A method and system for metallogenic prediction based on geological big data

Abstract

The present application relates to the technical field of metallogenic prediction method, and particularly relates to a metallogenic prediction method and system based on geological big data, which comprises the following steps: obtaining multi-source geological data and performing grid processing on the multi-source geological data to obtain multiple types of data channels, and calculating the spatial variation rate vector of each position in the data channel; for each position, dividing the local area into two sub-areas along multiple test directions, calculating the vector sum of all spatial variation rate vector fields in each sub-area, calculating the direction significance of the test direction by using the vector sum of the two sub-areas, and taking the test direction corresponding to the maximum direction significance as the tectonic direction of the position; performing direction enhancement on the data channel to obtain a direction-enhanced data channel, and calculating the metallogenic potential index of each position based on the direction-enhanced data channel, so as to evaluate the metallogenic possibility of the corresponding position. The present application improves the accuracy and reliability of finding metallogenic areas.

Description

Technical Field

[0001] This invention relates to the field of mineralization prediction methods, and in particular to a mineralization prediction method and system based on geological big data. Background Technology

[0002] In modern mineral resource exploration, mineralization prediction methods based on geological big data play a crucial role. These methods comprehensively analyze data from various sources, including geochemistry, geophysics, remote sensing imagery, and regional geological structures. They utilize statistical or machine learning models to uncover hidden mineralization patterns within the data. Then, by constructing predictive models, they quantitatively assess the mineralization potential of unknown areas, ultimately generating mineralization data channels. This technique combines the experience and knowledge of geological experts with the objective laws of massive amounts of data, effectively narrowing the target area for mineral exploration, significantly reducing exploration costs and risks, and thus improving the overall efficiency and success rate of mineral exploration.

[0003] However, existing mineralization prediction models fail to fully consider the directionality and multi-scale nature of geological anomaly fields. Mineralization processes are often controlled by geological structures with specific orientations, such as fault zones and fold axes, resulting in geochemical or geophysical anomalies exhibiting a banded, linear distribution along the structure—a phenomenon known as directionality. However, traditional models often use symmetrical square analysis windows for feature extraction. These operators treat all directions equally, failing to effectively identify and enhance continuous anomalies in specific directions, and easily misclassifying genuine structural anomalies as background noise. Simultaneously, the structural systems controlling mineralization are multi-scale, ranging from large-scale deep faults to localized small fractures; structures at different scales collectively constitute the favorable space for ore body formation. Existing methods typically analyze data only at a single spatial scale or use simple data channel scaling for coarse fusion, lacking a mechanism to systematically integrate directional features at different scales, ignoring prior knowledge of geological structures, and reducing the accuracy of mineralization prediction model results. Summary of the Invention

[0004] To address the issue of low accuracy in existing mineralization prediction models, this invention provides a mineralization prediction method and system based on geological big data.

[0005] In a first aspect, the present invention provides a mineralization prediction method based on geological big data, employing the following technical solution:

[0006] Multi-source geological data is acquired and rasterized to obtain multiple data channels. The spatial rate of change vector at each location in the data channel is calculated, and all spatial rate of change vectors at the same location are superimposed to obtain a spatial rate of change vector field. For each location, the local area is divided into two sub-regions along multiple test directions. The vector sum of all spatial rate of change vector fields in each sub-region is calculated. The directional significance of the test direction is calculated using the vector sum of the two sub-regions. The directional significance is negatively correlated with the dot product of the two vector sums. The test direction corresponding to the maximum directional significance is taken as the structural direction of the location. The data channels are directionally enhanced to obtain directionally enhanced data channels. Based on the directionally enhanced data channels, the mineralization potential index of each location is calculated to assess the mineralization probability of the corresponding location.

