CN117828379A

CN117828379A - Underground resource detection method based on multi-source data fusion

Info

Publication number: CN117828379A
Application number: CN202410244563.9A
Authority: CN
Inventors: 徐东来; 魏海东; 董彦刚; 刘朝聚; 陈东亮; 李希乐
Original assignee: Shandong Institute of Geological Surveying and Mapping
Current assignee: Shandong Institute of Geological Surveying and Mapping
Priority date: 2024-03-05
Filing date: 2024-03-05
Publication date: 2024-04-05
Anticipated expiration: 2044-03-05
Also published as: CN117828379B

Abstract

The invention relates to the technical field of data clustering, in particular to an underground resource detection method based on multi-source data fusion. According to the method, an initial clustering center point is obtained through the distance distribution and the density degree of sampling points in a detection area of the earth surface and a geological multi-dimensional coordinate system formed by geological property data; in the current iterative clustering of k-means clustering, obtaining a mineral distribution characteristic value of a sampling point through clustering distribution of the sampling point in a geological multidimensional coordinate system and a detection area and neighborhood data difference conditions of different geological properties, and obtaining a mineral weight of the sampling point by combining a historical iterative change value obtained according to the difference between every two adjacent iterative clusters before the current iterative clustering; optimizing the next iterative clustering, and finally obtaining a final class clustering result so as to obtain a detection strategy. According to the invention, the distribution characteristics of the sampling points are comprehensively analyzed, so that the clustering results of different geology are more accurate, and the detection strategy obtained according to the clustering results is more reliable.

Description

Underground resource detection method based on multi-source data fusion

Technical Field

The invention relates to the technical field of data clustering, in particular to an underground resource detection method based on multi-source data fusion.

Background

Underground mineral resource exploration is a technique for finding, locating and evaluating underground mineral resources by using geophysical prospecting methods. These exploration methods infer characteristics of subsurface structures by measuring physical properties of the subsurface, such as electrical, magnetic, and seismic wave propagation velocities, etc., thereby helping exploration personnel to find potential mineral reserves.

In the detection process, underground physical property data are required to be collected and geological conditions are required to be subjected to clustering analysis so as to better detect mineral areas, in the common k-means clustering process, mineral distribution information is usually obtained from clustering results according to physical differences of minerals and geology, but under different types of geological structures, unmineralized rock strata can possibly cause abnormal physical properties such as an electric method and a magnetic method and the like, so that point cluster distribution is directly carried out according to Euclidean distance between data points, similarity of the minerals and the special geology in the collected physical properties is not considered, clustering abnormality can exist in the clustering results, the local point clusters are distributed into the mineral point clusters, the clustering results of geological analysis are inaccurate, and accordingly the detection strategy obtained according to the clustering results is unreliable.

Disclosure of Invention

In order to solve the technical problem that in the prior art, a clustering result of geological analysis is inaccurate, so that a detection strategy obtained according to the clustering result is unreliable, the invention aims to provide an underground resource detection method based on multi-source data fusion, and the adopted technical scheme is as follows:

the invention provides an underground resource detection method based on multi-source data fusion, which comprises the following steps:

in a detection area of the earth surface, different kinds of geological property data of each sampling point position are obtained; the sampling points are uniformly distributed in the detection area of the earth surface; obtaining a geological multidimensional coordinate system of the sampling point based on different kinds of geological property data;

when k-means clustering is carried out on sampling points in a geological multi-dimensional coordinate system, an initial clustering center point is obtained according to the distance distribution and the density degree among the sampling points in the geological multi-dimensional coordinate system; iterative clustering is carried out based on the initial clustering center point, and a category cluster under the current iterative clustering is obtained;

according to the clustering distribution condition of the category cluster where each sampling point is located in the geological multidimensional coordinate system and the detection area and the different category geological property data difference condition of the corresponding sampling point and the adjacent sampling point in the detection area, obtaining the mineral distribution characteristic value of each sampling point under the current iterative cluster;

Between every two adjacent iterative clusters before the current iterative cluster, according to the volume distribution condition of the category cluster where each sampling point is located in the geological multidimensional coordinate system and the deviation condition of the mineral distribution characteristic value of the corresponding sampling point, obtaining a historical iterative change value of each sampling point under the current iterative cluster;

according to the mineral distribution characteristic value and the historical iteration change value of each sampling point under the current iteration cluster, obtaining the mineral weight of each sampling point under the current iteration cluster; optimizing the next iterative clustering process based on the mineral weight of each sampling point under the current iterative clustering until a final class cluster is obtained; and obtaining a detection strategy according to the distribution of the final class cluster in the detection area.

Further, the method for acquiring the initial cluster center point comprises the following steps:

obtaining a k value of a k-means cluster, and determining a first initial cluster center point in a geological multidimensional coordinate system according to the distribution position of the sampling points;

taking other sampling points except the initial clustering center point in the geological multidimensional coordinate system as analysis points, and taking the distance between each analysis point and other analysis points closest to the analysis point as the adjacent distance of each analysis point; taking the average value of all the adjacent distances of the analysis points as the average adjacent distance;

For any analysis point, counting the number of other analysis points of the analysis point as the adjacent distribution value of the analysis point in the average adjacent distance range taking the analysis point as the center, and calculating the accumulated value of the adjacent distances of all the other analysis points to carry out negative correlation mapping to obtain the adjacent dense value of the analysis point; normalizing the product of the adjacent distribution value and the adjacent dense value of the analysis point to obtain the dense index of the analysis point;

calculating the distance between the analysis point and each initial clustering center point to be used as the distribution distance of the analysis point; normalizing the accumulated values of the analysis points corresponding to all the distribution distances to obtain a dispersion index of the analysis points;

obtaining a central point characteristic index of the analysis point according to the dense index and the disperse index of the analysis point; the dense index and the disperse index are positively correlated with the characteristic index of the central point;

taking the analysis point with the maximum central point characteristic index in all the analysis points as an initial clustering central point; and recalculating the central characteristic index of the residual analysis points according to the initial clustering central points, and iteratively obtaining the initial clustering central points until the number of the initial clustering central points is equal to the k value.

