A method for monitoring mineral exploration based on Beidou satellite positioning

Abstract

This invention relates to the field of mineralized exploration monitoring technology, and discloses a mineralized exploration monitoring method based on BeiDou satellite positioning. The method collects BeiDou satellite positioning data of the target area to obtain the positioning coordinate sequence and accuracy characteristics. Based on the accuracy characteristics, it selects the optimal solution model for spatiotemporal decoupling, generating multi-dimensional data containing elevation and planar components. It extracts anomalous fluctuation components from the data and calculates geological deformation feature vectors in conjunction with geological structural constraints. It integrates vein distribution and stratum stability indicators from historical exploration data to construct a geological evolution trend prediction model. Based on this model, it divides the monitoring priority grid, dynamically adjusts the BeiDou sampling frequency to update data acquisition and correct deformation vectors. It performs multi-scale fusion analysis on the corrected vectors to extract significant mineralization-related geological activity patterns, thereby generating exploration path planning instructions to drive equipment to perform directional exploration, effectively improving the accuracy and efficiency of exploration.

Description

Technical Field

[0001] This invention relates to the field of mineral exploration monitoring technology, specifically a mineral exploration monitoring method based on BeiDou satellite positioning. Background Technology

[0002] Mineral exploration is a crucial step in mineral resource exploration and development. Traditional exploration methods primarily rely on geological mapping, geophysical exploration, and geochemical sampling. These methods often suffer from long cycles, high costs, and low efficiency, and have limited ability to detect deep ore bodies. With the development of satellite navigation technology, Global Navigation Satellite Systems (GNSS) have provided high-precision technical means for monitoring crustal deformation. The BeiDou Navigation Satellite System, as my country's independently developed satellite navigation system, features wide coverage and high positioning accuracy, providing a new data source for regional crustal deformation monitoring.

[0003] The application of BeiDou positioning technology in geological exploration is mainly concentrated in static topographic surveying and point positioning. In-depth and systematic research on dynamic geological deformation monitoring and mineralization potential analysis is still lacking. Existing technologies typically use BeiDou positioning data as independent spatial coordinate information, failing to fully exploit its inherent geological evolution information. Issues such as fluctuations in positioning data accuracy and multipath effects also affect its reliability in precise geological analysis. Furthermore, traditional mineralization prediction models are mostly based on historical geological data and statistical methods, making it difficult to reflect the dynamic changes in crustal activity in real time. The prediction results are updated late, failing to meet the timeliness requirements of modern mineral exploration.

[0004] There is a close spatiotemporal correlation between geological deformation and mineralization. Minor crustal deformation often foreshadows changes in deep geological activity, potentially indicating the formation or alteration of ore bodies. However, how to effectively extract mineralization-related geological deformation signals from BeiDou positioning data, how to organically combine real-time deformation monitoring with historical mineralization patterns, and how to dynamically adjust exploration strategies based on monitoring results are all problems that current technology has not yet adequately solved. Therefore, a fully intelligent method is needed that can fully utilize BeiDou positioning data to achieve a complete process from deformation monitoring to mineralization prediction and exploration optimization. Summary of the Invention

[0005] The purpose of this invention is to provide a mineral exploration monitoring method based on BeiDou satellite positioning to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a mineral exploration monitoring method based on BeiDou satellite positioning, the method comprising:

[0007] Collect BeiDou satellite positioning data of the target area to obtain positioning coordinate sequences and corresponding positioning accuracy characteristics;

[0008] The optimal positioning solution model is selected based on the positioning accuracy characteristics, and the positioning coordinate sequence is subjected to spatiotemporal decoupling processing to generate multi-dimensional positioning data containing elevation and planar components.

[0009] Extract the abnormal fluctuation components from the multi-dimensional positioning data, and calculate the geological deformation feature vector by combining them with preset geological structural constraints.

[0010] Obtain the distribution characteristics of veins and the stability index of rock strata from historical mineralization exploration data, and construct a geological evolution trend prediction model;

[0011] The geological deformation feature vector is input into the geological evolution trend prediction model, and a probability distribution map of the mineralization potential area is output.

[0012] Based on the probability distribution map, a monitoring priority grid is divided, a dynamically adjusted BeiDou sampling frequency strategy is generated, the positioning data acquisition process is updated according to the BeiDou sampling frequency strategy, and the geological deformation feature vector is corrected in real time.

[0013] Multi-scale fusion analysis is performed on the corrected geological deformation feature vector to extract geological activity patterns with significant mineralization correlation. Based on the geological activity patterns, exploration path planning instructions are generated to drive monitoring equipment to perform directional exploration operations.

[0014] Preferably, the selection of the optimal positioning solution model includes:

[0015] Calculate the sum of squared residuals and the confidence interval width for the positioning coordinate sequence using different solution models;

[0016] A dynamic weighted algorithm is used to fuse the residual sum of squares and the confidence interval width to generate a model fit score.

[0017] The solution model with the highest score is selected as the optimal localization solution model, and its corresponding coordinate transformation parameters are recorded.

[0018] Preferably, the calculation of the geological deformation feature vector includes:

[0019] The abnormal fluctuation components are convolved with the geological structural constraints to generate an initial deformation matrix.

[0020] Principal component analysis was performed on the initial deformation matrix, and the first three principal components were extracted as basic deformation features.

[0021] The basic deformation features are compared with historical data from the same location to obtain a standardized deformation feature vector.

[0022] Preferably, the construction of the geological evolution trend prediction model includes:

[0023] Convert the vein distribution characteristics in historical metallogenic exploration data into a spatial density field;

[0024] The stability index of the rock strata was spatially discretized using the Kriging interpolation method.

[0025] By integrating the spatial density field and discretized rock strata indicators through grey relational analysis, trend prediction weight coefficients are generated.