[0007] By dividing a local area into regions along multiple test directions, calculating and comparing the vector sum of the spatial rate of change vector fields, directional saliency is defined, and the most probable structural direction is determined. Based on this structural direction, data channels are enhanced, effectively amplifying linear anomalous signals along specific geological structural directions while suppressing randomly distributed background noise. Quantifying and utilizing the directionality of the geological anomaly field as a key parameter enables the model to simulate the control of geological structures on mineralization, thereby significantly improving the sensitivity and accuracy of identifying real structural anomalies from massive geological data and enhancing the reliability of mineralization potential prediction.

[0008] Preferably, the spatial rate of change vector is calculated using the Sobel operator.

[0009] The Sobel operator not only considers the influence of neighboring pixels, but also assigns greater weight to grids closer to the center point, which can better smooth noise. This provides more stable and reliable basic data for subsequent calculation of directional saliency, ensuring the effectiveness of spatial change rate calculation.

[0010] The preferred expression for directional significance is:

[0011] ;

[0012] in, Indicates the location Upward directional salience, Angle indicating the test direction , They represent local areas as follows: Centered on, direction is All adjacent locations on one side and the other side of the straight line, This represents the sum of the spatial rate of change vectors of all neighboring locations on one side within a local region. This represents the sum of the spatial rate of change vectors of all neighboring locations on the other side of the local region. This represents the magnitude of the vector.

[0013] By utilizing the properties of the vector dot product, the dot product of the vector sums of two sub-regions is negative, ensuring a positive and large significance when the two vector sums are in opposite directions. Simultaneously, multiplying the formula by the magnitude of the two vector sums ensures a large value is only obtained when significant changes exist on both sides, avoiding misjudgments caused by weak changes on one side. This allows for precise quantification and identification of potential geological structural directions.

[0014] The preferred expression for the directional enhancement data channel is:

[0015] ;

[0016] in, Indicates the location Enhance the channel in the direction of data channel k. Indicates The local area centered on Indicates the location Data channel k at location, , These represent the preset distance weights and direction weights, respectively. , Indicates the position coordinates.

[0017] By combining the original data, distance weights, and direction weights, selective amplification of information in specific directions is achieved, enhancing the continuity and intensity of linear anomaly signals along the structural direction. This makes the characteristics of mineral exploration targets more prominent and provides a data foundation with a higher signal-to-noise ratio for the subsequent calculation of the mineralization potential index.

[0018] Preferably, the expressions for distance weight and direction weight are:

[0019] ;

[0020] ;

[0021] in, Indicates distance weight, This represents the standard deviation of the squared Euclidean distances from all locations within a local region to the center location. Indicates directional weights. This represents the distance from all locations within a local area to the center location. standard deviation Indicates from point to The direction angle, Indicates position The structural direction, , The coordinates represent the position coordinates, and exp represents an exponential function with base e.

[0022] The distance weights employ a Gaussian function to ensure locality, meaning that closer points have a greater influence, consistent with the principle of spatial autocorrelation. The orientation weights also use a Gaussian function, with the variable being the angle between the orientation of neighboring points and the construction orientation. This results in a sharp peak in the weight distribution along the construction orientation, ensuring that only information aligned with the construction orientation is effectively amplified, significantly improving the accuracy and effectiveness of orientation enhancement.

[0023] Preferably, before calculating the mineralization potential index, the method further includes: performing multi-level downsampling on data channels of the same type to form data at different scales; performing directional enhancement on data at different scales to obtain directional enhancement channels at different scales; amplifying the directional enhancement channels calculated at low scales back to the original scale through interpolation; and merging all directional enhancement data channels at all scales according to data channel type to obtain a comprehensive dataset, which includes multiple types of features.

[0024] By performing multi-level downsampling of the data, repeating orientation identification and enhancement at different scales, and finally fusing the results from each scale, this method can simultaneously capture structural information at large, medium, and small scales, comprehensively depicting favorable mineralization spaces. This overcomes the limitations of single-scale analysis, enabling the prediction model to comprehensively consider the influence of multi-level geological structures, thereby obtaining more comprehensive and accurate prediction results than existing technologies.

