Mineral exploration information generation method and device and storage medium

Abstract

The invention discloses a mineral exploration information generation method and device and a storage medium, and relates to the technical field of geological information, and the method comprises the steps: carrying out the cloud integration and standardization preprocessing of multi-source exploration data, generating a three-dimensional and multi-dimensional data cube, reading information from a geoscience knowledge base, employing the natural language processing and entity relation extraction technology, and carrying out the cloud integration and standardization preprocessing of the multi-source exploration data; constructing a geoscience knowledge graph; based on the three-dimensional multi-dimensional data cube and in combination with a geoscience knowledge graph, training a lightweight explainable causal reasoning model at a cloud end, and setting a causal contribution degree threshold value and an evidence support strength threshold value of geological variables; and deploying the lightweight explainable causal reasoning model to field intelligent terminal equipment, and collecting field investigation data. According to the method, the problem that only static prediction results can be provided and internal basis cannot be revealed and converted into pain points of specific actions in the prior art is solved, and fundamental spanning from data prediction to decision enabling is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geological information, and in particular to a mineral exploration information generation method, device and storage medium. BACKGROUND

[0002] With the deep integration of information technology and geoscience, intelligent mineral exploration technology has become the core driving force for the development of the industry. Multi-source exploration data such as geology, geophysics, geochemistry, remote sensing, etc. are integrated in the cloud. Through data cleaning, gridding and other preprocessing methods, a unified three-dimensional geological space database is constructed. Machine learning algorithms are applied to train the database to establish a mapping relationship from multi-dimensional features to mineralization probability, and generate a mineralization prospect prediction map. This kind of system significantly improves the efficiency of identifying mineralization rules from complex data by combining massive data with artificial intelligence.

[0003] However, the above-mentioned prior art still has an optimization space for the key link in realizing the deep generation and accurate conversion of exploration information. The results generated by the existing system are usually mineralization probability maps or classification results, which are more presented as a static and general conclusion. Although the conclusion points out the favorable area for mineralization, it does not reveal the internal basis for the model to make this judgment, nor can it convert this information into clear instructions to guide the next specific exploration action. SUMMARY

[0004] In view of the above-mentioned existing problems, the present application is proposed.

[0005] Therefore, the present application provides a mineral exploration information generation method to solve the problem that the static conclusion generated by the prior art cannot reveal the prediction basis and convert it into specific exploration action instructions.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a mineral exploration information generation method, which includes cloud integration and standardized preprocessing of multi-source exploration data to generate a three-dimensional multi-dimensional data cube, reading information from a geoscience knowledge base, using natural language processing and entity relationship extraction technology to construct a geoscience knowledge graph; Based on the three-dimensional multi-dimensional data cube combined with the geoscience knowledge graph, a lightweight interpretable causal reasoning model is trained in the cloud, and a causal contribution threshold and an evidence support intensity threshold of geological variables are set; Deploy the lightweight interpretable causal reasoning model to a field intelligent terminal device to collect field exploration data; Use the deployed lightweight interpretable causal reasoning model to analyze the field exploration data in real time to obtain the causal contribution of the geological variables and the current evidence support intensity; The causal contribution degree of the geological variable is compared with the evidence support strength of the geological variable, and a contradictory geological variable is identified, in which the causal contribution degree of the geological variable is higher than the contribution degree threshold of the geological variable, and the evidence support strength of the geological variable is lower than the evidence strength threshold of the geological variable. The geoscience knowledge graph is queried, and an associated geological phenomenon related to the causal correlation of the contradictory geological variable and acquired by the field intelligent terminal device is retrieved from the graph to generate a specific instruction indicating that the associated geological phenomenon is verified preferentially.

[0007] As a preferred scheme of the mineral exploration information generation method, the multi-source exploration data is cloud-integrated and standardized preprocessed to generate a three-dimensional multi-dimensional data cube, information is read from a geoscience knowledge base, a geoscience knowledge graph is constructed by using natural language processing and entity relationship extraction technology, and the following steps are included: Satellite remote sensing images, aerial geophysical measurement data, ground geochemical sampling data, and geological mapping vector data are collected and combined to obtain multi-source exploration data, and the multi-source exploration data is subjected to coordinate system unification and gridding resampling to form a three-dimensional multi-dimensional data cube; The three-dimensional multi-dimensional data cube provides a spatial reference framework for knowledge extraction, and geological survey reports are read from the geoscience knowledge base based on the spatial range; The geoscience knowledge base text information is subjected to geological entity recognition and relationship extraction by natural language processing technology to obtain the results of entity relationship extraction; The results of natural language processing and entity relationship extraction are organized into structured knowledge in the form of entity-relation-entity triples to construct the geoscience knowledge graph.

