Mineral prediction method based on multi-element geoscience information superposition identification

Abstract

The invention relates to the field of mineral resource exploration, and discloses a mineral prediction method based on multivariate geoscience information superposition recognition, which comprises the following steps: collecting multivariate three-dimensional geoscience information through multiple modes such as unmanned aerial vehicle remote sensing and ground survey, classifying and sorting, and carrying out deep preprocessing and feature extraction such as denoising and correction to obtain a multi-element three-dimensional geoscience information model; a fusion layer is generated by adopting an adaptive fusion algorithm, a mineralization prediction model is constructed in combination with machine learning, and after training optimization, prediction regions of different grades are delineated and the resource quantity is estimated. And a verification feedback mechanism is also set, and each link is optimized and improved according to a field exploration result. The method improves the mineral product prediction accuracy and adaptability, improves the exploration efficiency, reduces the cost, promotes the technical development of the mineral exploration field, and has an important application value.

Description

Technical Field

[0001] The present invention relates to the field of mineral resource exploration, and particularly to a mineral prediction method based on the superposition and identification of multi-source geological information. Background Art

[0002] With the rapid development of modern computer technology, its application in various industries has become increasingly in-depth and plays a crucial role. The progress of computer technology has brought unprecedented opportunities and advantages to the processing of multi-source and complex data in mineral exploration. The significant improvement in computer hardware performance, such as the wide application of multi-core processors, high-performance graphics processing units (GPUs), and distributed computing technology, has greatly accelerated the data processing speed. Even in the face of massive and complex exploration data, it can quickly complete the analysis and processing and timely generate various analysis results and visualization charts for decision-making reference, strongly supporting the exploration decision-making work. The field of mineral exploration is no exception. The booming development of advanced software algorithms such as machine learning and data mining enables the computer to accurately extract valuable information and discover hidden laws from multi-source and complex data. With these technologies, it is possible to build a more accurate and reliable ore-forming prediction model based on multi-disciplinary multi-source and complex data, accurately predict key elements such as the potential distribution location and reserve scale of mineral resources, thereby significantly improving the accuracy of prediction, pointing out a more valuable direction for exploration work, effectively reducing exploration risks, and enhancing the success rate of ore prospecting. With the in-depth development of mineral exploration work.

[0003] Some existing multi-source geological information superposition prediction methods, although integrating multi-faceted information to a certain extent, still have deficiencies in terms of information comprehensiveness, fusion depth, and dynamic adaptability of the prediction model. Some methods fail to fully cover the geoscience data obtained from multiple sources, do not deeply mine the complex relationships between various types of information during the fusion process, and cannot adjust the prediction model in real time according to new data, resulting in limited accuracy and timeliness of mineral prediction. Therefore, there is an urgent need for a more perfect and intelligent mineral prediction method based on the superposition and identification of multi-source geological information to improve the efficiency and quality of mineral exploration. Summary of the Invention

[0004] The object of the present invention is to provide a mineral prediction method based on the superposition and identification of multi-source geological information, which solves the technical problems that some methods fail to fully cover the geoscience data obtained from multiple sources, do not deeply mine the complex relationships between various types of information during the fusion process, and cannot adjust the prediction model in real time according to new data, resulting in limited accuracy and timeliness of mineral prediction.

[0005] In order to achieve the above object of the invention, the technical solution adopted by the present invention is as follows:

[0006] A mineral prediction method based on the superposition and recognition of multi-source geological information, characterized by including the following steps: Step 1 Multi-source geological information collection step: Use various methods such as unmanned aerial vehicle (UAV) aerial remote sensing, satellite remote sensing, ground geological survey, geophysical exploration, geochemical analysis, and drilling to obtain multi-source three-dimensional geological information data of the target ore-producing area, and classify and organize them according to remote sensing factors, geological factors, geochemical factors, and geophysical factors. At the same time, record the metadata information of each data;

[0007] Step 2 Deep preprocessing and feature extraction step: Perform deep preprocessing operations such as denoising, calibration, registration, and fusion on the collected multi-source geological information data in sequence, and extract feature information with ore-forming indication significance from it for quantitative representation to form a feature data set;

