A mineral exploration system and method oriented to distributed seismic sensing

Abstract

The present application provides a kind of mineral exploration system and method for distributed vibration sensing, its system includes: vibration data acquisition subsystem, for based on the distributed optical fiber laid in exploration area, the original vibration data of geological body is collected;Vibration data processing subsystem, for the original vibration data is preprocessed, obtain target vibration data, and extract the seismic attribute parameters of target vibration data;Vibration data analysis subsystem, for the seismic attribute parameters are respectively input into supervised identification model and unsupervised identification model, corresponding obtain mineralization probability graph and anomaly distribution graph, and mineralization probability graph and anomaly distribution graph are cross-validated, generate three-dimensional ore body detection result;Interpretation and visualization subsystem, for three-dimensional ore body detection result is visualized and interactively corrected, generate mineral exploration report.The present application improves the efficiency and accuracy of mineral exploration.

Description

Technical Field

[0001] This invention relates to the field of mineral exploration technology, specifically to a mineral exploration system and method for distributed vibration sensing. Background Technology

[0002] Traditional mineral exploration methods, especially seismic exploration, play a crucial role in finding underground mineral deposits. The basic principle involves using seismic waves generated by artificial sources to propagate, reflect, and refract in the underground medium before being received by surface or underground geophones. Analysis of the received seismic wave data allows for the inference of underground geological structures and lithological distribution, thereby delineating mineral-bearing areas. However, the channel density of traditional geophone arrays is limited by cost and technological bottlenecks, making it difficult to achieve ultra-high density continuous coverage. This results in insufficient spatial sampling of seismic data, potentially missing subtle geological anomalies or deep, hidden ore bodies. To address this technical problem, Distributed Acoustic Sensing (DAS) technology has been proposed. DAS systems measure the phase change of backscattered light along an optical fiber, enabling high-precision, long-distance continuous monitoring of dynamic strain or vibration at any point along the fiber path. Its advantages lie in the fact that the optical fiber itself is a sensor, requiring no external power supply, resisting electromagnetic interference, and having strong environmental adaptability. It can achieve continuous, high-density measurements over a range of several kilometers or even tens of kilometers, significantly reducing the workload of on-site deployment and environmental impact.

[0003] However, due to the massive amount of data generated by DAS systems (TB / day) and the fact that the raw data typically has a low signal-to-noise ratio and contains a large amount of environmental noise, direct use for geological interpretation faces challenges. Current post-processing mainly relies on traditional seismic data processing methods such as filtering, denoising, stacking, and migration. These methods are inefficient when processing massive amounts of distributed fiber optic data with low signal-to-noise ratios and struggle to automatically identify complex seismic response characteristics of ore bodies. Especially when identifying hidden ore bodies with weak seismic anomalies or features similar to the surrounding rock, traditional methods often fall short, requiring time-consuming manual interpretation by a large number of experienced geophysicists. This limits the full realization of the potential of distributed fiber optic seismic sensing technology in mineral exploration. Specifically, it suffers from the following technical problems: 1. Low data processing efficiency: The massive amount of data generated by distributed fiber optic sensors is difficult to process efficiently using traditional methods. 2. Limited improvement in signal-to-noise ratio: Complex geological and environmental noise severely interferes with the identification of weak signals from ore bodies, and traditional denoising methods have limited effectiveness. 3. Insufficient interpretation accuracy and automation: It is difficult to automatically identify the seismic response of complex ore bodies, and it relies heavily on human experience. The interpretation results are affected by subjective factors, resulting in low efficiency and a high risk of errors.

[0004] Therefore, there is an urgent need to provide a mineral exploration system and method for distributed vibration sensing, which can efficiently process distributed fiber optic vibration sensing data and improve its signal-to-noise ratio, thereby achieving high-precision and high-reliability mineral exploration. Summary of the Invention

[0005] In view of this, it is necessary to provide a mineral exploration system and method for distributed vibration sensing to solve the technical problems in the existing technology, such as the inability to efficiently process fiber optic vibration sensing data and the poor signal-to-noise ratio of fiber optic vibration sensing data, which leads to low accuracy and efficiency in mineral exploration.

[0006] To address the aforementioned technical problems, in a first aspect, the present invention provides a mineral exploration system for distributed vibration sensing, comprising: a vibration data acquisition subsystem, a vibration data processing subsystem, a vibration data analysis subsystem, an interpretation and visualization subsystem, and a control and management subsystem; The vibration data acquisition subsystem is used to acquire raw vibration data of geological bodies based on distributed optical fibers deployed in the exploration area. The seismic data processing subsystem is used to preprocess the raw seismic data to obtain target seismic data and extract the seismic attribute parameters of the target seismic data. The seismic data analysis subsystem is used to input the seismic attribute parameters into the supervised identification model and the unsupervised identification model respectively, to obtain the mineralization probability map and the anomaly distribution map, and to cross-validate the mineralization probability map and the anomaly distribution map to generate a three-dimensional ore body detection result. The interpretation and visualization subsystem is used to visualize and interactively correct the three-dimensional ore body detection results, and generate a mineral exploration report. The control and management subsystem is used to control the working order of the vibration data acquisition subsystem, vibration data processing subsystem, vibration data analysis subsystem, and interpretation and visualization subsystem.

[0007] In one possible implementation, the vibration data processing subsystem includes a denoising module, which includes a data format conversion unit, a primary denoising unit, and an adaptive denoising unit. The data format conversion unit is used to convert the format of the raw vibration data from the original format to the target format. The primary denoising unit performs bandpass filtering on the original vibration data after format conversion to obtain coarse-filtered vibration data, and determines whether the coarse-filtered vibration data has single-frequency interference or coherent noise. When single-frequency interference exists, notch filtering is performed on the coarse-filtered vibration data to obtain primary-filtered vibration data. When coherent noise exists, FK filtering is performed on the coarse-filtered vibration data to obtain primary-filtered vibration data. The adaptive denoising unit is used to determine whether there is a usable reference noise channel in the primary filtered vibration data. If there is, adaptive noise cancellation processing is performed on the primary filtered vibration data. If not, it is determined whether there is an independent noise source in the primary filtered vibration data. If there is an independent noise source, independent component analysis is performed on the primary filtered vibration data to obtain fine filtered vibration data. If there is no independent noise source, the primary filtered vibration data is denoised based on principal component analysis, sparse transform domain denoising, or a deep learning denoising model to obtain fine filtered vibration data.

[0008] In one possible implementation, the vibration data processing subsystem further includes a gain compensation module, which includes a spherical diffusion compensation unit, a Q-value attenuation compensation unit, and an automatic gain unit. The spherical diffusion compensation unit is used to determine the spherical diffusion compensation gain factor, and the product of the fine-filtered vibration data and the spherical diffusion compensation gain factor is used as the first compensated vibration data. The Q-value attenuation compensation unit is used to construct a Q-value attenuation compensation filter based on a pre-established Q-value model, and input the first compensated vibration data into the Q-value attenuation compensation filter to obtain the second compensated vibration data. The automatic gain control unit is used to normalize the second compensated vibration data based on a preset sliding time window to obtain gain vibration data.

[0009] In one possible implementation, the vibration data processing subsystem further includes an interpolation module, which is used to perform interpolation processing on the gain vibration data based on an interpolation algorithm to obtain the target vibration data; the interpolation algorithm is at least one of linear interpolation, spline interpolation, K-nearest neighbor interpolation, sparse reconstruction-based interpolation, and deep learning-based interpolation.