[0008] As a preferred scheme of the mineral exploration information generation method, a lightweight interpretable causal reasoning model is trained in the cloud based on the three-dimensional multi-dimensional data cube and the geoscience knowledge graph, and a causal contribution degree threshold and an evidence support strength threshold of the geological variable are set, and the following steps are included: Based on the feature data in the three-dimensional multi-dimensional data cube and the causal relationship constraints in the geoscience knowledge graph, a lightweight interpretable causal reasoning model is trained in the cloud by using a causal reasoning algorithm; The trained lightweight interpretable causal reasoning model is used to perform feedback verification on the training data set to obtain the causal contribution degree of each geological variable to the prediction result, and variable contribution degree distribution data is generated; The lightweight interpretable causal reasoning model is used to predict the training data set to obtain a mineralization probability value, a binary real label is obtained by using the exploration engineering results of the mining area, a difference analysis is performed based on the mineralization probability value and the binary real label to obtain a verification result, and domain priori knowledge is obtained by using the geological law constraints in the geoscience knowledge graph; Based on the verification result and the prior knowledge of the field, the evidence support strength threshold of the geological variable is set.

[0009] As a preferred scheme of the mineral exploration information generation method, the lightweight interpretable causal reasoning model is deployed to the field intelligent terminal device, and the field exploration data is collected, including the following steps: The trained lightweight interpretable causal reasoning model is compressed and converted in format to generate a deployment package suitable for mobile devices; The lightweight interpretable causal reasoning model package is deployed to the field intelligent terminal device, and the completed lightweight interpretable causal reasoning model receives the field exploration data from the field intelligent terminal device collection interface; The built-in sensors and collection functions of the field intelligent terminal device are started to collect the field exploration data.

[0010] As a preferred scheme of the mineral exploration information generation method, the lightweight interpretable causal reasoning model is deployed to the field intelligent terminal device, and the field exploration data is collected, including the following steps: The field exploration data is standardized and formatted to generate a standardized field exploration data package; The standardized field exploration data package is input into the lightweight feature extraction algorithm to generate a multi-modal feature vector representing geological features; The multi-modal feature vector is input into the deployed lightweight interpretable causal reasoning model to activate the causal attention mechanism inside the lightweight interpretable causal reasoning model; Based on the causal attention mechanism inside the lightweight interpretable causal reasoning model, the original attention weight is output and normalized by Softmax to convert into the causal contribution degree of the geological variable; The cosine similarity between the multi-modal feature vector and the feature template of the lightweight interpretable causal reasoning model is calculated to quantify the current evidence support of the geological variable; The causal contribution degree of the geological variable and the current evidence support strength of the geological variable are combined and output to form the real-time analysis result.

[0011] As a preferred scheme of the mineral exploration information generation method, the causal contribution degree of the geological variable is compared with the evidence support strength of the geological variable to identify the contradictory geological variable whose causal contribution degree is higher than the contribution degree threshold of the geological variable and whose evidence support strength is lower than the evidence strength threshold of the geological variable, including the following steps: Based on the causal contribution degree list and the evidence support intensity list of the geological variables in the real-time analysis result, the causal contribution degree list of the geological variables is traversed, the causal contribution degree value of each geological variable is compared with the contribution degree threshold of the geological variable, and the geological variable whose causal contribution degree value is greater than the contribution degree threshold of the geological variable is screened out to form a high-contribution geological variable candidate list; Each geological variable in the high-contribution geological variable candidate list is traversed, the evidence support intensity is extracted from the real-time analysis result, the evidence support intensity of each high-contribution geological variable is compared with the evidence intensity threshold of the geological variable, and the geological variable whose evidence support intensity is less than the evidence intensity threshold of the geological variable is screened out. The geological variable that meets the conditions of the causal contribution degree being higher than the contribution degree threshold of the geological variable and the evidence support intensity being lower than the evidence intensity threshold of the geological variable is marked as a contradictory geological variable.

[0012] As a preferred scheme of the mineral exploration information generation method, the following steps are included: querying the geological knowledge graph, retrieving the associated geological phenomena that have a causal correlation with the contradictory geological variable and are obtained by the field intelligent terminal device from the graph, and generating a specific instruction indicating that the associated geological phenomena are verified preferentially. The list query condition of the contradictory geological variable is input into the geological knowledge graph, and the query interface retrieves the geological entities that have a direct causal correlation with the contradictory geological variable according to the entity relationship triple stored in the geological knowledge graph. The list of associated geological entities returned by the knowledge graph query is matched with the list of sensors and collection functions built in the field intelligent terminal device, the associated geological entities that have a causal correlation with the contradictory geological variable and are observed by the sensors on the field intelligent terminal device are screened out, and a candidate list of associated geological phenomena is formed. The candidate list of associated geological phenomena is prioritized according to the correlation strength and detection difficulty to obtain an optimal associated geological phenomenon verification target, and a specific verification instruction text is generated.