[0008] Step 3 Dynamic fusion superposition and model construction step: Use an adaptive data fusion method to fuse the feature data set to generate a fused geological information layer. Based on this layer and the regional ore-forming law, combine machine learning algorithms to construct an ore-forming prediction model, use the training set and test set to train and test the model, and quantitatively evaluate and optimize the model performance according to the test set;

[0009] Step 4 Precise prediction and resource volume estimation step: Input the data of the area to be predicted into the ore-forming prediction model, circle ore-forming prediction areas of different grades according to the output ore-forming probability values, and estimate the predicted mineral resources volume in each prediction area in combination with existing data and resource volume estimation methods;

[0010] Step 5 Verification feedback and continuous improvement step: Conduct on-site exploration and verification on some ore-forming prediction areas, compare and analyze the verification results and prediction results, and feedback and adjust and optimize the relevant links in the prediction method according to the differences to improve the accuracy and reliability of mineral prediction.

[0011] As an improvement, the classification and organization need to record in detail the metadata information such as the collection time, spatial coordinates, and accuracy range of each data. The aircraft used for UAV aerial remote sensing is a multi-rotor UAV or a fixed-wing UAV, and the remote sensing equipment carried by it has multi-spectral and hyperspectral imaging capabilities to ensure accurate identification of fine geological features and alteration information.

[0012] As an improvement, for feature extraction, professional software and algorithms are required to extract and quantify feature information with ore-forming indication significance, such as fold curvature, element enrichment coefficient, image texture features, etc., from various data. For denoising processing, algorithms such as median filtering and wavelet transform are used for remote sensing image data, and methods such as moving average filtering and statistical filtering are used for geophysical and geochemical data. Moreover, the filtering parameters can be adaptively adjusted according to the data noise level to achieve the best denoising effect. For the registration operation, ground control points and image matching algorithms are used to ensure that the error in the spatial position of data from different data sources is controlled within a certain range, guaranteeing the accuracy of subsequent fusion analysis.

[0013] As an improvement, in the step of dynamic fusion superposition and model construction, the adaptive data fusion methods include adaptive weighted fusion algorithms, deep learning fusion architectures, etc. When using the adaptive weighted fusion algorithm, the initial weights of various types of geoscience information data are preset according to expert experience and the statistical analysis results of historical ore-forming data. And during the fusion process, according to the correlation analysis of the data and the real-time feedback on the prediction results, the weight coefficients are dynamically adjusted to ensure the optimal fusion effect. For the random forest algorithm in the machine learning algorithm, the range of the number of decision trees is set, the maximum depth of each decision tree is set, and during the training process, sample sampling is used to improve the generalization ability and prediction accuracy of the model.

[0014] As an improvement, in the step of precise prediction and resource volume estimation, different grades of ore-forming prediction areas are delineated by comprehensively analyzing and setting thresholds according to the ore-forming probability value and regional ore-forming conditions. When using the geostatistical method to estimate the resource volume, a suitable variogram model needs to be selected according to the spatial distribution characteristics of the ore body, and its model parameters are determined by fitting and analyzing the experimental variogram of the exploration data of known ore deposits to improve the accuracy of resource volume estimation. When considering the ore body shape factor, for solid ore bodies, geometric features such as strike, dip, dip angle, and thickness change need to be analyzed; for liquid ore bodies, physical properties parameters such as reservoir porosity and permeability need to be considered; for gas ore bodies, occurrence conditions such as pressure and temperature need to be combined to more accurately estimate the resource volume of ore bodies with different shapes.

[0015] As an improvement, in the step of verification feedback and continuous improvement, actual geological, ore body, mineralization, and resource volume information, etc., are obtained through engineering means such as drilling, adit exploration, and trenching during the on-site exploration verification. During the on-site exploration verification process, the drilling holes of the drilling project need to follow the geostatistical principle and the predicted distribution of the ore body, and reasonable hole layout methods such as regular grids and irregular triangular meshes are adopted. Moreover, the drilling depth is determined according to the predicted buried depth of the ore body and the complexity of the geological structure to ensure that the key information required for verification can be effectively obtained.

