A method and system for full-chain traceability management of geological exploration work surfaces

Abstract

This invention discloses a method and system for full-chain traceability management of geological exploration work surfaces, belonging to the field of geological exploration information technology. The management method sequentially includes multimodal data acquisition and edge preprocessing, blockchain digital imprint generation and storage, data lineage map construction and automatic verification, traceability verification and 3D visualization, with verification results and traceability requests triggering reverse acquisition and review, forming a closed-loop full-chain system. The system includes intelligent acquisition terminals, edge aggregation nodes, a cloud management platform, and a blockchain storage network. This invention effectively solves the problems of traditional exploration data being easily tampered with, inefficient verification, difficult traceability, unreliable offline data, and lack of automatic contract fulfillment, achieving full-cycle, full-chain, reliable, and traceable control of geological exploration work surface data, significantly improving the level of geological exploration quality supervision and intelligent management.

Description

Technical Field

[0001] This invention relates to the field of information technology in geological exploration, specifically to a method and system for full-chain traceability management of geological exploration work surfaces. Background Technology

[0002] The geological exploration work site is the core site of engineering geological exploration. The authenticity of the data collected, the standardization of process control, and the traceability of the results directly determine the quality of engineering exploration and the safety of subsequent engineering construction. Current traditional geological exploration work site management models have many prominent technical shortcomings, making it difficult to meet the intelligent and standardized requirements of modern engineering exploration: The collection of various types of data, such as spatial positioning, on-site images, environmental parameters, and manual recordings, lacks unified collection standards and pre-processing mechanisms. The data is characterized by chaotic timing, excessive redundancy and noise, and inconsistent formats, failing to form a standardized data carrier that accurately represents the real-time status of the work site. Exploration data is mostly stored on centralized servers, lacking unique and reliable identification and tamper-proof evidence. Data can be easily modified or replaced without detection, making it impossible to accurately trace the responsible link and data source in the event of engineering quality problems. The verification of exploration specifications and quality requirements relies entirely on manual work, lacking automated and standardized verification logic. Not only is the verification efficiency low, but it is also prone to omissions and errors due to insufficient human experience. Furthermore, the verification results cannot promptly guide on-site data collection, making it difficult to form a verification-rectification closed loop. The presentation of exploration data is mostly two-dimensional, lacking three-dimensional visualization methods that integrate with the actual site conditions. This results in low efficiency for management personnel to trace and verify data on-site, making it impossible to intuitively grasp the changes in the work area's status throughout its entire lifecycle. Simultaneously, traditional exploration systems operate independently at different levels, with intelligent acquisition terminals, edge nodes, cloud platforms, and blockchain networks unable to collaborate. Offline data from areas without network coverage is difficult to securely retain and reliably upload. Cross-project data cannot be shared and optimized due to privacy protection restrictions. Exploration labor settlement and deliverables rely on manual review, resulting in cumbersome processes, long cycles, and a high risk of disputes. These problems, combined, prevent the formation of a complete closed-loop management system for geological exploration work areas, encompassing data collection, processing, storage, verification, and traceability review. Consequently, data credibility, regulatory efficiency, and the level of intelligent management fail to meet the high-quality development requirements of modern engineering exploration. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for full-chain traceability management of geological exploration work surfaces, solving the problems of easy tampering, inefficient verification, difficulty in traceability, unreliability of offline data, and lack of automatic contract fulfillment in traditional exploration methods.

[0004] To achieve the above objectives, the embodiments of this invention provide the following technical solutions:

[0005] This application provides a method for full-link traceable management of geological exploration work faces, including the following steps: S1, Multimodal data acquisition and edge preprocessing of the work face: A multimodal raw data stream containing spatial positioning data, work image data, environmental sensor data, and manually recorded data is acquired in real time using an intelligent acquisition terminal deployed on the work face. Timestamp alignment, data cleaning, and feature extraction are performed at the edge of the intelligent acquisition terminal to generate a real-time state vector of the work face; S2, Generation and storage of a blockchain-based digital imprint: The real-time state vector of the work face is combined with the unique identifier of the work face, the digital identity of the workers, and timestamp information. A hash algorithm is used to generate a unique digital imprint representing the current state of the work face, and the digital imprint is stored on the blockchain distributed ledger; S3, Construction and automatic verification of the full-link data lineage map of the work face: Based on the digital imprint, the same work face is linked across different exploration sites... All related data generated during the observation phase are used to construct a data lineage map with digital imprints as the root node. The consistency between the stage data and the preset observation specifications is automatically verified when data is entered into the database at each stage, and the verification results are marked. If the verification fails, a rectification task is automatically generated and pushed to the relevant responsible person's terminal. The rectification requirements are also fed back to the intelligent acquisition terminal of S1 in real time to guide on-site adjustments. Simultaneously, the rectification process is recorded as a new node in the lineage map. S4, Source tracing verification and visualization based on digital imprints: In response to source tracing query requests, the corresponding evidence information is retrieved from the blockchain based on the digital imprint of the target work surface. This information is then compared with the work data in the local database using hash comparison. The verification results, the lineage map, and the trajectory of the work surface's full-cycle state changes are displayed in three dimensions. If the verification results are inconsistent, a review of the multimodal raw data stream in S1 is triggered, and abnormal nodes are marked.

[0006] Furthermore, the edge preprocessing in S1 includes: running a lightweight target detection model through an edge computing unit to automatically identify and extract features from the core, sampling location, and cataloging labels in the operational image data. The lightweight target detection model replaces the YOLOv5s CSPDarknet53 backbone network with a lightweight feature extraction network based on ShuffleNet v2, introduces a lightweight channel-spatial attention module in the feature pyramid network, introduces a hollow spatial pyramid pooling module in front of the detection head, and uses a channel-level knowledge distillation strategy to compress and train the original YOLOv5l as the teacher model.

[0007] Furthermore, the generation of the digital imprint in S2 adopts a dynamic hash chain mechanism based on the BLAKE3 hash algorithm: for multiple state vectors v1, v2, ..., vi, ... vn generated at different time points 1, 2, ..., i, ..., n for the same work surface, the initial digital imprint H0 is defined as the hash value of the work surface's unique identifier and the initial timestamp; for the i-th time point, its digital imprint... This forms a time-series hash chain (H0, H1, ..., Hi, ... Hn), where A digital imprint of a previous point in time. Let i be the state vector at time point i. Let i be the timestamp of the i-th time point. This is a data concatenation operator.

[0008] Furthermore, the digital imprints in S2 are uploaded to the blockchain using a batch aggregation signature mechanism based on BLS signatures: within a preset time period, the cloud collects all digital imprints generated by all work surfaces within the preset time period, calculates the hash value of each digital imprint, and the consensus nodes of the consortium blockchain perform BLS signatures on the Merkle root of this batch of hash values. After aggregating all node signatures into a short signature, it is uploaded to the blockchain along with the Merkle root.

[0009] Furthermore, the automatic verification in S3 includes: constructing a knowledge graph of exploration specifications based on the Neo4j graph database, transforming the design requirements, specifications, and quality control points in the exploration plan into computable verification rules stored in the form of "entity-relationship-attribute" triples; triggering a rule engine based on graph pattern matching at each stage of data entry, constructing the data to be verified as a subgraph and performing subgraph isomorphic matching with the rule patterns in the knowledge graph, performing multi-dimensional verification of data integrity, spatial compliance, temporal rationality, and content standardization; if the verification fails, automatically generating a rectification task and pushing it to the relevant responsible person's terminal, recording the rectification process as a new node in the lineage graph, and providing real-time feedback to the acquisition terminal to guide on-site operation adjustments.

[0010] Furthermore, the 3D visualization display in S4 employs an improved ORB-SLAM3 algorithm to achieve augmented reality overlay display. This improved ORB-SLAM3 algorithm includes: introducing multi-sensor tightly coupled initialization, fusing GPS / RTK positioning data to construct a visual-inertial-GPS joint cost function for global optimization, with an initialization time of 3 seconds; introducing a dynamic feature removal mechanism based on a lightweight semantic segmentation network like Fast-SCNN, performing double verification on feature points in dynamic object regions before removal to reduce trajectory errors under dynamic interference; employing a texture-adaptive feature extraction strategy, dynamically adjusting the FAST corner detection threshold based on the image mesh texture complexity, and adding geometric context information for weighted matching based on the ORB descriptor to increase the number of feature points in sparse texture regions; and achieving spatiotemporal alignment of digital imprints with map fusion, inserting digital imprints as virtual nodes in the SLAM map and participating in graph optimization together with visual landmarks to ensure that the spatial alignment error between digital imprints and the 3D map is less than 0.2 meters.