[0026] Preferably, the strategy for generating dynamically adjusted BeiDou sampling frequencies includes:

[0027] The probability distribution map is divided into three potential level regions: high, medium, and low.

[0028] The maximum sampling frequency is assigned to high-potential areas, the baseline sampling frequency is assigned to medium-potential areas, and the minimum sampling frequency is assigned to low-potential areas.

[0029] The frequency gradient of the transition zone is set according to the difference in potential levels between adjacent areas.

[0030] Preferably, the real-time correction of the geological deformation feature vector includes:

[0031] Establish a sliding time window comparison mechanism between location data streams and historical deformation characteristics;

[0032] When the deformation difference over three consecutive time windows exceeds a preset threshold, the feature vector recalculation process is triggered.

[0033] The Kalman filter algorithm is used to smooth the recalculated results.

[0034] Preferably, the geological activity patterns with significant mineralization correlation include:

[0035] The time dimension is divided into three scales: short cycle, medium cycle, and long cycle.

[0036] Calculate the correlation coefficient matrix between deformation characteristics and vein distribution at each scale;

[0037] By fusing multi-scale correlation coefficients using the tensor decomposition algorithm, cross-period consistency feature patterns are extracted.

[0038] Preferably, the instructions for generating exploration route planning include:

[0039] Geological activity patterns are mapped as hotspot intensity fields in three-dimensional space;

[0040] The ant colony optimization algorithm is used to search for the optimal exploration path in the hotspot intensity field;

[0041] Generate a sequence of motion control parameters for the equipment based on the intensity values ​​of the hotspots along the path.

[0042] Preferably, the search for the optimal exploration path includes:

[0043] Initialize the random distribution state of the ant colony in the hotspot intensity field;

[0044] Ants are guided to gather in high-intensity areas by using pheromone concentration update rules;

[0045] The search terminates when the overlap of consecutive iterations reaches 95%, and the final set of path nodes is output.

[0046] Preferably, the present invention also includes a mineral exploration monitoring system based on BeiDou satellite positioning, comprising a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the above-mentioned mineral exploration monitoring method based on BeiDou satellite positioning.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] This invention achieves high-precision monitoring of crustal deformation through deep decoupling and multi-dimensional analysis of BeiDou satellite positioning data. Traditional applications of positioning data often remain at the level of simple coordinate acquisition, while this method, through spatiotemporal decoupling of coordinate sequences, separates elevation and planar components, enabling a more refined characterization of the three-dimensional deformation features of the Earth's crust and providing a richer data foundation for geological analysis.

[0049] By organically combining real-time deformation monitoring with historical mineralization patterns, the scientific rigor and timeliness of mineralization prediction have been enhanced. The geological evolution trend prediction model not only considers the distribution of historical veins and the stability of rock strata, but also incorporates real-time deformation feature vectors, enabling the prediction results to reflect the current state of geological activity. This achieves a shift from static to dynamic prediction and enhances the practical guiding significance of the prediction results.

[0050] A dynamic sampling strategy based on probability distribution maps enables optimized allocation of monitoring resources. By dividing the system into monitoring priority grids, it can identify key areas with high mineralization potential and automatically adjust the sampling frequency of BeiDou data in these areas, achieving precise deployment of monitoring resources. This adaptive sampling mechanism ensures monitoring accuracy in key areas while avoiding the average distribution and waste of resources, thus improving overall monitoring efficiency.

[0051] Multi-scale fusion analysis and the generation of exploration path planning instructions enable closed-loop management from monitoring to exploration. The system goes beyond data analysis and prediction; it can directly generate actionable exploration instructions based on the analysis results, driving monitoring equipment to perform directional exploration. This data-driven decision-making model shortens the time cycle from anomaly detection to on-site verification, improving the targeting and success rate of exploration work.

[0052] This method constructs a continuously optimizing intelligent monitoring system. By real-time correction of geological deformation feature vectors and dynamic adjustment of sampling strategies, the system can continuously optimize its parameters and strategies based on the latest monitoring data, forming a virtuous cycle of self-improvement. This adaptive capability enables the monitoring system to maintain high-efficiency operation over the long term, adapting to constantly changing geological conditions and exploration needs. Attached Figure Description

[0053] Figure 1 This is a schematic diagram illustrating the working principle of the mineral exploration monitoring method based on BeiDou satellite positioning described in this invention.

[0054] Figure 2 A flowchart for selecting the optimal localization solution model;

[0055] Figure 3 A flowchart for constructing a geological evolution trend prediction model. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Please see Figure 1This invention provides a method for monitoring mineralized exploration based on BeiDou satellite positioning. The method includes: when collecting BeiDou satellite positioning data of a target area, recording the positioning coordinate sequence using a high-precision BeiDou receiving device, and simultaneously extracting positioning accuracy features such as horizontal accuracy factor, vertical accuracy factor, and signal-to-noise ratio. Based on these features, the optimal positioning solution model is selected, and an adaptive evaluation algorithm is used to compare the output stability of multiple models, selecting the appropriate model for subsequent processing. The positioning coordinate sequence is decoupled spatiotemporally, decomposing it into an elevation component reflecting topographic changes and a planar component representing horizontal displacement, generating a multi-dimensional positioning dataset. Abnormal fluctuation components are extracted from the dataset, and fluctuation points are identified through time series analysis. Spatial correlation calculations are performed using geological structural constraints such as fault distribution and lithological boundaries to generate a geological deformation feature vector. The distribution characteristics of veins (ore body density, strike trend) and strata stability indicators (compressive strength, joint development degree) from historical mineralized exploration data are obtained, and a geological evolution trend prediction model is constructed using machine learning algorithms. The model inputs geological deformation feature vectors to output a probability distribution map of mineralization potential areas. Based on this, a monitoring priority grid is defined, a dynamically adjusted BeiDou sampling frequency strategy is generated, positioning data acquisition is updated, and the geological deformation feature vectors are corrected in real time. Multi-scale fusion analysis is performed on the corrected vectors to integrate temporal trends and extract geological activity patterns with significant mineralization correlations. Exploration path planning instructions are generated based on these patterns, mapped to a spatial heat map, and used to drive directional exploration by monitoring equipment, achieving efficient resource exploration.