[0025] The preferred expression for the mineralization potential index is:

[0026] ;

[0027] in, Indicates position The mineralization potential index This indicates the first element in the merged aggregate dataset. Class features at location The value, Indicates the first Weights of class features.

[0028] The use of a product model, where each feature value is incremented by 1 before weighted exponent calculation, reflects the principle that mineralization is a multi-factor coupling process. This means that multiple favorable conditions must coexist to form a high-potential area, and the absence of any key feature will significantly lower the final potential index. Using weights as the index ensures that the influence of features more closely related to mineralization on the final result grows exponentially. This better reflects the dominant role of key ore-controlling factors than simple linear weighting, making the final potential evaluation more consistent with geological mineralization laws.

[0029] Preferably, the ratio of the average value of the features in the mining area to the average value of the features in the non-mining area is used as the weight of the feature.

[0030] Preferably, the local region used to calculate the construction direction is an M×N window region.

[0031] Secondly, the present invention provides a mineralization prediction system based on geological big data, which adopts the following technical solution:

[0032] A mineralization prediction system based on geological big data includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a mineralization prediction method based on geological big data as described above is implemented.

[0033] The aforementioned mineralization prediction method based on geological big data is generated into a computer program and stored in a memory so that it can be loaded and executed by a processor. Thus, a system can be created based on the memory and processor for convenient use.

[0034] The present invention has the following technical effects:

[0035] This invention automatically identifies and quantifies linear structural directions related to mineralization from multi-source geological data, and uses these directions to selectively enhance data at different scales, thereby effectively amplifying the mineral exploration information in the data background. This overcomes the fundamental defect of traditional methods that suffer from low prediction accuracy due to neglecting directionality, and improves the accuracy and reliability of finding mineralized areas. Attached Figure Description

[0036] Figure 1 This is a flowchart of a mineralization prediction method based on geological big data according to the present invention. Detailed Implementation

[0037] 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, not all, of the embodiments of the present invention. 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.

[0038] This invention discloses a mineralization prediction method based on geological big data, referring to... Figure 1 This includes the following steps:

[0039] S1: Acquire multi-source geological data and calculate the spatial rate of change field.

[0040] When predicting the mineralization potential of a study area, multi-source geological data for that area is first collected from a geological database. For example, multi-source geological data includes geochemical data, aeromagnetic anomaly data, gravity anomaly data, and remote sensing alteration information. The area is then divided into multiple grid regions. To facilitate unified processing, the collected multi-source geological data undergoes coordinate registration and spatial rasterization to ensure that all data are aligned to the same grid, forming a set of data channels. The data at each location within each data channel is then normalized, and its value range is mapped to... The intervals are ultimately combined to form a three-dimensional data set, denoted as... ,in, Indicates the grid location in geospatial space. This indicates the data channel type, such as gold content, magnetic anomaly value, etc.

[0041] Subsequently, for each data channel ,application The operator calculates each position The spatial rate of change vector. Then, the different data channels are placed in the same location. The spatial rate of change vectors are superimposed to generate a total spatial rate of change vector field. This vector field comprehensively reflects the overall trend and intensity of all geological variables at their respective locations.

[0042] S2: Calculate the directional significance of each position in the spatial rate of change vector field to obtain the constructed direction.

[0043] Linear structures, such as fault zones or fold axes, are often a decisive factor influencing mineral deposit formation. The presence of linear structures increases the likelihood of mineralization. Linear structures often exhibit distinct linear or zonal characteristics. For example, fault zones in anomalous data channels often appear as ridges, with opposite spatial rates of change on either side, both pointing away from the ridge. It should be noted that, here, a ridge refers to a continuous region in a two-dimensional data field that exhibits linear or curved patterns and locally high values. Similarly, a continuous region in a two-dimensional data field that exhibits linear or curved patterns and locally low values ​​is called a valley.