[0016] The beneficial effects of the present invention are as follows: The present invention shows remarkable beneficial effects in aspects such as the accuracy of mineral prediction, adaptability, exploration efficiency, and technological development. It can provide strong technical support for the exploration, development, and rational utilization of mineral resources, and has important practical application value and industry development significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a process framework diagram of a mineral prediction method based on the overlay and identification of multi-source geological information of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] As Figure 1 shown, a mineral prediction method based on the overlay and identification of multi-source geological information, characterized by comprising the following steps: Step 1: Multi-source geological information collection step: Use various methods such as unmanned aerial vehicle (UAV) aerial remote sensing, satellite remote sensing, ground geological survey, geophysical exploration, geochemical analysis, and drilling to obtain multi-source three-dimensional geological information data of the target ore-producing area, and classify and sort them according to remote sensing factors, geological factors, geochemical factors, and geophysical factors. At the same time, record the metadata information of each data. The classification and sorting need to record in detail the metadata information such as the collection time, spatial coordinates, and accuracy range of each data. The UAV used for UAV aerial remote sensing is a multi-rotor UAV or a fixed-wing UAV, and the remote sensing equipment it carries has multi-spectral and hyperspectral imaging capabilities to ensure accurate identification of subtle geological features and alteration information;

[0019] Step 2: Deep preprocessing and feature extraction step: Perform deep preprocessing operations such as denoising, calibration, registration, and fusion on the collected multi-source geological information data in sequence, and extract feature information with ore-forming indication significance from it for quantitative representation to form a feature dataset. The feature extraction needs to use professional software and algorithms to extract feature information with ore-forming indication significance such as fold curvature, element enrichment coefficient, and image texture features from each data and quantify them. The denoising process uses algorithms such as median filtering and wavelet transform for remote sensing image data, and uses methods such as moving average filtering and statistical filtering for geophysical and geochemical data, and the filtering parameters can be adaptively adjusted according to the data noise level to achieve the best denoising effect. The registration operation uses ground control points and image matching algorithms to ensure that the error of data from different data sources in spatial position is controlled within a certain range to ensure the accuracy of subsequent fusion analysis;

[0020] Step 3: Dynamic fusion superposition and model construction step: An adaptive data fusion method is used to fuse the feature data set to generate a fused geoscience information layer. Based on this layer and the regional metallogenic law, a metallogenic prediction model is constructed by combining machine learning algorithms. The model is trained and tested using the training set and the test set, and the performance of the model is quantitatively evaluated and optimized according to the test set. In the dynamic fusion superposition and model construction step, the adaptive data fusion method includes an adaptive weighted fusion algorithm, a deep learning fusion architecture, etc. When using the adaptive weighted fusion algorithm, the initial weights of various types of geoscience information data are preset according to expert experience and the statistical analysis results of historical metallogenic data, and during the fusion process, according to the correlation analysis of the data and the real-time feedback on the prediction results, the weight coefficients are dynamically adjusted to ensure the optimal fusion effect. For the random forest algorithm in the machine learning algorithm, the number of decision trees is set within a certain range, the maximum depth of each decision tree is set to a certain value, and sample sampling is used during the training process to improve the generalization ability and prediction accuracy of the model. The adaptive algorithm first assigns initial weights to various types of geoscience information such as remote sensing factors, geological factors, geochemical factors, geophysical factors, etc. according to expert experience and the statistical analysis of historical metallogenic data. These initial weights reflect the relative importance degree of each type of data for metallogenic prediction under normal circumstances. During the actual fusion process, the matching degree between the fusion result and the distribution of known ore deposits, mineralization points, and actual geological conditions will be monitored in real time. For example, it is evaluated by calculating indicators such as the coincidence degree between the potential metallogenic area indicated by the fused geoscience information layer and the location of the proven ore deposit, and the rationality of the geological structure and metallogenic relationship reflected. Then, according to quantitative indicators such as mutual information and correlation coefficient between the data, the weight coefficients of each data are dynamically adjusted according to specific weight adjustment rules. For example, if the geochemical anomaly data shows a stronger correlation with the distribution of known ore bodies in a certain area, it indicates that the importance of the geochemical data in this area for judging the metallogenic situation has increased, then the weight of the geochemical factor will be appropriately increased. On the contrary, if a certain type of data has a weak indication effect on actual mineralization, its weight will be correspondingly reduced, so as to ensure that the fused geoscience information layer can more comprehensively and accurately reflect the comprehensive metallogenic characteristics of the region;