[0058] Example 1: See Figure 2 In the specific operation of selecting the optimal positioning solution model, the first step is to systematically calculate the residual sum of squares and confidence interval width of different solution models for the collected positioning coordinate sequence. This process relies on in-depth statistical analysis of the original observations. The calculation of the residual sum of squares requires comparing the theoretical value calculated by the candidate solution model for each coordinate point with the actual received observation value one by one. The difference is squared and then summed to obtain the total, thereby quantifying the degree of fit between the model and the original data. A smaller residual sum of squares usually means that the model can better reproduce the real observation situation. The calculation of the confidence interval width involves more complex statistical inference. It is necessary to determine the possible fluctuation range of the parameter estimates based on the selected confidence level (e.g., 95%) and the standard error of the solution model output. This width value reflects the stability and reliability of the model solution results. A narrower confidence interval indicates that the model has higher estimation accuracy under the current observation conditions.

[0059] The core step in model selection is to use a dynamic weighted algorithm to fuse the sum of squared residuals and the confidence interval width. This algorithm does not assign fixed weights but instead employs a mechanism that automatically adjusts the weight coefficients based on real-time data quality characteristics. For example, during periods of poor observation conditions and significant data noise, the algorithm appropriately increases the weight of the confidence interval width, as model stability is more critical than fit in such situations. By normalizing the two indicators, multiplying them by their respective dynamic weights, and then summing them, a comprehensive model fit score is generated. This score becomes the single quantitative standard for measuring model performance. The solution model with the highest score is selected as the optimal positioning solution model for the current data stream. This selection forms the basis for all subsequent spatial data processing. Simultaneously with model selection, the complete set of corresponding coordinate transformation parameters must be recorded. These parameters include the rotation matrix, translation vector, and scaling factor required for the transformation from the satellite coordinate system to the local measurement coordinate system, as well as parameters such as the central meridian and projection plane elevation involved in map projection. Recording these parameters ensures the traceability and repeatability of the transformation process from the original observations to the final applied coordinates.

[0060] In the stage of calculating the geological deformation feature vector, the work begins by performing a convolution operation on the pre-processed anomalous fluctuation components and the pre-set geological structural constraints. The anomalous fluctuation components are high-frequency signal components separated from multi-dimensional positioning data through a bandpass filter. They may contain information on minute deformations caused by geological activities. The geological structural constraints are input in the form of a digital raster map, where each pixel value represents the probability intensity of the presence of faults, folds or other structures at that location. The convolution operation is essentially a sliding weighted summation of the fluctuation signal along the structural constraint map in the spatial domain to generate a brand-new initial deformation matrix, which highlights the deformation response pattern under the control of geological structures.

[0061] Principal component analysis (PCA) of the initial deformation matrix is ​​a key step in data dimensionality reduction and feature extraction. This mathematical method can transform deformation matrix variables with multicollinearity into a new set of uncorrelated composite variables, namely principal components. By calculating the eigenvalues ​​and eigenvectors of the matrix, and selecting the three principal components with the highest contribution rates in descending order of eigenvalues ​​as basic deformation features, these three principal components can usually explain most of the variance information in the original deformation matrix. They represent the main extension direction, compression direction, and vertical motion characteristics of deformation in space, respectively, thus capturing the core elements of a complex deformation field with a small number of composite variables.

[0062] The implementation of differential operations begins with precise access to and alignment of historical databases. These databases store continuous observation records from the same monitoring area over many years. Each record point contains precise latitude and longitude coordinates, elevation data, and a corresponding observation timestamp. When accessing data, the system retrieves corresponding data from all historical periods for the same point in the historical database based on the spatial coordinates of the current basic deformation characteristic data point, ensuring complete spatial matching and comprehensive time series coverage. To handle potential data gaps or coordinate deviations, the system employs spatial interpolation or nearest neighbor matching algorithms for data repair, ensuring that each current data point can find a comparable historical benchmark value. Specifically, the system accesses the historical database to retrieve historical observation records that match the spatial coordinates of the current monitoring point. When data gaps or coordinate deviations are detected, for spatial interpolation algorithms, the system uses an inverse distance weighted interpolation method based on the spatial distribution of surrounding known historical data points to calculate the estimated value of the missing point. This method assigns higher weights to neighboring points, making the interpolation result more closely match local geological features. For the nearest neighbor matching algorithm, the system calculates the Euclidean distance between the current point and all historical points, selects the historical point with the smallest distance as the benchmark, and directly uses its data value for patching. During the data alignment process, the system also verifies the coordinate transformation parameters to ensure that the patched data is consistent with the current coordinate system, thereby guaranteeing the comparability and continuity of historical benchmark values. This patching mechanism effectively eliminates deviations introduced by sampling intervals or measurement errors, providing a reliable data foundation for subsequent difference operations.