[0011] Accordingly, this application also provides a full-chain traceable management system for geological exploration work surfaces, including: an intelligent acquisition terminal layer: deployed at each exploration work surface, integrating multimodal sensors and edge computing units, used to execute S1 and run the lightweight target detection model; an edge aggregation node layer: deployed at the center of the exploration site area, communicating with multiple intelligent acquisition terminals, used to aggregate and cache real-time status vectors of work surfaces within the area, run local data copies based on a time-series database, and provide local query and verification services when the cloud connection is interrupted; and a cloud management service platform: including a data aggregation management module, a digital imprint generation module, a data lineage map construction module, and an automatic calibration module. The verification engine and visualization service module are used to execute S2 to S4, wherein the visualization service module integrates the improved ORB-SLAM3 algorithm; the blockchain evidence storage network layer consists of a consortium chain composed of multiple exploration participant nodes, deploying smart contracts that support batch aggregation signatures, used to receive and store the digital imprints, and providing tamper-proof evidence storage services and hash comparison verification interfaces; wherein, the intelligent acquisition terminal layer is connected to the cloud management service platform through the edge aggregation node layer to form a forward data acquisition link, and the cloud management service platform is also connected to the intelligent acquisition terminal layer through the command issuance interface to push the problems found by automatic verification to the field terminal in real time to guide the operation adjustment.

[0012] Furthermore, the intelligent data acquisition terminal layer includes an offline operation terminal for areas without network signal coverage. The offline operation terminal has a built-in trusted execution environment and encrypted database based on ARM TrustZone technology. It is used to continuously collect data in offline mode, generate timestamps based on the local clock, run a lightweight hash algorithm to generate digital imprints, and encrypt and store the data and imprints in the isolated area of ​​the trusted execution environment. After the network is restored, the offline data is automatically synchronized and uploaded through the breakpoint resume and data verification protocol, and the on-chain evidence storage is triggered.

[0013] Furthermore, the cloud management service platform integrates a cross-project knowledge transfer engine based on a federated averaging algorithm. The cross-project knowledge transfer engine performs joint modeling with cloud platform instances of multiple projects: each project instance trains anomaly detection models on local data, and only uploads encrypted gradients of the anomaly detection models to the federated server. The federated server aggregates and updates the global model and then distributes it to each project instance and edge node, so as to optimize the anomaly detection capability of automatic verification rules without aggregating the original data.

[0014] Furthermore, the blockchain evidence storage network layer is also deployed with an automatic settlement and performance module driven by smart contracts based on Hyperledger Fabric. The automatic settlement and performance module is used to automatically trigger the settlement of labor services, confirmation of deliverables, and issuance of digital certificates to the survey participants after the digital imprint has completed the full-link closed-loop verification and all stage data has passed the compliance review.

[0015] The beneficial effects of this invention are as follows: By acquiring multimodal data from the work surface and performing edge preprocessing, the temporal consistency, noise removal, and feature extraction of multiple sets of core, multi-source data are achieved, effectively solving the problems of inconsistent data quality and formats in traditional methods. This improves the quality and standardization of raw data, providing a reliable data foundation for subsequent digital imprint generation and evidence storage. Through blockchain-based digital imprint generation and evidence storage, a unique and tamper-proof identity is assigned to the state of the work surface at each moment. Utilizing the decentralized and distributed storage characteristics of blockchain, the authenticity and credibility of the data are guaranteed from the source, preventing illegal tampering and replacement, and ensuring data verifiability. Through data lineage mapping and automatic verification, the correlation and connection of data throughout the entire exploration process are achieved, providing a clear and intuitive representation of the data. The entire lifecycle of geological exploration, from collection, processing, and storage to application, is streamlined, replacing manual verification and enabling automated checks on compliance. This significantly improves verification efficiency and accuracy, reduces errors caused by human intervention, and avoids missed or incorrect detections. Through digital imprint traceability verification and 3D visualization, the traceability process is reliably verified and intuitively displayed. Managers can quickly verify data authenticity and trace data flow, reducing the difficulty of monitoring and verifying exploration data and improving management efficiency. By using verification results and traceability requests to trigger data collection review and rectification, a closed-loop management system is formed, covering the entire data collection process from the source to the final data usage stage. This ensures that the entire geological exploration operation is controllable, verifiable, and traceable, comprehensively improving the intelligence and standardization of geological exploration management. Attached Figure Description

[0016] Figure 1 A flowchart illustrating a method for end-to-end traceability management of geological exploration work surfaces, provided in an embodiment of this application;

[0017] Figure 2 This is a schematic diagram of a full-chain traceable management system for geological exploration work surfaces, provided as an embodiment of this application. Detailed Implementation

[0018] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0019] In this invention, the terms "system" and "network" are used interchangeably. "Multiple" refers to two or more; therefore, in this invention, "multiple" can also be understood as "at least two." "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / ", unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, it should be understood that in the description of this invention, terms such as "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance or order.

[0020] Existing geological exploration workface management suffers from a lack of unified acquisition entry and edge preprocessing procedures for multi-source heterogeneous data. The four core data types—spatial positioning, work images, environmental sensing, and manual recording—cannot achieve temporal alignment and feature standardization, making it difficult to form a data carrier that can uniformly represent the workface status. This leads to problems in subsequent data processing and evidence preservation. Furthermore, data generated at different exploration stages (preliminary exploration, detailed exploration, sampling, testing, etc.) are isolated from each other, making it impossible to build a correlation relationship throughout the data lifecycle. Exploration standard verification can only be completed manually, resulting in a high probability of missed or incorrect detections. Moreover, the verification results cannot be fed back to the acquisition end in a timely manner, failing to guide adjustments in on-site operations.

[0021] Example 1

[0022] like Figure 1As shown, this application provides a method for full-link traceable management of geological exploration work faces, including the following steps: S1, Multimodal data acquisition and edge preprocessing of the work face: Through intelligent acquisition terminals deployed on the work face, multimodal raw data streams containing spatial positioning data, work image data, environmental sensor data, and manually recorded data are acquired in real time, and timestamp alignment, data cleaning, and feature extraction are performed at the edge of the intelligent acquisition terminal to generate a real-time state vector of the work face; S2, Generation and storage of digital imprints based on blockchain: The real-time state vector of the work face is combined with the unique identifier of the work face, the digital identity of the workers, and timestamp information to generate a unique digital imprint representing the current state of the work face through a hash algorithm, and the digital imprint is stored on the blockchain distributed ledger; S3, Construction and automatic verification of the full-link data lineage map of the work face: Based on the digital imprint, the same work face is linked together to form a data lineage map. All related data generated during the exploration phase are used to construct a data lineage map with digital imprints as the root node. The consistency between the stage data and the preset exploration specifications is automatically verified when data is entered into the database at each stage, and the verification results are marked. If the verification fails, a rectification task is automatically generated and pushed to the relevant responsible person's terminal. The rectification requirements are also fed back to the intelligent acquisition terminal of S1 in real time to guide on-site adjustments. Simultaneously, the rectification process is recorded as a new node in the lineage map. S4, Source tracing verification and visualization based on digital imprints: In response to source tracing query requests, the corresponding evidence information is retrieved from the blockchain based on the digital imprint of the target work surface. This information is then compared with the work data in the local database using hash comparison. The verification results, the lineage map, and the trajectory of the work surface's full-cycle state changes are displayed in three dimensions. If the verification results are inconsistent, a review of the multimodal raw data stream in S1 is triggered, and abnormal nodes are marked.