[0013] In the second aspect, the present application provides a mineral exploration information generation device, which includes a construction module, a training module, and an intelligent terminal deployment module. The training module trains a lightweight interpretable causal reasoning model in the cloud based on the three-dimensional multi-dimensional data cube and the geological knowledge graph, and sets the causal contribution degree threshold and the evidence support intensity threshold of the geological variable. The intelligent terminal deployment module deploys the lightweight interpretable causal reasoning model to the field intelligent terminal device to collect field exploration data. The analysis module uses the deployed lightweight interpretable causal reasoning model to perform real-time analysis on the field investigation data, and obtains the causal contribution degree and the current evidence support strength of the geological variable; The comparison module compares the causal contribution degree of the geological variable with the evidence support strength of the geological variable, and identifies a contradictory geological variable whose causal contribution degree is higher than a contribution degree threshold of the geological variable and whose evidence support strength is lower than an evidence strength threshold of the geological variable. The instruction module queries a geoscience knowledge graph, retrieves, from the graph, an associated geological phenomenon that is causally related to the contradictory geological variable and is acquired by the field intelligent terminal device, and generates a specific instruction indicating that the associated geological phenomenon is to be preferentially verified.

[0014] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program, when executed by the processor, implements any step of the mineral exploration information generation method according to the first aspect of the present application.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any step of the mineral exploration information generation method according to the first aspect of the present application.

[0016] The present application has the following beneficial effects: by deploying a lightweight interpretable causal reasoning model to a field terminal and innovatively introducing a contribution-evidence strength double-threshold comparison mechanism, the uncertainty of the prediction result of the lightweight interpretable causal reasoning model is traced back, and key contradictions are located, so that a contradictory geological variable with high contribution and low evidence can be accurately identified, an executable verification instruction for a specific associated geological phenomenon can be automatically generated by querying a geoscience knowledge graph that is causally related to the contradictory variable, the insight of artificial intelligence is directly converted into decision-making wisdom for guiding field exploration actions, the pain points of the prior art that can only provide static prediction results but cannot reveal the internal basis and convert it into specific actions are solved, and a fundamental leap from data prediction to decision empowerment is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0018] Fig. 1 The flowchart of the mineral exploration information generation method.

[0019] Fig. 2A schematic diagram of a mineral exploration information generation device.

[0020] Fig. 3 A schematic diagram of a geoscience knowledge graph.

[0021] Fig. 4 A flowchart of a process of acquiring causal contribution degree and evidence support strength in real time analysis. DETAILED DESCRIPTION

[0022] In order to make the above objectives, features and advantages of the present application more apparent, specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0023] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details given herein, that the present application can be practiced with other than the described implementations, and that the present application can use other applications. Therefore, the scope of the present application is indicated by the appended claims rather than the specific embodiments disclosed and given hereinafter.

[0024] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. The "in one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0025] REFERENCE Figs. 1-4 For one embodiment of the present application, the embodiment provides a mineral exploration information generation method, comprising the following steps: S1, cloud integration and standardized pretreatment of multi-source exploration data are performed to generate a three-dimensional multi-dimensional data cube, information is read from a geoscience knowledge base, natural language processing and entity relationship extraction technology are used to construct a geoscience knowledge graph.

[0026] S1.1, satellite remote sensing images, aerial geophysical measurement data, ground geochemical sampling data and geological mapping vector data are collected and combined to obtain multi-source exploration data, the multi-source exploration data are subjected to coordinate system unification and gridding resampling to form a three-dimensional multi-dimensional data cube.

[0027] Further, satellite remote sensing image data, airborne geophysical survey data, ground geochemical sampling data and geological mapping vector data are collected through data interfaces to form multi-source exploration data. The multi-source exploration data undergoes coordinate system processing to convert the plane coordinates and elevation datum of different sources to a unified geographic coordinate system, and the multi-source exploration data after coordinate unification is then subjected to gridding resampling, and the point, line and surface data are converted into attribute values in regular three-dimensional grid cells according to a preset three-dimensional grid size. Finally, a three-dimensional multi-dimensional data cube containing multi-source attribute values in each grid cell is generated.

[0028] S1.2, providing a spatial reference framework for knowledge extraction by the three-dimensional multi-dimensional data cube, reading geological survey reports from the geosciences knowledge base based on the spatial range.

[0029] Further, the geographical spatial range covered by the three-dimensional multi-dimensional data cube is extracted as a spatial reference framework. Based on the boundary coordinates of the spatial reference framework, a query condition is automatically generated and a query is initiated to the geosciences knowledge base, and the geosciences knowledge base retrieves and returns all the text contents of the geological survey reports located within the spatial range according to the query condition.

[0030] S1.3, the geosciences knowledge base text information is subjected to geological entity recognition and relationship extraction through natural language processing technology to obtain the results of entity relationship extraction.

[0031] Further, the text content of the geological survey report read from the geosciences knowledge base is input into the natural language processing process. The natural language processing process first scans the text using named entity recognition technology to identify and label geological entity names such as granite body, fault zone and potassium feldspar alteration, and then uses relationship extraction technology to analyze the sentence structure to determine the semantic relationship between the identified geological entity names, for example, to extract the coexistence relationship between the granite body and the potassium feldspar alteration. The output results of named entity recognition and relationship extraction are the results of entity relationship extraction.

[0032] S1.4, the results of natural language processing and entity relationship extraction are organized into structured knowledge in the form of entity-relation-entity triples to construct a geosciences knowledge graph.

