Intelligent construction system of deposit model space data sample library

Abstract

The application provides a deposit model space data sample library intelligent construction system, and the system comprises: a deposit model sample library management module, which is used for constructing a source space database, publishing tasks to a sample library construction workstation module, monitoring the published tasks, reprocessing the tasks delivered by the sample library construction workstation module, and constructing a deposit model sample library; the sample library construction workstation module is used for processing the published tasks and delivering the completed tasks to the deposit model sample library management module; and a delineation of a favorable prospecting area module is used for obtaining data in the deposit model sample library for processing, delineating a favorable prospecting area, and guiding the tasks of the sample library construction workstation module. Through standardized processing, structured storage and intelligent management of multi-dimensional geological data of typical deposits, a standardized sample set which can be used for machine learning, deep learning or expert system training is formed, and data support is provided for the delineation of a favorable prospecting area in an unknown area.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of spatial data sample library construction, and particularly relates to a deposit model spatial data sample library intelligent construction system. BACKGROUND

[0002] Under the background of increasing global resource demand, the exploration and development of mineral resources are crucial to ensuring national economic security and social sustainable development. Prospecting prediction, as an important preliminary work of mineral exploration, aims to delineate favorable areas with potential ore prospecting value through analysis of geological, geophysical, geochemical and other multi-source data. However, the traditional prospecting prediction method has problems such as scattered data, chaotic management, unscientific task allocation and inaccurate ore prospecting target positioning, which seriously restricts the accuracy and efficiency of prospecting prediction. SUMMARY

[0003] The present application aims to provide a deposit model spatial data sample library intelligent construction system, which standardizes, structures and intelligently manages multi-dimensional geoscience data of typical deposits, forms a standardized sample set for machine learning, deep learning or expert system training, and provides data support for delineating favorable areas in unknown regions.

[0004] To achieve the above-mentioned purpose, the present application provides a deposit model spatial data sample library intelligent construction system, which comprises:

[0005] A deposit model sample library management module is used to construct a source spatial database, publish tasks to a sample library construction workbench module, monitor the published tasks, reprocess the tasks delivered by the sample library construction workbench module, and construct a deposit model sample library.

[0006] A sample library construction workbench module is used to process the published tasks and deliver the completed tasks to the deposit model sample library management module.

[0007] A delineation of favorable ore areas module is used to acquire data in the deposit model sample library for processing, delineate favorable ore areas, and guide the tasks of the sample library construction workbench module.

[0008] Further, the deposit model sample library management module comprises:

[0009] A data acquisition unit is used to acquire typical deposit spatial data, analyze and identify heterogeneous data in the typical deposit spatial data through a multi-modal data adaptive acquisition engine, and align the typical deposit spatial data to a unified geographic coordinate system.

[0010] A source space database unit is configured to manage typical deposit space data in a deposit genetic type classification system, including establishing a classification system and data maintenance and update;

[0011] A task publishing unit is configured to publish space data sample production tasks to the sample library construction workbench module for the typical deposit space data in the source space database unit;

[0012] A task monitoring unit is configured to monitor the progress and workload of all sample production tasks in the sample library construction workbench module;

[0013] A task delivery detection unit is configured to receive tasks delivered by the sample library construction workbench module and perform quality detection on the delivered tasks;

[0014] A deposit model sample warehouse unit is configured to reprocess the typical deposit space data after sample production, construct a deposit model sample library, and perform visual display.

[0015] Further, the source space database unit includes:

[0016] A hierarchical classification management subunit is configured to classify typical deposit space data according to a classification system established based on deposit genetic types;

[0017] A space data lake warehouse integrated management subunit is configured to establish a unified data storage format and standard, and disperse the data in the hierarchical classification management unit for storage in multiple nodes through distributed storage technology;

[0018] A data dynamic real-time update and maintenance subunit is configured to record the update history of typical deposit space data by establishing a data version control mechanism, and repair and update errors and abnormalities in the typical deposit space data in the nodes using a data detection algorithm.

[0019] Further, the task publishing unit is specifically configured to perform the following operations:

[0020] Analyzing the data in the source space database and extracting key features of the data;

[0021] Processing the key features through a machine learning algorithm to obtain processing task requirements for each data;

[0022] Assigning the processing task requirements to the corresponding processing units of the sample library construction workbench module through a matching algorithm.

[0023] Further, the deposit model sample warehouse unit includes:

[0024] A sample temporary management subunit is configured to classify and manage the typical deposit space data after sample production;

[0025] The sample data processing subunit is configured to perform format conversion on all typical deposit spatial data in the sample temporary management subunit by using a vector-to-raster algorithm, and perform block segmentation on the converted data according to a unified size specification, so as to form a single-channel sample set and a multi-channel sample set.

[0026] The sample data enhancement subunit is configured to calculate a covariance matrix of data in each single-channel sample set and multi-channel sample set respectively, apply noise distribution disturbance to each data, and perform whitening / anti-whitening processing through the covariance matrix, and finally perform geometric transformation synchronously.

[0027] The deposit model sample management subunit is configured to construct a deposit model sample library, establish data description and metadata information for each single-channel data set, integrate and associate the multi-channel data set by using a data fusion technology, and form a multi-channel block data set.

[0028] The deposit model sample visualization subunit is configured to perform visual display of the deposit model sample library by using cloud computing and big data processing technology, in combination with a GIS and a Web visualization engine.