[0010] In one possible implementation, the seismic attribute parameters include basic seismic attributes and rock physical attributes. The basic seismic attributes include amplitude, frequency, phase, instantaneous amplitude, instantaneous frequency, and instantaneous phase. The rock physical attributes include coherence attributes, spectral decomposition attributes, Q-value attributes, AVO / AVA attributes, curvature attributes, texture attributes, and elastic parameter inversion attributes.

[0011] In one possible implementation, the seismic data analysis subsystem includes a feature dimensionality reduction module. This module performs principal component analysis on the seismic attribute parameters, determines the variance explanation rate of each principal component, and determines whether the sum of the variance explanation rates of the first k principal components is greater than a variance explanation rate threshold. If not, it performs nonlinear feature dimensionality reduction on the seismic attribute parameters based on a nonlinear feature dimensionality reduction model to obtain dimensionality-reduced attribute parameters.

[0012] In one possible implementation, the vibration data analysis subsystem further includes a ore body detection module; The orebody detection module is used to determine overlapping and complementary regions based on the mineralization probability map and the anomaly distribution map, and to determine the comprehensive confidence level of the complementary region based on the first confidence level of the mineralization probability map and the second confidence level of the anomaly distribution map. The ore body detection module is also used to determine the boundary lines and morphological parameters of the overlapping region and the complementary region based on threshold segmentation and morphological processing. The overlapping region is the region where the results of the mineralization probability map and the anomaly distribution map are consistent, and the complementary region is the region where the results of the mineralization probability map and the anomaly distribution map are inconsistent.

[0013] In one possible implementation, the vibration data analysis subsystem is further used to optimize and train the supervised recognition model and the unsupervised recognition model based on the data interactively corrected by the interpretation and visualization subsystem.

[0014] In one possible implementation, the control management subsystem further includes a log management module and a user permission management module; The log management module is used to record the output data of the vibration data acquisition subsystem, vibration data processing subsystem, vibration data analysis subsystem, and interpretation and visualization subsystem. The user permission management module is used to define the access permissions and operation scope of different users.

[0015] Secondly, the present invention also provides a mineral exploration method for distributed vibration sensing, comprising: Raw seismic data of geological bodies were acquired using distributed optical fibers deployed in the exploration area. The raw seismic data is preprocessed to obtain target seismic data, and the seismic attribute parameters of the target seismic data are extracted. The seismic attribute parameters are input into the supervised and unsupervised identification models respectively to obtain mineralization probability maps and anomaly distribution maps. The mineralization probability maps and anomaly distribution maps are then cross-validated to generate three-dimensional ore body detection results. The results of the three-dimensional ore body detection are visualized and interactively corrected to generate a mineral exploration report.

[0016] The beneficial effects of this invention are as follows: The mineral exploration system for distributed vibration sensing provided by this invention integrates a vibration data acquisition subsystem, a vibration data processing subsystem, a vibration data analysis subsystem, an interpretation and visualization subsystem, and a control and management subsystem, forming a complete automated exploration process. This greatly reduces manual intervention and repetitive labor, shortening the manual interpretation cycle in traditional mineral exploration, which takes weeks or even months, to a few days or even real-time interpretation, thus improving the efficiency of mineral exploration. Furthermore, based on the raw vibration data of geological bodies acquired by distributed optical fibers deployed in the exploration area, ultra-high-density spatial continuous sampling is achieved, avoiding spatial frequency distortion and information loss caused by traditional array geophones, providing a data foundation for refined exploration. Simultaneously, the distributed optical fiber deployment can cover a large area with a single deployment, and it is highly durable and has low maintenance costs, reducing the high costs associated with the repeated deployment and retrieval of traditional geophone arrays.

[0017] Furthermore, this invention generates three-dimensional ore body detection results by setting up a vibration data analysis subsystem based on cross-validation of supervised and unsupervised identification models. At the same time, it utilizes the advantages of both types of models. For example, the unsupervised identification model supplements the limitations of supervised learning, that is, it discovers "new" or "unknown" geological anomalies that do not match the known patterns in the training data. These anomalies may indicate new ore bodies or unexpected patterns, thereby improving the accuracy of mineral exploration.

[0018] Furthermore, by setting up an interpretation and visualization subsystem to interactively correct the three-dimensional ore body detection results, this invention combines the efficiency of artificial intelligence with the experience and wisdom of human experts, thus ensuring both exploration efficiency and the accuracy of the final results, thereby further improving the accuracy of mineral exploration. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of an embodiment of the mineral exploration system for distributed vibration sensing provided by the present invention. Figure 2 A schematic diagram of an embodiment of the noise reduction module in the vibration data processing subsystem provided by the present invention; Figure 3 A schematic diagram of an embodiment of the gain compensation module in the vibration data processing subsystem provided by the present invention; Figure 4This is a schematic flowchart of an embodiment of the mineral exploration method for distributed vibration sensing provided by the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] It should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this invention illustrate operations implemented according to some embodiments of the invention. It should be understood that the operations in the flowcharts may be implemented out of order, and steps without logical contextual relationships may be reversed or performed simultaneously. Furthermore, those skilled in the art, guided by the content of this invention, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0023] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0024] This invention provides a mineral exploration system and method for distributed vibration sensing, which will be described below.

[0025] Figure 1 This is a schematic flowchart of an embodiment of the mineral exploration system for distributed vibration sensing provided by the present invention, as shown below. Figure 1 As shown, the mineral exploration system 10 for distributed vibration sensing includes: a vibration data acquisition subsystem 100, a vibration data processing subsystem 200, a vibration data analysis subsystem 300, an interpretation and visualization subsystem 400, and a control and management subsystem 500. The seismic data acquisition subsystem 100 is used to acquire raw seismic data of geological bodies based on distributed optical fibers deployed in the exploration area.

[0026] Specifically, distributed optical fibers can be standard single-mode communication optical fibers, with an outer reinforcing protective layer to adapt to the field environment. Distributed optical fibers are directly buried in shallow underground layers or laid on the surface to form linear or two-dimensional arrays of several kilometers to tens of kilometers, used to acquire wide-area seismic data.

[0027] Distributed fiber optics can be deployed in a variety of scenarios: Surface array deployment: Distributed optical fibers are directly buried in shallow underground layers or laid on the surface to form linear or two-dimensional arrays of several kilometers to tens of kilometers, used to acquire wide-area seismic data.

[0028] Downhole vertical seismic profile (VSP) deployment: Distributed optical fibers are fixed inside the borehole wall to form a downhole vertical sensor array for high-resolution detection of the geological structure around the well.

[0029] Water area deployment: Watertight distributed optical fibers are submerged on the bottom of the water or suspended in the water for underwater seismic exploration.

[0030] It should be noted that the raw vibration data refers to the data acquired through distributed optical fiber acquisition and then demodulated by the demodulator. Specifically, the demodulator employs a distributed vibration sensing principle, such as phase-sensitive optical time-domain reflectometry, injecting laser pulses into the optical fiber and sensing the strain or vibration of the fiber by detecting the phase change of the reverse Rayleigh scattered light along the fiber. This demodulator has multi-channel output capability and can process data from multiple optical fibers or multiple optical fiber segments simultaneously.

[0031] The seismic data processing subsystem 200 is used to preprocess the raw seismic data to obtain the target seismic data and extract the seismic attribute parameters of the target seismic data.

[0032] It should be understood that the mineral exploration system 10 for distributed vibration sensing also includes a high-performance data transmission and storage subsystem 600, which is used to transmit raw vibration data from the vibration data acquisition subsystem 100 to the vibration data processing subsystem 200.