[0023] In another possible embodiment, corresponding intelligent data acquisition terminals are first deployed at various work sites within the geological exploration operation. These intelligent data acquisition terminals are dedicated devices integrating various data acquisition components and edge computing functions. These terminals are primarily dedicated devices obtained by coupling multiple data acquisition components from existing technologies, and therefore will not be elaborated upon here. These intelligent data acquisition terminals can acquire multimodal raw data in real time that comprehensively reflects the state of the work surface. Spatial positioning data is used to determine the actual geographical location of the work surface; operational image data is used to record real-time scenes of drilling, core sampling, and on-site operations; and environmental sensing data is used to collect data from the surrounding environment of the work surface. Environmental parameters such as temperature, humidity, vibration, and tilt angle are collected. Manually compiled data includes geological stratification, lithological descriptions, and sampling information entered by on-site personnel based on actual conditions. After collection, the raw data does not need to be directly uploaded to the cloud. Instead, the various data types are processed uniformly at the edge of the intelligent acquisition terminal. This edge-side processing is a lightweight local processing mechanism, independent of cloud computing power. First, all data is timestamped according to the collection time to ensure different data types correspond to the same collection time. Then, a preset algorithm removes redundant information, abnormal noise, and invalid data, completing the data processing. The data is cleaned, and then feature information representing the core state of the work face is extracted from the cleaned data, such as the morphological characteristics of the rock core, the spatial location characteristics of the work face, and the variation characteristics of environmental parameters. This results in the generation of a real-time state vector that uniformly represents the real-time situation of the work face. This real-time state vector serves as a carrier for transforming multi-source heterogeneous data into standardized data, facilitating subsequent storage, computation, and comparison. Following this, a blockchain-based digital signature generation and notarization step is performed. This involves combining the previously generated real-time state vector with a unique identifier assigned to each work face (a unique identity code for each work face, used to distinguish different work faces) and the unique digital identity of each worker. The information (electronic identity credentials of workers, used to trace work responsibility) and the timestamp information of data collection are combined, and the combined information is encrypted and calculated using a hash algorithm to generate a digital imprint that uniquely corresponds to the current state of the work surface. This digital imprint is equivalent to an "electronic ID card" for the state of the work surface at a certain moment, which can uniquely identify all the state data of the work surface at that moment. After the digital imprint is generated, it is immediately uploaded and stored in the blockchain distributed ledger. The blockchain distributed ledger is a database jointly maintained by multiple nodes. There is no single control node, which can ensure that the data cannot be tampered with once it is stored, thereby completing the trusted evidence storage of the data.Next, a complete data lineage map of the work area is constructed and automatically verified. Using the generated digital imprint as the core root node, all related data generated at different exploration stages (preliminary exploration, detailed exploration, sampling, testing, logging, etc.) of the same work area are sequentially linked to construct a data lineage map that completely records the data source, flow, and changes. This data lineage map is like a "data family tree," intuitively showing the source, processing, and relationships of each data point. When data is entered into the database at each stage, the system automatically performs consistency checks according to preset geological exploration standards, verifying the data's completeness (whether there are missing items), spatial coordinate rationality (whether it meets exploration design requirements), temporal logic correctness (whether data from different times is continuous), and content standardization (whether it conforms to the standard description). The system clearly marks the results of passing or failing the checks and records the verification process. Finally, the system performs source tracing verification and visualization based on the digital imprint. When a source tracing query request is received from a user (such as a manager or supervisor), the system uses the digital imprint corresponding to the target work area as the retrieval basis, from the area... The system retrieves the corresponding evidence information from the blockchain and then compares it with the job data stored in the local database using a hash comparison algorithm. If the hash values ​​match, the data has not been tampered with; otherwise, it has been tampered with. This verifies the authenticity and integrity of the data. After verification, the hash comparison results, the constructed data lineage map, and the entire lifecycle of the job from initial data collection to final completion are visually displayed in 3D. 3D visualization uses 3D modeling and augmented reality technologies to transform abstract data into intuitive 3D images for easy viewing. Simultaneously, failed results marked during the verification process and abnormal requests discovered during source tracing trigger the system to issue review or rectification instructions to the data collection end. These instructions clearly inform on-site personnel of the content requiring review and the rectification requirements, guiding adjustments and corrections to the on-site data collection operation. This forms a closed-loop management system from the data collection source to the final data usage stage, ensuring that every piece of data is traceable, verifiable, and rectifiable.

[0024] By acquiring multimodal data from the work site and performing edge preprocessing, the temporal consistency, noise removal, and feature extraction of multiple sets of core, multi-source data are achieved. This effectively solves the problems of inconsistent data quality and formats in traditional methods, improving the quality and standardization of raw data and providing a reliable data foundation for subsequent digital imprint generation and storage. Through blockchain-based digital imprint generation and storage, a unique and tamper-proof identity is assigned to the state of the work site at each moment. Leveraging the decentralized and distributed storage characteristics of blockchain, the authenticity and credibility of the data are guaranteed from the source, preventing illegal tampering and replacement, and ensuring data verifiability. Through data lineage mapping and automatic verification, the correlation and connection of data throughout the entire exploration process are realized, intuitively demonstrating the data's progression from acquisition to... The entire lifecycle of data processing, storage, and application is automated, replacing manual verification and enabling automatic checks on compliance. This significantly improves verification efficiency and accuracy, reduces errors caused by human intervention, and avoids missed or incorrect checks. Through digital imprint traceability verification and 3D visualization, the traceability process is reliably verified and intuitively displayed. Managers can quickly verify data authenticity and trace data flow links, reducing the difficulty of supervision and verification of exploration data and improving management efficiency. By using verification results and traceability requests to trigger data collection review and rectification, a closed-loop management system is formed from the source of data collection to the final data use stage. This ensures that the entire process of geological exploration operations is controllable, verifiable, and traceable, comprehensively improving the intelligence and standardization of geological exploration management.

[0025] Traditional image recognition often employs heavy-duty deep learning models with large parameter counts and high computational consumption. However, the hardware computing power of edge acquisition terminals is limited, making it impossible to support the real-time operation of heavy-duty models. This results in high data processing latency, low recognition efficiency, and even the inability to complete real-time recognition. Conventional YOLOv5s models have complex backbone network structures, insufficient feature extraction efficiency, and lack targeted attention enhancement mechanisms. This leads to low recognition accuracy for small targets such as core samples, sampling locations, and cataloging labels, making them prone to omissions and errors. Furthermore, these models are not lightweight and compressed, resulting in large file sizes that are unsuitable for the storage and operating environments of edge terminals. This makes it difficult to achieve automatic recognition and feature extraction of key information from operational images, requiring manual annotation and extraction of image data, which is inefficient and prone to errors.

[0026] In this embodiment of the application, the edge preprocessing in S1 includes: running a lightweight target detection model through an edge computing unit to automatically identify and extract features from the core, sampling location, and cataloging labels in the operational image data. The lightweight target detection model replaces the YOLOv5s CSPDarknet53 backbone network with a lightweight feature extraction network based on ShuffleNet v2, introduces a lightweight channel-spatial attention module in the feature pyramid network, introduces a hollow spatial pyramid pooling module in front of the detection head, and uses a channel-level knowledge distillation strategy to compress and train the original YOLOv51 as the teacher model.

[0027] In another possible embodiment, during the multimodal data acquisition and edge preprocessing process at the work surface, the edge computing unit, as the core processing component on the terminal side, is responsible for running a specially optimized lightweight target detection model. This edge computing unit is a small computing module integrated within the intelligent acquisition terminal, capable of rapid local data processing without relying on cloud computing power. This lightweight model is based on the YOLOv5s architecture and undergoes targeted optimization. Firstly, the computationally intensive and structurally complex CSPDarknet53 backbone network in the original model is replaced with the lightweight ShuffleNet v2 feature extraction network. The v2 lightweight feature extraction network significantly reduces the model's parameter and computational load without sacrificing core feature extraction capabilities through grouped convolutions and channel shuffling, making it adaptable to edge computing environments. Simultaneously, a lightweight channel-spatial attention module is introduced into the model's feature pyramid network. This module automatically filters channel features and spatial regions in the image that are crucial for recognition results, allowing the model to focus on key targets such as core samples, sampling locations, and catalog labels, while automatically ignoring irrelevant background and debris. A dilated spatial pyramid pooling module is introduced before the model's detection head. This module captures feature information from different receptive fields through dilated convolutions with varying dilation rates, achieving… Multi-scale feature fusion extraction addresses the uneven recognition of targets of different sizes by traditional models. During model training, a channel-level knowledge distillation strategy is employed, using a similar model with higher accuracy and larger scale as a teacher model. The learning ability and recognition experience of the teacher model are transferred to the lightweight model, completing model compression and optimization. After training, the model is deployed in an edge computing unit, which can directly process the collected operational image data in real time, automatically identify key targets such as cores, sampling locations, and cataloging labels in the images, and accurately extract corresponding feature information, such as the size and shape of the core, the coordinates of the sampling location, and the content of the cataloging labels. This provides reliable image feature support for subsequent data processing and verification. The entire process requires no manual intervention, realizing the automation and intelligence of image data processing.