[0021] Step 4 Precise Prediction and Resource Quantity Estimation Step: Input the data of the area to be predicted into the metallogenic prediction model. Circle metallogenic prediction areas of different grades according to the output metallogenic probability values, and estimate the predicted mineral resources in each prediction area by combining existing data and resource quantity estimation methods. In the precise prediction and resource quantity estimation step, the threshold for setting the metallogenic prediction areas of different grades needs to be determined through comprehensive analysis based on the metallogenic probability value and regional metallogenic conditions. When using the geostatistical method to estimate the resource quantity, a suitable variogram model needs to be selected according to the spatial distribution characteristics of the ore body, and its model parameters are determined by fitting and analyzing the experimental variogram of the exploration data of known ore deposits to improve the accuracy of resource quantity estimation. When considering the ore body shape factor, for solid ore bodies, geometric characteristics such as strike, dip, dip angle, and thickness change need to be analyzed; for liquid ore bodies, physical property parameters such as reservoir porosity and permeability need to be considered; for gas ore bodies, occurrence conditions such as pressure and temperature need to be combined to more precisely estimate the resource quantity of ore bodies of different shapes;

[0022] Step 5 Verification Feedback and Continuous Improvement Step: Conduct on-site exploration and verification on some metallogenic prediction areas, compare and analyze the verification results with the prediction results, and feedback and adjust and optimize the relevant links in the prediction method based on the differences to improve the accuracy and reliability of mineral prediction. In the verification feedback and continuous improvement step, the on-site exploration and verification need to obtain actual geological, ore body, mineralization, and resource quantity information through engineering means such as drilling, adit, and trenching. During the on-site exploration and verification process, the borehole layout of the drilling project needs to follow the geostatistical principle and the predicted distribution of the ore body, and reasonable borehole layout methods such as regular grids and irregular triangular meshes are adopted, and the borehole depth is determined according to the buried depth of the predicted ore body and the complexity of the geological structure to ensure that key information required for verification can be effectively obtained.

[0023] The median filtering algorithm is adopted in Step 2: Median filtering is a non-linear filtering method based on the theory of order statistics. For each pixel in the image or data sequence, the pixel values within a certain neighborhood range are selected for sorting, and then the middle value after sorting is used to replace the original value of the pixel, which can effectively remove isolated noise points such as salt-and-pepper noise, make the image or data smoother, and can better retain details such as edges;

[0024] Wavelet transform algorithm: Wavelet transform can decompose a signal into sub-band components of different scales and frequencies. By performing threshold processing on the wavelet coefficients in the high-frequency sub-bands, setting the coefficients below or above a specific threshold to zero or performing shrinkage adjustment, and then performing wavelet reconstruction, the purpose of removing noise can be achieved. It can remove noise while better retaining the detailed features in the signal, such as fine but ore-forming indication-related information such as texture and edges in the image;

[0025] Moving average filtering algorithm: For a data sequence, a sliding window with a fixed length is set, the average value of the data within the window is calculated, and this average value is used to replace the data value at the center position of the window. As the window slides along the data sequence in turn, such processing is performed on each data point to smooth out the fluctuating noise in the data and highlight the overall change trend of the data;

[0026] Statistical filtering algorithm: Based on statistical principles, it is usually assumed that the data follows a certain probability distribution, and then whether a data point is an abnormal noise point is judged according to statistical characteristics such as the mean and standard deviation of the data. For example, if a data point deviates from the overall mean of the data by more than a certain multiple of the standard deviation, it is considered that this data point may be abnormal noise, and then it is corrected or removed, etc. to ensure the data quality and facilitate the subsequent mining of effective information related to mineralization;