[0033] Specifically, high-bandwidth network interfaces (such as 10 Gigabit Ethernet and Fibre Channel) and optimized data transmission protocols are used to ensure that the raw vibration data can be transmitted efficiently and stably.

[0034] Furthermore, the high-performance data transmission and storage subsystem 600 is also used with high-capacity, high-I / O performance storage arrays (such as RAID systems, NVMe-over-Fabric), and can be integrated with cloud storage solutions to achieve long-term, reliable storage and fast access to raw vibration data.

[0035] The seismic data analysis subsystem 300 is used to input seismic attribute parameters into the supervised and unsupervised identification models respectively, obtain mineralization probability maps and anomaly distribution maps, and cross-validate the mineralization probability maps and anomaly distribution maps to generate three-dimensional ore body detection results.

[0036] In specific embodiments of the present invention, the supervised recognition model may be at least one of convolutional neural networks, long short-term memory networks, recurrent neural networks, support vector machines, random forests, etc. The unsupervised recognition model may be at least one of isolation forests, Local Outlier Factor (LOF), single-class support vector machines, generative adversarial networks, predictive coding, etc.

[0037] Specifically, convolutional neural networks (CNNs), with their powerful local feature extraction and spatial pattern recognition capabilities, are particularly suitable for processing seismic data, which has a grid-like structure (2D profiles or 3D volumetric data). Through multiple convolutional kernels, CNNs can automatically learn and identify various complex spatial patterns related to ore bodies in seismic data, such as reflection interface morphology, fracture patterns, spatial distribution of amplitude anomalies, and texture features. For 2D seismic profiles, 2D CNNs can be used; for 3D seismic data volumes, 3D CNNs can be used for semantic segmentation, directly outputting the classification results for each voxel. This is used to identify ore body boundaries, internal structures, and alteration zones or surrounding rock features associated with the ore body.

[0038] Recurrent Neural Networks (RNNs) / Long Short-Term Memory (LSTMs) excel at processing sequential data and capturing long-range dependencies. In seismic data processing, each seismic trace can be viewed as a time series or depth series. RNNs / LSTMs can learn longitudinal variation characteristics along the seismic trace, such as stratigraphic sequence, lithological variations, or vertical mineralization patterns. Combined with the local features output by CNNs, they can further analyze the evolution of seismic signals in the time / depth dimensions, making them particularly suitable for processing downhole fiber optic VSP data.

[0039] Support Vector Machines (SVMs) are powerful supervised learning classifiers designed to find the optimal hyperplane in a high-dimensional feature space to maximize the margin between different classes. When feature engineering is sufficient and the data dimensionality is reduced, SVMs perform well on small datasets and have good generalization ability. They can be used for binary or multi-class classification of ore bodies and non-ore bodies on limited, high-quality feature samples.

[0040] Random forests are ensemble learning methods based on decision trees. They improve classification accuracy and robustness by constructing multiple decision trees and performing ensemble voting or weighted summation. They have a strong ability to model nonlinear relationships between features and are less prone to overfitting. They also provide feature importance assessments, helping to understand which seismic attributes are most critical for orebody identification. When features are abundant and interpretability is critical, these models can provide efficient and reliable classification results.

[0041] The supervised identification model training process employs optimization algorithms such as Adam and SGD (Stochastic Gradient Descent). Through the backpropagation mechanism, the model's weights and biases are iteratively updated to minimize a predefined loss function (such as the cross-entropy loss function), thereby enabling the model to learn the optimal mapping relationship from seismic features to geological classification.

[0042] Hyperparameter tuning employs methods such as grid search, random search, or Bayesian optimization to optimize model hyperparameters including learning rate, batch size, number of network layers, and kernel size to achieve optimal model performance. Cross-validation uses techniques such as K-fold cross-validation to evaluate the model's generalization ability and avoid overfitting.

[0043] It should be understood that the samples used in the training were high-quality samples that had been annotated by experts, and the labels included geological entity classification markers, such as “target ore body area”, “surrounding rock”, “fault”, “aquifer”, “non-mineralized alteration zone”, etc.

[0044] For scenarios with scarce data, models pre-trained on large public datasets (such as ImageNet or general geophysical datasets) can be used as a starting point. Based on transfer learning techniques, the models can be fine-tuned to adapt to specific ore body identification tasks, thereby accelerating model convergence and improving performance.

[0045] The output of the supervised recognition model is abstract data. In order to achieve the visualization of the data, in a specific embodiment of the present invention, the recognition results are superimposed on the geological body to obtain a mineralization probability map that characterizes the three-dimensional morphology of the ore body and its mineralization probability.

[0046] Similarly, the results of the unsupervised identification model are superimposed on the geological body to obtain an anomaly distribution map that characterizes the three-dimensional morphology of the ore body and its anomaly probability.

[0047] The interpretation and visualization subsystem 400 is used to visualize and interactively correct the results of three-dimensional ore body detection, and generate mineral exploration reports.

[0048] Specifically, the interpretation and visualization subsystem 400 provides a modern, intuitive, and user-friendly graphical user interface (GUI) as a portal for geological experts to interact with the entire intelligent system. The interface design follows ergonomic principles, ensuring a clear menu structure and customizable toolbars, integrating functions such as data loading, display control, processing parameter settings, interpretation tool selection, model training initiation, result saving, and report generation. Simultaneously, the GUI supports multi-window and multi-view modes, allowing users to simultaneously view raw data, data from different processing stages, seismic attribute volumes, interpretation results, etc., and enabling synchronous navigation between different views (such as mouse interaction and simultaneous display of the same geological location).

[0049] Furthermore, the graphical user interface provides a clear parameter setting panel, allowing users to adjust display attributes (such as color mapping, transparency, and slice direction), filtering parameters, interpretation thresholds, etc., and preview the adjustment effects in real time. In addition, the graphical user interface can also display system running status, task progress, warnings and error messages, as well as key operation logs in real time, facilitating user monitoring and troubleshooting.

[0050] This invention, through its graphical user interface, significantly lowers the barrier to entry for users and improves work efficiency. The user-friendly interactive interface allows users to easily and quickly access and control various system functions, achieving seamless human-machine collaboration.

[0051] Since two-dimensional seismic profiles are sometimes of greater interest in practical applications, in some embodiments of the present invention, the interpretation and visualization subsystem 400 can also achieve three-dimensional-two-dimensional switching display.

[0052] Specifically, it provides high-resolution seismic profile display capabilities, supporting the generation of 2D seismic slices along any direction (such as the survey line direction, perpendicular to the survey line direction, or the tilt direction). Multi-attribute slice display: In addition to the conventional amplitude profile, it also supports the simultaneous or independent display of various extracted seismic attribute profiles, such as instantaneous frequency, instantaneous phase, coherence, Q value, and curvature. Through color mapping and transparency adjustment, the indicative role of different attributes in ore body identification is highlighted.

[0053] For example, users can generate two-dimensional cross-sectional views of ore bodies, faults, or strata along any plane (horizontal, vertical, or inclined), and can overlay seismic backgrounds. Users can also render identified ore bodies, strata, and fault structures as three-dimensional models. Users can observe the spatial distribution, shape, thickness, and interrelationships of underground geological bodies from any angle through rotation, scaling, and sectioning operations. Different geological bodies can be distinguished using different colors and transparency levels.

[0054] Furthermore, the interpretation and visualization subsystem 400 supports zooming, panning, and rotating of the profile, as well as dynamic sliding viewing along depth / time and distance, and can perform operations such as tracking and picking along the same phase axis.