[0028] By replacing the original backbone network with the ShuffleNet v2 lightweight feature extraction network, the model's parameter count and computational power consumption are effectively reduced, enabling it to perfectly adapt to the hardware environment of edge terminals and ensuring smooth operation at the edge, avoiding processing delays and stuttering. By introducing a lightweight channel-spatial attention module into the feature pyramid network, the model can automatically focus on key areas such as core samples, sampling locations, and cataloging labels in the image, weakening the interference of irrelevant background information and significantly improving the feature extraction accuracy of key targets, reducing omissions and errors. By introducing a hollow spatial pyramid pooling module before the detection head, multi-scale feature fusion extraction is achieved, enhancing the model's ability to recognize targets of different sizes and shapes, enabling accurate recognition of everything from small cataloging labels to larger core samples and sampling areas. By employing a channel-level knowledge distillation strategy to perform compressed training based on a large model, the model size is further reduced while maintaining recognition accuracy, decreasing the model's storage and computational power requirements at the edge terminal. Ultimately, this achieves real-time, high-precision automatic recognition and feature extraction of key targets in operational images at the edge, replacing manual operation and improving data processing efficiency and accuracy.

[0029] Traditional hash-based evidence storage only performs independent hash calculations on single data points, resulting in no temporal correlation between work surface status data at different times. This forms a single-point evidence storage model, where even if the state vector at a certain moment is illegally tampered with, it cannot be quickly detected through chain relationships. Tampering is highly concealed, and the credibility and traceability of the evidence cannot be guaranteed. At the same time, single-hash evidence storage cannot reflect the temporal changes of data and cannot trace the dynamic changes in the work surface status, making it difficult to meet the high credibility and traceability requirements of geological exploration data. Once a data dispute occurs, it is impossible to reconstruct the true trajectory of data changes through the evidence storage information.

[0030] In this embodiment, the generation of the digital imprint in step S2 adopts a dynamic hash chain mechanism based on the BLAKE3 hash algorithm: for multiple state vectors v1, v2, ..., vi, ... vn generated at different time points 1, 2, ..., i, ..., n for the same work surface, the initial digital imprint H0 is defined as the hash value of the unique identifier of the work surface and the initial timestamp; for the i-th time point, its digital imprint... This forms a time-series hash chain (H0, H1, ..., Hi, ... Hn), where A digital imprint of a previous point in time. Let i be the state vector at time point i. Let i be the timestamp of the i-th time point. This is a data concatenation operator.

[0031] In another possible embodiment, an initial digital imprint is first defined for each work surface. This initial digital imprint is generated by combining the unique identifier of the work surface with the timestamp of the initial data collection moment, and then performing an encrypted calculation using the BLAKE3 hash algorithm. The initial digital imprint is the starting point of the entire hash chain and is used to identify the initial state of the work surface. Subsequently, when generating new real-time state vectors at different data collection moments of the work surface, independent hash calculations are no longer performed. Instead, the digital imprint generated at the previous moment, the current real-time state vector, and the current data collection timestamp are combined and again encrypted using the BLAKE3 hash algorithm to generate the digital imprint for the current moment. This method iteratively calculates and forms a continuous time-series hash chain. Each digital imprint in this hash chain is closely associated with the previous imprint, forming an inseparable temporal relationship. When verifying data, it is only necessary to check the consistency of each digital imprint in sequence according to the hash chain. If the state vector at a certain moment is illegally tampered with, its corresponding digital imprint will change, which will cause all subsequent digital imprints to be inconsistent with the information stored on the chain. This allows for a quick determination that the data has been tampered with. At the same time, through the temporal relationship of the hash chain, the dynamic changes of the work surface status can be clearly traced, and the work surface status data at each moment can be restored, providing a reliable basis for data traceability and responsibility determination.

[0032] By employing the BLAKE3 hash algorithm to generate digital imprints, compared to traditional hash algorithms, the speed and security of hash calculations are significantly improved. While ensuring data encryption, it also enhances data processing efficiency, meeting the need for rapid generation of digital imprints across multiple time points and work areas. By constructing a dynamic hash chain mechanism, digital imprints from different time points on the same work area form a tight temporal correlation. Each subsequent digital imprint depends on the imprint generated at the previous time point, ensuring that any alteration of the state vector at any time point will invalidate all subsequent digital imprints. This allows for rapid detection of data tampering and eliminates the concealment of such actions. Simultaneously, the dynamic hash chain can completely record the temporal changes in the work area's state. Through the hash chain, the entire cycle of state changes from initial collection to final formation of the work area can be traced, significantly improving the tamper-proof nature, temporal traceability, and overall credibility of data evidence, providing a reliable basis for data traceability and responsibility determination.

[0033] Traditional digital imprinting on the blockchain uses a single, independent on-chain method. Each digital imprint requires a separate signature verification by a consensus node in the blockchain consortium chain. This signature processing is inefficient and consumes a lot of computing power. When multiple work areas generate digital imprints simultaneously and there is a concurrent on-chain demand, it can easily cause blockchain network congestion and affect on-chain efficiency. At the same time, the on-chain storage of a single signature and data consumes a lot of on-chain storage resources. As the exploration work progresses, the number of digital imprints continues to increase, which will significantly increase the cost of evidence storage and system load. Furthermore, subsequent data verification requires retrieving each individual digital imprint, resulting in low verification efficiency and failing to meet the evidence storage and verification needs of large-scale exploration operations.

[0034] In this embodiment of the application, the digital imprints on the blockchain in S2 adopt a batch aggregation signature mechanism based on BLS signature: within a preset time period, the cloud collects all digital imprints generated by all work surfaces within the preset time period, calculates the hash value of each digital imprint, and the consensus nodes of the consortium blockchain perform BLS signatures on the Merkle root of this batch of hash values. After aggregating all node signatures into a short signature, it is uploaded to the blockchain along with the Merkle root.

[0035] In another possible embodiment, during the digital imprint on-chain storage stage, the cloud platform collects digital imprints generated from all exploration work surfaces within a preset time period (set according to the collection frequency of the exploration operation to ensure both batch data processing and data real-time performance). Each collected digital imprint is hashed to obtain its corresponding hash value. Then, all hash values ​​are constructed into a Merkle tree according to the Merkle tree construction rules. A Merkle tree is a hash binary tree that, through layer-by-layer hash calculations, ultimately generates a unique Merkle root, which represents the hash information of all digital imprints. Finally, the hash information is processed by the consensus nodes in the blockchain consortium chain. The Merkle root is then subjected to BLS signing. BLS signing is a high-efficiency digital signature algorithm that can aggregate multiple signatures. The signatures generated by all nodes are aggregated into a short aggregate signature, which represents the verification results of all consensus nodes, ensuring the security of batch notarization. Finally, the aggregate signature and the Merkle root are uploaded to the blockchain for storage, completing the on-chain notarization of batch digital imprints. When verifying data in the future, only the Merkle root and aggregate signature stored on the chain need to be retrieved. The integrity and authenticity of the batch digital imprints can be verified through the Merkle root, without having to retrieve each individual digital imprint, which greatly improves verification efficiency and reduces the occupation of on-chain storage resources, thus lowering the notarization cost.

[0036] By employing a batch aggregation signature mechanism using BLS signatures, signatures of multiple digital imprints within the same time period can be merged into a single short signature, significantly reducing the computational power consumption and network data transmission volume of signature processing, improving signature processing efficiency, and avoiding network congestion caused by concurrent on-chain storage. By using Merkle root on-chain storage, only the root hash value of the batch of digital imprints is uploaded to the chain, eliminating the need to upload each digital imprint individually, significantly reducing on-chain storage requirements and saving on-chain storage costs. Through the aggregation signature of consensus nodes in the consortium blockchain, the security and credibility of batch notarization are guaranteed, ensuring that the data uploaded in batches is not tampered with or forged. Furthermore, during subsequent data verification, the batch data can be verified simply through the Merkle root, eliminating the need to retrieve individual digital imprints, significantly improving verification efficiency and meeting the notarization and verification needs of large-scale exploration operations.