[0027] Texture feature extraction algorithm based on gray-level co-occurrence matrix (GLCM): It is mainly used for the quantitative extraction of texture features of image data such as remote sensing images. It calculates the co-occurrence probability of pixel gray values in different directions and at different distances in the image, and then obtains multiple texture feature parameters. Common ones include contrast, correlation, energy, entropy, etc. Contrast reflects the contrast degree of local gray changes in the image; correlation reflects the linear correlation degree of pixel gray values in space; energy represents the uniformity of the image gray distribution and the texture thickness; entropy measures the randomness or complexity of the image texture;

[0028] Fold curvature calculation algorithm based on curve fitting: In geological data processing, for the analysis of fold structures, the curve shape of the fold is first digitally sampled to obtain a series of discrete coordinate points to represent the fold curve. Then, mathematical methods are used to perform curve fitting on these discrete points to obtain the mathematical expression of the fold curve. After that, through mathematical operations such as taking the derivative of this expression, the curvature values at different positions on the curve are calculated to quantify the fold curvature feature, which can reflect geological information related to mineralization such as the stress concentration situation in the fold area;

[0029] Element enrichment coefficient calculation algorithm: For geochemical data, the calculation of the element enrichment coefficient is achieved by comparing the content of a certain element in the target area with the average content of this element in the regional background. The specific calculation formula is: Element enrichment coefficient = (Content of the element in the target area / Content of the element in the regional background); The content of the element in the target area is obtained by collecting samples in the field (such as rock, soil samples, etc.) and then using laboratory chemical analysis methods or spectroscopic analysis instruments, etc.; The content of the element in the regional background is the average value obtained through statistical analysis of a large number of samples collected from a larger unmineralized area.

[0030] The algorithm adopted in Step 3 is the adaptive weighted fusion algorithm: Based on expert experience and statistical analysis of historical ore-forming data, initial weights are assigned to various types of geoscience information (such as remote sensing factors, geological factors, geochemical factors, geophysical factors, etc.). These initial weights reflect the relative importance of each type of data for ore-forming prediction under normal circumstances. For example, if past experience shows that geochemical data is more relevant in predicting a certain type of mineral, a relatively higher initial weight may be assigned to it;

[0031] Deep learning fusion architecture: Taking the convolutional neural network as an example, it includes structures such as convolutional layers, pooling layers, and fully connected layers. The convolutional layer performs convolutional operations by sliding the convolutional kernel over the input data (such as different types of geoscience image data), extracting features at different levels, which gradually progress from low-level edge and texture features to high-level abstract features such as geological structures. The pooling layer then downsamples the features after convolution, reducing the data volume while retaining key features and improving computational efficiency. The fully connected layer integrates the features extracted previously and finally outputs the fused result representation;

[0032] The above are only the preferred embodiments of the present invention patent, and are not intended to limit the present invention patent. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention patent shall be included within the protection scope of the present invention patent.

Claims

1. A mineral prediction method based on the superposition recognition of multi-source geological information, characterized in that It includes the following steps: Step 1 Multi-source geoscience information collection step: Use multiple methods such as unmanned aerial vehicle (UAV) aerial remote sensing, satellite remote sensing, ground geological survey, geophysical exploration, geochemical analysis, and drilling to obtain multi-dimensional three-dimensional geoscience information data of the target ore-producing area, and classify and organize them according to remote sensing factors, geological factors, geochemical factors, and geophysical factors. At the same time, record the metadata information of each data; Step 2 Deep preprocessing and feature extraction step: Perform deep preprocessing operations such as denoising, calibration, registration, and fusion on the collected multi-source geoscience information data in sequence, and extract characteristic information with ore-forming indication significance from it for quantitative representation to form a characteristic data set; Step 3 Dynamic fusion superposition and model construction step: Use an adaptive data fusion method to fuse the characteristic data set to generate a fused geoscience information layer. Based on this layer and the regional ore-forming law, combined with machine learning algorithms, construct an ore-forming prediction model, use the training set and test set to train and test the model, and quantitatively evaluate and optimize the model performance according to the test set; Step 4 Precise prediction and resource volume estimation step: Input the data of the area to be predicted into the ore-forming prediction model, delineate ore-forming prediction areas of different grades according to the output ore-forming probability values, and estimate the predicted mineral resources volume in each prediction area in combination with existing data and resource volume estimation methods; Step 5 Verification feedback and continuous improvement step: Conduct on-site exploration and verification on some ore-forming prediction areas, compare and analyze the verification results and prediction results, and feedback and adjust and optimize the relevant links in the prediction method according to the differences to improve the accuracy and reliability of mineral prediction.