Further, the method for acquiring the mineral distribution characteristic value comprises the following steps:

sequentially taking each sampling point as a measured point, and taking a class cluster in which the measured point is positioned as a measured cluster;

calculating the distance between the sampling point closest to the measured point and the measured point in the measured cluster in each preset direction in the geological multidimensional coordinate system, and obtaining the direction distance of the measured point in each preset direction; according to the uniform distribution condition of all direction distances of the measured point, acquiring a cluster distribution characteristic index of the measured point;

obtaining neighborhood distribution characteristic indexes of the measured points according to the numerical deviation condition and the numerical discrete condition of the measured points and other sampling points on each geological property data in the preset neighborhood range of the measured points in the detection area;

obtaining the distribution characteristic degree of the measured points according to the cluster distribution characteristic index and the neighborhood distribution characteristic index of the measured points; the cluster distribution characteristic index and the neighborhood distribution characteristic index are positively correlated with the distribution characteristic degree;

counting the number of other sampling points in the cluster to be detected in the preset neighborhood range of the detection area to obtain the cluster aggregation number of the detected points; taking the ratio of the cluster aggregation number of the measured points to the total number of other sampling points of the measured points in the preset neighborhood range of the detection area as the aggregation characteristic of the measured points;

And calculating the product of the distribution characteristic degree and the aggregation characteristic degree of the measured points to obtain the mineral distribution characteristic value of the measured points.

Further, the obtaining the cluster distribution characteristic index of the measured point according to the uniform distribution condition of all the directional distances of the measured point comprises:

calculating the accumulated value of all direction distances of the measured point to obtain the direction compactness of the measured point; calculating variances of all direction distances of the measured point to obtain the direction uniformity of the measured point;

and calculating the product of the direction compactness and the direction uniformity of the measured point, performing negative correlation mapping and normalization processing, and obtaining the cluster distribution characteristic index of the measured point.

Further, the method for obtaining the neighborhood distribution characteristic index comprises the following steps:

calculating the data value difference of the measured point and other sampling points on each geological property data for any one other sampling point in the preset neighborhood range of the detected area, and accumulating the data value difference of the measured point and the other sampling points on all geological property data to obtain the numerical value difference index of the measured point and the other sampling points; calculating the accumulated value of the number difference indexes of the measured point and all other sampling points to obtain the numerical deviation degree of the measured point;

Calculating the variance of the data values of all sampling points on each geological property data in the preset neighborhood range of the detected point to obtain the numerical value confusion index of the detected point on each geological property data; calculating the average value of numerical value chaotic indexes of the measured points on all geological property data to obtain the numerical value dispersion of the measured points;

and calculating the numerical deviation degree and the product of the numerical deviation degree of the measured point, performing negative correlation mapping and normalization processing, and obtaining the neighborhood distribution characteristic index of the measured point.

Further, the method for obtaining the historical iteration change value comprises the following steps:

if the current iterative clustering is the first iterative clustering, setting a historical iterative change value of each sampling point under the current iterative clustering as a preset historical iterative change value;

otherwise, each two adjacent iterative clusters before the current iterative cluster are used as iterative pairs; taking each sampling point as a target point in sequence, and for any iteration pair, calculating the difference of mineral distribution characteristic values of the target point among the iteration clusters of the iteration pair to obtain the mineral distribution difference of the target point in the iteration pair;

acquiring the volume of a class cluster in which the target point is located in each iterative cluster of the iterative pair in a geological multidimensional coordinate system; taking the difference of the volumes of the target points among the iterative clusters of the iterative pair as the volume distribution difference of the target points in the iterative pair;

According to the mineral distribution difference and the volume distribution difference of the target point in the iteration pair, obtaining an iteration distribution index of the target point in the iteration pair; the mineral distribution difference and the volume distribution difference are positively correlated with the iteration distribution index;

and carrying out negative correlation mapping on accumulated values of iteration distribution indexes of the target points in all iteration pairs to obtain historical iteration change values of the target points under the current iteration clusters.

Further, the method for acquiring the mineral weight comprises the following steps:

and calculating the average value of the mineral distribution characteristic value and the historical iteration change value of each sampling point under the current iteration cluster, carrying out negative correlation mapping and normalization processing, and obtaining the mineral weight of each sampling point under the current iteration cluster.

Further, the optimizing the next iterative clustering process based on the mineral weight of each sampling point under the current iterative clustering until a final class cluster is obtained includes:

in the next iterative clustering process, calculating the distance between each sampling point and the center point of the class cluster of the current iterative cluster as a clustering distance;

taking the product of the mineral weight and the clustering distance of each sampling point under the current iterative clustering as the distance measurement of the next iterative clustering of each sampling point; and carrying out next iterative clustering based on the distance measurement of each sampling point until the class cluster converges to obtain a final class cluster.

Further, the obtaining the detection strategy according to the distribution of the final class cluster in the detection area includes:

and inputting the final class cluster into the trained neural network model, and outputting the detection strategy.

Further, the geological multi-dimensional coordinate system for obtaining the sampling point based on the geological property data of different kinds comprises:

and taking each geological property data as a dimensional coordinate axis, obtaining a geological multi-dimensional coordinate system, and mapping each sampling point position into the geological multi-dimensional coordinate system according to the corresponding geological property data.