[0055] This invention provides an interpretation and visualization subsystem 400 to visualize the results of three-dimensional ore body detection, transforming abstract numerical classification results into concrete three-dimensional geological entities. This greatly improves geological experts' understanding of the spatial distribution, geometric morphology, and relationship with the surrounding rock / structure of the ore body.

[0056] It should be noted that the graphical user interface can also overlay and display various other geological, geophysical, and borehole data for comprehensive interpretation. For example, gravity anomaly maps, magnetic anomaly maps, electrical resistivity profiles, and high-resolution remote sensing images can be loaded and overlaid. This cross-verification and supplementation of multi-source data improves the reliability of the interpretation. Furthermore, the location, trajectory, core description, well logging curves (such as gamma, resistivity, and density), and mineralization intersections of boreholes can be accurately displayed in a 3D scene. This is crucial for verifying seismic interpretation results and providing geological truth constraints. Simultaneously, regional geological maps, mineral distribution maps, and other 2D Geological Information System (GIS) data can also be overlaid.

[0057] This invention, by integrating data from multiple independent sources, performing cross-validation and comprehensive analysis, overcomes the limitations of a single data source and greatly improves the accuracy and reliability of geological interpretation.

[0058] It should be further explained that interactive correction refers to the ability of users to review, evaluate, correct, and optimize 3D orebody detection results via a graphical user interface. Specifically, interactive editing tools include, but are not limited to, boundary editing, attribute reclassification, structure loss, deletion and addition, and confidence assessment. Boundary editing: Modify the three-dimensional boundaries of ore bodies, strata, or faults by dragging, clicking, and erasing with the mouse.

[0059] Attribute reclassification: Manually label or reclassify a region (e.g., relabel a region of surrounding rock that the AI ​​misidentified as an ore body, or vice versa).

[0060] Structure Picking: Allows experts to manually pick key geological structures such as phase axes, faults, and unconformities to compensate for the limitations of AI in localized complex areas.

[0061] Deletion and addition: Delete erroneous explanatory elements or manually add new explanatory elements.

[0062] Confidence assessment: The system displays the confidence level (probability or uncertainty output by the machine learning model) of each explanation result, guiding experts to focus on low-confidence areas for manual review and correction.

[0063] It should also be noted that the content of a mineral exploration report includes, but is not limited to: an overview of the exploration area, geographical location, an overview of the distributed fiber optic seismic exploration method, data acquisition parameters, ore body identification results (such as ore body location, spatial coordinates, burial depth, geometric shape, and volume estimation), predicted ore body type, relationship with surrounding rock, interpretation of major fault structures and stratigraphy, confidence assessment and uncertainty analysis of interpretation results, recommended follow-up exploration work (such as borehole verification location), and map integration: automatically embedding high-quality 2D / 3D visualization maps (such as seismic profiles, ore body 3D models, attribute volume rendering maps, and multi-source data overlay maps), supporting multiple output formats (such as PDF, Word, and image files).

[0064] The embodiments of the present invention can standardize and automate the report writing process by generating mineral exploration reports, ensuring the accuracy, consistency and timeliness of the reports, greatly reducing the workload of geological personnel and accelerating the exploration decision-making process.

[0065] The control and management subsystem 500 is used to control the working order of the vibration data acquisition subsystem, vibration data processing subsystem, vibration data analysis subsystem, and interpretation and visualization subsystem.

[0066] It should be noted that the control and management subsystem 500 can decompose and schedule tasks to be executed in parallel on different nodes in a high-performance computing cluster (such as a GPU server array or a CPU server cluster), achieving load balancing and maximizing resource utilization. For example, resource scheduling can be performed based on cluster management systems such as Kubernetes or YARN.

[0067] The control and management subsystem 500 ensures the automated, efficient, and orderly operation of the entire system workflow. It significantly improves data processing throughput and turnaround time, minimizes manual intervention, and enhances overall system efficiency.

[0068] Compared with existing technologies, the mineral exploration system 10 for distributed vibration sensing provided in this invention integrates a vibration data acquisition subsystem 100, a vibration data processing subsystem 200, a vibration data analysis subsystem 300, an interpretation and visualization subsystem 400, and a control and management subsystem 500, forming a complete automated exploration process. This greatly reduces manual intervention and repetitive labor, shortening the manual interpretation cycle that traditionally takes weeks or even months in mineral exploration to a few days or even achieving real-time interpretation, thus improving the efficiency of mineral exploration. Furthermore, based on the raw vibration data of geological bodies acquired by distributed optical fibers deployed in the exploration area, ultra-high-density spatial continuous sampling is achieved, avoiding spatial frequency distortion and information loss caused by traditional array geophones, providing a data foundation for refined exploration. Simultaneously, the distributed optical fiber deployment can cover a large area with a single deployment, and it is highly durable and has low maintenance costs, reducing the high costs associated with the repeated deployment and retrieval of traditional geophone arrays.

[0069] Furthermore, in this embodiment of the invention, a vibration data analysis subsystem 300 is set up to perform cross-validation based on a supervised recognition model and an unsupervised recognition model to generate three-dimensional ore body detection results. At the same time, the advantages of the two types of models are utilized. For example, the unsupervised recognition model supplements the limitations of supervised learning, that is, it discovers "new" or "unknown" geological anomalies that do not match the known patterns in the training data. These anomalies may indicate new ore bodies or unexpected patterns, thereby improving the accuracy of mineral exploration.

[0070] Furthermore, by setting up an interpretation and visualization subsystem 400 to interactively correct the three-dimensional ore body detection results, this embodiment of the invention combines the efficiency of artificial intelligence with the experience and wisdom of human experts, thus ensuring both exploration efficiency and the accuracy of the final results, thereby further improving the accuracy of mineral exploration.

[0071] To improve the signal-to-noise ratio of target vibration data, in some embodiments of the present invention, such as Figure 1 and Figure 2 As shown, the vibration data processing subsystem 200 includes a noise reduction module 210, which includes a data format conversion unit 211, a primary noise reduction unit 212, and an adaptive noise reduction unit 213. The data format conversion unit 211 is used to convert the format of the original vibration data from the original format to the target format.

[0072] Specifically, the original format is binary, and the target format is SEG-Y. The advantage of the SEG-Y format lies in its good interoperability, which facilitates integration with existing seismic processing software and interpretation systems.

[0073] The primary denoising unit 212 performs bandpass filtering on the original vibration data after format conversion to obtain coarse-filtered vibration data, and determines whether there is single-frequency interference or coherent noise in the coarse-filtered vibration data. When single-frequency interference exists, notch filtering is performed on the coarse-filtered vibration data to obtain primary-filtered vibration data. When coherent noise exists, FK filtering is performed on the coarse-filtered vibration data to obtain primary-filtered vibration data.

[0074] Specifically, bandpass filtering, based on the characteristic frequency range of the seismic response of the target ore body, applies digital bandpass filters (such as Butterworth and Chebyshev filters) to filter out noise above and below the target frequency band, such as high-frequency random noise and low-frequency environmental background noise (such as ocean waves and ground vibrations). Notch filtering precisely removes periodic interference at specific frequencies, such as 50Hz / 60Hz power frequency noise and instrument harmonic noise. FK filtering (frequency-wavenumber filtering) utilizes the characteristic that seismic waves have a specific slope (apparent velocity) in the frequency-wavenumber domain (FK domain). By designing a two-dimensional filter, it separates and suppresses coherent noise with different apparent velocities (such as surface waves, air waves, and scattered waves). It is particularly suitable for suppressing high-energy linear noise commonly found in DAS data.