[0063] After data alignment, time window standardization is required. Since historical observations may be conducted in different years and seasons with varying data collection frequencies, the system resamples historical data to the same time resolution as the current data. For example, if the current data represents monthly deformation characteristics, historical data will also be aggregated into a monthly average to eliminate biases introduced by different sampling intervals. Considering the long-term nature of geological processes, the system sets a reference time baseline, such as using data from the past five or ten years as the background trend, and calculates its moving average to represent the long-term stable state of that point. The core of the difference operation is to calculate the residual of the current value relative to the historical baseline value. For each data point, the system extracts the basic deformation characteristic value of the current period and subtracts the characteristic value corresponding to the historical baseline period of the same point. This historical baseline value is usually taken as the median or mean of the long-term observation series to reduce the influence of outliers. The difference result directly reflects the recent change after removing the long-term trend. Positive values ​​may indicate uplift or expansion, while negative values ​​may suggest subsidence or contraction. During the operation, the confidence interval of the change is also calculated, and the significance of the current change is assessed based on the fluctuation range of historical data.

[0064] To enhance the reliability of the results, the system performs smoothing filtering on the difference results, employing moving averages or low-pass filters to eliminate high-frequency noise. This noise may originate from short-term meteorological factors, measurement errors, or other non-geological interferences. The filtered difference values ​​more clearly reveal the true deformation signals. Subsequently, these values ​​are organized into standardized geological deformation feature vectors, where each dimension represents the standardized variation of a deformation feature. The generation of standardized vectors also includes a normalization step, scaling the difference values ​​to a uniform numerical range, for example, using min-max normalization or Z-score normalization, making different features comparable. The normalized vectors not only quantify the magnitude of the change but also retain the directional information, thus providing consistent data input for subsequent models. Finally, the standardized geological deformation feature vectors are input into the geological evolution trend prediction model. Their purity allows the model to more accurately identify anomalous deformation patterns related to mineralization, avoiding the misleading influence of long-term background noise and improving the accuracy and practicality of predictions.

[0065] Example 2: See Figure 3 The core step in constructing a geological evolution trend prediction model begins with converting the vein distribution characteristics in historical mineralization exploration data into a continuous spatial density field. This conversion process relies on the kernel density estimation method in spatial point pattern analysis. Historical data usually exists in the form of discrete points, each containing the precise geographic coordinates of the vein outcrop or borehole encounter and its associated ore reserves or grade information. Kernel density estimation calculates the density contribution of each point to the surrounding space by setting a kernel function with a predefined bandwidth for each discrete point. Then, the contributions of all points are superimposed and summed on regular grid nodes to generate a smooth and continuous density surface covering the entire study area. This density field intuitively reflects the degree of spatial aggregation and distribution trend of veins, with high-density areas indicating favorable mineralization locations.

[0066] Spatial discretization of rock strata stability indices using Kriging interpolation is a key step in integrating geological attribute data. Rock strata stability indices originate from field geological surveys, borehole core mechanical tests, and geophysical exploration results. These data are often irregularly distributed. As a geostatistical method, Kriging interpolation can fully utilize the spatial autocorrelation characteristics of sample points. It quantifies the spatial correlation structure by calculating the semi-variogram between sample points, and then performs unbiased optimal estimation of the attribute values ​​of the grid points to be estimated. This process not only generates a numerical surface of rock strata stability on a regular grid, but also provides estimation variance information, thereby realizing the spatial normalization expression of multi-source heterogeneous stability indices such as lithology, joint development degree, and compressive strength.

[0067] The implementation of grey relational analysis begins with constructing a data sequence for comparison for each grid cell. Each grid cell, regularly divided within the study area, is treated as an independent analysis unit. Constructing Sequence 1 requires extracting the spatial density values ​​of veins within the cell and all surrounding grids within a certain window. This window typically uses a rectangular or circular neighborhood centered on the current cell. The window size must consider the scale of the regional geological structure; for example, a 3x3 or 5x5 grid window might be used. Arranging the density values ​​of all grids within the window in spatial order constitutes Sequence 1, reflecting the distribution characteristics of veins around the cell. Sequence 2 is constructed using the same spatial window, extracting rock stability index values ​​generated by Kriging interpolation within the corresponding region, and arranging them in the same order to form Sequence 2. These two sequences are spatially completely corresponding, aiming to analyze the similarity of their variation patterns.

[0068] Data standardization is a necessary preprocessing step before calculating grey relational analysis. Since the spatial density of ore veins and the stability index of rock strata typically have different dimensions and numerical ranges, directly comparing their absolute values ​​is meaningless. Therefore, it is necessary to perform dimensionless processing on the data within each sequence. Initialization or mean normalization methods are commonly used. Initialization involves dividing each data point in the sequence by the first data point of that sequence, while mean normalization involves dividing each data point by the average value of the sequence. After processing, both sequences are transformed into purely numerical sequences, highlighting their variation patterns and facilitating comparisons of shape similarity. Calculating the grey relational coefficient is the core step in the analysis. The correlation coefficient reflects the degree of geometric similarity between two sequences at corresponding points. The absolute difference between sequence one and sequence two at each corresponding point is calculated, and the maximum and minimum values ​​among all differences are identified. The calculation of the correlation coefficient relies on a resolution coefficient, which is typically set to 0.5 to adjust for environmental differences in the calculation. The correlation coefficient at each point is determined by the minimum, maximum, absolute difference at that point, and resolution coefficient. The correlation coefficient ranges from 0 to 1; a larger value indicates a more consistent trend between the two sequences at that point.

[0069] The final grey relational coefficient is synthesized by averaging the correlation coefficients calculated at all points in the sequence. This average value represents the overall correlation between vein distribution and strata stability within the grid cell. This coefficient is a macroscopic similarity measure, comprehensively reflecting the overall closeness of the curve shapes of the two sequences throughout the entire analysis window. Geological interpretation of the calculated grey relational coefficients is the final step in completing the fusion. Grid cells with correlation coefficients close to 1 indicate that the spatial distribution density of veins and the stability of strata are highly synchronized in their variation patterns within that cell and its neighborhood. This synchronization may suggest the existence of some common geological controlling factors. For example, the area may be located in a special tectonic position, where areas with high strata stability are conducive to vein deposition and preservation. Conversely, cells with lower correlation coefficients indicate that the occurrence and development of the two phenomena are relatively independent, lacking a significant spatial coupling relationship. The resulting correlation coefficient field provides an important weighting basis for the subsequent construction of geological evolution trend prediction models, highlighting areas where vein distribution is more significantly controlled by rock mass stability.