[0029] Further, the sample library construction workbench module comprises:

[0030] The data visualization unit is configured to perform visualization on the deposit model spatial data in the publishing task, and perform visualization on the data in the positive sample manufacturing, negative sample manufacturing and labeling processes.

[0031] The positive sample manufacturing unit is configured to manufacture positive samples based on the typical deposit spatial data in the publishing task.

[0032] The negative sample manufacturing unit is configured to manufacture negative samples based on the typical deposit spatial data in the publishing task.

[0033] The quality detection unit is configured to perform quality inspection on the completed typical deposit spatial data, and if the quality inspection result is unqualified, the positive sample manufacturing or the negative sample manufacturing is performed again, and if the quality inspection result is qualified, the completed typical deposit spatial data is delivered to the deposit model sample library management module.

[0034] Further, the positive sample manufacturing unit comprises:

[0035] The manual labeling subunit is configured to perform manual labeling on the typical deposit spatial data in the publishing task.

[0036] The automatic labeling subunit is configured to perform automatic labeling on the typical deposit spatial data in the publishing task.

[0037] The intelligent labeling subunit is configured to perform intelligent labeling on the typical deposit spatial data in the publishing task.

[0038] Further, the positive sample manufacturing unit is specifically configured to perform the following operations:

[0039] According to the preset rule, the typical deposit spatial data in the publishing task is allocated to different labeling units;

[0040] The typical deposit spatial data with clear structure and clear rules is allocated to the automatic labeling subunit, and the allocated data is labeled by an algorithm based on the preset rule;

[0041] The typical deposit spatial data remaining and with automatic labeling confidence lower than the preset threshold is allocated to the intelligent labeling subunit, and the allocated data is labeled by a deep learning model;

[0042] The typical deposit spatial data with intelligent labeling subunit labeling confidence lower than the preset threshold is allocated to the manual labeling subunit, and the allocated data is labeled by an expert;

[0043] The results of manual labeling, automatic labeling and intelligent labeling are integrated, the area of each positive sample region is calculated, and the positive sample total area is obtained.

[0044] Further, the negative sample manufacturing unit is specifically configured to perform the following operations:

[0045] Determine the distribution range of the abnormal area of each layer of the typical deposit spatial data in the publishing task, and fuse the abnormal area distribution of each layer to construct a multi-dimensional geological feature space;

[0046] For each layer, use a preset anomaly recognition algorithm to identify the abnormal area in each layer, and perform spatial superposition sum on the abnormal areas of all layers to obtain the comprehensive coverage area of all abnormal areas;

[0047] Perform spatial difference operation on the overall range of the study area and the comprehensive coverage area to obtain the blank area not covered by the abnormal area, and the blank area is the potential spatial distribution range of the negative sample;

[0048] According to the total area of the positive sample, determine the target value of the total area of the negative sample, set the range of the negative sample data in the blank area, and automatically calculate the spatial distribution of the negative sample in the blank area using a random function.

[0049] Further, the delineation of the ore-prospecting favorable area module includes:

[0050] The intelligent algorithm library unit is configured to train the data in the deposit model sample library by machine learning algorithms and deep learning algorithms to obtain a plurality of AI ore-prospecting favorable area delineation models, and guide the sample library construction workbench module by the AI ore-prospecting favorable area delineation model;

[0051] The deposit model sample processing unit is configured to analyze and process the data in the deposit model sample library to obtain a prediction data set;

[0052] The ore-prospecting favorable area delineation unit is used for inputting a prediction data set into an AI ore-prospecting favorable area delineation model for reasoning to delineate an ore-prospecting favorable area.

[0053] Compared with the prior art, the present application has the following advantages:

[0054] The mineral deposit model spatial data sample library intelligent construction system provided by the present application has the advantages that the mineral deposit model sample library management module is dedicated to the construction of a source spatial database, task publishing and monitoring, and post-processing work after task delivery, and the entire sample library construction process is macroscopically controlled through centralized management. The sample library construction workbench module is dedicated to the specific processing of tasks, and the results are accurately delivered through the operation of various complex data processing work, thereby improving the efficiency of sample library construction and providing a high-quality data basis for subsequent mineral deposit model construction. The ore-prospecting favorable area delineation module uses the high-quality data processed in the mineral deposit model sample library management module, combines analysis algorithms, and quickly delineates an area with potential mineral resources, thereby shortening the ore-prospecting target area prediction cycle and improving the scientificity and reliability of the prediction results. Meanwhile, the task guidance function can feed back the data requirements found in the delineation process to the sample library construction workbench module, thereby providing a direction for the optimization of the subsequent sample library. The present application solves the problems of low efficiency, dispersion, and disconnection in the construction of a mineral deposit model knowledge base, establishes a knowledge closed-loop feedback mechanism, significantly improves the intelligent ore-prospecting prediction efficiency and precision, and promotes the intelligentization and knowledge upgrading of mineral exploration. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of the provided drawings.

[0056] Figure 1 A structure schematic diagram of the mineral deposit model spatial data sample library intelligent construction system provided by the present application is shown in the accompanying drawings. DETAILED DESCRIPTION

[0057] The present application will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present application, and not to limit the present application. In addition, it should be noted that, for the convenience of description, only the parts related to the present application are shown in the drawings, rather than all the structures.

[0058] Reference Figure 1 The present embodiment provides a mineral deposit model spatial data sample library intelligent construction system, which comprises:

[0059] The deposit model sample library management module is configured to construct a source space database, publish tasks to the sample library construction workstation module, monitor the published tasks, and reprocess the tasks delivered by the sample library construction workstation module to construct the deposit model sample library.