[0075] The adaptive denoising unit 213 is used to determine whether there is a usable reference noise channel in the primary filtered vibration data. If there is, the primary filtered vibration data is processed by adaptive noise cancellation (ANC). If not, it is determined whether there is an independent noise source in the primary filtered vibration data. If there is an independent noise source, independent component analysis is performed on the primary filtered vibration data to obtain fine filtered vibration data. If there is no independent noise source, the primary filtered vibration data is denoised based on principal component analysis, sparse transform domain denoising, or deep learning denoising model to obtain fine filtered vibration data.

[0076] Specifically, sparse transform domain denoising can be any of the wavelet transform, Curvelet transform, or Shearlet transform.

[0077] This invention, through data format conversion, ensures data universality and processability, providing a unified and identifiable data interface for all subsequent modules. Simultaneously, the primary denoising unit 212 effectively mitigates common background noise and periodic interference using various filtering algorithms, initially improving the signal-to-noise ratio of the vibration signal. Building upon the primary denoising unit 212, deeper and more refined noise suppression is performed on the more complex and nonlinear environmental noise unique to distributed optical fiber data (such as wind disturbance, vehicle vibration, water flow impact, and ground micro-tremors), significantly improving the signal-to-noise ratio of the target ore body signal.

[0078] Since vibration data attenuates with propagation distance, in order to clearly present deep and weak signals, in some embodiments of the present invention, such as... Figure 1 and Figure 3 As shown, the vibration data processing subsystem 200 also includes a gain compensation module 220, which includes a spherical diffusion compensation unit 221, a Q-value attenuation compensation unit 222, and an automatic gain unit 223. The spherical diffusion compensation unit 221 is used to determine the spherical diffusion compensation gain factor and use the product of the fine-filtered vibration data and the spherical diffusion compensation gain factor as the first compensated vibration data.

[0079] Among them, the spherical diffusion compensation gain factor can be determined based on the two-way travel time of the seismic wave and the formation velocity corresponding to the seismic wave at a certain moment.

[0080] Q-value attenuation compensation unit 222 is used to construct a Q-value attenuation compensation filter based on a pre-established Q-value model, input the first compensation vibration data into the Q-value attenuation compensation filter, and obtain the second compensation vibration data.

[0081] The Q-value model can be obtained from drilling data or seismic inversion.

[0082] The automatic gain unit 223 is used to normalize the second compensated vibration data based on a preset sliding time window to obtain the gain vibration data.

[0083] Specifically, for each sliding time window, the root mean square amplitude or average absolute amplitude is calculated for all sampling points therein. The ratio of the predefined target amplitude to the root mean square amplitude or average absolute amplitude is used as the gain factor of the sliding time window. Based on the gain factor, the second compensated vibration data is amplified to obtain the amplified vibration data.

[0084] This invention restores the energy loss caused by geometric diffusion by setting a spherical diffusion compensation unit 221, compensates for the amplitude attenuation or frequency narrowing caused by the inelastic absorption of the underground medium by setting a Q-value attenuation compensation unit 222, and solves the technical problem of huge energy difference between shallow and deep signals that the previous two units could not achieve by setting an automatic gain unit 223, so that deep and weak signals can be clearly presented and ensure that there is a sufficiently identifiable signal strength throughout the entire exploration depth range.

[0085] During field data acquisition, sensor malfunctions, fiber optic cable interruptions, or limitations in deployment in specific areas may lead to missing or unevenly sampled seismic traces. To address this technical problem, in some embodiments of the present invention, such as... Figure 1As shown, the vibration data processing subsystem 200 also includes an interpolation module 230, which is used to perform interpolation processing on the gain vibration data based on an interpolation algorithm to obtain the target vibration data; the interpolation algorithm is at least one of linear interpolation, spline interpolation, K-nearest neighbor interpolation, sparse reconstruction-based interpolation, and deep learning-based interpolation.

[0086] It should be understood that the gain vibration data can be sorted before interpolation, specifically according to the coordinates of the receiving point, offset, etc.

[0087] As is well known to those skilled in the art, the more comprehensive the parameters input into the supervised and unsupervised recognition models, the more accurate the recognition results. Therefore, in a specific embodiment of the present invention, the seismic attribute parameters include basic seismic attributes and rock physical attributes. The basic seismic attributes include amplitude, frequency, phase, instantaneous amplitude, instantaneous frequency, and instantaneous phase. The rock physical attributes include coherence attributes, spectral decomposition attributes, Q-value attributes, AVO / AVA attributes, curvature attributes, texture attributes, and elastic parameter inversion attributes.

[0088] Specifically: Amplitude reflects the intensity of the seismic wave reflection interface and is related to the difference in wave impedance between the rock layers on both sides of the interface. High amplitude anomalies may indicate ore bodies or high-density rocks. Frequency is the periodic characteristic of seismic waves and is related to the thickness, physical properties, and attenuation characteristics of the rock layers. Ore bodies may cause local frequency anomalies. Phase is the initial state of the seismic wave and is related to the polarity of the reflection interface. Anomalous phase changes may indicate changes in interface properties. Instantaneous amplitude / envelope reflects the intensity of seismic energy and can highlight the continuity and faulting of the seismic phase axis. Instantaneous frequency reflects local frequency changes and is sensitive to thin layers and fluids. Instantaneous phase can highlight the continuity of the phase axis, eliminate the interference of amplitude in interpretation, and is sensitive to faults and stratigraphic pinch-outs. Energy is the total energy of seismic waves within a certain time window; high-energy areas may be related to ore bodies or tectonic anomalies.

[0089] In other words, the basic properties of an earthquake provide the most intuitive and fundamental physical quantitative characteristics of earthquake signals.

[0090] Coherence attributes quantify the continuity of seismic reflections by measuring the similarity or correlation between adjacent seismic traces. Commonly used algorithms include cross-correlation-based coherence volumes and eigenstructure-based coherence volumes (such as semblance and eigen-coherence). Low coherence zones typically indicate stratigraphic discontinuities, such as faults, fracture zones, intrusive body boundaries, or mineralization alteration zones.

[0091] Coherence properties can effectively identify underground geological structures (such as faults, fissures, and folds) and discontinuities, which are often favorable structures for the occurrence of ore bodies.

[0092] Spectral decomposition attributes decompose seismic signals into different frequency components, thereby analyzing the contribution of different frequency components to the formation response. Commonly used methods include Short-Time Fourier Transform (STFT), Continuous Wavelet Transform (CWT), and S-transform. By analyzing the amplitude, phase, and other properties of different frequencies, frequency anomalies caused by thin-layer tuning effects, fluid influences, or special lithologies can be identified. For example, certain ore bodies may exhibit unique responses within specific frequency ranges.

[0093] Spectral decomposition properties can improve the ability of seismic data to identify changes in thin geological bodies and fluids (or ore bodies), and are particularly important for identifying concealed ore bodies.

[0094] Q-factor attributes are parameters that measure the inelastic absorption and attenuation characteristics of subsurface media. Rocks containing fluids (such as oil and gas, water), gas, or specific minerals (such as sulfide ore bodies) often have low Q-factors (high attenuation). Q-factor attributes can be extracted from seismic data using inversion methods (such as spectral ratio method and instantaneous frequency attenuation method).

[0095] Q-values ​​can directly indicate anomalies in underground fluids, lithology, or mineralization zones, and are of great significance for identifying certain types of ore bodies (such as hydrothermal ore bodies and stratabound sulfide ore bodies).