[0033] Further, the results of entity relationship extraction are sent to the knowledge structuring process. The knowledge structuring process organizes the results of entity relationship extraction into the subject-predicate-object format to form entity relationship entity triples, and the entity relationship entity triples are batch imported into the graph database, in which the geological entities are nodes and the relationships between entities are edges, to construct a geosciences knowledge graph.

[0034] S2, based on a three-dimensional multidimensional data cube combined with a geoscience knowledge graph, train a lightweight interpretable causal reasoning model in the cloud, and set the causal contribution threshold and evidence support intensity threshold of the geological variables.

[0035] S2.1, based on the feature data in the three-dimensional multidimensional data cube and the causal relationship constraints in the geoscience knowledge graph, train a lightweight interpretable causal reasoning model in the cloud using a causal reasoning algorithm.

[0036] Further, the feature data in the three-dimensional multidimensional data cube is extracted as training samples, and the causal relationship constraints stored in the geoscience knowledge graph are converted into regularization terms in the training process. In the cloud computing environment, a lightweight interpretable causal reasoning model is trained using an algorithm based on a causal attention mechanism. The training process aims to enable the lightweight interpretable causal reasoning model to learn the ability to accurately predict ore-forming potential from the feature data in the three-dimensional multidimensional data cube while meeting the causal laws revealed by the geoscience knowledge graph.

[0037] S2.2, based on the trained lightweight interpretable causal reasoning model, feedback verification is performed on the training data set, the causal contribution of each geological variable to the prediction result is obtained, and variable contribution distribution data is generated.

[0038] Further, the trained lightweight interpretable causal reasoning model performs forward propagation reasoning on the entire training data set. During the reasoning process, the attention weight assigned by the lightweight interpretable causal reasoning model's internal causal attention mechanism to each geological variable is recorded. After normalization by the Softmax function, the attention weight is quantified as the causal contribution of each geological variable to the prediction result. The causal contribution values of the same geological variable for all samples in the training data set are collected to form a contribution distribution data set for the geological variable.

[0039] S2.3, through the prediction of the training data set by the lightweight interpretable causal reasoning model, obtain the ore-forming probability value, use the exploration engineering results of the mining area to obtain the binary real label, based on the difference analysis between the ore-forming probability value and the binary real label, obtain the verification result, through the geological law constraint in the geoscience knowledge graph, obtain the domain priori knowledge.

[0040] Further, the prediction output of the lightweight interpretable causal reasoning model on the training data set is a continuous ore-forming probability value. The exploration engineering results of the mining area, such as drilling results, are used to perform binary labeling on each sample area to generate binary real labels of non-mineral or mineral. Through the difference between the ore-forming probability value and the binary real label, a deviation analysis report of the model prediction and the actual situation is obtained as the verification result. Qualitative descriptions about the correlation of ore-forming elements and the reliability of evidence are directly extracted from the geoscience knowledge graph and used as domain priori knowledge.

[0041] S2.4, based on the verification result and the domain prior knowledge, set the evidence support strength threshold of the geological variable.

[0042] Further, based on the analysis report of the difference between the model prediction and the binary real label in the verification result, and the qualitative description of the correlation of ore-forming elements and the reliability of evidence in the domain prior knowledge, the evidence support strength threshold of the geological variable is set. The verification result reveals the actual influence distribution of different geological variables on prediction accuracy under different evidence support strengths, while the domain prior knowledge provides a geological basis for judging the reliability of evidence. By comprehensively analyzing the quantitative distribution law in the verification result and the qualitative constraint of the domain prior knowledge, a critical value that can effectively distinguish reliable evidence from unreliable evidence is determined, which is set as the evidence support strength threshold of the geological variable.

[0043] S3, deploy the lightweight interpretable causal reasoning model to the field intelligent terminal device to collect field exploration data.

[0044] S3.1, the trained lightweight interpretable causal reasoning model is compressed and converted in format to generate a deployment package suitable for mobile devices.

[0045] Further, the network structure with low contribution is removed by pruning algorithm, the model weight is converted from 32-bit floating point number to 8-bit integer by using quantization technology, and the model structure and parameters are packaged into specific file format by using the conversion tool of mobile terminal reasoning framework. The purpose of this series of model compression and format conversion operations is to significantly reduce the model size and improve the calculation speed under the premise of controllable loss of model prediction accuracy, so as to generate a lightweight interpretable causal reasoning model deployment package suitable for mobile device computing resources and operating system.

[0046] S3.2, deploy the lightweight interpretable causal reasoning model package to the field intelligent terminal device, and the deployed lightweight interpretable causal reasoning model receives the field exploration data from the collection interface of the field intelligent terminal device.

[0047] Further, the lightweight interpretable causal reasoning model package is transmitted to the storage directory of the field intelligent terminal device through wired or wireless connection, and the model file is decompressed and registered to the local reasoning engine of the terminal device through the installation script, completing the deployment and configuration of the lightweight interpretable causal reasoning model. The deployed lightweight interpretable causal reasoning model enters standby state immediately, continuously listens and prepares to receive the field exploration data stream transmitted from the collection interface of the field intelligent terminal device.

[0048] S3.3, Start the built-in sensor and collection function through the field intelligent terminal equipment to collect the field investigation data.