[0060] The sample library construction workstation module is configured to process the published tasks and deliver the completed tasks to the deposit model sample library management module.

[0061] The favorable ore-prospecting area delineation module is configured to obtain data in the deposit model sample library, process the data, delineate a favorable ore-prospecting area, and guide the sample library construction workstation module.

[0062] In this embodiment, the deposit model sample library management module is configured to manage the source space database and the deposit model sample library, including constructing the source space database, publishing and monitoring tasks, and delivering and reprocessing tasks to construct the deposit model sample library. First, typical deposit spatial data is collected and standardized to construct the source space database. Then, tasks are planned and published to the sample library construction workstation module, and the task execution progress and status are monitored in real time. Finally, the delivered task results are received, data quality is checked and reprocessed, and the deposit model sample library is constructed to provide accurate data support for subsequent ore-prospecting prediction.

[0063] The sample library construction workstation module is configured to receive and process tasks, process and analyze data, and complete and deliver tasks. First, the tasks published by the deposit model sample library management module are received, and the required data is extracted and prepared from the source space database. Then, the data is processed and analyzed according to the task requirements, and the data is processed using corresponding algorithms. Finally, the processed and analyzed results are delivered to the deposit model sample library management module to assist in the construction and improvement of the deposit model sample library.

[0064] The favorable ore-prospecting area delineation module is configured to obtain data and preprocess data, delineate a favorable ore-prospecting area, and guide and feedback tasks. The module first obtains multi-source data from the deposit model sample library and preprocesses the data. Then, the data is analyzed using data mining algorithms to delineate a favorable ore-prospecting area. Finally, task guidance is provided based on the delineation results for the sample library construction workstation module, and feedback information is collected to optimize the delineation method, thereby improving the accuracy and efficiency of ore-prospecting work.

[0065] As a preferred embodiment, the deposit model sample library management module includes:

[0066] The data collection unit is configured to collect typical deposit spatial data, analyze and identify heterogeneous data in the typical deposit spatial data through a multi-modal data adaptive collection engine, and align the typical deposit spatial data to a unified geographic coordinate system.

[0067] A source space database unit is configured to manage typical deposit space data in a deposit genetic type classification system, including establishing a classification system and data maintenance and update.

[0068] A task publishing unit is configured to publish space data sample production tasks to the sample library construction workbench module for the typical deposit space data in the source space database unit.

[0069] A task monitoring unit is configured to monitor the progress and workload of all sample production tasks in the sample library construction workbench module.

[0070] A task delivery detection unit is configured to receive tasks delivered by the sample library construction workbench module and perform quality detection on the delivered tasks.

[0071] A deposit model sample warehouse unit is configured to reprocess the typical deposit space data after sample production, construct a deposit model sample library, and perform visual display.

[0072] In this embodiment, the data collection unit collects typical deposit space data into a database, including structure, stratum, rock mass, geochemical exploration, geophysical exploration, remote sensing, heavy sand, ore body, mineralization, alteration, and lithology. A multi-modal data adaptive collection engine is used to automatically identify points, lines, surface elements, and raster data of different structures through heterogeneous data analysis, to intelligently register coordinates in space, to automatically match a unified map projection coordinate system, to ensure the accuracy and integrity of the collected data, and to improve data quality.

[0073] The source space database unit manages typical deposit space data, and the typical deposit model data is classified and managed in units of deposit genetic types, including classification establishment, data maintenance and update, and management of data types including vector data and raster data.

[0074] The task publishing unit first obtains space data of typical deposits from the source space database unit, plans and defines specific space data sample production tasks according to the requirements of deposit model sample library construction, clearly defines the objectives and requirements of the tasks, and then publishes these tasks to the sample library construction workbench module through a system interface or interface.

[0075] The task monitoring unit includes progress monitoring and workload monitoring. The progress monitoring unit tracks the completion of all sample production tasks in real time, counts the number of completed sample production tasks, and calculates the completion rate. The number of completed sample production tasks and the completion rate are counted for different types of samples, which are used to analyze the progress of sample production work for different types of samples, to find possible differences in progress, and to adjust resource allocation in a timely manner.

[0076] Workload monitoring, statistics of sample production task quantity of each ore, and calculation of its proportion in total task quantity, clear understanding of the distribution of sample production workload of different ores. Similarly, according to the combination of ore and genetic type, the number statistics and proportion calculation are carried out, and the distribution of sample production workload of different types of ore deposit is analyzed, so as to reasonably arrange resources and adjust task plan.

[0077] Delivery task detection unit ensures high quality delivery of sample production, improves data reliability, including:

[0078] Intelligent acceptance, firstly, machine learning algorithm is adopted to automatically check sample production space data, compare sample production space data with original data in source space database, detect data integrity, accuracy and consistency. For example, check whether the annotation of structure, stratum and other space data is consistent with the actual geological characteristics, whether the spatial distribution of data is reasonable, etc.

[0079] Expert assisted review: on the basis of automatic preliminary inspection, expert assisted review mechanism is introduced. Geological experts sample review sample production space data, focus on sample production of complex geological features and key data, and put forward review opinions.

[0080] Feedback and correction: for the problems found in the acceptance process, detailed feedback report is automatically generated, and the report is sent to the sample production personnel. The sample production personnel correct the sample production space annotation data according to the feedback report, and submit it for acceptance again until it passes the acceptance.