[0096] AVO / AVA attributes (Amplitude Versus Offset / Angle Attributes) are used to analyze the variation of seismic wave amplitude with source-receiver offset (AVO) or incident angle (AVA). Different lithologies, pore fluids, and porosity have a significant impact on the AVO / AVA response. By extracting AVO attributes such as intercept and gradient, rock physical parameters such as Poisson's ratio, density, and elastic modulus can be calculated. Some ore bodies, especially those with elastic properties significantly different from the surrounding rocks, may exhibit unique AVO / AVA anomalies.

[0097] AVO / AVA properties provide physical parameters that are more sensitive to lithology, fluids, and fractures, helping to distinguish ore bodies from surrounding rocks and assess the physical properties of ore bodies.

[0098] Curvature attributes measure the degree of curvature of seismic phase axes and are highly sensitive to identifying subtle structural features such as faults, fractures, fold axes, and salt dome boundaries. These structures are often favorable locations for ore body formation and enrichment.

[0099] Curvature properties can highlight fine structural features and help delineate structural ore-controlling factors related to ore bodies.

[0100] Texture attributes describe the local spatial variation patterns of seismic data using statistical methods (such as the Gray-Level Co-occurrence Matrix, GLCM), including contrast, energy, entropy, and homogeneity. These attributes can reflect the roughness, uniformity, and other characteristics of seismic facies, while different mineralization types and surrounding rocks may have unique seismic texture features.

[0101] Texture properties can capture subtle spatial variations in seismic response, providing additional information for identifying seismic facies and potential mineralization zones.

[0102] Elastic parameter inversion attributes are rock physical parameters such as P-wave impedance, S-wave impedance, density, Poisson's ratio, and Young's modulus that are directly obtained from seismic data using seismic inversion techniques (such as synchronous inversion and pre-stack inversion). These parameters are closer to the true physical properties of geological bodies than traditional seismic attributes, and are more sensitive to lithology, mineral composition, and fluids, making them strong indicative parameters for ore body identification.

[0103] Elastic parameter inversion properties can provide more direct and quantitative rock physics information, significantly improving the accuracy and reliability of ore body identification.

[0104] In other words, by setting earthquake attributes including basic earthquake attributes and time-delay physical attributes, this embodiment of the invention constructs a high-dimensional, highly discriminative feature space, comprehensively characterizing the seismic response features of underground geological bodies, and providing sufficient and high-quality input for subsequent intelligent interpretation.

[0105] Faced with the large number of seismic attribute parameters generated by the seismic data processing subsystem 200, directly applying them to supervised and unsupervised identification models would lead to the curse of dimensionality, increase computational complexity, and reduce the model's generalization ability. Therefore, to learn the most essential and discriminative feature representations from the data, in some embodiments of this invention, such as... Figure 1As shown, the seismic data analysis subsystem 300 includes a feature dimensionality reduction module 310. The feature dimensionality reduction module 310 is used to perform principal component analysis on the seismic attribute parameters, determine the variance explanation rate of each principal component, and determine whether the sum of the variance explanation rates of the first k principal components is greater than the variance explanation rate threshold. If not, the seismic attribute parameters are subjected to nonlinear feature dimensionality reduction based on a nonlinear feature dimensionality reduction model to obtain the dimensionality-reduced attribute parameters.

[0106] This invention first performs linear dimensionality reduction, followed by nonlinear dimensionality reduction, which improves dimensionality reduction efficiency while effectively removing redundant features. Furthermore, by implementing nonlinear dimensionality reduction, more abstract and expressive essential features than the original attributes can be learned, while filtering out redundant information and noise, thus enhancing the robustness of the features.

[0107] The nonlinear feature dimensionality reduction model can be at least one of the following: autoencoder, stacked autoencoders, denoising autoencoders, and variational autoencoders (VAEs).

[0108] An autoencoder is an unsupervised deep neural network consisting of an encoder and a decoder. The encoder maps high-dimensional seismic attribute data (input layer) to a low-dimensional latent space (bottleneck layer), forming a compressed representation or "encoder" of the data; the decoder then attempts to reconstruct the original data from this low-dimensional encoding. Through training, the network learns how to efficiently compress and decompress data, and the output of the bottleneck layer is a low-dimensional, nonlinear feature representation of the original data.

[0109] Stacked autoencoders are stacked layer by layer to form deeper nonlinear features.

[0110] Denoising autoencoders intentionally introduce noise into the input data and train the network to recover clean, original data from the noisy input, making the learned features more robust to noise.

[0111] Variational autoencoders learn the probability distribution of input data and can generate new features that are similar to the original data distribution, which helps to understand the feature space.

[0112] In some embodiments of the present invention, such as Figure 1 As shown, the vibration data analysis subsystem 300 also includes a ore body detection module 320; The orebody detection module 320 is used to determine overlapping and complementary regions based on the mineralization probability map and the anomaly distribution map, and to determine the comprehensive confidence level of the complementary region based on the first confidence level of the mineralization probability map and the second confidence level of the anomaly distribution map. The orebody detection module 320 is also used to determine the boundary lines and morphological parameters of overlapping and complementary regions based on threshold segmentation and morphological processing; Among them, the overlapping area is the area where the results of the mineralization probability map and the anomaly distribution map are consistent, and the complementary area is the area where the results of the mineralization probability map and the anomaly distribution map are inconsistent.

[0113] Specifically, mineralization probability thresholds and anomaly thresholds are set separately. Regions in the mineralization probability map that exceed the mineralization probability threshold and regions in the anomaly distribution map that exceed the anomaly threshold are segmented using these thresholds and marked as potential ore bodies. Then, morphological operations in image processing (such as dilation, erosion, opening, and closing operations) are applied to optimize the segmentation results, removing isolated noise points, connecting fractured ore body regions, and smoothing boundaries to make them more consistent with the morphological characteristics of geological bodies.

[0114] For each region with the obtained boundary line, calculate its three-dimensional volume, centroid coordinates, major axis direction, flattening ratio and other morphological parameters to provide basic data for resource estimation.

[0115] Specifically, the overall confidence level can be the weighted sum of the first confidence level and the second confidence level.

[0116] To maximize the advantages of expert experience, in some embodiments of the present invention, such as Figure 1 As shown, the vibration data analysis subsystem 300 is also used to optimize and train supervised and unsupervised recognition models based on the data interactively corrected by the interpretation and visualization subsystem 400.

[0117] This invention collects interactively corrected data as new, high-quality "human-annotated data." This new labeled data can be integrated into the training dataset for retraining or incremental learning of machine learning / deep learning models, forming an "expert-AI" closed-loop feedback mechanism. This allows the machine learning model to continuously learn from the expert's knowledge and experience, constantly improving its interpretive accuracy and generalization ability. Furthermore, when geological conditions in the exploration area are complex or new types of ore bodies appear, the system can quickly adapt to new geological conditions through expert correction and model retraining.

[0118] In some embodiments of the present invention, such as Figure 1 As shown, the control management subsystem 500 also includes a log management module 510 and a user permission management module 520; Log management module 510 is used to record the output data of the vibration data acquisition subsystem, vibration data processing subsystem, vibration data analysis subsystem, and interpretation and visualization subsystem; The User Permission Management Module 520 is used to define the access permissions and operation scope of different users.

