data within the system through the data immune system; and provide dedicated interfaces with analytical capabilities and data visualization tools through compliance audit and regulatory interfaces.

[0209] In some embodiments, steps 61 to 64 can be explained in conjunction with other embodiments of this application, and will not be repeated here.

[0210] In summary, the unexpected technical effects of the smart contract-based geospatial data element circulation method and system of this application are as follows: Specific manifestations in geospatial data circulation: 1. Paradigm Shift 1: A Qualitative Change from “Data Trading Platform” to “Geospatial Element Operating System”.

[0211] Traditional application scenario: A city planning department wants to purchase the latest high-precision remote sensing imagery and mobile signaling data for a certain area to analyze the city's job-housing balance. They need to search, inquire about prices, sign contracts on data trading platforms, and then download the data to their local server. Local GIS engineers and data scientists need to solve a series of problems such as data format conversion, secure storage, and computing environment configuration themselves. If the analysis algorithms they want to use require specific software libraries, they also need to install and debug them themselves. The whole process is cumbersome, inefficient, and carries security risks.

[0212] The specific manifestations of emergent effects: In the "Trusted Geospatial Feature Operating System," this process will be entirely different. When the hardware-level trusted computing endpoint 13 is combined with the "Dynamic Genealogical Map Engine," the workflow of urban planning departments will undergo a revolutionary change: From "buying data" to "calling services": Planning departments no longer need to purchase and download raw data. They can find a "Trusted Application (TA) for Urban Job-Housing Balance Analysis" developed by a third-party geospatial analytics company in the operating system's "app store." They only need to pay for this service call.

[0213] Secure and automated execution: When they click "Run," this "Analysis TA" is securely loaded into their trusted computing endpoint. This TA automatically requests a one-time data usage authorization from the data owner's (such as satellite companies and telecommunications operators) smart contract via the operating system interface.

[0214] Intelligent data scheduling and preprocessing: The operating system's "dynamic genealogical map engine" comes into play here. It already knows that the remote sensing imagery and mobile signaling data have a strong correlation in content. Therefore, when "Analyze TA" is scheduled to execute, the operating system kernel can intelligently preload these two data sets into the cache of trusted endpoints and automatically complete format alignment and coordinate system registration. Developers do not need to write a single line of preprocessing code.

[0215] Results are output, not raw data exposure: the entire analysis process is completed in a hardware-isolated, trusted environment. Ultimately, only the analysis results are output—for example, a heat map of the job-housing imbalance and an analysis report—to the planning department. The raw remote sensing imagery and mobile phone signaling data are never exposed to the planning department's general computing environment.

[0216] An unexpected effect: This is not just about simplifying the transaction process. It signifies a paradigm shift in geospatial data analytics from "data-centric" to "application-centric." Developers can focus on creating valuable analytical applications, while users can access complex data analytics capabilities with a single click, much like using a smartphone app, without worrying about the underlying data storage, security, and processing details. The core of geospatial data circulation has changed from "buying and selling data" to "the instant delivery of data analytics capabilities as a service."

[0217] 2. Paradigm Shift Two: The Upgrade from "Static Data Files" to "Active Programmable Assets".

[0218] Traditional application scenario: An insurance company possesses vector data on the extent of historical flooding across the country. They wish to license this data to a research institution for climate change research and to a real estate company for risk assessment. They would need to sign complex legal contracts with both institutions, detailing the scope, purpose, and duration of use, and rely on the institutions' compliance and subsequent audits to ensure compliance.

[0219] The specific manifestations of emergent effects: When "deep data identity" is combined with "asymmetric programmable rights contracts", this flood data will become a "living" "element intelligent agent" with autonomous management capabilities.

[0220] Endogenous, differentiated equity contracts: Insurance companies will forge two distinct "equity contracts" for this flood data: Contract A (for research institutions): The trusted application of the nominated research institution is designated as the "Nominee". The rights and benefits terms stipulate that it "can only be used for non-commercial climate model training", "is valid for three years", and "can be called an unlimited number of times".