[0044] Furthermore, in step S1, although the types of data collected are different, the physical prior of the linear structure of the fault zone or fold axis exhibiting obvious linear or banded characteristics is satisfied on different data channels. This is because, for aeromagnetic anomaly data channels, when a fault cuts through rock masses with different magnetic properties, or when the fault zone itself is intruded by later dikes, a very clear linear or banded anomaly will form on the aeromagnetic data channel. The profile of this anomaly is a typical ridge or valley, and the spatial rate of change of magnetic field intensity on both sides must be in opposite directions.

[0045] Similarly, for gravity anomaly data channels, if the rock densities on both sides of a fault are different, or if high-density / low-density materials intrude / alter along the fault zone, linear high or low gravity anomalies will form on the gravity data channel. These anomalies will also resemble ridges or valleys, and the spatial rate of change of gravity on both sides will necessarily be opposite. For geochemical data channels, when a fault serves as a channel for the migration of ore-forming hydrothermal fluids, specific ore-forming elements or associated indicator elements will accumulate along the fault zone. On the element content contour data channel generated by spatial interpolation of the collected discrete sample point data, this enrichment zone will appear as a distinct ridge-shaped high-value area. From the center of the ridge towards both sides, the element concentration gradually decreases; therefore, the spatial rate of change of concentration must point towards the low-value areas on both sides, perfectly conforming to this physical prior.

[0046] For remote sensing data channels, during the migration of ore-forming hydrothermal fluids, they typically react with the surrounding rocks to form specific alteration mineral assemblages, such as pyritization, sericitization, and silicification. These alteration zones can be identified and delineated on remote sensing images through spectral analysis. If alteration is strictly controlled by fault structures, then in the final generated alteration intensity data channel, the area along the fault zone will appear as a high-intensity anomaly. The morphology and spatial variation rate characteristics of this anomaly zone also conform to the ridge model.

[0047] Specifically, for the spatial rate of change vector field Every position on Calculate the saliency of the direction Used to assess direction Whether it is suitable as a construction direction for this location is expressed by the following expression:

[0048]

[0049] in, Indicates the location Upward directional salience, The angle representing the test direction has a range of values. , Indicates a local area Centered on, direction is All adjacent positions on one side of the straight line, This represents all adjacent locations on the other side of the line within a local area. This represents the sum of the spatial rate of change vectors of all neighboring locations on one side within a local region, reflecting the direction of the total spatial rate of change to that side. This represents the sum of the spatial rate of change vectors of all neighboring locations on the other side of the local region, reflecting the direction of the total spatial rate of change on the other side. This represents the vector magnitude. The size of the local region can be set according to the research scale; for example, it can be expressed in terms of location. Construct a 7x7 window centered on the location; the window area represents the position. A local area.

[0050] For each position Define a local region around it, for each possible direction The local region is divided into two parts. The sum of the spatial rate of change vectors of all locations in each part is calculated separately, yielding... and Calculate the dot product and take its negative value. When the two vectors are in opposite directions, the dot product is negative; when the two vectors are in the same direction, the dot product is positive. After taking the negative sign, when the two vectors are in opposite directions, the term is positive and the value is large, indicating that there is a significant geological anomaly in that direction. It is the product of the intensity of the total spatial change rate on both sides. When there are strong spatial change rates on both sides, the product is larger, ensuring that the significance of direction not only considers the opposite direction, but also the intensity of the spatial change rate on both sides, thus avoiding misjudgment due to the weak spatial change rate on one side.

[0051] Multiplying the negative of the dot product by the product of the two vector lengths yields the directional salience for a fixed position. Using different The output shows different directional saliences. A higher directional salience indicates a greater likelihood of a linear structure at that location, and thus a higher probability of mineralization. Further, among all directions, the direction with the highest directional salience is selected as the structural direction for that location, denoted as […]. The construction direction reflects the extension direction of the linear structure.