[0070] The generation of trend prediction weight coefficients is a further refinement based on the results of grey relational analysis. The grey relational coefficients calculated for each grid cell directly constitute the initial weights of that cell in the prediction model, reflecting the importance of geological factors to mineralization prediction. However, the initial weights still need to be regionalized to eliminate the influence of possible spatial anomalies. The mean filtering method is used to smooth the initial weight coefficient map, making the spatial variation of the weight coefficients more reasonable and avoiding abrupt extreme points. The final generated trend prediction weight coefficient matrix will be used as a key input parameter and embedded into the subsequent geological evolution trend prediction model. This model usually uses weighted logistic regression or neural network algorithms, taking the spatial density field, rock layer stability field and their weight coefficients as independent variables, and the known mineral point distribution as the dependent variable for training and learning.

[0071] The construction of the entire geological evolution trend prediction model is a process of continuous fusion and optimization of multi-source information. It is not a simple superposition of geological data, but rather an in-depth exploration of the inherent spatial correlation between vein distribution and strata stability through spatial statistics, geoscientific calculations, and grey system theory. This correlation is quantified and expressed as weighted coefficients. The resulting prediction model has the ability to infer the spatial distribution of future mineralization potential from current geological deformation characteristics. The model output is a probability distribution map, where the value of each pixel represents the probability of mineralization based on historical evolution trends and current geological conditions, providing a scientific basis for subsequent dynamic monitoring and exploration deployment. The advantage of this method lies in its ability to handle the uncertainties and incompleteness commonly found in geological data, effectively coupling quantitative and qualitative geological information through grey relational analysis.

[0072] Example 3: The strategy for generating dynamically adjusted BeiDou sampling frequencies begins with the regional division of a probability distribution map. This probability distribution map is output by a geological evolution trend prediction model, where each pixel value represents the mineralization potential probability of the corresponding geographical location. Based on a preset probability threshold, the entire study area is divided into three potential levels: high, medium, and low. For example, grid cells with a probability value greater than 0.7 are classified as high-potential areas, cells with a probability value between 0.3 and 0.7 are classified as medium-potential areas, and cells with a probability value less than 0.3 belong to low-potential areas. This division is not a simple mechanical segmentation, but is supplemented by the closing operation processing in spatial morphology to eliminate overly fragmented classification patches within the area, ensuring the spatial contiguousness and boundary smoothness of each level of area.

[0073] The core of the strategy is to allocate differentiated sampling frequencies to areas with different potential levels. High-potential areas are assigned the maximum sampling frequency, the value of which is set according to the highest performance limit of the BeiDou receiving equipment, aiming to capture fast or weak deformation signals that may exist in the area. Medium-potential areas use a reference sampling frequency, which is usually determined with reference to the regular monitoring cycle of historical surveys. Low-potential areas are assigned the minimum sampling frequency to meet basic environmental monitoring and background value acquisition needs. The frequency assignment process considers not only the potential level but also the overall constraints of the equipment's total power consumption and communication bandwidth to avoid excessive concentration of resources in local areas. Setting a transition band frequency gradient based on the difference in potential levels between adjacent areas is key to achieving a smooth transition in the sampling strategy. At the boundary between two different level areas, a direct jump in sampling frequency may cause data discontinuity. Therefore, a frequency gradient buffer needs to be established. Within this buffer, the sampling frequency... Depending on spatial location The relationship between the distance to the high-potential zone boundary and the linear transition can be expressed by the following formula:

[0074]

[0075] in: The sampling frequency representing the high-potential area. The sampling frequency represents the sampling frequency on one side of the adjacent lower-level potential zone. This represents the distance of the calculation point from the boundary of the high-potential zone. This is the preset total width of the transition zone.

[0076] In the sub-process of real-time correction of geological deformation feature vectors, a sliding time window comparison mechanism needs to be established between the location data stream and historical deformation features. This mechanism sets a fixed-length time window, which slides forward as new data continuously flows in. Within each time window, the system calculates the current geological deformation feature vector and compares it with the feature vectors of historical windows corresponding to the same time period stored in the database. The comparison operation usually uses cosine similarity or Euclidean distance to measure the degree of difference between the two in the feature space. When the system detects that the deformation difference of three consecutive time windows exceeds a preset threshold, it will trigger the recalculation process of the feature vector. The threshold is usually determined based on the background fluctuation range obtained from long-term statistical analysis of historical data, such as taking the mean of historical difference values ​​plus a certain number of standard deviations. The triggering condition emphasizes "continuity" to avoid false triggering due to single accidental errors or noise interference, thereby ensuring the necessity and accuracy of the recalculation action. The recalculation process will backtrack the original location data of the current and previous windows and re-execute the entire set of calculations from solution to feature extraction using possibly updated parameters or models.

[0077] Smoothing the recalculated results using the Kalman filter algorithm is the final step in the correction process. Kalman filtering is an optimal estimation algorithm suitable for noisy dynamic systems. In this application scenario, the change of the geological deformation feature vector is regarded as a dynamic process. The system state equation describes the model of the vector's evolution over time, while the observation equation corresponds to the new vector value obtained in each recalculation. The filtering process continuously predicts and updates, weighting and fusing the noisy observation values ​​(i.e., the recalculated results) with the predicted values ​​of the system model. The weights are determined by their respective uncertainties, and finally, a smoothed, optimally estimated sequence of geological deformation feature vectors is output. This effectively suppresses the influence of high-frequency random noise in the data and highlights the true deformation trend.