The invention has the following beneficial effects:

according to the invention, through comprehensive analysis of the sampling points in the detection area of the earth surface and the distribution situation in the geological multi-dimensional coordinate system formed by geological property data, the sampling points which are more similar to the geological characteristics of the special geology and the minerals can be more accurately distinguished, and the initial clustering center point is obtained according to the distance distribution and the density degree of the sampling points in the geological multi-dimensional coordinate system, so that the local optimal solution of k-means clustering is avoided, and the special geology and the mineral geological areas which are more difficult to distinguish can be clustered better. Further, in the iterative clustering process, the characteristic that sampling points of mineral features are more tightly concentrated in distribution and are uniformly distributed is considered, and the mineral distribution feature value of each sampling point is obtained through the clustering distribution condition of each sampling point in a geological multidimensional coordinate system and a detection area and the neighborhood data difference condition of different types of geological properties in the detection area, so that the mineral feature significance of each sampling point is reflected. Meanwhile, the stability difference of special geology and mineral geology in the clustering process is considered, the distribution difference between every two adjacent iterative clusters before the current iterative clustering is analyzed, the historical iterative change value is further obtained, and the mineral characteristics of each sampling point position are reflected to be obvious. Finally, the mineral distribution characteristic value and the historical iteration change value are integrated to obtain the mineral weight, the distance measurement of the next iteration clustering process of k-means clustering is optimized, and finally, a more accurate clustering result is obtained, and then a detection strategy is obtained. According to the invention, the distribution characteristics of the sampling points are comprehensively analyzed, so that the difference between mineral geology and non-mineral geology is improved, the clustering results of different geology are more accurate, and the detection strategy obtained according to the clustering results is more reliable.

Drawings

In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flowchart of an underground resource detection method based on multi-source data fusion according to an embodiment of the present invention.

Detailed Description

In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following detailed description refers to the specific implementation, structure, characteristics and effects of the underground resource detection method based on multi-source data fusion according to the invention by combining the accompanying drawings and the preferred embodiment. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

The following specifically describes a specific scheme of the underground resource detection method based on multi-source data fusion provided by the invention with reference to the accompanying drawings.

Referring to fig. 1, a flowchart of a method for detecting underground resources based on multi-source data fusion according to an embodiment of the invention is shown, and the method includes the following steps:

s1: in a detection area of the earth surface, different kinds of geological property data of each sampling point position are obtained; the sampling points are uniformly distributed in the detection area of the earth surface; and obtaining a geological multidimensional coordinate system of the sampling point based on different kinds of geological property data.

Based on the difference of the geological properties of the common geology and the mineral geology, different categories can be obtained through clustering for analysis, the detection direction strategy is adjusted, and the like, so that the detection area is required to be determined on the ground surface for clustering analysis. In the embodiment of the invention, the detection area is the area which needs to be analyzed first, a rectangular area with a side length of one kilometer can be set as the detection area, the number of sampling points is set to be 200, and a specific numerical value implementation person can adjust according to specific implementation conditions without limitation. In the detection area of the earth surface, different kinds of geological property data of each sampling point are acquired, wherein the sampling points are represented as sampling positions on the earth surface, and the sampling points are uniformly distributed in the detection area of the earth surface so as to ensure complete detection of the detection area.

In the embodiment of the present invention, the geological properties are mainly physical properties of geology, and the types of geological property data include: the electric conductivity, the magnetic field intensity, the propagation speed of the vibration wave and the like can be acquired according to specific implementation conditions, each sampling point is acquired by each geological property data, for example, electric exploration is carried out by using instruments such as an electrode, a current source, a potentiometer and the like, the electrode is arranged on the ground surface, and the current is introduced into the ground to obtain the electric conductivity of the sampling point. And measuring the intensity data of the geomagnetic field of the sampling point by using a magnetometer to obtain the magnetic field intensity. The earthquake wave is transmitted at the sampling point by using the earthquake device, the underground vibration is generated, and the propagation velocity of the earthquake wave under the sampling point is recorded by the earthquake receiver. It should be noted that, the method for collecting different kinds of geological property data is a technical means well known to those skilled in the art, and will not be described in detail herein.

When the sampling point is subjected to geologic category analysis, the sampling point is required to be combined with different types of geologic property data for similarity difference condition analysis, so that the multi-source data is required to be placed in a multi-dimensional space for comprehensive analysis and clustering, and a geologic multi-dimensional coordinate system of the sampling point is obtained based on the different types of geologic property data. In one embodiment of the invention, each geological property data is used as a dimensional coordinate axis, a geological multi-dimensional coordinate system is obtained, and each sampling point position is mapped into the geological multi-dimensional coordinate system according to the corresponding geological property data.

S2: when k-means clustering is carried out on sampling points in a geological multi-dimensional coordinate system, an initial clustering center point is obtained according to the distance distribution and the density degree among the sampling points in the geological multi-dimensional coordinate system; and carrying out iterative clustering based on the initial clustering center point to obtain a category cluster under the current iterative clustering.

After the space dimension unification of the multi-source data is completed, clustering can be carried out according to the distribution of each sampling point in a geological multi-dimensional coordinate system, and the main process of a common k-means clustering algorithm is as follows: firstly, randomly selecting k initial clustering center points; secondly, clustering is carried out through the distance from other data points to the initial clustering center point, and a cluster is obtained; thirdly, readjusting the center point according to the cluster; and fourthly, repeating the process of the second step and the third step according to the adjusted central point to perform iterative clustering until convergence or a stopping condition is met, and obtaining a final clustering result.