[0052] S3: Perform directional enhancement on the data channel to obtain directional enhancement channel data.

[0053] The formation of ore bodies is often influenced by geological structures in specific directions, and information from these directions is more important than information from other directions. By enhancing the information in these directions, the location of potential ore deposits can be identified more clearly, interference from irrelevant information can be reduced, and thus the accuracy of mineralization prediction can be improved.

[0054] For each data channel, a directional enhancement data channel is generated; specifically, for each type of data channel in the original data... Each position The expression for the directional enhancement data channel is:

[0055] ;

[0056] ;

[0057] ;

[0058] in, Indicates the location Enhance the channel in the direction of data channel k. Indicates A local area centered on the window, such as a 21×21 window area. Indicates the location Data channel k at location, This represents the distance weight; values ​​closer to the center point have higher weights. This represents the standard deviation of the squared Euclidean distances from all locations within a local region to the center location. Indicates directional weights. This represents the distance from all locations within a local area to the center location. standard deviation Indicates from point to The direction angle. Indicates position The structural direction, , Represents position coordinates, and exp represents an exponential function with base e. If the value is 0, then let The value is 1, if If the value is 0, then let The value is 1.

[0059] This ensures locality, meaning that positions closer to the center point have higher weights, which reduces the influence of irrelevant information from distant locations. Direction awareness is implemented, meaning that only points whose directions align with the construction direction receive higher weights. This allows the direction-enhanced values ​​to incorporate information along the construction direction, thereby strengthening linear anomaly signals. Direction enhancement is applied to the data channels to create direction-enhanced channels, reducing noise interference.

[0060] S4: Calculate the mineralization potential index for each location.

[0061] The formation of ore bodies is influenced not only by local small-scale structures but also by larger-scale geological structures. By extracting and fusing directional features at multiple spatial scales, a comprehensive understanding of the mineralization environment can be achieved, leading to more accurate mineralization predictions.

[0062] First, the original three-dimensional data structure Multi-level downsampling is performed to generate data at different scales. For example, downsampling can be performed sequentially by 2x, 4x, and 8x to obtain three sets of data at the original scale, mesoscale, and macroscale. Next, for each scale of data, the calculation processes S2 and S3 are repeated to obtain directional enhancement channels at different scales.

[0063] Then, all the orientation enhancement channels calculated at low scale (coarse scale) are amplified back to the original scale using interpolation. The orientation enhancement data channels at all scales are then merged according to their data channel type to form a comprehensive dataset F containing rich multi-scale directional features. Each feature class in the comprehensive dataset F is denoted as... .

[0064] Using the linear normalization algorithm on the comprehensive dataset Each feature in the dataset is normalized, and the location of the mining site is calculated using the known locations of mining sites in historical data. The average value of each characteristic in the mining area and the average value in non-mineralized areas The ratio of the two is used as the weight of the feature. .

[0065] The mineralization potential index for each location is calculated using the following expression:

[0066] ;

[0067] in, Indicates position The mineralization potential index Represents all merged feature categories Perform a series of multiplications. This indicates the first element in the merged aggregate dataset. Class features at location The value, Indicates the first Weights of class features.

[0068] For each feature, calculate its weight based on known mineral deposits. For each position Each feature value is incremented by 1, and then multiplied by its weight as an exponent. The results of multiplying all the features together are used to obtain the mineralization potential index for that location. This incorporates the importance of all features; adding 1 prevents a single feature value of 0 from causing the entire product to be zero. Weights are used. As an index, the stronger the positive correlation between a feature and mineralization, the greater its influence on the final potential index. This reflects the principle of multi-factor coupling in mineralization, meaning that multiple favorable conditions must coexist for a high mineralization potential to form. For each location within the study area, a higher mineralization potential index indicates a greater likelihood of ore presence, and vice versa.

[0069] S5: Assess mineralization potential.