[0221] Contract B (for real estate companies): The real estate company's "risk control TA" is designated as the "nominated party". The terms and conditions stipulate that the contract "can only be used for risk inquiries on commercial land parcels", "charged per inquiry, at RMB 0.1 per inquiry", and "the contract is valid for one year".

[0222] Self-managed access control: When a research institution's trusted application requests access to data, it presents its identity to contract A. Contract A verifies the identity and generates an execution token for model training. Conversely, when a real estate company's application requests access, it must present its identity to contract B and pay a fee before contract B generates an execution token for a single query. Both contracts automatically reject access attempts by an unauthorized third party.

[0223] Intelligent interaction based on "deep genes": An agricultural company wants to assess the impact of flood risk in a certain area on its crop yield. When their "agricultural model TA" queries the operating system, it not only finds the flood data but also automatically discovers another set of "regional historical rainfall data" released by the meteorological department due to the similarity of "deep genes." These two "factor agents" can securely and automatically perform a temporary, cross-institutional combined calculation through their embedded equity contracts, without the need for cumbersome manual coordination among the three parties.

[0224] An unexpected benefit: Data owners are freed from the burden of cumbersome, trust-based contract management. Data compliance no longer relies on legal texts and post-audit, but is automatically enforced beforehand by cryptographically guaranteed, tamper-proof code. More importantly, data itself gains the ability to "autonomously socialize" and "combine value" within a trusted ecosystem, enabling it to securely and spontaneously create new value opportunities with other data elements based on its own inherent appeal.

[0225] 3. Paradigm Shift Three: From "Elite AI R&D" to "Decentralized Collaborative Innovation"

[0226] Traditional application scenario: A small to medium-sized mining company with unique geological exploration data wants to use a state-of-the-art, deep learning-based mineral resource prediction model to analyze its data. However, they lack the capability to train such a large model from scratch and are concerned that uploading their highly confidential exploration data to a third-party cloud platform for analysis might lead to the leakage of trade secrets.

[0227] The specific manifestations of emergent effects: When the "parameter-efficient model-as-a-service engine" and the "LLM-driven federated intelligent collaboration framework" work together, the mining company's predicament is easily resolved.

[0228] Instead of "buying" the model, the company "rented" the capability: they didn't need to purchase or train the entire large model. They downloaded a LoRA adapter, only a few megabytes in size, that came with the "Global Mineral Resources Prediction Base Model" from the operating system's model library to their trusted computing endpoint.

[0229] Secure fine-tuning of data without leaving the domain: They use their own unique, limited geological exploration data to fine-tune and train the LoRA adapter in a completely isolated, locally trusted environment. This process may only require a high-performance workstation and take a few hours. Their most critical data assets never leave their physical control.

[0230] Personalized realization of world-class AI capabilities: After fine-tuning, this small adapter is combined with the vast capabilities of the basic model to form a unique "customized AI model" with extremely high predictive accuracy for the geological features of their region.

[0231] Quantification and Rewards of Contribution: In another scenario, the company participated in a federated learning task initiated by multiple institutions to improve the accuracy of national iron ore reserve predictions. They contributed their data for joint modeling. The LLM coordinator precisely calculated their "contribution score" based on the contribution of their data to the improvement of the final model's accuracy, and automatically allocated a portion of the federated task's revenue (e.g., the fee share for other institutions using the federated model) to them accordingly.

[0232] An unexpected effect: Cutting-edge geospatial AI technology is no longer the exclusive domain of a few giants. Any small or medium-sized enterprise or individual with high-quality, unique scenario data can safely and cost-effectively leverage top-tier AI capabilities to solve their problems. More importantly, for the first time, the value of data can be objectively and accurately quantified and directly translated into economic returns, greatly incentivizing data sharing and value co-creation.

[0233] 4. Paradigm Shift Four: The Emergence of “Passive Security Protection” to “Ecosystem Self-Healing”.

[0234] Traditional application scenario: A geographic information data sharing platform discovers that a user has uploaded a malicious data packet containing a virus, or that a large number of web crawlers are scraping data, causing a decline in the platform's service quality. The platform administrator needs to intervene manually, banning accounts, deleting data, and fixing vulnerabilities. The process is severely delayed and always involves "locking the stable door after the horse has bolted."