[0119] Specifically, the log management module 510 is also used for detailed logging: recording all critical operations, data transmission, processing procedures, algorithm execution, parameter modifications, user login / logout, and all warning and error messages in the system. Log classification and filtering: supporting the classification, searching, and filtering of logs by time, module, and log level (information, warning, error), facilitating rapid problem location. Log storage and archiving: log data is securely stored and can be archived according to policies for long-term auditing and fault analysis. Visual analysis: providing log visualization analysis tools, such as trend charts and event distribution charts, to help identify system performance bottlenecks or potential risks.

[0120] By setting up a log management module 510, this embodiment of the invention enables system administrators to promptly identify and resolve potential problems through comprehensive monitoring and detailed log recording, thereby improving system availability and reliability and reducing operation and maintenance costs.

[0121] The user access control module 520 is responsible for defining, allocating, and managing the access permissions and operational scopes of different users in the system, ensuring data security and operational compliance. Specifically, it has the following functions: Role and Permission Definitions: The system has preset or allows administrators to define different user roles (such as system administrator, project manager, geophysical expert, geological engineer, operator, etc.), and each role is assigned different operating permissions: Data access permissions: Controls user permissions to read, modify, and delete raw data, processing results, and interpretation models.

[0122] Functional module access permissions: Controls whether users can access and use specific processing modules (such as model training and parameter optimization).

[0123] Operation permissions: Controls whether users can start / stop tasks, modify system configurations, generate reports, etc.

[0124] User authentication and authorization: Provides a secure login authentication mechanism (such as username / password, LDAP integration) and authorizes users based on their roles, restricting them to only performing permitted operations.

[0125] Audit trail: Records each user's actions, including time, content of the action, and objects modified, to facilitate post-event auditing and tracing.

[0126] Data encryption and security: Ensure data is encrypted during transmission and storage to prevent unauthorized access and data leakage.

[0127] This invention, through the user permission management module 520, ensures the security and integrity of system data, preventing unauthorized access and misoperation. Fine-grained permission control ensures that each user can only perform operations within their assigned scope of responsibility, improving system stability and management standardization.

[0128] In addition to the aforementioned functions, the control and management subsystem 500 also provides a centralized platform for managing all configurable parameters within the system and integrates an intelligent optimization mechanism to improve processing performance. Specifically, it allows users to centrally configure parameters for all submodules via a graphical interface or configuration files, including: Acquisition parameters: sampling rate, pulse repetition frequency, and measurement point spacing of the fiber optic demodulator.

[0129] Preprocessing parameters: filter type, cutoff frequency, denoising algorithm threshold, gain curve, etc.

[0130] Attribute extraction parameters: frequency range of spectral decomposition, offset range of AVO analysis, etc.

[0131] Machine learning / deep learning model parameters: learning rate, batch size, model architecture (number of layers, number of neurons), optimizer selection, regularization parameters, training epochs, convergence threshold, etc.

[0132] Visualization parameters: color mapping, transparency, slice orientation, etc.

[0133] Preset parameters and templates: Supports saving and loading commonly used parameter configuration templates, making it convenient to apply them quickly in different projects or under different geological conditions.

[0134] Adaptive parameter adjustment and optimization: Integrates advanced optimization algorithms to achieve intelligent recommendation and adaptive adjustment of processing parameters. Bayesian optimization: For hyperparameter tuning of deep learning models, Bayesian optimization can intelligently select the next set of hyperparameters to try based on historical evaluation results, so as to find the optimal configuration more quickly.

[0135] Genetic algorithms / particle swarm optimization: can be used to optimize complex geophysical inversion parameters or denoising algorithm parameters to maximize a certain objective function (such as signal-to-noise ratio, model fit).

[0136] Rule-based expert systems: Based on empirical rules or preset conditions input by geological experts, the system can automatically adjust certain processing parameters.

[0137] This configuration significantly enhances the system's flexibility and adaptability. Through intelligent optimization algorithms, the system can automatically find the optimal combination of processing parameters, eliminating the need for extensive manual trial and error, thereby improving processing efficiency and reducing reliance on expert experience.

[0138] To monitor the system's operational status, the control and management subsystem 500 also performs real-time monitoring and records all critical events and operations in detail to ensure stable system operation and troubleshooting. Specifically, the following states can be monitored in real time: Hardware resource monitoring: Real-time display of hardware resource usage such as CPU utilization, GPU utilization, memory usage, hard disk I / O, and network bandwidth.

[0139] Software process monitoring: Monitor the running status (startup, running, stopping, abnormal) of each subsystem and module, and identify zombie processes or crashed modules.

[0140] Task progress monitoring: Displays a list of currently executing tasks, completed tasks, and pending tasks, and provides detailed progress bars or percentage displays.

[0141] Data flow monitoring: Tracks the flow of data in the system in real time, displaying data volume, transmission rate, etc.

[0142] Alarms and notifications: When system resource usage reaches a threshold, a task fails, an abnormal error occurs, or a specific event occurs, the system can send real-time alarms to administrators or relevant users via email, SMS, screen notifications, etc.

[0143] In summary, the Control and Management Subsystem 500, with its powerful task scheduling, parameter management, status monitoring, and access control capabilities, provides a stable, efficient, and secure operating environment for the entire distributed fiber optic vibration sensing intelligent data processing and interpretation system. It is a key support for achieving system automation, intelligence, and maintainability, ensuring that all functional modules can work together consistently to serve mineral exploration.

[0144] In summary, the mineral exploration system for distributed seismic sensing proposed in this invention utilizes optical fiber as a continuous linear sensor to achieve ultra-high-density spatial sampling, acquiring detailed subsurface medium response data that is unmatched by traditional geophone arrays. This ensures that even weak or localized seismic response signals of ore bodies can be completely captured, avoiding data omissions and spatial aliasing problems caused by discrete sampling. It significantly improves the spatial resolution and sampling density of seismic data, providing richer and more accurate raw information for subsequent intelligent analysis, which is a prerequisite for identifying hidden ore bodies. In the raw data preprocessing stage, advanced algorithms such as adaptive filtering, sparse representation denoising, and deep learning denoising networks (such as U-Net) are employed to more effectively suppress inherent environmental noise, non-seismic interference, and system noise in distributed optical fiber data, significantly improving the signal-to-noise ratio of the target signal. In the seismic attribute extraction stage, in addition to traditional attributes such as amplitude, frequency, and phase, methods such as wavelet transform, spectral decomposition, and Q-value inversion are used to extract deeper and more sensitive seismic attributes related to the physical properties of ore bodies (such as density, wave velocity, and elastic modulus), forming rich feature vectors. This invention significantly improves the signal-to-noise ratio of the original data, highlighting weak orebody signals that were previously obscured by noise. Simultaneously, the extracted multi-dimensional, high-discrimination seismic attributes provide high-quality input features for subsequent intelligent interpretation, forming the foundation for accurate orebody identification by machine learning models. In actual exploration, unsupervised learning is used to identify anomalous areas significantly different from known patterns in the training data; these areas may indicate unknown orebody types or occurrence patterns. This achieves high automation, high precision, and high efficiency in the mineral exploration interpretation process. It significantly reduces reliance on expert experience, overcoming the subjectivity and limitations of manual interpretation. It can identify concealed orebody, thin orebody, or orebody within complex structures that are difficult to discover using traditional methods, significantly improving the exploration success rate and shortening the interpretation cycle from weeks / months to days or even real-time. In summary, this invention provides an unprecedented high-resolution, high-efficiency, high-precision, and highly intelligent solution for mineral exploration, with significant positive effects in terms of economic, resource, and environmental benefits.