[0078] Example 4: Extracting geological activity patterns with significant mineralization correlations requires multi-timescale division. This division is based on the inherent rhythms of geological processes and the accumulation cycle of monitoring data, dividing the continuous time stream into three scales: short-term, medium-term, and long-term. Short-term typically covers several hours to several days and aims to capture rapid deformation responses caused by seismic activity, groundwater fluctuations, or human mining. Medium-term extends to several weeks or even months and is suitable for observing continuous deformation caused by seasonal precipitation loads, slow release of tectonic stress, etc. Long-term spans several years and is used to analyze macroscopic geological evolution trends such as regional crustal uplift and subsidence. This division enables the analysis to distinguish between instantaneous events, medium-term trends, and long-term background fields.

[0079] The first step in the computational work is data preparation and time scale alignment. Deformation feature data comes from continuous observations of the BeiDou positioning system. After processing, the elevation change, east-west displacement, and north-south displacement of each monitoring point at different time points are obtained. Further, indicators such as elevation change rate and combined displacement rate can be calculated. The vein distribution data is based on historical geological mapping, borehole core records, and geophysical and geochemical anomaly information. The spatial density value of the vein in each grid cell is generated through spatial interpolation. This is a relatively static background field. Before the calculation, it is necessary to correlate the dynamic deformation feature time series with the static vein density data. For each grid cell, the deformation feature observation values ​​in the short, medium, and long time scales are extracted to form the deformation feature time series of the cell at different scales. Since the vein density value is a constant in the cell, it is necessary to construct a sequence with the same length as the deformation feature sequence for each scale, but all element values ​​are equal to the vein density of the cell.

[0080] The Pearson correlation coefficient is calculated independently for each grid cell. For a given grid cell and a selected time scale, the system obtains two sequences of equal length: Sequence A is a sequence of deformation characteristic values ​​of the cell arranged in chronological order at this scale, and Sequence B is a sequence formed by repeatedly filling the vein density values ​​of the cell. To calculate the Pearson correlation coefficient between these two sequences, the mean of each sequence is calculated first, then the deviation of each element of Sequence A from the mean is calculated, and the deviation of each element of Sequence B from the mean is calculated. The sum of the products of deviations at all corresponding points is calculated, and the sum of the squares of the deviations of each sequence is calculated separately. The Pearson correlation coefficient is equal to the sum of the products of deviations divided by the square root of the product of the two sums of squares of deviations. The numerical range of the calculated result is between -1 and 1.

[0081] The calculated correlation coefficients are assigned to corresponding grid cells. Once all grid cells in the study area have completed the correlation coefficient calculation at the current time scale, these coefficient values ​​are arranged according to the spatial coordinates of the grid to generate a spatialized correlation coefficient matrix. This matrix clearly reveals the linear correlation pattern between surface deformation and vein distribution at this time scale. Positive correlation areas indicate that enhanced deformation and high vein density are synchronous, while negative correlation areas indicate that the two are inversely related.

[0082] The tensor decomposition algorithm, which fuses multi-scale correlation coefficients, aims to extract consistency features beyond a single period. The correlation coefficient matrix calculated for short, medium, and long periods is treated as a three-dimensional tensor, with its three dimensions corresponding to the row index, column index, and time period index of the spatial grid, respectively. Tensor decomposition employs the classic Canonical Polyacrylic Decomposition (CPD) model, approximating the original high-order tensor as a product of several low-rank components. This decomposition process can uncover shared patterns behind data at different scales. For example, a factor might spatially correspond to a linear tectonic zone, showing high correlation across all time scales in the period dimension. Such cross-period consistency patterns are considered to reveal continuously active tectonic weak zones or material migration channels controlled by deep mineralization processes, and their mineralization indication is far more significant than single-scale anomalies. Specifically, in the tensor decomposition process, the system uses the classic CPD model to approximate the original high-order tensor as a product of several low-rank components to uncover shared patterns behind data at different scales. The specific implementation is as follows: The system organizes the correlation coefficient matrices calculated for short, medium, and long time periods into a three-dimensional tensor, where the three dimensions correspond to the row index, column index, and time period index of the spatial grid, respectively. For the CPD model, the system decomposes this three-dimensional tensor into the product of three factor matrices, representing the spatial dimension, temporal dimension, and feature dimension, respectively. The decomposition process uses alternating least squares for iterative optimization, solving for the loading values ​​of the factor matrices by minimizing the error between the reconstructed tensor and the original tensor. These loading values ​​reveal the hidden linear structure in the tensor. For example, in the spatial factor matrix, grid cells with high loading values ​​may correspond to a linear tectonic zone, while in the temporal factor matrix, this tectonic zone exhibits high correlation across all time scales, thus identifying cross-period consistent feature patterns. This pattern indicates continuously active tectonic weak zones or material migration channels controlled by deep mineralization processes, enhancing the ability to identify mineralization-related geological activities. The entire decomposition process emphasizes computational efficiency and stability, ensuring that the extracted patterns have geological interpretation significance and provide a reliable basis for subsequent exploration path planning. Referring to Table 1, the multi-scale correlation coefficient data structure used for tensor decomposition is shown. The study area is divided into a 3x3 grid, and a correlation coefficient value between deformation characteristics and vein density is calculated for each grid at different time scales.