[0037] Traditional geological exploration data verification relies entirely on manual review against exploration standards. The abstract standards cannot be translated into calculable logical rules. Verification of multiple dimensions, such as data integrity, spatial compliance, temporal rationality, and content standardization, depends solely on manual experience. This not only results in low verification efficiency but also a high risk of missed or incorrect checks. When verification fails, rectification tasks cannot be automatically generated and pushed to relevant personnel. The rectification process is unrecorded and untraceable, and the results cannot be fed back to the data acquisition station in real time to guide on-site operations. Furthermore, manual verification lacks standardized procedures; inconsistent verification standards among different personnel lead to poor consistency in results. This hinders the formation of a closed-loop process for verification, rectification, and feedback, ultimately failing to guarantee the standardization and integrity of the exploration data.

[0038] In this embodiment, the automatic verification in S3 includes: constructing a knowledge graph of exploration specifications based on the Neo4j graph database, converting the design requirements, specifications, and quality control points in the exploration plan into computable verification rules stored in the form of "entity-relationship-attribute" triples; triggering a rule engine based on graph pattern matching at each stage of data entry, constructing the data to be verified as a subgraph and performing subgraph isomorphic matching with the rule patterns in the knowledge graph, performing multi-dimensional verification of data integrity, spatial compliance, temporal rationality, and content standardization; if the verification fails, automatically generating a rectification task and pushing it to the relevant responsible person's terminal, recording the rectification process as a new node in the lineage graph, and providing real-time feedback to the acquisition terminal to guide on-site operation adjustments.

[0039] In another possible embodiment, during the automatic verification of work surface data, a knowledge graph of exploration specifications is first built based on the Neo4j graph database. This knowledge graph decomposes the design requirements, national and industry standards, and quality control points in the geological exploration plan into triples composed of "entity-relationship-attribute" pairs and stores them. Entities represent specific objects in the exploration operation, such as boreholes, cores, sampling points, and personnel; relationships represent the constraint logic between different entities, such as the positional relationship between boreholes and sampling points, and the correspondence between cores and geological strata; attributes represent specific specification requirements, such as borehole depth requirements and core logging specifications. In this way, abstract specifications are transformed into a verification rule base that the system can recognize and compute. When exploration data from each stage is entered into the database, the system automatically triggers a rule engine based on graph pattern matching. The rule engine is the core module responsible for executing verification rules in the system, automatically reading the triple rules in the verification rule base and simultaneously processing the data to be verified. The data is constructed into corresponding data subgraphs in the same format, which are the relationship graphs of the data to be verified. Then, the data subgraphs are matched with the rule patterns in the exploration specification knowledge graph. Subgraph isomorphic matching determines whether the data subgraph completely matches the rule pattern, thereby completing the automatic verification of the data from multiple dimensions such as data integrity, spatial coordinate compliance, temporal logic rationality, and content entry standardization. After verification, the system will clearly mark the corresponding results. If the verification passes, the data is normally stored in the database. If the verification fails, the system will automatically generate a rectification task containing information such as rectification requirements, rectification deadlines, and responsible persons, and push it directly to the work terminals of the relevant responsible persons. At the same time, the rectification process and rectification results are added as new nodes to the data lineage graph to ensure that the rectification process is traceable. Rectification information will also be sent to the work face acquisition terminal in real time to guide on-site personnel to adjust the acquisition operation in a timely manner, such as supplementing missing data and correcting erroneous compilation content, to ensure that the data collected subsequently meets the specification requirements, forming a closed loop of verification, rectification, and feedback.

[0040] By constructing a knowledge graph of exploration specifications based on the Neo4j graph database, abstract exploration design requirements, national and industry standards, and quality control points are transformed into computable structured triple rules. This digitizes and standardizes exploration specifications, enabling the system to automatically identify and execute verification rules, replacing manual verification. A rule engine based on graph pattern matching performs subgraph isomorphic matching, achieving multi-dimensional automatic data verification. This simultaneously checks data integrity, spatial compliance, temporal rationality, and content standardization, significantly improving verification accuracy and efficiency and avoiding omissions and errors in manual verification. Furthermore, by automatically generating rectification tasks and pushing them to the responsible personnel's terminals, while recording the rectification process and results in the data lineage graph, the rectification process is traceable. Rectification information can also be fed back to the acquisition terminal in real time, guiding on-site personnel to adjust acquisition operations promptly. This forms a closed-loop management system encompassing verification, rectification, and feedback, ensuring that exploration data meets specifications and improving data quality.

[0041] Traditional 3D visualization relies solely on a single visual SLAM algorithm without integrating GPS / RTK positioning data, resulting in insufficient global positioning accuracy, easy positional shifts, and inability to accurately match the actual work site. The algorithm cannot eliminate feature interference from dynamic objects such as personnel and equipment on site, and the movement of dynamic objects will lead to large feature matching errors, affecting the accuracy of visualization. For different texture areas on site (such as weakly textured rock surfaces and strongly textured vegetation areas), it cannot adaptively adjust the feature extraction strategy, resulting in insufficient feature extraction in weakly textured areas and redundant features in strongly textured areas, leading to unstable feature extraction.

[0042] In this embodiment, the 3D visualization display in step S4 uses an improved ORB-SLAM3 algorithm to achieve augmented reality overlay display. The improved ORB-SLAM3 algorithm includes: introducing multi-sensor tightly coupled initialization, fusing GPS / RTK positioning data to construct a visual-inertial-GPS joint cost function for global optimization, with an initialization time of 3 seconds; introducing a dynamic feature removal mechanism based on a lightweight semantic segmentation network Fast-SCNN, performing double verification on feature points in dynamic object regions before removal to reduce trajectory errors under dynamic interference; adopting a texture adaptive feature extraction strategy, dynamically adjusting the FAST corner detection threshold according to the image mesh texture complexity, and adding geometric context information to the ORB descriptor for weighted matching to increase the number of feature points in sparse texture regions; and achieving spatiotemporal alignment of digital imprints and map fusion, inserting digital imprints as virtual nodes in the SLAM map and participating in graph optimization together with visual landmarks to ensure that the spatial alignment error between digital imprints and the 3D map is less than 0.2 meters.

[0043] In another possible embodiment, during the 3D visualization process, an improved ORB-SLAM3 algorithm is specifically used to achieve augmented reality overlay display. ORB-SLAM3 is a vision-based simultaneous localization and mapping (SLAM) algorithm capable of real-time positioning and map generation. This technology has been optimized in several aspects. First, the algorithm undergoes multi-sensor tight coupling initialization optimization. Visual acquisition information (including work surface images), inertial measurement data (including equipment attitude and acceleration data), and GPS / RTK spatial positioning data (including the precise geographical location of the work surface) collected by the intelligent acquisition terminal are deeply fused to construct a vision-inertial-GPS joint cost function. This joint cost function is used to globally optimize the positioning data, solving the problem of insufficient positioning accuracy from a single sensor, ensuring positioning accuracy, and avoiding positional shifts in the visualization. Simultaneously, a lightweight Fast-SCNN semantic segmentation network is introduced. This network is a lightweight image segmentation algorithm that can quickly identify dynamic object regions in the work images, such as on-site workers and drilling equipment. The identified dynamic region feature points are double-verified (verified using two different algorithms) before being removed to avoid... The algorithm improves stability and visualization accuracy by matching the moving interference features of dynamic objects. It also employs a texture-adaptive feature extraction strategy. The system automatically analyzes the texture complexity of different grid regions in the image. For sparse, weak-texture regions (such as smooth rock surfaces), the FAST corner detection threshold is appropriately lowered to ensure sufficient feature point extraction. For dense, strong-texture regions (such as vegetation-covered areas), the FAST corner detection threshold is appropriately increased to avoid extracting too many redundant feature points. Furthermore, geometric context information is added to the ORB descriptor to further refine the feature point extraction process. Weighted matching is performed to improve the accuracy of feature matching. Finally, digital imprints are inserted as virtual nodes into the SLAM map and participate in graph optimization together with visual landmarks (i.e., feature points extracted from images). This achieves precise spatiotemporal alignment between digital imprints and the real-world map. The work surface status data corresponding to the digital imprints will be superimposed on the real-world image in augmented reality form. Managers can intuitively see the real-world image of the work surface, digital imprint information, data lineage map, and the trajectory of status changes throughout the entire cycle through a visualization terminal. This achieves accurate and intuitive visualization of the work surface status throughout the entire cycle, facilitating rapid traceability and verification.