[0145] On the other hand, embodiments of the present invention also provide a mineral exploration method for distributed vibration sensing, such as... Figure 4 As shown, mineral exploration methods based on distributed vibration sensing include: S401. Acquire raw seismic data of geological bodies based on distributed optical fibers deployed in the exploration area; S402. Preprocess the raw seismic data to obtain the target seismic data, and extract the seismic attribute parameters of the target seismic data. S403. Input the seismic attribute parameters into the supervised and unsupervised identification models respectively to obtain the mineralization probability map and anomaly distribution map. Then, cross-validate the mineralization probability map and anomaly distribution map to generate the three-dimensional ore body detection results. S404. Visualize and interactively correct the three-dimensional ore body detection results to generate a mineral exploration report.

[0146] The mineral exploration method for distributed vibration sensing provided in the above embodiments can realize the technical solutions described in the above embodiments of the mineral exploration system for distributed vibration sensing. The specific implementation principles of each module or unit can be found in the corresponding content in the above embodiments of the mineral exploration system for distributed vibration sensing, which will not be repeated here.

[0147] The above provides a detailed description of a mineral exploration system and method for distributed vibration sensing provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A mineral exploration system for distributed vibration sensing, characterized in that, include: Vibration data acquisition subsystem, vibration data processing subsystem, vibration data analysis subsystem, interpretation and visualization subsystem, and control and management subsystem; The vibration data acquisition subsystem is used to acquire raw vibration data of geological bodies based on distributed optical fibers deployed in the exploration area. The seismic data processing subsystem is used to preprocess the raw seismic data to obtain target seismic data and extract the seismic attribute parameters of the target seismic data. The seismic data analysis subsystem is used to input the seismic attribute parameters into the supervised identification model and the unsupervised identification model respectively, to obtain the mineralization probability map and the anomaly distribution map, and to cross-validate the mineralization probability map and the anomaly distribution map to generate a three-dimensional ore body detection result. The interpretation and visualization subsystem is used to visualize and interactively correct the three-dimensional ore body detection results, and generate a mineral exploration report. The control and management subsystem is used to control the working order of the vibration data acquisition subsystem, vibration data processing subsystem, vibration data analysis subsystem, and interpretation and visualization subsystem.

2. The mineral exploration system for distributed vibration sensing according to claim 1, characterized in that, The vibration data processing subsystem includes a noise reduction module, which includes a data format conversion unit, a primary noise reduction unit, and an adaptive noise reduction unit. The data format conversion unit is used to convert the format of the raw vibration data from the original format to the target format. The primary denoising unit performs bandpass filtering on the original vibration data after format conversion to obtain coarse-filtered vibration data, and determines whether the coarse-filtered vibration data has single-frequency interference or coherent noise. When single-frequency interference exists, notch filtering is performed on the coarse-filtered vibration data to obtain primary-filtered vibration data. When coherent noise exists, FK filtering is performed on the coarse-filtered vibration data to obtain primary-filtered vibration data. The adaptive denoising unit is used to determine whether there is a usable reference noise channel in the primary filtered vibration data. If there is, adaptive noise cancellation processing is performed on the primary filtered vibration data. If not, it is determined whether there is an independent noise source in the primary filtered vibration data. If there is an independent noise source, independent component analysis is performed on the primary filtered vibration data to obtain fine filtered vibration data. If there is no independent noise source, the primary filtered vibration data is denoised based on principal component analysis, sparse transform domain denoising, or a deep learning denoising model to obtain fine filtered vibration data.

3. The mineral exploration system for distributed vibration sensing according to claim 2, characterized in that, The vibration data processing subsystem also includes a gain compensation module, which includes a spherical diffusion compensation unit, a Q-value attenuation compensation unit, and an automatic gain unit. The spherical diffusion compensation unit is used to determine the spherical diffusion compensation gain factor, and the product of the fine-filtered vibration data and the spherical diffusion compensation gain factor is used as the first compensated vibration data. The Q-value attenuation compensation unit is used to construct a Q-value attenuation compensation filter based on a pre-established Q-value model, and input the first compensated vibration data into the Q-value attenuation compensation filter to obtain the second compensated vibration data. The automatic gain control unit is used to normalize the second compensated vibration data based on a preset sliding time window to obtain gain vibration data.

4. The mineral exploration system for distributed vibration sensing according to claim 3, characterized in that, The vibration data processing subsystem further includes an interpolation module, which is used to perform interpolation processing on the gain vibration data based on an interpolation algorithm to obtain the target vibration data; the interpolation algorithm is at least one of linear interpolation, spline interpolation, K-nearest neighbor interpolation, sparse reconstruction-based interpolation, and deep learning-based interpolation.

5. The mineral exploration system for distributed vibration sensing according to claim 1, characterized in that, The seismic attribute parameters include basic seismic attributes and rock physical attributes. The basic seismic attributes include amplitude, frequency, phase, instantaneous amplitude, instantaneous frequency, and instantaneous phase. The rock physical attributes include coherence attributes, spectral decomposition attributes, Q-value attributes, AVO / AVA attributes, curvature attributes, texture attributes, and elastic parameter inversion attributes.

6. The mineral exploration system for distributed vibration sensing according to claim 1, characterized in that, The seismic data analysis subsystem includes a feature dimensionality reduction module, which is used to perform principal component analysis on the seismic attribute parameters, determine the variance explanation rate of each principal component, and determine whether the sum of the variance explanation rates of the first k principal components is greater than the variance explanation rate threshold. If not, nonlinear feature dimensionality reduction is performed on the seismic attribute parameters based on a nonlinear feature dimensionality reduction model to obtain dimensionality-reduced attribute parameters.

7. The mineral exploration system for distributed vibration sensing according to claim 1, characterized in that, The vibration data analysis subsystem also includes a ore body detection module; The orebody detection module is used to determine overlapping and complementary regions based on the mineralization probability map and the anomaly distribution map, and to determine the comprehensive confidence level of the complementary region based on the first confidence level of the mineralization probability map and the second confidence level of the anomaly distribution map. The ore body detection module is also used to determine the boundary lines and morphological parameters of the overlapping region and the complementary region based on threshold segmentation and morphological processing. The overlapping region is the region where the results of the mineralization probability map and the anomaly distribution map are consistent, and the complementary region is the region where the results of the mineralization probability map and the anomaly distribution map are inconsistent.

8. The mineral exploration system for distributed vibration sensing according to claim 1, characterized in that, The vibration data analysis subsystem is also used to optimize and train the supervised and unsupervised recognition models based on the data interactively corrected by the interpretation and visualization subsystem.

9. The mineral exploration system for distributed vibration sensing according to claim 1, characterized in that, The control and management subsystem also includes a log management module and a user permission management module; The log management module is used to record the output data of the vibration data acquisition subsystem, vibration data processing subsystem, vibration data analysis subsystem, and interpretation and visualization subsystem. The user permission management module is used to define the access permissions and operation scope of different users.

10. A mineral exploration method for distributed vibration sensing, characterized in that, include: Raw seismic data of geological bodies were acquired using distributed optical fibers deployed in the exploration area. The raw seismic data is preprocessed to obtain target seismic data, and the seismic attribute parameters of the target seismic data are extracted. The seismic attribute parameters are input into the supervised and unsupervised identification models respectively to obtain mineralization probability maps and anomaly distribution maps. The mineralization probability maps and anomaly distribution maps are then cross-validated to generate three-dimensional ore body detection results. The results of the three-dimensional ore body detection are visualized and interactively corrected to generate a mineral exploration report.

Images