[0235] The specific manifestations of emergent effects: When the "self-supervised out-of-distribution data immune system" works in conjunction with the "dynamic phylogenetic engine," the operating system exhibits remarkable self-healing and self-regulating capabilities.

[0236] Active "Immune Cell" Elimination: A hacker, in an attempt to contaminate the data market, meticulously crafted a "Trojan horse" dataset that appeared to be normal city footage but actually contained subtle adversarial perturbations. When this data attempted to register with the system, the self-supervised decoder of the "data immune system," during its reconstruction test, detected reconstruction errors far exceeding normal thresholds. The system immediately classified it as high-risk out-of-distribution data, automatically rejecting its registration and adding its "genetic" characteristics to a temporary "immune watchlist." The entire process was completed in milliseconds, requiring no manual intervention.

[0237] Macro-level regulation by the "central nervous system": At the same time, the "dynamic genealogy engine" detected at the macro level that the transaction volume and query requests for a certain niche data category (e.g., "Antarctic ice sheet subsurface topography data") were suddenly abnormally active. It did not know the specific reason, but this "abnormal signal" would trigger the system's macro-level early warning mechanism.

[0238] Linked Response: The system will combine this macro-level early warning with the micro-level intelligence from the immune system. It may automatically increase the immune review level for all newly registered "Antarctic" related data, requiring them to pass more rigorous reconstruction tests. Simultaneously, it will push a report to the regulatory interface and system administrator, stating: "Abnormal trading activity has occurred in the Antarctic ice sheet data category, potentially indicating speculation or data misuse risks. Furthermore, the micro-immune system has intercepted multiple attempts to inject self-supervised out-of-distribution data into this category, and a thorough investigation is recommended." An unexpected result: the system's security has achieved a leap from "passive response" to "prediction and self-healing." It is no longer a fragile system that requires constant human monitoring, but a complex adaptive system with an inherent homeostatic regulation mechanism. It can automatically eliminate lesions at the micro level, perceive risks at the macro level, and intelligently self-regulate, thereby maintaining long-term health, stability, and order in a constantly changing external environment.

[0239] In the several embodiments provided in this application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0240] If the integrated units in the other embodiments described above are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0241] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A geospatial data element circulation system based on smart contracts, characterized in that, The geospatial data element circulation system includes: The perception and execution layer is used to establish a trust benchmark for the entire system through a distributed trusted ledger network, conduct multi-party secure collaboration through AI-assisted computing nodes, and provide scalable and ultimate secure execution assurance through hardware-level trusted computing endpoints. The element-based and rights-based layer is used to endow data with rich connotations and unique identities through deep identity-based identification, enabling it to be deeply understood and recognized by machines; and to bind clear and rigid rights rules to data through asymmetric programmable contracts, enabling it to flow and interact securely in open networks. The computation and discovery layer is used to understand the structure and evolution of the data ecosystem through a dynamic genealogical graph engine; and to empower every participant with AI technology through a parameter-efficient model-as-a-service engine. The immunity and compliance layer is used to automatically identify and remove substandard and malicious data within the system through a data immune system; and to provide dedicated interfaces with analytical capabilities and data visualization tools through compliance audit and regulatory interfaces.

2. The geospatial data element circulation system according to claim 1, characterized in that, The perception and execution layer includes: A distributed trusted ledger network is used to build a high-performance distributed ledger using consortium blockchain or access-controlled public blockchain technology to record meta-state information with the highest level of trust. The meta-state information includes element identity history, equity contracts and state machines, and key operation and transaction logs. Artificial intelligence collaborative computing nodes are used to build a large model-driven federated intelligent collaboration framework. On the hardware level, they support homomorphic encryption, secure multi-party computation, and various privacy computing protocols. On the software level, a large model, either centrally deployed or maintained by consensus among multiple nodes, plays the role of intelligent commander. Hardware-level trusted computing endpoints are used to build a complete trusted application lifecycle management system, providing each participant with a standardized, secure vault running on their own devices.