[0049] Further, the control program on the field intelligent terminal equipment sends instructions to start the built-in high-resolution camera for image shooting and the short-wave infrared spectrometer for spectrum scanning, and the GPS module built in the equipment synchronously records the current spatial coordinate information. The raw data collected by the camera, the spectrometer and the GPS module are received by the data collection interface of the equipment in real time and are preliminarily formatted and packaged to form a standardized field investigation data package.

[0050] S4, Real-time analysis of the field investigation data using the deployed lightweight interpretable causal reasoning model to obtain the causal contribution degree of the geological variable and the current evidence support strength.

[0051] S4.1, Standardized formatting processing of the field investigation data to generate a standardized field investigation data package.

[0052] Further, the field investigation data includes high-resolution images from the camera, spectral curves from the spectrometer and coordinate information from the GPS module. These data are sent to a formatting processing flow, which adjusts the image data to a uniform pixel size and normalizes the pixel value, performs baseline correction on the spectral curve and interpolates to a standard wavelength sequence, and converts the coordinate information to a uniform map projection coordinate to obtain a standardized field investigation data package.

[0053] S4.2, Input the standardized field investigation data package into the lightweight feature extraction algorithm to generate a multi-modal feature vector representing the geological features.

[0054] Further, a pre-trained MobileNetV2 convolutional neural network is used for the image data part to extract a high-dimensional texture feature vector, and the absorption depth and area of the spectral data part in the specific altered mineral characteristic waveband are calculated to form a spectral feature vector. The texture feature vector and the spectral feature vector are spliced to generate a multi-modal feature vector representing the geological features of the current point, The most representative, compact and numerical features of the geological meaning are extracted from the original investigation data, which greatly reduces the data dimension and improves the processing efficiency of the subsequent lightweight interpretable causal reasoning model.

[0055] S4.3, Input the multi-modal feature vector into the deployed lightweight interpretable causal reasoning model to activate the causal attention mechanism inside the lightweight interpretable causal reasoning model.

[0056] Further, based on the multi-modal feature vector being fed into the inference interface of the lightweight interpretable causal reasoning model deployed on the intelligent terminal device in the field, after the multi-modal feature vector is received by the lightweight interpretable causal reasoning model, the causal attention mechanism in the lightweight interpretable causal reasoning model which takes causal inference as the core starts to work. The causal attention mechanism regards the multi-modal feature vector as keys and values, and interacts with a query vector representing the current prediction task.

[0057] S4.4, based on the causal attention mechanism inside the lightweight interpretable causal reasoning model, the output original attention weight is processed by Softmax normalization to convert into the causal contribution degree of the geological variable.

[0058] Further, the causal attention mechanism inside the lightweight interpretable causal reasoning model calculates an original attention weight for each geological feature dimension in the multi-modal feature vector. The attention weight is then sent to a Softmax normalization function for processing. The Softmax function converts each original attention weight into a probability value between 0 and 1. The probability value is the causal contribution degree of the corresponding geological variable. The sum of the causal contribution degrees of all geological variables is 1.

[0059] S4.5, by performing cosine similarity calculation between the multi-modal feature vector and the feature template of the lightweight interpretable causal reasoning model, the current evidence support strength of the geological variable is quantified.

[0060] The current evidence support strength expression is: ; Wherein, is the current evidence support strength, is the geological variable, is the weight coefficient of the th feature mode, is the index of the feature mode, is the feature sub-vector of the th feature mode, is the ideal feature template sub-vector on the th feature mode.

[0061] Further, the multi-modal feature vector is compared with the pre-stored feature template in the lightweight interpretable causal reasoning model. For each geological variable, the feature sub-vector in the corresponding dimension of the multi-modal feature vector is extracted, and the ideal feature template sub-vector of the variable is obtained from the feature template of the lightweight interpretable causal reasoning model. The cosine similarity between the two vectors is quantified as the current evidence support strength of the geological variable.

[0062] S4.6, the causal contribution degree of the geological variable and the current evidence support strength of the geological variable are merged and output, forming a real-time analysis result.

[0063] Further, the causal contribution degree of the geological variable and the current evidence support strength of all geological variables are collected and organized into a structured data object, which contains the identifier of each geological variable, its causal contribution degree value and its current evidence support strength. The structured data object is output as the result of real-time analysis.

[0064] S5, compare the causal contribution degree of the geological variable with the evidence support strength of the geological variable, and identify the contradictory geological variable whose causal contribution degree is higher than the contribution degree threshold of the geological variable and whose evidence support strength is lower than the evidence strength threshold of the geological variable.

[0065] S5.1, based on the causal contribution degree list and the evidence support strength list of the geological variable in the real-time analysis result, traverse the causal contribution degree list of the geological variable, compare the causal contribution degree value of each geological variable with the contribution degree threshold of the geological variable, and filter out the geological variable whose causal contribution degree value is greater than the contribution degree threshold of the geological variable, forming a high-contribution geological variable candidate list.