When k-means clustering is carried out on sampling points in a geological multidimensional coordinate system, the positions of initial clustering centers are required to be determined, the traditional initial clustering centers are selected to avoid the initial clustering centers from being trapped into local optimum, the distribution requirements among the centers are generally as far as possible, namely, the distribution is more scattered, but in the geological data analysis of underground minerals, the distribution of the clustering centers is only required to be ensured, the local optimum is not required to be additionally avoided, the geological property data difference between the special geology and the mineral geology is smaller, and more initial clustering centers are required to be distributed as much as possible in a data-dense area on the premise that the clustering centers are scattered, so that the mineral resource points and the geological points of similar characteristics in the dense area are conveniently divided. Therefore, according to the distance distribution and the density degree between the sampling points in the geological multidimensional coordinate system, an initial clustering center point is obtained.

Preferably, a k value of k-means clustering is obtained, a first initial clustering center point is determined according to a sampling point distribution position in a geological multidimensional coordinate system, in the embodiment of the present invention, sampling points located at the center positions of all sampling points in the geological multidimensional coordinate system are used as the first initial clustering center point, and an implementer can adjust according to a specific implementation scene, which is not described herein. It should be noted that, the method for obtaining the center point position and the method for obtaining the k value of k-means clustering are all technical means well known to those skilled in the art, for example, an elbow method is adopted to obtain the k value, which is not described herein.

And taking other sampling points except the initial clustering center point in the geological multidimensional coordinate system as analysis points, and analyzing the characteristics of the distribution points which are possible to be the center points. The distance between each analysis point and other analysis points closest to the analysis point is first used as the adjacent distance of each analysis point, that is, the minimum distance between each analysis point and other analysis points is obtained as the adjacent distance of each analysis point. And taking the average value of all the adjacent distances of the analysis points as the average adjacent distance to reflect the average distribution distance of the analysis points.

For any one analysis point, counting the number of other analysis points of the analysis point as the adjacent distribution value of the analysis point in the average adjacent distance range taking the analysis point as the center, namely in the circular range corresponding to the analysis point taking the average adjacent distance as the radius, wherein the adjacent distribution value reflects the local distribution density degree of the analysis point. And calculating accumulated values of adjacent distances of all other analysis points to carry out negative correlation mapping, obtaining the adjacent dense value of the analysis point, reflecting the local aggregation degree of the analysis point, and indicating that the local aggregation degree of the analysis point is higher when the distance between each analysis point and other analysis points is closer in the corresponding range of the analysis point. And carrying out normalization processing on the product of the adjacent distribution value and the adjacent dense value of the analysis point to obtain the dense index of the analysis point, and reflecting the dense degree of the local distribution of the analysis point through the dense index.

Further, the distance between the analysis point and each initial clustering center point is calculated and used as the distribution distance of the analysis point, and the distribution distance reflects the position distribution situation of the analysis point and the determined initial clustering center points. And carrying out normalization processing on accumulated values of all distribution distances corresponding to the analysis points to obtain a dispersion index of the analysis points, and reflecting the distribution dispersion degree of the analysis points and the initial clustering center points through the position distribution condition between the analysis points and all the initial clustering center points.

When the overall dispersion degree between the analysis point and the initial clustering center point is larger, the distribution density degree of the local analysis point of the analysis point is higher, which indicates that the analysis point can be used as the initial clustering center point, the center point characteristic index of the analysis point is further obtained according to the density index and the dispersion index of the analysis point, the center point characteristic index reflects the optimization degree of the analysis point as the clustering center point, and the density index and the dispersion index are positively correlated with the center point characteristic index, and in the embodiment of the invention, the expression of the center point characteristic index is as follows:

，

in the method, in the process of the invention,denoted as +.>Center point feature index of each analysis point, +.>Expressed as total number of initial cluster center points, +.>Denoted as +.>Analysis Point and->Distribution distance of initial clustering center points, +.>Denoted as +.>The number of adjacent distribution of the individual analysis points, +.>Indicated as in->First +.within the average proximity distance range centered at the analysis point>The proximity distance of the other analysis points, +.>Expressed as an exponential function based on natural constants, < ->Represented as a normalization function, it should be noted that normalization is well known to those skilled in the artThe choice of normalization function can be linear normalization or standard normalization by known technical means, and the specific normalization method is not limited herein.

Wherein,denoted as +.>The neighboring dense values of the individual analysis points,denoted as +.>Dense index of individual analysis points, +.>Denoted as +.>The larger the dispersion index is, the more the distribution of the analysis points and all initial clustering center points is dispersed, and the larger the density index is, the higher the distribution density degree of the local analysis points of the analysis points is, the higher the optimization degree of the analysis points serving as the clustering center points is, so that the characteristic index of the center points is larger. In other embodiments of the present invention, other basic mathematical operations may be used to reflect that the dense index and the disperse index are both positively correlated with the central point characteristic index, such as multiplication or power operation, without limitation.

Finally, taking the analysis point with the maximum center point characteristic index in all the analysis points as an initial clustering center point, obtaining a new optimal initial clustering center point under the condition that the initial clustering center is determined, and recalculating the center characteristic index according to the initial clustering center point for the rest analysis points to obtain the initial clustering center points in an iterative manner until the number of the initial clustering center points is equal to the k value, namely, the process of determining the initial clustering center point for calculating the center characteristic index needs to iterate k-1 times, and finishing the selection of the initial clustering center points in the k-means clustering process.

On the basis of avoiding sinking into local optimum, the characteristic of small difference of geological property data of special geology and mineral geology is combined, sampling points in a denser area are selected, and an initial clustering center point which is most favorable for clustering classification of different geology is obtained, so that the method is a basis for a subsequent iterative analysis clustering result.