3. The geospatial data element circulation system according to claim 2, characterized in that, The functions of the intelligent commander, as portrayed by the large model, include: The contribution assessment and dynamic weighting of intelligent systems; Proactive adversarial training and model hardening; Adaptive privacy-efficiency balance scheduling.

4. The geospatial data element circulation system according to claim 2, characterized in that, The work content of the trusted application lifecycle management system includes: In this operating system, a trusted application is a standardized, securely encapsulated software module that runs in a hardware-enclosed memory space. After developing a new trusted application, perform automated auditing and system re-signing of the new trusted application; When a user needs to use a new feature, the trusted computing endpoint requests the download of the double-signed trusted application from the distribution center; Once downloaded, the viable application is placed into the trusted execution environment of the device hardware.

5. The geospatial data element circulation system according to claim 1, characterized in that, The elementization and rights layer includes: The geospatial data deep identity module is used to perform multi-level feature fusion and deep representation of geospatial data to obtain deep genes, and to perform structure-aware comparative hash encoding on the deep genes to obtain gene digests. The asymmetric programmable contract module for data rights provides contract casting and rights nomination functions, rights activation and verifiable execution functions, and rights non-transferability functions.

6. The geospatial data element circulation system according to claim 1, characterized in that, The computation and discovery layer includes: The data ecosystem dynamic genealogy graph engine is used to measure distances based on the deep genes generated for each data asset according to the elementization and equity layers, and to perform clustering using an incremental update mechanism when a new data asset is registered in the system; as well as to provide assisted intelligent pricing and proactive value discovery and recommendation for each piece of data. A parameter-efficient Model-as-a-Service engine provides a centralized base model and a lightweight LoRA adapter for users to download and train in a local trusted environment. When inference is required, the user's trusted endpoint dynamically combines the output of the centralized base model with the output of the locally trained LoRA adapter.

7. The geospatial data element circulation system according to claim 6, characterized in that, The data ecosystem dynamic genealogy map engine is also used to assist in intelligent pricing for each piece of data by utilizing a geospatial element dynamic pricing model.

8. The geospatial data element circulation system according to claim 1, characterized in that, The immunity and compliance layer includes: A self-supervised off-distribution data immune system is used to perform an immune check on any new data asset when it attempts to register into the system. The transparent compliance audit and regulatory interface is used to assign different query and monitoring permissions to different regulators, and automatically pulls and visualizes the complete lifecycle of abnormal data transactions from the blockchain when regulators click on them.

9. The geospatial data element circulation system according to claim 8, characterized in that, The self-supervised out-of-distribution data immune system is equipped with a trained encoder-decoder model, which is used to reconstruct new data assets and calculate errors, and to perform graded anomaly response when the error is judged to be an anomaly; wherein, graded anomalies include low-risk anomalies and high-risk anomalies.

10. A method for the circulation of geospatial data elements based on smart contracts, characterized in that, The method is applied to the geospatial data element circulation system as described in any one of claims 1-9, and the method includes: By leveraging the perception and execution layers, a trust benchmark for the entire system is established through a distributed trusted ledger network. Multi-party secure collaboration is achieved through AI-assisted computing nodes, and scalable, ultimate secure execution assurance is provided through hardware-level trusted computing endpoints. By leveraging element-based and rights-based layers, data is endowed with rich connotations and unique identities through deep identity-based methods, enabling it to be deeply understood and recognized by machines; and by binding clear and rigid rights rules to data through asymmetric programmable contracts, it has the ability to securely flow and interact in open networks. By leveraging the computation and discovery layers and a dynamic genealogical graph engine, we can understand the structure and evolution of the data ecosystem; and by using a parameter-efficient Model-as-a-Service engine, we can empower every participant with AI technology. The system leverages an immune and compliance layer to automatically identify and remove substandard and malicious data from within the system through a data immune system; and provides dedicated interfaces with analytical capabilities and data visualization tools through compliance audit and regulatory interfaces.

Images