2. According to the ore prediction method based on multi-source geoscience information superposition recognition described in claim 1, the classification and organization need to record in detail the metadata information such as the collection time, spatial coordinates, and accuracy range of each data. The aircraft used for UAV aerial remote sensing is a multi-rotor UAV or a fixed-wing UAV, and the remote sensing equipment carried by it has multi-spectral and hyperspectral imaging capabilities to ensure accurate identification of subtle geological features and alteration information.

3. A mineral prediction method based on the overlay and recognition of multivariate geological information according to claim 1, characterized in that The feature extraction needs to use professional software and algorithms to extract and quantify characteristic information with ore-forming indication significance such as fold curvature, element enrichment coefficient, and image texture features from each data. The denoising process uses algorithms such as median filtering and wavelet transform for remote sensing image data, and uses methods such as moving average filtering and statistical filtering for geophysical and geochemical data, and the filtering parameters can be adaptively adjusted according to the data noise level to achieve the best denoising effect. The registration operation uses ground control points and image matching algorithms to ensure that the error of data from different data sources in the spatial position is controlled within a certain range to ensure the accuracy of subsequent fusion analysis.

4. A mineral prediction method based on the overlay recognition of multivariate geological information according to claim 1, characterized in that, In the dynamic fusion superposition and model construction steps, the adaptive data fusion method includes an adaptive weighted fusion algorithm, a deep learning fusion architecture, etc. When using the adaptive weighted fusion algorithm, the initial weights of various types of geological information data are preset according to expert experience and the statistical analysis results of historical ore-forming data, and during the fusion process, according to the correlation analysis of the data and the real-time feedback on the prediction results, the weight coefficients are dynamically adjusted to ensure the optimal fusion effect. For the random forest algorithm in the machine learning algorithm, the number of decision trees is set within a certain range, the maximum depth of each decision tree is set to a certain value, and sample sampling is adopted during the training process to improve the generalization ability and prediction accuracy of the model.

5. A mineral prediction method based on the overlay recognition of multi-source geological information according to claim 1, characterized in that, In the precise prediction and resource volume estimation steps, the delineation of ore-forming prediction areas at different grades needs to be set by comprehensively analyzing and setting thresholds according to the ore-forming probability values and regional ore-forming conditions. When using the geostatistical method to estimate the resource volume, an appropriate variogram model needs to be selected according to the spatial distribution characteristics of the ore body, and its model parameters are determined by fitting and analyzing the experimental variogram of the exploration data of known ore deposits to improve the accuracy of resource volume estimation. When considering the ore body shape factor, for solid ore bodies, it is necessary to analyze their geometric characteristics such as strike, dip, dip angle, and thickness change; for liquid ore bodies, it is necessary to consider physical property parameters such as reservoir porosity and permeability; for gas ore bodies, it is necessary to combine their occurrence conditions such as pressure and temperature to more accurately estimate the resource volume of ore bodies with different shapes.

6. A mineral prediction method based on the superposition and recognition of multivariate geological information according to claim 1, characterized in that, In the verification feedback and continuous improvement steps, the on-site exploration verification needs to obtain actual geological, ore body, mineralization, and resource volume information through engineering means such as drilling, adit exploration, and trenching. During the on-site exploration verification process, the drilling holes of the drilling project need to follow the principles of geostatistics and the predicted distribution of the ore body, and reasonable hole layout methods such as regular grids and irregular triangular meshes are adopted, and the drilling depth is determined according to the predicted burial depth of the ore body and the complexity of the geological structure to ensure that the key information required for verification can be effectively obtained.

Images