[0044] By introducing multi-sensor tight coupling initialization, visual acquisition information, inertial measurement data, and GPS / RTK spatial positioning data are deeply fused to construct a vision-inertial-GPS joint cost function and complete global optimization, significantly improving global positioning accuracy, avoiding positional offset, and ensuring accurate matching between the visualized image and the actual work site. By introducing a lightweight semantic segmentation network, Fast-SCNN is used to achieve dynamic feature removal, which can accurately identify and remove feature interference from dynamic objects such as personnel and equipment on site, improving the stability of the algorithm and the accuracy of visualization. By adopting a texture adaptive feature extraction strategy, the feature detection threshold is dynamically adjusted according to the texture complexity of different grid areas in the image, optimizing the feature extraction and matching effect under different texture scenes, ensuring stable and accurate feature extraction in both weak and strong texture areas. By realizing spatiotemporal alignment of digital imprints and map fusion, digital imprints are integrated into the SLAM map as virtual nodes, ensuring the accuracy of augmented reality overlay display, allowing digital information to be perfectly combined with the real scene, realizing an intuitive and accurate visualization of the entire life cycle status of the work site, and improving the efficiency and accuracy of traceability and verification.

[0045] Example 2

[0046] Reference Figure 2 This application also provides a full-chain traceable management system for geological exploration work surfaces, including: an intelligent acquisition terminal layer: deployed at each exploration work surface, integrating multimodal sensors and edge computing units, used to execute S1 and run the lightweight target detection model; an edge aggregation node layer: deployed at the center of the exploration site area, communicating with multiple intelligent acquisition terminals, used to aggregate and cache real-time status vectors of work surfaces within the area, run local data copies based on a time-series database, and provide local query and verification services when the cloud connection is interrupted; and a cloud management service platform: including a data aggregation management module, a digital imprint generation module, a data lineage map construction module, and an automatic calibration module. The verification engine and visualization service module are used to execute S2 to S4, wherein the visualization service module integrates the improved ORB-SLAM3 algorithm; the blockchain evidence storage network layer consists of a consortium chain composed of multiple exploration participant nodes, deploying smart contracts that support batch aggregation signatures, used to receive and store the digital imprints, and providing tamper-proof evidence storage services and hash comparison verification interfaces; wherein, the intelligent acquisition terminal layer is connected to the cloud management service platform through the edge aggregation node layer to form a forward data acquisition link, and the cloud management service platform is also connected to the intelligent acquisition terminal layer through the command issuance interface to push the problems found by automatic verification to the field terminal in real time to guide the operation adjustment.

[0047] In another possible embodiment, this system adopts an end-edge-cloud-chain collaborative architecture to achieve end-to-end traceable management of geological exploration work surfaces. First, there is the intelligent acquisition terminal layer, deployed at each geological exploration work surface. Each work surface deploys at least one intelligent acquisition terminal according to requirements. The terminal integrates multimodal sensors and edge computing units. The multimodal sensors are used to acquire four core types of data: spatial positioning, work images, environmental sensing, and manual recording. The edge computing unit is used to perform edge preprocessing and lightweight target detection, responsible for real-time acquisition and preliminary processing of multimodal data, while simultaneously running a lightweight target detection model to automatically identify key targets in the images, providing a foundation for subsequent data processing. Secondly, there is the edge... The aggregation node layer, deployed at the regional center of the survey site (e.g., the on-site project headquarters), establishes stable communication connections with multiple intelligent data acquisition terminals. It is responsible for aggregating and caching real-time status vectors from all work surfaces within the area, avoiding network pressure caused by directly uploading data to the cloud. Simultaneously, it builds a local data copy based on a time-series database. This database is specifically designed for storing time-series data, enabling efficient storage and querying of work surface data at different times. When the cloud network connection is interrupted, this layer can directly provide local data query and verification services to the site, ensuring uninterrupted on-site operations. Once the network is restored, the cached data is synchronized to the cloud. Following this is the cloud management service platform, which serves as the core of the system, integrating data... The platform comprises modules for data aggregation and management, digital imprint generation, data lineage mapping construction, automatic verification engine, and visualization services. These modules are responsible for executing core processing logic such as digital imprint generation, data lineage mapping construction, automatic verification, and traceability verification. The visualization service module integrates an improved ORB-SLAM3 algorithm to achieve 3D visualization of work surface data. The platform can receive data uploaded from edge aggregation nodes, complete core processing, push the digital imprint to the blockchain, and simultaneously receive blockchain-stored evidence information for traceability verification. Finally, there is the blockchain evidence storage network layer. This layer consists of a consortium blockchain composed of nodes from various relevant entities, including surveying units, supervision units, and construction units. Only authorized nodes are allowed to participate in the consortium blockchain to ensure data security. In addition to privacy, smart contracts supporting batch aggregation signatures are deployed. These smart contracts are automated execution programs on the blockchain that can automatically complete the reception, storage, and verification of digital imprints. They are responsible for receiving and storing digital imprints pushed from the cloud, providing tamper-proof evidence storage services and hash comparison verification interfaces for the cloud platform to retrieve for traceability verification. During collaborative work, the intelligent acquisition terminal layer connects to the cloud management service platform through the edge aggregation node layer, forming a forward data acquisition link from the field to the cloud, ensuring that data can be uploaded quickly and stably. The cloud management service platform establishes a two-way connection with the blockchain evidence storage network layer, which can both upload digital imprints to the blockchain for evidence storage and retrieve evidence storage information from the blockchain for traceability verification.Meanwhile, the cloud-based management service platform connects directly to the intelligent data acquisition terminal layer via a command distribution interface, pushing automatically detected issues and rectification requirements to the on-site terminals in real time. This guides operators to adjust data acquisition operations promptly, achieving closed-loop management across the entire process.

[0048] By adopting a four-layer collaborative architecture of end-edge-cloud-chain, a two-way collaborative mechanism is formed, enabling data to be collected, processed, and uploaded to the blockchain from the field terminal to the edge node, then to the cloud platform, and finally to the blockchain. This includes the reverse transmission, rectification, and feedback of instructions from the cloud platform to the edge node and back to the field terminal. The edge aggregation node layer enables the aggregation, caching, and local service of field data, reducing network bandwidth consumption and preventing data transmission delays and loss, while ensuring operational continuity during network instability or interruptions. The cloud management service platform integrates core processing logic, achieving unified control over data management, digital imprint generation, map construction, automatic verification, and visualization, thus improving management efficiency. The blockchain evidence storage network layer provides tamper-proof evidence storage and verification services, ensuring data credibility and traceability. This four-layer collaborative architecture ultimately constructs a complete end-to-end control system for geological exploration operations, achieving intelligent and standardized management of the entire process from data acquisition to results application.

[0049] In this embodiment, the intelligent data acquisition terminal layer includes an offline operation terminal for areas without network signal coverage. The offline operation terminal has a built-in trusted execution environment and encrypted database based on ARM TrustZone technology. It is used to continuously collect data in offline mode, generate timestamps based on the local clock, run a lightweight hash algorithm to generate digital imprints, and encrypt and store the data and imprints in the isolated area of ​​the trusted execution environment. After the network is restored, the offline data is automatically uploaded synchronously and triggered on-chain evidence storage through breakpoint resume and data verification protocols.