[0083] Table 1: Correlation coefficients between multi-scale geological deformation and vein distribution

[0084]

[0085] Geological interpretation of the factor matrix obtained after tensor decomposition is crucial for ultimately extracting geological activity patterns. The spatial factor matrix reveals grid clusters with similar spatial correlation patterns. These clusters often coincide closely with known fault zones, rock mass contact zones, or specific lithological stratigraphic boundaries. The time-period factor matrix reveals the contribution weights of different time scales to the overall correlation pattern, which helps determine the dominant period of mineralization correlation. By delineating the areas with the highest loading values ​​in the spatial factor map and combining them with their corresponding time-period weights, spatial units that exhibit significant and stable positive correlations across short, medium, and long periods can be identified. These units are interpreted as geological activity patterns with significant mineralization correlations, i.e., "mineralization hotspots." They indicate currently active tectonic-deformation systems that control the enrichment of ore-forming materials, providing precise target areas for subsequent exploration path planning.

[0086] Example 5: The starting point for generating exploration path planning instructions is to map the identified geological activity patterns into a hotspot intensity field in three-dimensional space. Spatially, this pattern is represented by each grid cell having an intensity value that indicates the significance of its mineralization association. In order to construct a continuous three-dimensional field, these discrete spatial grid point data need to be interpolated. The radial basis function interpolation method is used, with the center coordinates of each grid cell and its intensity value as control points. The hotspot intensity of any point in the entire study area (including the elevation dimension) is calculated to generate a smooth and continuous three-dimensional scalar field. The intensity isosurface of this field can intuitively show the spatial morphology and distribution range of mineralization association activities.

[0087] The core computational component employs an ant colony optimization algorithm to search for the optimal exploration path within a 3D hotspot intensity field. This algorithm simulates the collective behavior of real ant colonies seeking the shortest path through pheromone communication. In this application scenario, the 3D space is discretized into a grid composed of cubic units, each considered a traversable node. During algorithm initialization, a certain number of virtual "ants" are randomly deployed at a preset starting point in the 3D space (e.g., the current location of the exploration equipment or its base). Each ant represents a potential exploration path. The movement of ants within the 3D grid follows a probability transition rule. The probability of an ant moving from its current node to an adjacent node (26 adjacent directions) is determined by two factors: first, the hotspot intensity of the adjacent node itself (higher intensity means greater attraction); and second, the pheromone concentration accumulated along the path connecting the current node and adjacent nodes (higher concentration means a higher probability of selection). This probabilistic selection mechanism ensures that the ant colony is both motivated to move towards high-intensity areas and guided by the collective experience (pheromone) of the explored paths.

[0088] The implementation of the pheromone concentration update rule begins with the local update process, which is closely coupled with the individual movement behavior of each ant. When an ant moves from one node in the 3D mesh to another adjacent node, it immediately releases a certain amount of pheromone on the path it just passed. The amount of pheromone released is not a fixed value, but is determined by the normalized hotspot intensity value of the current node and the intensity value of the previous node (i.e., the predecessor node). The specific calculation usually takes the average or maximum value of the two node intensities and multiplies it by a preset base pheromone release coefficient. This means that when the ant passes through an area with higher hotspot intensity, it will leave a denser pheromone trace, thereby marking a valuable path segment in the local area. The pheromone evaporation design is intended to simulate the natural evaporation of pheromones over time in nature, preventing the infinite accumulation of pheromones on a certain path from causing the algorithm to converge prematurely to a local optimum. In each global update phase, the system traverses all edges (i.e., connections between nodes) in the 3D grid and multiplies the current pheromone concentration on each edge by a evaporation coefficient between 0 and 1, such as 0.7. This means that after each iteration, about 70% of the pheromones on all paths will be retained, while the remaining 30% will be evaporated. This mechanism ensures that the influence of paths explored in the early stages will gradually weaken, allowing the ant colony to continuously try new possibilities and preventing the entire search process from becoming too rigid prematurely.

[0089] The pheromone enhancement operation focuses on rewarding high-quality paths discovered in the current iteration. After all ants have completed the path construction from the starting point to the end point, the system evaluates the quality of each path. The evaluation criteria comprehensively consider the sum of the total hotspot intensities of the nodes traversed by the path and the total length of the path. The fitness value of each path is calculated through a weighted scoring function, and the optimal path for this iteration is selected. The system adds extra pheromone to each connection of this optimal path. The amount added is proportional to the fitness value of the path. The better the path (higher total intensity and relatively shorter length), the more pheromone enhancement it receives. This is equivalent to broadcasting a "recommended route" in the group, guiding ants in subsequent iterations to be more inclined to explore this path and its surrounding areas. The coordinated operation of local and global updates forms a regulatory mechanism at different time scales. Local updates occur in the microscopic steps of ant movement, marking the ants' immediate discoveries in real time and dynamically, similar to the scent trails left by ants as they crawl. This type of update is fast and fine-grained, quickly reflecting the value of local areas. Global updates occur at the macroscopic level of each iteration, summarizing and adjusting based on the exploration results of all ants in this round. By volatile and targeted enhancement of global pheromones, it macroscopically guides the search direction. These two updates occur simultaneously, jointly shaping the search behavior of the entire ant colony in three-dimensional space, ultimately causing the ant colony to gradually converge from random exploration to a high-value exploration path that traverses the main hotspot intensity areas.

[0090] The search process terminates when the overlap of consecutive iterations reaches a preset threshold of 95%. The determination of path overlap requires calculating the degree of overlap between the optimal path generated in the current iteration and the optimal paths from previous iterations in three-dimensional space. Specifically, this is done by comparing the node sets traversed by the two paths and calculating the ratio of the number of intersection nodes to the number of union nodes. When this ratio remains at a high level of over 95% for multiple consecutive iterations, it indicates that the path discovered by the ant colony has become stable and the algorithm has converged. At this point, the search is terminated and the final optimal path node sequence is output.