[0081] Data submission: if the sample production space data passes the acceptance, the system will automatically submit the sample production completed data package to the "sample temporary management sub unit" by using data privacy calculation encryption technology, ensure the safety and non tamperability of the data, and record the submission time and related information.

[0082] Ore deposit model sample warehouse unit, classified management of completed sample production data, using ore, ore deposit genetic type for classification and organization management, including sample rasterization standard block, sample data enhancement, ore deposit model sample library management and visualization processing.

[0083] As a preferred embodiment, the source space database unit includes:

[0084] Hierarchical classification management sub unit, used for establishing classification system according to ore deposit genetic type and classifying typical ore deposit space data.

[0085] Space data lake warehouse integrated management sub unit, used for establishing unified data storage format and standard, and dispersing the data in hierarchical classification management unit in multiple nodes through distributed storage technology.

[0086] The data dynamic real-time updating and maintaining subunit is configured to record the updating history of the typical deposit spatial data by establishing a data version control mechanism and repair and update errors and abnormalities in the typical deposit spatial data in the node by using a data detection algorithm.

[0087] In this embodiment, the hierarchical classification management subunit establishes a classification system according to different deposit genetic types, such as endogenic deposits, exogenic deposits and metamorphic deposits, and accurately classifies various spatial data.

[0088] The spatial data lake warehouse integrated management subunit realizes efficient storage and management of different types of data by establishing a unified data storage format and standard. At the same time, the distributed storage technology is used to store the data on multiple nodes, improve the storage capacity and access speed of the data, and enhance the security and reliability of the data.

[0089] The data dynamic real-time updating and maintaining subunit realizes comprehensive management of the data structure and content of the spatial data layers such as structure and stratum for the typical deposit model spatial data, including updating and maintaining work such as data addition, modification and deletion. By establishing a data version control mechanism, the updating history of the data is recorded to ensure the traceability of the data. At the same time, by using an automatic data detection algorithm, errors and abnormalities in the data are found in time, and automatic repair and update are performed, thereby improving the efficiency and accuracy of data updating and maintenance.

[0090] As a preferred embodiment, the task publishing unit is specifically configured to perform the following operations:

[0091] Analyzing the data in the source spatial database and extracting key features of the data.

[0092] Processing the key features by a machine learning algorithm to obtain processing task requirements of each data.

[0093] Assigning the processing task requirements to the processing units corresponding to the sample library construction workbench module by a matching algorithm.

[0094] In this embodiment, the task publishing unit first extracts key features of the typical deposit model spatial data set in the source spatial data warehouse, such as data type, data volume, complexity, data correlation and the like. These features are processed by a machine learning algorithm to understand the unique requirements of each data set. Then, according to the data feature analysis result, a personalized sample making task is automatically generated. Considering the skill level, experience and work load of the sample makers, the task is reasonably assigned to the appropriate labeling personnel by using an optimization algorithm. At the same time, the sample makers obtain the task through encrypted identity verification to ensure data security and fairness of task allocation.

[0095] As a preferred embodiment, the deposit model sample warehouse unit comprises:

[0096] The sample temporary management subunit is used for classified management of typical deposit spatial data completed by sample making.

[0097] The sample data processing subunit is used for cutting and segmentation of all typical deposit spatial data in the sample temporary management subunit in a unified size after format conversion through a vector-to-raster algorithm, to form a single-channel sample set and a multi-channel sample set.

[0098] The sample data enhancement subunit is used for calculation of a covariance matrix of data in each single-channel sample set and multi-channel sample set, respectively, to apply noise distribution disturbance to each data, and to perform whitening / anti-whitening processing through the covariance matrix, and finally to perform geometric transformation synchronously.

[0099] The deposit model sample management subunit is used for construction of a deposit model sample library, establishment of data description and metadata information for each single-channel data set, integration and correlation of multi-channel data sets through data fusion technology, and formation of a multi-channel cutting data set.

[0100] The deposit model sample visualization subunit is used for visualization display of the deposit model sample library through cloud computing and big data processing technology, in combination with GIS and Web visualization engine.

[0101] In the embodiment, the sample temporary management subunit realizes classified management of typical deposit spatial data completed by sample making, and classified organization management is performed according to ore type and deposit genetic type.

[0102] The sample data processing subunit performs unified data analysis and processing on all typical deposit spatial data of the sample temporary management subunit, and realizes format conversion through a vector-to-raster algorithm of dynamic programming adaptive resolution by cooperative calculation of MPI scheduling CPU and GPU, and then performs cutting in a unified size (for example, 28*28 pixels) to form a single-channel sample set and a multi-channel sample set. The single-channel sample set is a cutting data set of a single raster element of certain typical deposit spatial data, for example, a structure single-channel data set or an Au element anomaly single-channel data set. The multi-channel sample set is a cutting data set of n elements of certain typical deposit spatial data, for example, a structure, rock mass, Au element anomaly, and magnetic anomaly four-channel cutting data set.

[0103] The sample data augmentation subunit targets input data with multiple dimensions or modalities. It generates high-quality augmented samples to improve the model's generalization ability by preserving inter-channel correlations, mining cross-channel features, or introducing novel transformation strategies. Using multi-factor sample data augmentation technology, based on the spatial data of the ore deposit model, simulated noise is added to the ore-forming elements of the model, and the correlation between elements is maintained through single-factor covariance matrix constraints. First, the covariance matrix of each ore-forming element data is calculated to capture the correlation between elements. Second, a perturbation conforming to the noise distribution is applied to each element, and covariance matrix whitening / dewhitening processing is used to ensure that the perturbed data still conforms to the real data distribution. Then, a geometric transformation is performed simultaneously to maintain spatial consistency. Finally, a cross-scale CycleGAN sample data augmentation technique is used to construct a dual-generator network to convert a small-scale data set of a certain ore-forming element to a large-scale data set, compensating for insufficient samples, and ensuring the reversibility of the transformation through cycle consistency loss.