[0050] In another possible embodiment, the intelligent data acquisition terminal layer includes an offline operation terminal specifically designed for areas without network signal coverage. This terminal, based on a regular intelligent data acquisition terminal, adds specialized security protection and offline processing functions. Firstly, the terminal has a built-in trusted execution environment based on ARM TrustZone technology. This trusted execution environment is a hardware-level secure area completely isolated from the terminal's normal operating area, equivalent to an internal security vault, effectively resisting unauthorized access and tampering. Simultaneously, the terminal carries a dedicated encrypted database that encrypts the stored data, allowing access only to authorized programs and personnel. Even in offline mode without network signal, the terminal can operate normally, continuously acquiring multimodal data from the work area. It generates standard timestamps using a built-in local high-precision clock to ensure the accuracy of time information. Simultaneously, it runs a lightweight hash algorithm to autonomously generate digital imprints corresponding to the time. The generation rules for these digital imprints are consistent with the online state, ensuring consistency between offline and online data. The collected data and generated digital imprints are encrypted and stored in an isolated area of ​​a trusted execution environment to prevent unauthorized access, tampering, or leakage. When the terminal moves to an area with a network signal, the terminal will automatically initiate a breakpoint resume and data verification protocol after the network is restored. The breakpoint resume protocol can identify data that was not uploaded while offline and only upload the missing data to avoid duplicate uploads. The data verification protocol can verify the integrity and authenticity of the uploaded data to ensure that the data is not lost or tampered with. After the data is uploaded to the cloud platform, the system will automatically trigger the digital imprint on-chain notarization process after the data verification is completed and the digital imprint of the offline data is uploaded to the blockchain for notarization, ensuring that offline operation data and online operation data have the same credibility and realizing full-link management of survey data in areas without network access.

[0051] By embedding a trusted execution environment based on ARM TrustZone technology into the offline operation terminal, hardware-level security isolation is achieved, separating the data storage area from the terminal's normal operating area. This prevents offline data from being illegally tampered with, accessed, or leaked, ensuring data storage security. Furthermore, by incorporating an encrypted database, the collected data and generated digital imprints are encrypted and stored, further enhancing data security. Through local timestamp generation and lightweight hash calculation in offline mode, it is ensured that offline data can generate trusted digital imprints, guaranteeing that offline data has the same credibility as online data. Through breakpoint resume and data verification protocols, automatic synchronization and on-chain storage of offline data are achieved after network recovery, eliminating the need for manual operation, improving data upload efficiency, preventing data loss and duplicate uploads, and ensuring the security, integrity, and trustworthiness of operational data in network-free areas. This enables comprehensive survey data management across all regions and scenarios.

[0052] In this embodiment of the application, the cloud management service platform integrates a cross-project knowledge transfer engine based on a federated averaging algorithm. The cross-project knowledge transfer engine performs joint modeling with cloud platform instances of multiple projects: each project instance trains anomaly detection models on local data, and only uploads the gradient of the anomaly detection model to the federated server with encryption. The federated server aggregates and updates the global model and then distributes it to each project instance and edge node, so as to optimize the anomaly detection capability of automatic verification rules without aggregating the original data.

[0053] In another possible embodiment, the cloud management service platform integrates a cross-project knowledge transfer engine based on a federated averaging algorithm. This engine is a core module specifically designed for joint optimization of cross-project models. It can establish stable joint modeling connections with cloud platform instances of multiple different exploration projects while strictly protecting the data privacy of each project. Each project's cloud platform instance trains anomaly detection models only on the exploration data of its local project. The anomaly detection models are used to identify abnormal information in the exploration data, such as missing data, data tampering, and data non-compliance with specifications. During the training process, no original exploration data is transmitted externally to ensure data privacy and security. After the model training is completed, each project's cloud platform instance only encrypts the model gradient information. The model gradient is the parameter update information generated during model training, which can reflect the training effect of the model. It is encrypted before being uploaded to the federated learning server. The federated learning server, as the core of joint modeling, receives encrypted gradient information uploaded by all projects. It aggregates this gradient information using a federated averaging algorithm, integrates the model training experience of each project, and updates and generates a globally optimized anomaly detection model. This global model combines the training advantages of multiple projects, resulting in stronger generalization capabilities. Subsequently, the federated learning server distributes the global model to the cloud platform instances and edge nodes of each project. Each node uses the global model to optimize its local automatic verification rules, adjust anomaly detection parameters and logic, and improve anomaly detection accuracy. Throughout the process, the original data of each project is always stored locally and not leaked externally, achieving a win-win situation for data privacy protection and model optimization. Without aggregating any original project data, the system continuously improves its ability to detect anomalies in survey data, adapting to the survey needs of different projects and scenarios.

[0054] By employing a cross-project knowledge transfer engine with a federated averaging algorithm, joint modeling can be completed without leaving the original data of each project locally, effectively protecting the data privacy and ownership of each project and preventing data leakage. Through local model training in each project, encrypted upload of model gradients, and cloud-aggregated updates of the global model, shared optimization of model capabilities is achieved without aggregating the original data. This integrates training experience from multiple projects and improves the model's generalization ability. By distributing the global model to each project and edge node, the anomaly detection capability of automatic verification rules is continuously improved, enabling accurate identification of new types of abnormal data and data from special scenarios, thus enhancing verification accuracy. Breaking down data silos, cross-project technical collaboration and resource sharing are achieved, reducing redundant computing and manpower investment, lowering costs, and promoting the overall improvement of exploration and management levels in the industry.

[0055] In this embodiment of the application, the blockchain evidence storage network layer is also deployed with an automatic settlement and performance module driven by smart contracts based on Hyperledger Fabric. The automatic settlement and performance module is used to automatically trigger the settlement of labor, confirmation of results delivery and issuance of digital certificates to the exploration participants after the digital imprint has completed the full-link closed-loop verification and all stage data has passed the compliance review.

[0056] In another possible embodiment, the blockchain evidence storage network layer deploys an automatic settlement and fulfillment module driven by smart contracts based on Hyperledger Fabric. Hyperledger Fabric is an open-source consortium blockchain framework with high security and scalability, suitable for multi-party collaborative scenarios. Here, the smart contracts are pre-written automatic execution programs that can automatically trigger corresponding fulfillment processes according to preset rules. This module monitors the full-link verification status of the work surface data in real time, obtains the verification results of the work surface digital imprint and the compliance review results of data at each stage through the blockchain interface, and determines whether the fulfillment conditions are met. When the digital imprint of the target work surface completes the full-link closed-loop verification, and all data from all exploration stages pass the compliance review, meeting the preset fulfillment conditions, the smart contract automatically triggers the preset fulfillment process. First, it automatically calculates the labor costs of the exploration participants, based on the exploration workload, etc. The system automatically generates a settlement list based on preset standards for work quality, then automatically completes the confirmation process for labor cost settlement and pushes settlement information to relevant parties. Next, it triggers the confirmation process for the delivery of survey results, automatically recording the delivery time, content, recipient, and other information to complete the delivery confirmation. Finally, it automatically issues a survey digital certificate, which is an electronic credential of the survey results' qualification, containing survey project information, result information, review information, etc., and is automatically uploaded to the blockchain for storage after issuance. All relevant data, including operation records, process information, settlement list, delivery confirmation, and digital certificate, are stored on the blockchain, making them tamper-proof and traceable at any time. This achieves automated and transparent management of the survey performance process, reduces manual intervention, avoids liability disputes, and improves performance efficiency.

[0057] By deploying a smart contract-driven module based on Hyperledger Fabric, the settlement, delivery, and certification processes are automatically linked with the data verification results across the entire chain. The performance process is only triggered when the exploration data passes the closed-loop verification across the entire chain, ensuring the quality of exploration results. The performance process is automatically triggered after passing the closed-loop verification, reducing manual intervention, avoiding human error, significantly improving process efficiency, and shortening the performance cycle. Blockchain notarization ensures that every step of the performance process is transparent and tamper-proof. All operation records and process information are stored on the blockchain and can be traced at any time, avoiding liability disputes. This enables intelligent and automated management of the entire exploration operation process, from data control to commercial performance, improving the performance efficiency and standardization of the exploration industry.

[0058] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention.

[0059] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not describe the various possible combinations separately.

[0060] Furthermore, various different implementations of the present invention can be combined arbitrarily, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed in the present invention.