[0091] Generating a sequence of motion control parameters based on the intensity values ​​of hotspots traversed along the path is the final step in transforming the digital path into actual exploration actions. The output path node sequence contains the three-dimensional coordinates of each node and its corresponding Gaussian-filtered normalized hotspot intensity value. The motion control parameters mainly include travel speed, turning angle, dwell time, and sampling instructions from specific sensors on the equipment (such as spectrometers and magnetometers). The travel speed is designed to be inversely proportional to the hotspot intensity at the current location. In the core area of ​​extremely high-intensity "hotspots," the equipment will significantly reduce its speed or even pause its movement to conduct more precise positioning measurements for a longer period of time. The turning angle is calculated from the spatial vector difference of continuous path nodes to ensure a smooth transition for the equipment. The dwell time is directly proportional to the local hotspot intensity value, with a longer data acquisition time set in high-intensity areas. The final generated control parameter sequence is encapsulated into a standard instruction set and sent to the flight control system or drive system of monitoring equipment such as UAVs and ground robots to drive them to autonomously execute this directional exploration operation covering the main mineralization-related activity areas.

[0092] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0093] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring mineralized exploration based on BeiDou satellite positioning, characterized in that, Includes the following steps: Collect BeiDou satellite positioning data of the target area to obtain positioning coordinate sequences and corresponding positioning accuracy characteristics; The optimal positioning solution model is selected based on the positioning accuracy characteristics, and the positioning coordinate sequence is subjected to spatiotemporal decoupling processing to generate multi-dimensional positioning data containing elevation and planar components. Extract the abnormal fluctuation components from the multi-dimensional positioning data, and calculate the geological deformation feature vector by combining them with preset geological structural constraints. Obtain the distribution characteristics of veins and the stability index of rock strata from historical mineralization exploration data, and construct a geological evolution trend prediction model; The geological deformation feature vector is input into the geological evolution trend prediction model, and a probability distribution map of the mineralization potential area is output. Based on the probability distribution map, a monitoring priority grid is divided, a dynamically adjusted BeiDou sampling frequency strategy is generated, the positioning data acquisition process is updated according to the BeiDou sampling frequency strategy, and the geological deformation feature vector is corrected in real time. Multi-scale fusion analysis is performed on the corrected geological deformation feature vector to extract geological activity patterns with significant mineralization correlation. Based on the geological activity patterns, exploration path planning instructions are generated to drive monitoring equipment to perform directional exploration operations. The optimal localization solution model includes: Calculate the sum of squared residuals and the confidence interval width for different solution models for the positioning coordinate sequence; use a dynamic weighted algorithm to fuse the sum of squared residuals and the confidence interval width to generate a model fit score; select the solution model with the highest score as the optimal positioning solution model and record its corresponding coordinate transformation parameters; The calculation of the geological deformation feature vector includes: The abnormal fluctuation components are convolved with geological structural constraints to generate an initial deformation matrix; principal component analysis is performed on the initial deformation matrix to extract the first three principal components as basic deformation features; the basic deformation features are then differentially analyzed with historical data from the same location to obtain a standardized deformation feature vector.

2. The mineral exploration monitoring method based on BeiDou satellite positioning according to claim 1, characterized in that, The geological evolution trend prediction model includes: Convert the vein distribution characteristics in historical metallogenic exploration data into a spatial density field; The stability index of the rock strata was spatially discretized using the Kriging interpolation method. By integrating the spatial density field and discretized rock strata indicators through grey relational analysis, trend prediction weight coefficients are generated.

3. The mineral exploration monitoring method based on BeiDou satellite positioning according to claim 1, characterized in that, The strategy for generating dynamically adjusted BeiDou sampling frequencies includes: The probability distribution map is divided into three potential level regions: high, medium, and low. The maximum sampling frequency is assigned to high-potential areas, the baseline sampling frequency is assigned to medium-potential areas, and the minimum sampling frequency is assigned to low-potential areas. The frequency gradient of the transition zone is set according to the difference in potential levels between adjacent areas.

4. The mineral exploration monitoring method based on BeiDou satellite positioning according to claim 1, characterized in that, The real-time correction of the geological deformation feature vector includes: Establish a sliding time window comparison mechanism between location data streams and historical deformation characteristics; When the deformation difference over three consecutive time windows exceeds a preset threshold, the feature vector recalculation process is triggered. The Kalman filter algorithm is used to smooth the recalculated results.

5. The mineral exploration monitoring method based on BeiDou satellite positioning according to claim 1, characterized in that, The geological activity patterns with significant mineralization correlations extracted include: The time dimension is divided into three scales: short cycle, medium cycle, and long cycle. Calculate the correlation coefficient matrix between deformation characteristics and vein distribution at each scale; By fusing multi-scale correlation coefficients using the tensor decomposition algorithm, cross-period consistency feature patterns are extracted.

6. The mineral exploration monitoring method based on BeiDou satellite positioning according to claim 1, characterized in that, The instructions for generating exploration route planning include: Geological activity patterns are mapped as hotspot intensity fields in three-dimensional space; The ant colony optimization algorithm is used to search for the optimal exploration path in the hotspot intensity field; Generate a sequence of motion control parameters for the equipment based on the intensity values ​​of the hotspots along the path.

7. The mineral exploration monitoring method based on BeiDou satellite positioning according to claim 6, characterized in that, The optimal exploration path includes: Initialize the random distribution state of the ant colony in the hotspot intensity field; Ants are guided to gather in high-intensity areas by using pheromone concentration update rules; The search terminates when the overlap of consecutive iterations reaches 95%, and the final set of path nodes is output.

8. A mineral exploration monitoring system based on BeiDou satellite positioning, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the mineral exploration and monitoring method based on BeiDou satellite positioning as described in any one of claims 1 to 7.

Images