Iterative clustering is carried out based on initial clustering center points, iterative clustering can be carried out according to initial clustering centers, clustering results can be obtained in each clustering, in the traditional clustering process, each clustering is carried out based on distribution distances in a geological multidimensional coordinate system, and the situation that the geological property difference between special geology and mineral geology is small exists in geology, so that position distance measurement when mineral characteristics are required to be further analyzed and adjusted for clustering is only inaccurate according to distance analysis, and therefore a class clustering cluster under the current iterative clustering is obtained, and the condition of sampling points in the class clustering cluster is further analyzed.

S3: according to the clustering distribution condition of the category cluster where each sampling point is located in the geological multidimensional coordinate system and the detection area and the different category geological property data difference condition of the corresponding sampling point and the adjacent sampling point in the detection area, obtaining the mineral distribution characteristic value of each sampling point under the current iterative cluster.

In general, geological strata and minerals have larger geological property differences, such as larger differences in conductivity, magnetic field strength and seismic wave propagation velocity, but data anomalies of conductivity, magnetic field strength and seismic wave propagation velocity may be caused in non-mineralized special geology, specifically, for example, if sandstone containing saline water exists in the depth detection process, conductivity is increased, magma activity in the underground may be caused, or underground cavities and karst in the detection process also cause anomalies of seismic wave propagation, and the like, which have similarity with characteristics of mineral anomalies in conductivity, magnetic field strength and seismic wave propagation velocity.

Therefore, the characteristic analysis acquisition is needed based on the difference expression of geological property data of special geology and mineral geology in the clustering process, so that more accurate clustering distinction is convenient in the subsequent iterative adjustment clustering process. Compared with a special geological stratum, the molecular structure property of mineral geology is tighter, so that the overall conductivity, the magnetic field intensity and the propagation speed of the seismic wave are more stable, geological property data of different sampling points are more similar, the sampling points are more concentrated in a geological multidimensional coordinate system, and the neighborhood distribution similarity of the geological property data in a detection area is high. The special geological strata, such as the strata with brine saturation, have conductivity, but the brine saturation in the strata is easy to generate non-uniformity, so that the conductivity difference of different sampling points is larger, the distribution of the saline saturation is more discrete and chaotic in a geological multidimensional coordinate system, the magma activity is dynamic fluid, the instability of the magnetic field intensity is also generated, and the differences of the seismic wave velocities in different sampling points are also caused by underground cavity karst cave, so that the characteristics of more discrete and chaotic distribution in the geological multidimensional coordinate system are comprehensively reflected, and the neighborhood distribution similarity of geological property data in a detection area is low.

Therefore, for each sampling point, the uniform distribution aggregation condition of the sampling points in the geological multi-dimensional coordinate system and the detection area can be analyzed to obtain the mineral distribution characteristic value of each sampling point which is mineral geology, so that the accuracy of clustering discrimination can be improved during subsequent clustering, and therefore, the mineral distribution characteristic value of each sampling point under the current iterative clustering can be obtained according to the clustering distribution condition of the category cluster where each sampling point is respectively located in the geological multi-dimensional coordinate system and the detection area and the different types of geological property data difference condition of the corresponding sampling point and the adjacent sampling point in the detection area.

Preferably, each sampling point is sequentially used as a measured point, the characteristics of each sampling point are sequentially analyzed, and the class cluster where the measured point is located is used as a measured cluster. In the embodiment of the invention, the preset directions are positive and negative directions of each dimension in the geological multi-dimensional coordinate system, for example, when the geological multi-dimensional coordinate system is a three-dimensional coordinate system, the preset directions are positive and negative directions of an x axis, positive and negative directions of a y axis and positive and negative directions of a z axis, and 6 preset directions are total. The distribution aggregation condition of the measured points in the multidimensional space can be reflected through the direction distance of the measured points in each preset direction in the measured cluster.

Further, according to the condition that the distribution of all direction distances of the measured points is uniform, a cluster distribution characteristic index of the measured points is obtained, in one embodiment of the invention, the accumulated value of all direction distances of the measured points is calculated, the direction compactness of the measured points is obtained, and when the accumulated sum of the direction distances is smaller, the local distribution of the measured points in a cluster space is more compact. And calculating the variance of all direction distances of the measured point to obtain the direction uniformity of the measured point, wherein the smaller the variance is, the more uniform the local distribution of the measured point in the cluster space is. Calculating the product of the direction compactness and the direction uniformity of the measured point, performing negative correlation mapping and normalization processing to obtain a cluster distribution characteristic index of the measured point, wherein the cluster distribution characteristic index is combined with the local distribution compactness degree and the distribution uniformity degree of the measured point in a geological multidimensional coordinate system to reflect the remarkable degree of the measured point embodied as mineral characteristics, and in the embodiment of the invention, the expression of the cluster distribution characteristic index is as follows:

，

in the method, in the process of the invention,denoted as +.>Clustering distribution characteristic indexes of sampling points, < + >>Denoted as +.>Total number of all preset directions corresponding to the sampling points, +. >Denoted as +.>The sampling point is at->Direction distance in a preset direction, +.>Denoted as +.>Average value of direction distances in all preset directions corresponding to the sampling points, < >>Represented as an exponential function with a base of natural constant.

Wherein,denoted as +.>The directional compactness of the individual sampling points, +.>Denoted as +.>The direction uniformity of each sampling point is smaller when the direction compactness and the direction uniformity are smaller, which means that the local characteristic distribution of the sampling points is more uniform and compact, so that the cluster distribution characteristic index is larger, and the sampling points represent the geological characteristics of minerals more obviously.

Further analyzing the neighborhood distribution similarity condition of the measured point in the detection area, and obtaining the neighborhood distribution characteristic index of the measured point according to the numerical deviation condition and the numerical discrete condition of the measured point and other sampling points on each geological property data in the preset neighborhood range of the detected point in the detection area.