[0104] The ore deposit model sample management subunit implements a classification management system and data management technology, enabling efficient management of various complex data types and datasets from different channels. This improves the efficiency of data retrieval, retrieval, and analysis, while a data access control mechanism ensures data security and privacy. Specifically, it includes:

[0105] The classification management system, which classifies and manages the labeled results, is based on mineral type and genetic type. For sample data of different mineral types and genetic types, separate storage directories and indexes are established.

[0106] Single-channel data sample set management establishes detailed data descriptions and metadata for each single-channel dataset, including data source, acquisition time, and data accuracy. Users can quickly locate and retrieve the required single-channel datasets through keyword search and conditional filtering.

[0107] Multi-channel data sample set management employs data fusion technology to integrate and correlate datasets from different channels, forming multi-channel chunked datasets. Users can intuitively view the relationships between multi-channel data through visualization tools and conduct comprehensive analysis.

[0108] To ensure data security and privacy, a data access control mechanism is introduced, which grants tiered access to the labeled data based on the user's identity and permissions.

[0109] The mineral deposit model sample visualization sub-unit, based on cloud computing and big data processing technologies, combined with GIS and Web visualization engines, enables data querying and visualization, including innovative technologies for global visualization, innovative technologies for visualization of source spatial data of typical mineral deposit models, and innovative technologies for visualization of sample library data.

[0110] As a preferred embodiment, the sample library construction workstation module comprises:

[0111] A data visualization unit is configured to visualize the deposit model spatial data in the publishing task and the data in the positive sample making, negative sample making and labeling process.

[0112] A positive sample making unit is configured to make positive samples based on the typical deposit spatial data in the publishing task.

[0113] A negative sample making unit is configured to make negative samples based on the typical deposit spatial data in the publishing task.

[0114] A quality detection unit is configured to check the quality of the labeled typical deposit spatial data, and if the quality is not qualified, the positive sample making or the negative sample making is performed again, and if the quality is qualified, the labeled typical deposit spatial data is delivered to the deposit model sample library management module.

[0115] In this embodiment, the general tools used by the sample library construction workstation in the processing of the publishing task include a structure data processing tool, a structure data processing tool, a stratum processing tool, a mineralization information extraction tool, a point conversion field gridding tool, an editing processing tool and a spatial analysis tool. The positive sample making and labeling tools include a manual positive sample making and labeling tool, an automated positive sample making and labeling tool and an intelligent positive sample making and labeling tool. After the labeling, the label management realizes the quick query, viewing, modification and deletion of the input label information through a scientific classification management system and efficient operation, improves the efficiency of data management, and provides flexible expansion ability for users, which can meet the individualized needs of different users. When inputting the label, optional item recommendations are provided, which reduces the time and error rate of manual label input, improves the accuracy and efficiency of labeling, and includes classification management, operation function, custom expansion and intelligent recommendation.

[0116] The data visualization unit uses professional visualization tools and technologies, such as geographic information system software and three-dimensional visualization engines, to display the deposit model spatial data in the publishing task and the data in the positive sample making, negative sample making and labeling process in the form of intuitive graphics, charts and maps. By setting different colors, symbols and layers, the spatial distribution of the deposit, the geological structure characteristics, the distribution of sample data and the labeling information are clearly presented, so that users can quickly understand and analyze the data, and provide intuitive data support and decision basis for the construction of the deposit model sample library and the subsequent prospecting prediction work.

[0117] The positive sample manufacturing unit realizes labeling of data through automatic labeling, intelligent labeling and manual labeling, including labeling of structural data space, stratum data space, rock mass data space, geochemical exploration data space, geophysical exploration data space, remote sensing data space, heavy sand data space, alteration data space and lithology data space.

[0118] The negative sample manufacturing unit generally requires the number of negative samples to be comparable to the number of positive samples. Through negative sample manufacturing based on spatial data, the strategy adopted is to make the total area of negative samples close to the total area of positive samples.

[0119] The quality detection unit checks the quality of the labeled deposit model, and the unqualified ones are reworked, and the qualified ones are delivered, including:

[0120] Multi-source data fusion analysis combines data from different sources, utilizes the complementarity of multi-source data, identifies the characteristics and attributes of the deposit, and improves the quality and reliability of the data.

[0121] Deep learning model construction, based on deep learning algorithm, constructs neural network model of deposit model data quality inspection, learns and trains a large number of labeled deposit model spatial data, automatically extracts features and patterns in the data, and realizes quality evaluation of the deposit model.

[0122] Automatic quality inspection process, combining multi-source data fusion analysis and deep learning model, realizes automatic quality inspection of labeled deposit model, including feature extraction, model prediction, result evaluation and other links, which can quickly and accurately detect errors and abnormalities in the deposit model, and improve the efficiency and automation of quality inspection.

[0123] Visual display and interaction, the quality inspection results are presented in an intuitive graphical way, the quality evaluation results of the deposit model are viewed through the platform, including accuracy, integrity, consistency and other indicators of the model, and interactive functions are provided.

[0124] As a preferred embodiment, the positive sample manufacturing unit comprises:

[0125] The manual labeling subunit is used for manually labeling typical deposit spatial data in the publishing task.

[0126] The automatic labeling subunit is used for automatically labeling typical deposit spatial data in the publishing task.

[0127] The intelligent labeling subunit is used for intelligently labeling typical deposit spatial data in the publishing task.

[0128] In this embodiment, the positive sample making refers to sample labeling of typical deposit spatial data. The general idea is to integrate manual positive sample spatial labeling, automatic positive sample spatial labeling and intelligent positive sample spatial labeling, and combine label management and labeling tools to realize comprehensive and efficient positive sample spatial labeling of typical deposit spatial data of typical deposit model. In the process of positive sample spatial labeling, the experience and knowledge of geologists and the advantages of artificial intelligence algorithms are fully considered, and the accuracy and efficiency of labeling are improved through human-computer interaction. At the same time, by using big data processing technology and cloud computing platform, rapid processing and storage of large-scale spatial data are realized.

[0129] As a preferred embodiment, the positive sample making unit is specifically configured to perform the following operations:

[0130] According to the preset rule, the typical deposit spatial data in the publishing task is allocated to different labeling units.

[0131] The typical deposit spatial data with clear structure and clear rules is allocated to the automatic labeling subunit, and the allocated data is labeled by an algorithm based on the preset rule.

[0132] The remaining typical deposit spatial data and the typical deposit spatial data with automatic labeling confidence lower than the preset threshold are allocated to the intelligent labeling subunit, and the allocated data is labeled by a deep learning model.

[0133] The typical deposit spatial data with intelligent labeling subunit labeling confidence lower than the preset threshold is allocated to the manual labeling subunit, and the allocated data is labeled by an expert.

[0134] The results of manual labeling, automatic labeling and intelligent labeling are integrated, the area of each positive sample region is calculated, and the areas are summarized to obtain the total area of the positive samples.

[0135] In this embodiment, through the hierarchical labeling process of man-machine cooperation, full-coverage accurate labeling of simple to complex data is realized, and the controllability and traceability of data quality are ensured through the confidence threshold mechanism, which provides high credibility positive sample support for subsequent intelligent prospecting model.

[0136] As a preferred embodiment, the negative sample making unit is specifically configured to perform the following operations:

[0137] Determine the distribution range of the abnormal area of each layer of the typical deposit spatial data in the publishing task, and fuse the abnormal area distribution of each layer to construct a multi-dimensional geological feature space.

[0138] For each layer, use a preset anomaly recognition algorithm to identify the abnormal area in each layer, and perform spatial superposition sum on the abnormal areas of all layers to obtain the comprehensive coverage area of all abnormal areas.

[0139] The overall range of the study area is spatially differenced with the comprehensive coverage area to obtain an abnormal uncovered blank area, which is the potential spatial distribution range of the negative samples.

[0140] According to the total area of the positive samples, the target value of the total area of the negative samples is determined, the range of the negative sample data is set in the blank area, and the spatial distribution of the negative samples is automatically calculated in the blank area by using a random function.

[0141] In this embodiment, the spatial distribution range of the negative samples is set in the abnormal area spatial superposition sum of the typical deposit spatial data. In the blank area, two random parameters of the number and area of the negative samples are set, and the distribution of the negative samples is automatically calculated by using a random function, and the requirement is that the total area covered by the finally calculated negative samples is approximately equal to the total area of the positive samples.

[0142] 1) Select typical deposit ore-forming element data

[0143] Select typical deposit spatial data of ore-forming element layers such as structure, stratum, lithology, remote sensing, and alteration to determine the distribution range of the abnormal area of each layer.

[0144] 2) Build a multi-dimensional geological feature space

[0145] Fuse the data of each ore-forming element layer to build a multi-dimensional geological feature space containing structure features, stratum features, lithology features, remote sensing features, and alteration features. Assuming that there are n ore-forming element layers, the feature vector of the i-th layer is where m is the feature dimension, and the fused multi-dimensional geological feature space can be represented as , which provides rich geological information for subsequent abnormal area identification.

[0146] 3) Abnormal area identification and spatial superposition sum

[0147] For each ore-forming element layer, an abnormal area in each layer is identified by using a preset abnormal identification algorithm. Taking a threshold method based on statistical analysis as an example, assuming that the feature value of the i-th layer is , the mean value is , the standard deviation is , and the threshold coefficient is k, then the judgment formula of the abnormal area is:

[0148]

[0149] If the above conditions are met, the area is an abnormal area. When using a classification algorithm based on machine learning, assuming that the classification model is , when , it is determined that the area is an abnormal area.

[0150] Then, the abnormal areas of all ore-forming element layers are overlapped and combined, and the abnormal area A is set as , which represents the abnormal area set of the i-th layer, and is obtained by The comprehensive coverage area of all abnormal areas in the study area is obtained.

[0151] 4) Determine the spatial distribution range of negative samples

[0152] The overall range R of the study area is spatially subtracted from the above-mentioned comprehensive abnormal coverage area A, that is , to obtain the blank area N of abnormal coverage, which is the potential spatial distribution range of negative samples.

[0153] 5) Set random parameters of the number and area of negative samples

[0154] According to the total area of positive samples , the target value of the total area of negative samples is determined , which satisfies , where is the area fluctuation coefficient, which can be adjusted according to actual conditions.

[0155] In the blank area N, the range of the number of negative samples is set as , and the random parameter range of the area of a single negative sample is . Assuming that the area of the blank area is , where represents rounding up, represents rounding down.

[0156] 6) Automatic distribution calculation of negative samples

[0157] The distribution of negative samples in the blank area is automatically calculated using a random function.

[0158] The blank area N is divided into s sub-areas using stratified random sampling method, and the area of the j-th sub-area is , and the number of negative samples generated in the j-th sub-area is , which satisfies:

[0159] and , where is the area of a single negative sample generated in the j-th sub-area, and . Specifically, according to the set random parameters of the number and area of negative samples, the position and range of negative samples are randomly generated in the blank area, and the total area of generated negative samples is calculated in real time during the generation process, and the generation is stopped when , where t is the number of generated negative samples.

[0160] 7) Negative sample verification and adjustment

[0161] The generated negative samples are verified to check whether their spatial distribution is within the blank area N and whether the total area is close to the total area of the positive samples. If there are cases that do not meet the requirements, adjust the random parameters and re-calculate the distribution of the negative samples until the requirements are met.

[0162] As a preferred embodiment, the delineation of the ore-prospecting favorable area module comprises:

[0163] The intelligent algorithm library unit is configured to train the data in the deposit model sample library by machine learning algorithms and deep learning algorithms to obtain a plurality of AI ore-prospecting favorable area delineation models, and to guide the task of the sample library construction workbench by the AI ore-prospecting favorable area delineation model.

[0164] The deposit model sample processing unit is configured to analyze and process the data in the deposit model sample library to obtain a prediction data set.

[0165] The ore-prospecting favorable area delineation unit is configured to input the prediction data set into the AI ore-prospecting favorable area delineation model for reasoning to delineate the ore-prospecting favorable area.

[0166] In this embodiment, the intelligent algorithm library unit is configured to collect machine learning and deep learning algorithms for ore-prospecting prediction, including random forest, SVM, CNN, UNet, AlexNet, etc. All algorithms in the "intelligent algorithm library" and the ore-prospecting prediction annotation data set in the "deposit model sample library" are used for training to form a plurality of "AI model files" for reasoning in the next step. The AI ore-prospecting favorable area delineation model guides the task of the sample library construction workbench, realizes the intelligent closed loop of "prediction-feedback-optimization", reversely identifies data defects (such as annotation blind area, rule failure) through model prediction results, automatically generates supplement, correction and optimization tasks, and upgrades the sample library from static storage to a continuously iterating "living knowledge engine"; based on the confidence diagnosis of the model feedback, the annotation resources are allocated in a targeted manner, the data with clear rules are automatically processed by the algorithm, and the difficult data are preferentially allocated for expert manual annotation; the new ore-forming regularity mined by the model is converted into annotation rules in real time and injected into the workbench.

[0167] The deposit model sample processing unit extracts corresponding spatial data from the deposit model sample library according to the user-set prediction area spatial coordinate range, ore type, genetic type and ore-prospecting prediction element layer data using fast spatial indexing and cutting analysis technology, including multi-scale and multi-element vector or raster spatial data related to geology, geophysics, geochemistry, remote sensing and ore. After extraction, the data form a "prediction area spatial layer data set", and the data content includes tectonic spatial data, stratigraphic spatial data, rock mass spatial data, geochemical spatial data, geophysical spatial data, remote sensing spatial data, heavy sand spatial data, etc.

[0168] The ore-favorable area delineation unit uses an algorithm model in the intelligent algorithm library and a corresponding "AI model file" to perform intelligent reasoning on a prediction data set formed after "spatial data rasterization cutting", and generates an ore-finding prediction favorable area after the intelligent reasoning is completed.

[0169] The above merely describes preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A deposit model spatial data sample library intelligent construction system, characterized in that, The system comprises: a deposit model sample library management module for constructing a source space database, issuing a task to a sample library construction workbench module, monitoring the issued task, and reprocessing the task delivered by the sample library construction workbench module to construct a deposit model sample library; the sample library construction workbench module for processing the issued task and delivering the completed task to the deposit model sample library management module; the sample library construction workbench module comprises: a data visualization unit for visualizing the deposit model space data in the issued task and visualizing the data in the positive sample making, negative sample making and labeling processes; a positive sample making unit for making positive samples based on the typical deposit space data in the issued task; a negative sample making unit for making negative samples based on the typical deposit space data in the issued task; a quality detection unit for quality checking the typical deposit space data after labeling, re-making positive samples or negative samples if unqualified, and delivering to the deposit model sample library management module if qualified; the positive sample making unit comprises: a manual labeling sub-unit for manually labeling the typical deposit space data in the issued task; an automatic labeling sub-unit for automatically labeling the typical deposit space data in the issued task; an intelligent labeling sub-unit for intelligently labeling the typical deposit space data in the issued task; the positive sample making unit is specifically configured to: allocate the typical deposit space data in the issued task to different labeling units according to preset rules; allocate the typical deposit space data with clear structure and explicit rules to the automatic labeling sub-unit for labeling the allocated data based on an algorithm according to preset rules; allocate the remaining typical deposit space data and the typical deposit space data with automatic labeling confidence lower than a preset threshold to the intelligent labeling sub-unit for labeling the allocated data by a deep learning model; allocate the typical deposit space data with intelligent labeling sub-unit labeling confidence lower than a preset threshold to the manual labeling sub-unit for labeling the allocated data by an expert; integrate the results of manual labeling, automatic labeling and intelligent labeling, calculate the area of each positive sample region, and summarize to obtain the total area of the positive samples; the negative sample making unit is specifically configured to: determine the distribution range of the abnormal area of each layer of the typical deposit space data in the issued task, fuse the abnormal area distribution of each layer to construct a multi-dimensional geological feature space; for each layer, identify the abnormal area in each layer by using a preset abnormality recognition algorithm, and perform spatial superposition sum on the abnormal areas of all layers to obtain a comprehensive coverage area of all abnormal areas; perform spatial difference operation on the overall range of the study area and the comprehensive coverage area to obtain a blank area not covered by the abnormal area, which is a potential spatial distribution range of the negative sample; determine the target value of the total area of the negative sample according to the total area of the positive sample, set the range of the negative sample data in the blank area, and automatically calculate the spatial distribution of the negative sample in the blank area by using a random function; the quality detection unit is configured to: perform quality checking on the typical deposit space data after labeling, re-make positive samples or negative samples if unqualified, and deliver to the deposit model sample library management module if qualified; the negative sample making unit is specifically configured to: determine the distribution range of the abnormal area of each layer of the typical deposit space data in the issued task, fuse the abnormal area distribution of each layer to construct a multi-dimensional geological feature space; for each layer, identify the abnormal area in each layer by using a preset abnormality recognition algorithm, and perform spatial superposition sum on the abnormal areas of all layers to obtain a comprehensive coverage area of all abnormal areas; perform spatial difference operation on the overall range of the study area and the comprehensive coverage area to obtain a blank area not covered by the abnormal area, which is a potential spatial distribution range of the negative sample; determine the target value of the total area of the negative sample according to the total area of the positive sample, set the range of the negative sample data in the blank area, and automatically calculate the spatial distribution of the negative sample in the blank area by using a random function. The module is used for obtaining data in the deposit model sample library, delineating a favorable ore prospecting area, and guiding the task of the sample library construction workbench module.

2. The intelligent construction system of deposit model spatial data sample library according to claim 1, characterized in that, The deposit model sample library management module comprises: The data acquisition unit is configured to acquire typical deposit spatial data, analyze and identify heterogeneous data in the typical deposit spatial data through a multi-modal data adaptive acquisition engine, and align the typical deposit spatial data to a unified geographic coordinate system; The source spatial database unit is configured to manage the typical deposit spatial data in a hierarchical classification manner according to a deposit genetic type, including establishing a classification system and data maintenance and update; The task publishing unit is configured to publish a spatial data sample production task to the sample library construction workbench module for the typical deposit spatial data in the source spatial database unit; The task monitoring unit is configured to monitor the progress and workload of all sample production tasks in the sample library construction workbench module; The task delivery detection unit is configured to receive a task delivered by the sample library construction workbench module and perform quality detection on the delivered task; The deposit model sample warehouse unit is configured to reprocess the typical deposit spatial data after sample production, construct a deposit model sample library, and perform visual display.

3. The intelligent construction system of deposit model spatial data sample library according to claim 2, characterized in that, The source spatial database unit comprises: The hierarchical classification management subunit is configured to classify the typical deposit spatial data according to a classification system established according to a deposit genetic type; The spatial data lake warehouse integrated management subunit is configured to establish a unified data storage format and standard, and store the data in the hierarchical classification management unit in multiple nodes in a distributed storage manner; The data dynamic real-time update and maintenance subunit is configured to record the update history of the typical deposit spatial data by establishing a data version control mechanism, and repair and update errors and abnormalities in the typical deposit spatial data in the nodes by using a data detection algorithm.

4. The intelligent construction system of deposit model spatial data sample library according to claim 2, characterized in that, The task publishing unit is specifically configured to perform the following operations: analyze the data in the source spatial database, extract key features of the data, process the key features by using a machine learning algorithm to obtain processing task requirements of each data, and assign the processing task requirements to processing units corresponding to the sample library construction workbench module by using a matching algorithm. The deposit model sample warehouse unit comprises: The sample temporary management subunit is configured to classify the typical deposit spatial data after sample production; 5. The intelligent construction system of deposit model spatial data sample library according to claim 2, characterized in that, The sample data processing subunit is configured to perform format conversion on all the typical deposit spatial data in the sample temporary management subunit by using a vector-to-raster algorithm, perform cutting and segmentation on the data in a unified specification and size, form a single-channel sample set and a multi-channel sample set, calculate a covariance matrix of data in each single-channel sample set and multi-channel sample set, apply noise distribution disturbance to each data, and perform whitening / anti-whitening processing through the covariance matrix, and finally perform geometric transformation; The deposit model sample management subunit is configured to construct a deposit model sample library, establish data description and metadata information for each single-channel data set, integrate and associate the multi-channel data set by using a data fusion technology, and form a multi-channel cutting data set. ​ ​ ​ The deposit model sample visualization subunit is used for visualizing the deposit model sample library by cloud computing and big data processing technology, in combination with GIS and a Web visualization engine.

6. The intelligent construction system of deposit model spatial data sample library according to claim 1, characterized in that, The delineation of the ore-prospecting favorable area module comprises: The intelligent algorithm library unit is used for training data in the deposit model sample library by machine learning algorithms and deep learning algorithms to obtain a plurality of AI ore-prospecting favorable area delineation models, and guiding the sample library construction workbench module by the AI ore-prospecting favorable area delineation model. The deposit model sample processing unit is used for analyzing and processing data in the deposit model sample library to obtain a prediction data set. The ore-prospecting favorable area delineation unit is used for inputting the prediction data set into the AI ore-prospecting favorable area delineation model for reasoning to delineate the ore-prospecting favorable area.

Images