[0070] Based on the mineralization potential index of each location in the study area, the areas with the most promising mineral exploration prospects can be identified. For example, by selecting the top 5% of the areas with the highest mineralization potential index, the areas with the most promising mineral exploration prospects can be directly delineated within the study area, thus providing guidance for subsequent high-cost exploration activities such as field verification and drilling projects.

[0071] This invention also discloses a mineralization prediction system based on geological big data, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a mineralization prediction method based on geological big data according to this invention.

[0072] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0073] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for metallogenic prediction based on geological big data, characterized in that, The method comprises the steps of: obtaining multi-source geological data and performing rasterization on the multi-source geological data to obtain multiple types of data channels, calculating a spatial variation rate vector of each position in the data channel, superimposing all the spatial variation rate vectors of the same position to obtain a spatial variation rate vector field; for each position, in its local region, the local region is divided into two sub-regions along multiple test directions, the vector sum of all the spatial variation rate vector fields in each sub-region is calculated, the directional significance of the test direction is calculated by using the vector sum of the two sub-regions, the directional significance is negatively correlated with the dot product of the two vector sums, and the test direction corresponding to the maximum directional significance is taken as the structural direction of the position; performing directional enhancement on the data channels to obtain a directional enhancement data channel, and calculating a metallogenic potential index of each position based on the directional enhancement data channel, so as to evaluate the metallogenic possibility of the corresponding position; The expression of the directional enhancement data channel is: ； represents a direction enhancement channel of the data channel k at position represents a direction enhancement channel of the data channel k at position represents a local region centered at represents a local region centered at represents the data channel k at position represents the data channel k at position , respectively represent a preset distance weight, a direction weight, , represents a position coordinate; The expression of the metallogenic potential index is: ； representing the position of the metallogenic potential index, representing the value of the class feature in the position of the merged comprehensive data set, representing the weight of the class feature. 2.The metallogenic prognosis method based on geological big data according to claim 1, characterized in that, The spatial variation rate vector is calculated by using a Sobel operator. 3.The metallogenic prognosis method based on geological big data according to claim 1, characterized in that, The expression of the directional significance is: ； wherein, denotes the direction saliency at position denotes the direction saliency at position denotes the direction saliency at position denotes the angle of the test direction, , denote all neighboring positions on one side and the other side of the straight line centered at with direction in the local region, respectively, denotes the sum of the spatial change rate vectors of all neighboring positions on one side in the local region, denotes the sum of the spatial change rate vectors of all neighboring positions on the other side in the local region, denotes the vector length.

4. The metallogenic prognosis method based on geological big data according to claim 1, characterized in that, The expressions of the distance weight and the direction weight are: ； ； wherein, denotes the distance weight, denotes the standard deviation of the squared Euclidean distance of all positions in the local region to the center position, denotes the direction weight, denotes the standard deviation of the Euclidean distance of all positions in the local region to the center position, , denotes the direction angle from to , denotes the construction direction of the position , , denotes the position coordinate, exp denotes the exponential function with base e.

5. The method according to claim 1, wherein, Before the metallogenic potential index is calculated, the method further comprises the steps of: performing multi-level down-sampling on data channels of the same type to form data of different scales, performing directional enhancement on the data of different scales to obtain directional enhancement channels at different scales, enlarging the directional enhancement channel obtained at a low scale to the original scale by using an interpolation method, and merging the directional enhancement data channels of all scales according to the types of the data channels to obtain a comprehensive data set, wherein the comprehensive data set comprises multiple types of features.

6. The method according to claim 5, wherein, The ratio of the average value of the ore point region feature to the average value of the feature of the non-ore point region is taken as the weight of the feature.

7. The method according to claim 1, wherein, The local region used for calculating the structural direction is a window region with a size of MxN.

8. A metallogenic prediction system based on geological big data, characterized in that, The method comprises the steps of: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a method for metallogenic prediction based on geological big data according to any one of claims 1-7 is implemented.

Images