Claims

1. A geological survey operation face full-link traceable management method, characterized in that, Includes the following steps: S1. Multimodal data acquisition and edge preprocessing of the work surface: Through the intelligent acquisition terminal deployed on the work surface, multimodal raw data streams containing spatial positioning data, work image data, environmental sensor data and manually recorded data are collected in real time. Timestamp alignment, data cleaning and feature extraction are performed at the edge of the intelligent acquisition terminal to generate a real-time status vector of the work surface. S2. Blockchain-based digital imprint generation and storage: The real-time status vector of the work surface is combined with the unique identifier of the work surface, the digital identity of the workers and the timestamp information to generate a unique digital imprint representing the current status of the work surface through a hash algorithm, and the digital imprint is stored on the blockchain distributed ledger. S3. Construction and Automatic Verification of the Full-Link Data Lineage Map of the Working Face: Based on the digital imprint, all related data generated at different exploration stages of the same working face are linked together to construct a data lineage map with the digital imprint as the root node. When data is entered into the database at each stage, the consistency between the stage data and the preset exploration specifications is automatically verified, and the verification result is marked. When the verification fails, a rectification task is automatically generated and pushed to the terminal of the relevant responsible person. The rectification requirements are fed back to the intelligent acquisition terminal of S1 in real time to guide the on-site operation adjustment. At the same time, the rectification process is recorded as a new node of the lineage map. S4. Traceability Verification and Visualization Based on Digital Imprints: In response to traceability query requests, the corresponding evidence information is retrieved from the blockchain based on the digital imprint of the target work surface, and hash comparison is performed with the work data in the local database for verification. The verification results, lineage map, and the trajectory of the work surface's full-cycle state change are displayed in three dimensions. When the verification results are inconsistent, the multimodal original data stream in S1 is reviewed, and abnormal nodes are marked.

2. The geological survey operation face full-link traceability management method according to claim 1, characterized in that, The edge preprocessing in S1 includes: running a lightweight target detection model through an edge computing unit to automatically identify and extract features from core samples, sampling locations, and cataloging labels in the operational image data. The lightweight target detection model replaces the YOLOv5s CSPDarknet53 backbone network with a lightweight feature extraction network based on ShuffleNet v2. A lightweight channel-spatial attention module is introduced into the feature pyramid network, a hollow spatial pyramid pooling module is introduced in front of the detection head, and a channel-level knowledge distillation strategy is used to compress and train the model with the original YOLOv51 as the teacher model.

3. The method for full-chain traceability management of geological exploration work surfaces according to claim 1, characterized in that, The generation of the digital imprint in S2 adopts a dynamic hash chain mechanism based on the BLAKE3 hash algorithm: For multiple state vectors v1, v2, vi, vn generated at different time points 1, 2, ..., i, ..., n for the same work surface, the initial digital imprint H0 is defined as the hash value of the unique identifier of the work surface and the initial timestamp. for the i-th time point, its digital fingerprint forms a time series hash chain (H0, H1,..., Hi,... Hn), where is the digital fingerprint of the previous time point, is the state vector of the i-th time point, is the time stamp of the i-th time point, is a data concatenation operator.

4. The method for full-chain traceability management of geological exploration work surfaces according to claim 1, characterized in that, The digital imprints in S2 are uploaded to the blockchain using a batch aggregation signature mechanism based on BLS signatures: within a preset time period, the cloud collects all digital imprints generated by all work surfaces within the preset time period, calculates the hash value of each digital imprint, and the consensus nodes of the consortium blockchain perform BLS signatures on the Merkle root of this batch of hash values. After aggregating all node signatures into a short signature, it is uploaded to the blockchain along with the Merkle root.

5. The method for full-chain traceability management of geological exploration work surfaces according to claim 1, characterized in that, The automatic verification in S3 includes: constructing a knowledge graph of exploration specifications based on the Neo4j graph database, transforming the design requirements, specifications, and quality control points in the exploration plan into computable verification rules stored in the form of "entity-relationship-attribute" triples; triggering a rule engine based on graph pattern matching at each stage of data entry, constructing a subgraph of the data to be verified and performing subgraph isomorphic matching with the rule patterns in the knowledge graph, performing multi-dimensional verification of data integrity, spatial compliance, temporal rationality, and content standardization; if the verification fails, automatically generating a rectification task and pushing it to the relevant responsible person's terminal, recording the rectification process as a new node in the lineage graph, and providing real-time feedback to the acquisition terminal to guide on-site operation adjustments.

6. The method for full-chain traceability management of geological exploration work surfaces according to claim 2, characterized in that, The 3D visualization in S4 uses an improved ORB-SLAM3 algorithm to achieve augmented reality overlay display. The improved ORB-SLAM3 algorithm includes: A multi-sensor tightly coupled initialization method is introduced, and a visual-inertial-GPS joint cost function is constructed by fusing GPS / RTK positioning data for global optimization, with an initialization time of 3 seconds; A dynamic feature removal mechanism based on a lightweight semantic segmentation network Fast-SCNN is introduced to remove feature points in dynamic object regions after double verification, thereby reducing trajectory errors under dynamic interference. A texture-adaptive feature extraction strategy is adopted, which dynamically adjusts the FAST corner detection threshold according to the complexity of the image mesh texture, and adds geometric context information to the ORB descriptor for weighted matching to increase the number of feature points in sparse texture regions. To achieve spatiotemporal alignment of digital imprints and map fusion, digital imprints are inserted as virtual nodes in the SLAM map and participate in map optimization together with visual landmarks, so that the spatial alignment error between digital imprints and the 3D map is less than 0.2 meters.

7. A geological exploration workface full-chain traceability management system for implementing the geological exploration workface full-chain traceability management method of claim 6, characterized in that, include: Intelligent acquisition terminal layer: Deployed on each exploration work surface, integrating multimodal sensors and edge computing units, used to execute S1 and run the lightweight target detection model; Edge aggregation node layer: Deployed at the center of the survey site area, it communicates with multiple intelligent acquisition terminals to aggregate and cache the real-time status vectors of the work surface within the area, runs a local data copy based on a time series database, and provides local query and verification services when the cloud connection is interrupted; The cloud management service platform includes a data aggregation and management module, a digital imprint generation module, a data lineage map construction module, an automatic verification engine, and a visualization service module, which are used to execute S2 to S4, wherein the visualization service module integrates the improved ORB-SLAM3 algorithm. Blockchain Evidence Storage Network Layer: Composed of multiple exploration participant nodes forming a consortium blockchain, deploying smart contracts that support batch aggregation signatures, used to receive and store the digital imprints, providing tamper-proof evidence storage services and hash comparison verification interfaces; The intelligent acquisition terminal layer is connected to the cloud management service platform through the edge aggregation node layer to form a forward data acquisition link. The cloud management service platform is also connected to the intelligent acquisition terminal layer through the command issuance interface to push the problems found by automatic verification to the field terminal in real time to guide the operation adjustment.

8. The full-chain traceable management system for geological exploration work surfaces according to claim 7, characterized in that, The intelligent data acquisition terminal layer includes an offline operation terminal for areas without network signal coverage. The offline operation terminal has a built-in trusted execution environment and encrypted database based on ARM TrustZone technology. It is used to continuously collect data in offline mode, generate timestamps based on the local clock, run a lightweight hash algorithm to generate digital imprints, and encrypt and store the data and imprints in the isolated area of ​​the trusted execution environment. After the network is restored, the offline data is automatically synchronized and uploaded through the breakpoint resume and data verification protocol, and the on-chain evidence storage is triggered.

9. The full-chain traceable management system for geological exploration work surfaces according to claim 7, characterized in that, The cloud management service platform integrates a cross-project knowledge transfer engine based on a federated averaging algorithm. The cross-project knowledge transfer engine performs joint modeling with cloud platform instances of multiple projects: each project instance trains anomaly detection models on local data, and only uploads encrypted gradients of the anomaly detection models to the federated server. The federated server aggregates and updates the global model and then distributes it to each project instance and edge node, so as to optimize the anomaly detection capability of automatic verification rules without aggregating the original data.

10. The full-chain traceability management system for geological exploration work surfaces according to claim 7, characterized in that, The blockchain evidence storage network layer is also equipped with an automatic settlement and performance module driven by smart contracts based on Hyperledger Fabric. The automatic settlement and performance module is used to automatically trigger the settlement of labor services, confirmation of results delivery, and issuance of digital certificates to the exploration participants after the digital imprint has completed the full-link closed-loop verification and all stage data has passed the compliance review.

Images