In one embodiment of the invention, for any one other sampling point in a preset neighborhood range of a detected point, calculating the data value difference of the detected point and the other sampling point on each geological property data, accumulating the data value differences of the detected point and the other sampling point on all geological property data to obtain a numerical value difference index of the detected point and the other sampling point, and reflecting the similarity degree of the detected point and the other sampling point through the numerical value difference condition of the other sampling point and the detected point on each geological property data. And calculating the accumulated value of the numerical difference indexes of the measured point and all other sampling points to obtain the numerical deviation degree of the measured point, and integrating the difference between the measured point and all the sampling points in the neighborhood range to reflect the neighborhood distribution similarity of the measured point on the numerical value of the geological property data.

Further, in the preset neighborhood range of the detected point, calculating the variance of the data values of all sampling points on each geological property data, and obtaining the numerical value chaotic index of the detected point on each geological property data, wherein the numerical value chaotic index is the overall data distribution confusion of each geological property data on the neighborhood. And calculating the average value of numerical value confusion indexes of the measured points on all geological property data, obtaining the numerical value dispersion of the measured points, and integrating the distribution confusion degree of all kinds of geological property data to reflect the neighborhood distribution similarity of the measured points on the uniform distribution of the geological property data in the neighborhood range.

Finally, integrating the data distribution of different types of geological property data on the neighborhood, calculating the product of the numerical deviation degree and the numerical dispersion degree of the measured point, carrying out negative correlation mapping and normalization processing to obtain the neighborhood distribution characteristic index of the measured point, wherein the neighborhood distribution characteristic index is combined with the distribution difference and the distribution stability degree of the geological property data of the measured point in the detection area to reflect the significance degree of the measured point embodied as mineral characteristics, and in the embodiment of the invention, the expression of the neighborhood distribution characteristic index is as follows:

，

In the method, in the process of the invention,denoted as +.>Neighborhood distribution characteristic indexes of sampling points, +.>Denoted as +.>Total number of other sampling points in preset neighborhood range of the sampling points, +.>Expressed as the total category number of geological property data, < >>Denoted as +.>The sampling point is at->Data values on the seed geological property data +.>Denoted as +.>The sampling points correspond to the +.>The other sampling points are +>Data values on the seed geological property data +.>Denoted as +.>The sampling point is at->Numerical confusion index on the seed geological property data, +.>Expressed as absolute value extraction function,/->Represented as an exponential function with a base of natural constant.

Wherein,denoted as +.>The sampling points are corresponding to the first +.>The other sampling points are +>Data value differences on the seed geological property data, +.>Denoted as +.>The sampling points are corresponding to the first +.>Numerical value difference index of other sampling points, +.>Denoted as +.>Numerical deviation of the individual sampling points, +.>Denoted as +.>Numerical dispersion of the individual sampling points. When the numerical deviation degree is smaller, which indicates that the neighborhood distribution of the geological property data of the sampling points in the detection area is more similar, so that the neighborhood distribution characteristic index is larger, and the sampling points represent the geological characteristics of minerals more obviously.

Further, according to the cluster distribution characteristic index and the neighborhood distribution characteristic index of the measured point, the distribution characteristic degree of the measured point is obtained, the distribution characteristic degree reflects the mineral characteristic distribution condition of the sampling point in a geological multidimensional coordinate system and a detection area, and the cluster distribution characteristic index and the neighborhood distribution characteristic index are positively correlated with the distribution characteristic degree.

Counting the number of other sampling points in a detected cluster in a preset neighborhood range of a detection area to obtain the cluster aggregation number of the detected points, taking the ratio of the cluster aggregation number of the detected points to the total number of other sampling points in the preset neighborhood range of the detection area as the aggregation characteristic of the detected points, wherein in the detection area, the more the number of the sampling points in the same cluster in the neighborhood range of the detected points is, the more the detected points are aggregated in the geographic neighborhood, and the higher the geological probability of the minerals is.

Finally, calculating the product of the distribution characteristic degree and the aggregation characteristic degree of the measured point to obtain a mineral distribution characteristic value of the measured point, wherein the mineral distribution characteristic value reflects the significance degree of the measured point as mineral ground particles, and in the embodiment of the invention, the expression of the mineral distribution characteristic value is as follows:

，

In the method, in the process of the invention,denoted as +.>Mineral distribution characteristic values of the sampling points under the current iterative clustering>Denoted as +.>Clustering distribution characteristic indexes of sampling points, < + >>Denoted as +.>Neighborhood distribution characteristic indexes of sampling points, +.>Denoted as +.>Total number of other sampling points in preset neighborhood range of the sampling points, +.>Denoted as +.>Cluster aggregation number of individual sampling points.

Wherein,denoted as +.>Aggregation feature of the sampling points, +.>Denoted as +.>In other embodiments of the present invention, other basic mathematical operations may be used to reflect that the clustered distribution feature index and the neighborhood distribution feature index are positively correlated with the distribution feature degree, such as multiplication or power operation, which is not described herein.

And finally, completing the distribution analysis of each sampling point under the current iterative clustering condition to obtain the mineral geological feature of each sampling point under the current iterative clustering.

S4: and acquiring a historical iteration change value of each sampling point under the current iteration cluster according to the volume distribution condition of the category cluster where each sampling point is located in the geological multidimensional coordinate system and the deviation condition of the mineral distribution characteristic value of the corresponding sampling point between every two adjacent iteration clusters before the current iteration cluster.