[0066] Further, the real-time analysis result contains a causal contribution degree list of the geological variable and an evidence support strength list of the geological variable. The processing logic first accesses the causal contribution degree list of the geological variable, reads the causal contribution degree value of each geological variable in the list in turn, compares each read causal contribution degree value of the geological variable with the contribution degree threshold of the geological variable, and records the name of each geological variable that meets the condition of the causal contribution degree value being greater than the contribution degree threshold of the geological variable in a new list, obtaining a high-contribution geological variable candidate list.

[0067] S5.2, traverse each geological variable in the high-contribution geological variable candidate list, extract the evidence support strength from the real-time analysis result, compare the evidence support strength of each high-contribution geological variable with the evidence strength threshold of the geological variable, and filter out the geological variable whose evidence support strength is less than the evidence strength threshold of the geological variable.

[0068] Further, the processing logic then traverses each of the high-contribution geologic variable candidates in the list, and for each of the high-contribution geologic variable candidates, finds and extracts the corresponding evidence support strength in the list of evidence support strengths of the geologic variables in the real-time analysis result, and compares the extracted evidence support strength of each of the high-contribution geologic variables with the evidence strength threshold of the geologic variable, and records all geologic variable names that satisfy the evidence support strength being less than the evidence strength threshold of the geologic variable.

[0069] S5.3, mark the geologic variables that satisfy the causal contribution degree being higher than the contribution degree threshold of the geologic variable and the evidence support strength being lower than the evidence strength threshold of the geologic variable as contradictory geologic variables.

[0070] Further, the geologic variables that have the value in the list of causal contribution degrees of the geologic variables being higher than the contribution degree threshold of the geologic variable and the value in the list of evidence support strengths of the geologic variables being lower than the evidence strength threshold of the geologic variable are marked with a specific state, and they are officially marked as contradictory geologic variables.

[0071] S6, retrieve the associated geologic phenomena that have causal correlation with the contradictory geologic variables and are acquired by the field intelligent terminal device from the knowledge graph, and generate a specific instruction indicating that the associated geologic phenomena are to be verified preferentially.

[0072] S6.1, input the list of contradictory geologic variables as a query condition into the geoscience knowledge graph, and the query interface retrieves geologic entities that have direct causal correlation with the contradictory geologic variables according to the entity relationship triples stored in the geoscience knowledge graph.

[0073] Further, the list of contradictory geologic variables is submitted to the query interface of the geoscience knowledge graph as a query condition, the query interface searches the set of entity relationship triples stored in the geoscience knowledge graph after analyzing the query condition, finds all triples with the contradictory geologic variable as the subject or the object, extracts other geologic entities that are directly connected to the contradictory geologic variable through causal relationship edges from the triples, for example, the quartz vein entity that has a paragenetic relationship with the contradictory geologic variable of silicified alteration.

[0074] S6.2, match the list of associated geologic entities returned by the knowledge graph query with the list of sensors and collection functions built in the field intelligent terminal device, filter out the associated geologic entities that have causal correlation with the contradictory geologic variables and are observed by the sensors on the field intelligent terminal device, form a list of associated geologic phenomenon candidates.

[0075] Further, the list of associated geological entities returned by the knowledge graph query is transmitted to a matching process. The process compares each entity in the list of associated geological entities with a list of sensors and acquisition functions built into the field intelligent terminal device, which records the types of geological phenomena that the device can detect, such as quartz vein outcrops that can be identified by high-resolution cameras and alteration minerals that can be identified by short-wave infrared spectrometers. The matching process filters out those geological entities that are both in the list of associated geological entities and can be directly or indirectly observed by the existing sensors on the field intelligent terminal device, forming a list of candidate associated geological phenomena.

[0076] S6.3, the list of candidate associated geological phenomena is prioritized according to the strength of the association and the difficulty of detection, obtaining the optimal associated geological phenomenon to be verified, and generating a specific verification instruction text.

[0077] Further, the sorting algorithm scores each item in the candidate list according to the strength of the causal association recorded in the geoscience knowledge graph and the difficulty of field detection. Associated geological phenomena with high association strength and easy detection receive higher scores. The sorting algorithm sorts the list of candidate associated geological phenomena in descending order based on the score results. The top-ranked associated geological phenomenon is selected as the optimal verification target, and a natural language instruction text is generated based on the specific attributes of the verification target.

[0078] The embodiment also provides a mineral exploration information generation device, comprising: a construction module that integrates and standardizes preprocessing of multi-source exploration data in the cloud to generate a three-dimensional multi-dimensional data cube, reads information from a geoscience knowledge base, and constructs a geoscience knowledge graph using natural language processing and entity relationship extraction technology; A training module that trains a lightweight interpretable causal reasoning model in the cloud based on the three-dimensional multi-dimensional data cube combined with the geoscience knowledge graph, and sets a causal contribution threshold and an evidence support strength threshold for geological variables; An intelligent terminal deployment module that deploys the lightweight interpretable causal reasoning model to a field intelligent terminal device to collect field exploration data; An analysis module that uses the deployed lightweight interpretable causal reasoning model to analyze field exploration data in real time to obtain the causal contribution of geological variables and the current evidence support strength; A comparison module that compares the causal contribution of geological variables with the evidence support strength of geological variables to identify contradictory geological variables whose causal contribution is higher than the contribution threshold and whose evidence support strength is lower than the evidence strength threshold; An instruction module that queries the geoscience knowledge graph to retrieve associated geological phenomena that are causally associated with the contradictory geological variables and are obtained by the field intelligent terminal device, and generates a specific instruction indicating that the associated geological phenomena should be verified first.

[0079] The embodiment further provides a computer device suitable for the case of the mineral exploration information generation method, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the mineral exploration information generation method proposed in the above embodiment.

[0080] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, an operator network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0081] The embodiment further provides a storage medium having a computer program stored thereon, the program being executed by a processor to realize the mineral exploration information generation method proposed in the above embodiment. The storage medium can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0082] To sum up, by deploying the lightweight interpretable causal reasoning model to the field terminal and innovatively introducing the contribution-evidence strength double threshold comparison mechanism, the uncertainty of the prediction result of the lightweight interpretable causal reasoning model is traced and the key contradiction is positioned, the contradictory geological variable with high contribution and low evidence can be accurately identified, the executable verification instruction for the specific associated geological phenomenon is automatically generated by querying the causal association of the contradictory variable in the geological knowledge graph, so as to directly convert the insight of artificial intelligence into decision-making wisdom guiding the field exploration action, the pain point that the prior art can only provide static prediction results but cannot reveal the internal basis and convert into specific actions is solved, and a fundamental breakthrough from data prediction to decision empowerment is realized.

[0083] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A method for generating mineral exploration information, characterized in that: This includes cloud-based integration and standardized preprocessing of multi-source exploration data to generate a three-dimensional multidimensional data cube, reading information from a geoscience knowledge base, and constructing a geoscience knowledge graph using natural language processing and entity relationship extraction technologies. Based on a three-dimensional multidimensional data cube combined with a geoscience knowledge graph, a lightweight interpretable causal reasoning model is trained in the cloud, and thresholds for the causal contribution of geological variables and the strength of evidence support are set. Deploy lightweight, interpretable causal reasoning models to smart terminal devices in the field to collect on-site survey data; Real-time analysis of field survey data is performed using deployed lightweight, interpretable causal reasoning models to obtain the causal contribution of geological variables and the strength of current evidence. By comparing the causal contribution of geological variables with the strength of evidence for geological variables, contradictory geological variables are identified where the causal contribution of a geological variable is higher than the contribution threshold of the geological variable, and the strength of evidence for the geological variable is lower than the strength of evidence threshold of the geological variable. The system queries the geoscience knowledge graph, retrieves related geological phenomena that are causally linked to contradictory geological variables and acquired by field intelligent terminal devices, and generates a specific instruction to prioritize the verification of the related geological phenomena.

2. The method for generating mineral exploration information as described in claim 1, characterized in that: Multi-source exploration data is integrated and standardized preprocessed in the cloud to generate a three-dimensional multidimensional data cube. Information is retrieved from the geoscience knowledge base, and natural language processing and entity relation extraction techniques are used to construct a geoscience knowledge graph. The process includes the following steps: Satellite remote sensing images, airborne geophysical measurement data, ground geochemical sampling data, and geological mapping vector data are collected and combined to obtain multi-source exploration data. The multi-source exploration data is then resampled using coordinate system 1 and gridding to form a three-dimensional multi-dimensional data cube. By providing a spatial reference framework for knowledge extraction using three-dimensional multidimensional data cubes, geological survey reports are read from the geoscience knowledge base based on spatial extent; Geological entity identification and relation extraction are performed on text information in the geoscience knowledge base using natural language processing technology to obtain the results of entity relation extraction. The results of natural language processing and entity relation extraction are organized into structured knowledge, and a geoscience knowledge graph is constructed in the form of entity-relation-entity triples.

3. The method for generating mineral exploration information as described in claim 2, characterized in that: Based on a three-dimensional multidimensional data cube combined with a geoscientific knowledge graph, a lightweight interpretable causal reasoning model is trained in the cloud, and thresholds for the causal contribution of geological variables and the strength of evidence are set. The process includes the following steps: Based on the feature data in the three-dimensional multidimensional data cube and the causal relationship constraints in the geoscience knowledge graph, a lightweight interpretable causal reasoning model is trained in the cloud using a causal reasoning algorithm. The training dataset is validated using a lightweight, interpretable causal inference model after training to determine the causal contribution of each geological variable to the prediction results and generate variable contribution distribution data. The mineralization probability value is obtained by predicting the training dataset using a lightweight interpretable causal reasoning model. Binarized real labels are obtained using the exploration engineering results of the mining area. Based on the difference analysis between the mineralization probability value and the binary real labels, the verification results are obtained. Domain prior knowledge is obtained through geological law constraints in the geoscience knowledge graph. Based on the validation results and prior knowledge of the domain, a threshold for the strength of evidence support for geological variables is set.