Because most of minerals are integral after being formed, the clustering distribution at the sampling points is easy to form an integral, namely, the clustering change degree of data points is small in the iterative process of multiple clustering, and the stability of the mineral characteristic distribution degree is good. The geological formation is usually corroded by water flow and air flow, the aggregation of the geological formation is unstable, the data points are unstable when clustered in the iterative process of multiple clustering, the change degree is large, and the distribution degree stability of mineral features is poor.

Therefore, between every two adjacent iterative clusters before the current iterative cluster, according to the volume distribution condition of the class cluster where each sampling point is located in the geological multidimensional coordinate system and the deviation condition of the mineral distribution characteristic value of the corresponding sampling point, the historical iterative change value of each sampling point under the current iterative cluster is obtained. Preferably, if the current iterative cluster is the first iterative cluster, there is no iterative cluster before the current iterative cluster, and then the historical iterative change value of each sampling point under the current iterative cluster is set to a preset historical iterative change value, in the embodiment of the present invention, the preset historical iterative change value is set to zero, and the specific numerical implementer can adjust according to specific implementation conditions.

Otherwise, each two adjacent iterative clusters before the current iterative cluster are used as iterative pairs, and the change condition in the adjacent iterative processes is analyzed, for example, if the current iterative cluster is a second iterative cluster, the iterative pairs are one, if the current iterative cluster is a third iterative cluster, the iterative pairs are two, the first iterative cluster and the second iterative cluster are one iterative pair, and the second iterative cluster and the third iterative cluster are one iterative pair.

And sequentially taking each sampling point as a target point, and for any iteration pair, calculating the difference of mineral distribution characteristic values of the target point between the iteration clusters of the iteration pair to obtain the mineral distribution difference of the target point in the iteration pair and reflect the change condition of the mineral characteristics. The volume of the class cluster in which the target point is located in each iteration cluster of the iteration pair in the geological multidimensional coordinate system is acquired, and it is to be noted that the acquisition of the volume in the space is a technical means well known to the skilled person in the field, such as a minimum circumscribed body, and the like, and is not described herein. And taking the difference of the volumes of the target points among the iterative clusters of the iterative pair as the difference of the volume distribution of the target points among the iterative pair, and reflecting the change condition of the distribution of the category cluster where the target points are among the iterations.

Further, according to the mineral distribution difference and the volume distribution difference of the target point in the iteration pair, iteration distribution indexes of the target point in the iteration pair are obtained, the change degree of the target point between iterations is reflected, and the mineral distribution difference and the volume distribution difference are positively correlated with the iteration distribution indexes. And carrying out negative correlation mapping on accumulated values of iteration distribution indexes of the target points in all iteration pairs to obtain historical iteration change values of the target points under the current iteration cluster, wherein the historical iteration change values reflect the overall change condition of the target points among all iterations before the current iteration cluster. In the embodiment of the invention, the expression of the historical iteration change value is as follows:

，/>

in the method, in the process of the invention,denoted as +.>Historical iteration change values of sampling points under current iteration cluster, < ->Denoted as the firstTotal number of iteration pairs under current iteration cluster of sampling points, +.>Denoted as +.>Sample point position +.>Mineral distribution characteristic value of one iteration cluster in iteration pair,/->Denoted as +.>Sample point position +.>Mineral distribution characteristic value of the other iteration cluster in the iteration pair,/->Denoted as +. >Sample point position +.>For the volume of one iteration cluster in the iteration pair, +.>Denoted as +.>Sample point position +.>For the volume of the other iteration cluster in the iteration pair, +.>Expressed as absolute value extraction function,/->Represented as an exponential function with a base of natural constant.

Wherein,denoted as +.>Sample point position +.>For differences in mineral distribution of iterative pairs, +.>Denoted as +.>Sample point position +.>For the volume distribution differences of iterative pairs, +.>Denoted as +.>Sample point position +.>For the iteration distribution index of the iteration pair, the mineral distribution difference and the volume distribution difference are reflected in a multiplication mode to be positively correlated with the iteration distribution index, and in other embodiments of the invention, other basic mathematical operations can be used to reflect that the mineral distribution difference and the volume distribution difference are positively correlated with the iteration distribution index, such as addition or power operation, and the like, and the method is not limited. When the iteration distribution index is smaller, the difference of sampling points between iteration pairs is smaller, so that the historical iteration change value is larger, the integral change difference of the sampling points in the process from the current iteration clustering is smaller, and the sampling points are more obvious in characteristics of mineral geology.

S5: according to the mineral distribution characteristic value and the historical iteration change value of each sampling point under the current iteration cluster, obtaining the mineral weight of each sampling point under the current iteration cluster; optimizing the next iterative clustering process based on the mineral weight of each sampling point under the current iterative clustering until a final class cluster is obtained; and obtaining a detection strategy according to the distribution of the final class cluster in the detection area.

And synthesizing the mineral feature significance of the sampling points to obtain the influence weight condition of each sampling point on the geological feature, namely obtaining the mineral weight of each sampling point under the current iterative cluster according to the mineral distribution feature value and the historical iterative change value of each sampling point under the current iterative cluster. Preferably, calculating an average value of a mineral distribution characteristic value and a historical iteration change value of each sampling point under the current iteration cluster, performing negative correlation mapping and normalization processing to obtain a mineral weight of each sampling point under the current iteration cluster, wherein in the embodiment of the invention, the expression of the mineral weight is as follows:

，

in the method, in the process of the invention,denoted as +.>Mineral weights of the sampling points under the current iterative cluster,/- >Denoted as +.>Historical iteration change values of sampling points under current iteration cluster, < ->Denoted as +.>Mineral distribution characteristic values of the sampling points under the current iterative clustering>Represented as an exponential function with a base of natural constant.