4. The method for generating mineral exploration information as described in claim 3, characterized in that: Deploying a lightweight, interpretable causal reasoning model to intelligent field terminal devices to collect on-site survey data includes the following steps: The trained lightweight, interpretable causal reasoning model is compressed and converted to generate a deployment package suitable for mobile devices. The lightweight, interpretable causal reasoning model deployment package is configured on the field intelligent terminal device. The deployed lightweight, interpretable causal reasoning model receives field survey data from the field intelligent terminal device's data acquisition interface. By activating the built-in sensors and data acquisition functions of the field intelligent terminal device, on-site survey data can be collected.

5. The method for generating mineral exploration information as described in claim 4, characterized in that: Real-time analysis of field survey data using deployed lightweight, interpretable causal reasoning models to obtain the causal contribution of geological variables and the current strength of evidence support includes the following steps: The field survey data is standardized and formatted to generate a standardized field survey data package; Standardized field survey data packages are input into a lightweight feature extraction algorithm to generate multimodal feature vectors representing geological features. Inputting multimodal feature vectors into the deployed lightweight interpretable causal reasoning model activates the causal attention mechanism within the lightweight interpretable causal reasoning model. Based on the causal attention mechanism within the lightweight interpretable causal inference model, the original attention weights output are transformed into the causal contribution of geological variables after Softmax normalization. By calculating the cosine similarity between multimodal feature vectors and feature templates of a lightweight interpretable causal reasoning model, the current evidence support strength for geological variables is quantified. The causal contribution of geological variables and the current strength of evidence supporting them are combined to form real-time analysis results.

6. The method for generating mineral exploration information as described in claim 5, characterized in that: The causal contribution of geological variables is compared with the strength of evidence supporting them to identify contradictory geological variables whose causal contribution is higher than a contribution threshold but whose strength of evidence is lower than a strength of evidence threshold. This process includes the following steps: Based on the list of causal contributions and the list of evidence support strength of geological variables in the real-time analysis results, the list of causal contributions of geological variables is traversed, and the causal contribution value of each geological variable is compared with the contribution threshold of the geological variable. Geological variables whose causal contribution value is greater than the contribution threshold of the geological variable are selected to form a candidate list of high contribution geological variables. Iterate through each geological variable in the candidate list of high-contribution geological variables, extract the strength of evidence support from the real-time analysis results, compare the strength of evidence support for each high-contribution geological variable with the strength of evidence support threshold of the geological variable, and filter out geological variables whose strength of evidence support is less than the strength of evidence support threshold of the geological variable. Geological variables that satisfy the conditions of having a causal contribution higher than the contribution threshold of the geological variable and having an evidence support strength lower than the evidence strength threshold of the geological variable are marked as contradictory geological variables.

7. The method for generating mineral exploration information as described in claim 6, characterized in that: The process involves querying a geoscience knowledge graph, retrieving related geological phenomena from the graph that are causally linked to contradictory geological variables and acquired by field intelligent terminal devices, and generating a specific instruction to prioritize the verification of these related geological phenomena. This includes the following steps: Input the query conditions of the contradictory geological variable list into the geoscience knowledge graph, and the query interface will retrieve the geological entities that have a direct causal relationship with the contradictory geological variables based on the entity relationship triples stored in the geoscience knowledge graph. The list of associated geological entities returned by the knowledge graph query is matched with the list of sensors and data collection functions built into the field intelligent terminal device. The associated geological entities that have a causal relationship with contradictory geological variables and are observed by the sensors on the field intelligent terminal device are selected to form a candidate list of associated geological phenomena. The candidate list of associated geological phenomena is prioritized according to the strength of association and the difficulty of detection. The optimal associated geological phenomenon is selected as the verification target, and a specific verification instruction text is generated.

8. A mineral exploration information generation device, based on the mineral exploration information generation method according to any one of claims 1 to 7, characterized in that: This includes building modules that integrate and standardize multi-source exploration data in the cloud to generate a three-dimensional multidimensional data cube, read information from a geoscience knowledge base, and use natural language processing and entity relationship extraction technologies to construct a geoscience knowledge graph; The training module, based on a three-dimensional multidimensional data cube combined with a geoscience knowledge graph, trains a lightweight interpretable causal reasoning model in the cloud and sets thresholds for the causal contribution of geological variables and the strength of evidence support. The intelligent terminal is equipped with a deployment module to deploy a lightweight, interpretable causal reasoning model to the intelligent terminal device in the field to collect on-site survey data. The analysis module uses a deployed lightweight, interpretable causal reasoning model to perform real-time analysis of field survey data, obtaining the causal contribution of geological variables and the strength of current evidence support. The comparison module compares the causal contribution of geological variables with the strength of evidence supporting geological variables, and identifies contradictory geological variables whose causal contribution is higher than the contribution threshold of geological variables and whose strength of evidence supporting geological variables is lower than the strength of evidence threshold of geological variables. The instruction module queries the geoscience knowledge graph, retrieves related geological phenomena that are causally linked to contradictory geological variables and acquired by field intelligent terminal devices, and generates a specific instruction to prioritize the verification of the related geological phenomena.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the mineral exploration information generation method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the mineral exploration information generation method according to any one of claims 1 to 7.

Images