When the mineral distribution characteristic value and the historical iteration change value are larger, the mineral characteristics of the sampling points in the current cluster are obvious, and in the subsequent iteration clusters, the distance measurement needs to be adjusted to be smaller when the distance between the sampling points and the cluster center point is calculated, so that the clustering effect of the sampling points is improved. And optimizing the next iterative clustering process based on the mineral weight of each sampling point under the current iterative clustering until a final class cluster is obtained.

In one embodiment of the invention, in the next iterative clustering process, the distance between each sampling point and the center point of the class cluster of the current iterative cluster is calculated and used as the clustering distance, namely the distance measure utilized in the traditional k-means clustering. And taking the product of the mineral weight and the clustering distance of each sampling point under the current iterative clustering as the distance measurement of the next iterative clustering of each sampling point, completing optimization of the distance measurement through weighting of the mineral weight, and carrying out the next iterative clustering based on the distance measurement of each sampling point until the category cluster converges, so as to obtain the final category cluster. It should be noted that, the k-means clustering iterative clustering and convergence stopping are known technical means known to those skilled in the art, and may be adjusted according to specific implementation conditions, which are not described herein.

Finally, a detection strategy is obtained according to the distribution of the final class clusters in the detection area, and the final class clusters reflect different geological class conditions, so that in one embodiment of the invention, the final class clusters are input into a trained neural network model, and the detection strategy is output. In the embodiment of the present invention, the neural network may adopt a CNN neural network to perform classification processing, and it should be noted that the CNN neural network is a technical means well known to those skilled in the art, and is not described herein in detail, but only a brief process of determining the detection policy by using the CNN neural network in one embodiment of the present invention is described briefly: the CNN neural network is trained by using the reference data set, the collected data of different geological categories and corresponding detection strategies are used as the reference data set, the data are input into the CNN neural network for training, cross entropy is used as a loss function, and the self-adaptive motion estimation algorithm is adopted for optimization, so that a trained neural network model is obtained, the final category cluster is input into the trained neural network model, and the detection strategies are output. The essential task of the neural network model is to classify and integrate the adaptive motion estimation algorithm, so the neural network structure that implements the task includes a variety. The specific neural network structure, training algorithm process and adaptive motion estimation algorithm are well known to those skilled in the art, and will not be described herein. In other embodiments of the present invention, the classification screening with high mineral characteristics and the principal component analysis may be performed according to the geological data corresponding to the final classification cluster, and the principal component direction is used as the detection adjustment direction to improve the mineral detection probability, which is not described herein.

In summary, the invention comprehensively analyzes the distribution situation of the sampling points in the detection area of the earth surface and the geological multi-dimensional coordinate system formed by geological property data, so that the sampling points which are more similar to the geological characteristics of special geology and minerals can be more accurately distinguished, firstly, the initial clustering center point is obtained according to the distance distribution and the density degree of the sampling points in the geological multi-dimensional coordinate system, and the special geology and minerals which are more difficult to distinguish can be clustered better while avoiding the local optimal solution of k-means clustering. Further, in the iterative clustering process, the characteristic that sampling points of mineral features are more tightly concentrated in distribution and are uniformly distributed is considered, and the mineral distribution feature value of each sampling point is obtained through the clustering distribution condition of each sampling point in a geological multidimensional coordinate system and a detection area and the neighborhood data difference condition of different types of geological properties in the detection area, so that the mineral feature significance of each sampling point is reflected. Meanwhile, the stability difference of special geology and mineral geology in the clustering process is considered, the distribution difference between every two adjacent iterative clusters before the current iterative clustering is analyzed, the historical iterative change value is further obtained, and the mineral characteristics of each sampling point position are reflected to be obvious. Finally, the mineral distribution characteristic value and the historical iteration change value are integrated to obtain the mineral weight, the distance measurement of the next iteration clustering process of k-means clustering is optimized, and finally, a more accurate clustering result is obtained, and then a detection strategy is obtained. According to the invention, the distribution characteristics of the sampling points are comprehensively analyzed, so that the difference between mineral geology and non-mineral geology is improved, the clustering results of different geology are more accurate, and the detection strategy obtained according to the clustering results is more reliable.

It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.

Claims

1. An underground resource detection method based on multi-source data fusion, which is characterized by comprising the following steps:

2. The underground resource detection method based on multi-source data fusion according to claim 1, wherein the method for acquiring the initial cluster center point comprises the following steps:

3. The underground resource detection method based on multi-source data fusion according to claim 1, wherein the method for acquiring the mineral distribution characteristic value comprises the following steps:

4. The method for detecting underground resources based on multi-source data fusion according to claim 3, wherein the obtaining the cluster distribution characteristic index of the measured point according to the uniform distribution of all direction distances of the measured point comprises:

5. The underground resource detection method based on multi-source data fusion according to claim 3, wherein the method for acquiring the neighborhood distribution characteristic index comprises the following steps:

6. The underground resource detection method based on multi-source data fusion according to claim 1, wherein the method for obtaining the historical iteration change value comprises the following steps:

7. The underground resource detection method based on multi-source data fusion according to claim 1, wherein the mineral weight acquisition method comprises the following steps:

8. The underground resource detection method based on multi-source data fusion according to claim 1, wherein the optimizing the next iterative clustering process based on the mineral weight of each sampling point under the current iterative clustering until the final class cluster is obtained comprises:

9. The method for detecting underground resources based on multi-source data fusion according to claim 1, wherein the obtaining the detection strategy according to the distribution of the final class cluster in the detection area comprises:

10. The method for detecting the underground resource based on the multi-source data fusion according to claim 1, wherein the obtaining the geological multi-dimensional coordinate system of the sampling point based on the geological property data of different kinds comprises: