Method for identifying ultra-basic rock by using support vector machine with improved geological factors

Abstract

The invention discloses a geological factor improved support vector machine ultra-basic rock identification method. The method comprises the following steps: acquiring aeromagnetic data of a to-be-detected area, carrying out chemical pole calculation and potential field separation, and carrying out chemical pole magnetic anomaly data and vertical first-order derivatives corresponding to residual magnetic anomaly data and analyzing signal amplitudes; performing standardization calculation on the pole-converting magnetic anomaly data, the residual magnetic anomaly data, the vertical first-order derivatives corresponding to the pole-converting magnetic anomaly data and the residual magnetic anomaly data and the analysis signal amplitudes to obtain first standardization data; obtaining radioactivity data of the to-be-detected area and carrying out standardization calculation to obtain second standardization data; taking the first standardized data and the second standardized data as a prediction set; distribution of known ultra-basic rocks in a geological map and radioactive data and magnetic survey data of a corresponding area serve as a sample set, the sample set is input into a genetic algorithm improved support vector machine constrained by geological factors, an optimal vector machine model is obtained through learning and training, and finally the distribution condition of the ultra-basic rocks is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of deep geological mineral prediction, and particularly to a method for identifying ultrabasic rocks by an improved support vector machine of geological factors. Background Art

[0002] Copper and nickel ores are national strategic mineral resources in short supply. Ultrabasic rock bodies are important ore-forming geological bodies for copper and nickel ores. In actual work, most of the identification and delineation of ultrabasic rocks are mainly manual operations, highly dependent on human experience, and with low work efficiency. In the prior art, based on aeromagnetic and radioactive data, simple artificial neural networks and naive support vector machine models are used to identify ultrabasic rocks, which can ensure the accuracy and precision of prediction, but cannot consider the geological distribution characteristics of actual ultrabasic rocks, which are small-scale, sporadic, and have a certain trend, and effectively identify the high magnetic susceptibility ultrabasic rocks that cause magnetic anomalies. Only one physical property characteristic of magnetic susceptibility is considered, and aeromagnetic data contains a large amount of underground geological information, while radioactive data mainly reflects the shallow surface conditions. The combination of the two with the actual ground physical properties needs to be further comprehensively considered.

[0003] Therefore, aiming at the deficiencies and defects existing in the prediction and delineation of ultrabasic rocks by the naive support vector machine model using only aeromagnetic and radioactive data, it is urgent to develop a rock body identification method that comprehensively ensures the prediction accuracy, considers the geological distribution characteristics of ultrabasic rocks, and has higher accuracy and geological interpretability in combining deep geology - shallow surface - ground physical properties, so as to solve the problems of low accuracy, weak geological interpretability, and inconsistent ultrabasic rock distribution characteristics with the actual situation in the process of identifying ultrabasic rocks by the naive support vector machine in aeromagnetic radioactivity due to the lack of geological characteristics and the binding of ground physical properties. Summary of the Invention

[0004] The present invention provides a method for identifying ultrabasic rocks by an improved support vector machine of geological factors, which is used to solve the problems of low accuracy, weak geological interpretability, and inconsistent ultrabasic rock distribution characteristics with the actual situation in the process of identifying ultrabasic rocks by the naive support vector machine in aeromagnetic radioactivity in the prior art due to the lack of geological characteristics and the binding of ground physical properties.

[0005] The present invention provides a method for identifying ultrabasic rocks by an improved support vector machine of geological factors, and the method includes:

[0006] Obtain the aeromagnetic data of the area to be measured; perform reduction to the pole calculation and potential field separation on the aeromagnetic data to obtain reduced-to-the-pole magnetic anomaly data and residual magnetic anomaly data respectively;

[0007] Calculate the vertical first derivative and analytic signal amplitude corresponding to the reduced-to-the-pole magnetic anomaly data and the residual magnetic anomaly data respectively;

[0008] Perform standardized calculations on the reduced pole magnetic anomaly data, residual magnetic anomaly data, vertical first derivative and analytical signal amplitude corresponding to the reduced pole magnetic anomaly data, and vertical first derivative and analytical signal amplitude corresponding to the residual magnetic anomaly data respectively to obtain first standardized data;

[0009] Obtain the radioactive data of the area to be measured and perform standardized calculations to obtain second standardized data;

[0010] Use the first standardized data and the second standardized data as the prediction set;

[0011] Use the distribution of known ultramafic rocks in the geological map and the corresponding radioactive data and magnetic survey data in its corresponding area as the sample set, input it into the improved support vector machine of the genetic algorithm constrained by geological factors, learn and train to obtain the best model parameters of the vector machine to obtain the best vector machine model; Geological factors include area factor, fragmentation factor, and shape factor, which respectively describe the distribution range, fragmentation degree, and strip aspect ratio characteristics of ultramafic rocks;

[0012] Input the prediction set into the best vector machine model to obtain the distribution of ultramafic rocks in the area to be measured.

[0013] Optionally, obtaining the first standardized data and the second standardized data includes:

[0014] Convert the aeromagnetic data to obtain the converted aeromagnetic data; denoise the radioactive data to obtain the denoised radioactive data;

[0015] Perform standardized processing on the converted aeromagnetic data and the denoised radioactive data respectively according to the following formula to obtain the first standardized data and the second standardized data:

[0016]

[0017] where X is the denoised radioactive data or the converted aeromagnetic data, Y is the first standardized data or the second standardized data, a is the preset data lower limit, and b is the preset data upper limit; where, when the denoised radioactive data or the converted aeromagnetic data is not higher than a, the first standardized data or the second standardized data is 0, and when the denoised radioactive data or the converted aeromagnetic data is not lower than b, the first standardized data or the second standardized data is 1.

[0018] Optionally, the preset data lower limit and the preset data upper limit are obtained in the following manner:

[0019] Group the distribution range of the denoised radioactive data or the converted aeromagnetic data, count the number of characteristic data points corresponding to the corresponding group, and obtain the characteristic data distribution curve;

[0020] Obtain the gradient curve of the characteristic data distribution curve, and obtain the maximum and minimum values of the gradient curve;

[0021] Use the denoised radioactive data or the converted aeromagnetic data corresponding to the maximum value as the preset data lower limit, and use the denoised radioactive data or the analyzed aeromagnetic data corresponding to the minimum value as the preset data upper limit.

[0022] Optionally, including the distribution of known ultrabasic rocks in the geological map and their corresponding radioactive data and magnetic survey data in the sample set:

[0023] Divide the distribution of the known ultrabasic rocks into ultrabasic rocks and non-ultrabasic rocks as the labels of the sample set, and combine them with the magnetic survey data and radioactive data in the corresponding area to construct the sample set;

[0024] Obtain the training set and the test set from the sample set.

[0025] Optionally, learning and training to obtain the best model parameters of the support vector machine to obtain the best support vector machine model includes:

[0026] Encode the kernel function parameters and penalty factors in the support vector machine as the population gene strings in the genetic algorithm to obtain the encoded parameters;

[0027] Substitute the encoded parameters into the support vector machine, after learning the training set to obtain the best support vector machine model, calculate the test set data, and count the confusion matrix parameters of the test set data. The confusion matrix parameters include: accuracy, recall rate, and F1 score;

[0028] Use the weighted average of the F1 score and the area factor in the geological factor as the fitness function of the genetic algorithm;

[0029] Calculate the fitness function of each population in the population of the genetic algorithm, and copy, select, cross, and mutate the population to obtain a new population for iterative loop;

[0030] When the fitness function meets the termination condition of the iterative loop, the optimization of the genetic algorithm ends, and the best parameter combination of the kernel function parameters and penalty factors is obtained.

[0031] Optionally, the method further includes:

[0032] The area factor is obtained by the following calculation formula:

[0033]

[0034] Among them, F_area is the calculated area factor, Area_ulramafic is the total area of the identified ultramafic rocks, and Area_all is the total area of the area to be measured;

[0035] The scatter factor is obtained by the following calculation formula:

[0036]

[0037] Among them, F_scatter is the calculated scatter factor, and Num_ulramafic is the total number of the identified ultramafic rocks;

[0038] The shape factor is obtained by the following calculation formula:

[0039]

[0040] Among them, F_shape is the calculated shape factor, length i is the length of the i-th ultramafic rock, and width i is the width of the i-th ultramafic rock.

[0041] Optionally, the method further includes:

[0042] The fitness function of the genetic algorithm is obtained by the following calculation formula:

[0043] Fittness = F1 + F_area * weight,

[0044] Among them, Fittness is the fitness function of the genetic algorithm, F1 is the F1 score, F_area is the area factor, and weight is the weight coefficient.

[0045] Optionally, obtaining the optimal parameter combination of the kernel function parameter and the penalty factor includes:

[0046] The initial population size of the genetic algorithm is 50, and the maximum number of iterations is 200 times;

[0047] The vector machine kernel function is the RBF function, the range of the penalty factor is 0 to 10, and the range of the kernel function parameter γ value is 0 to 10;

[0048] The precision of the genetic algorithm for calculating the penalty factor and the kernel function parameter is both 0.1.

[0049] Optionally, inputting the prediction set into the optimal vector machine model to obtain the distribution of ultramafic rocks in the area to be measured specifically includes:

[0050] Input the prediction set into the optimal vector machine model to obtain the classification of the distribution of ultramafic rocks at corresponding points in the area to be predicted.

[0051] The specific classification is: ultramafic rocks are of class 1, and non-ultramafic rocks are of class 0.

[0052] Optionally, the method further includes: screening the ultramafic rocks obtained after calculating the prediction set one by one according to known surface collected samples, retaining the ultramafic rock masses that do not meet the conditions, and removing the ultramafic rock masses that meet the conditions; wherein, the screening conditions are: the surface lithology is non-ultramafic rock, and the magnetic susceptibility value is greater than 2000×10-5SI.

[0053] In the present invention, the original aeromagnetic anomaly reduction to the pole calculation is performed on the aeromagnetic survey data to obtain the reduced-to-pole magnetic anomaly data, and the potential field separation is performed on the reduced-to-pole magnetic anomaly to obtain the residual magnetic anomaly data, which can eliminate the influence of inclined magnetization of the non-reduced-to-pole magnetic data and the influence of large-scale regional fields, extract weak deep information, and improve the accuracy of data feature extraction. The prediction results conform to the actual geological distribution characteristics of ultramafic rocks, have high geological interpretability, and integrate the measured geoscience data on the surface - shallow subsurface - underground, and can effectively predict ultramafic rocks. Description of the Drawings

[0054] Figure 1 is a flowchart of the method for identifying ultramafic rocks by the support vector machine with improved geological factors in the embodiment of the present invention;

[0055] Figure 2 is a schematic diagram of the distribution result of the identified ultramafic rocks by the support vector machine with improved geological factors in the embodiment of the present invention;

[0056] Figure 3 is a statistical table of the evaluation indexes of the calculation results of the method for identifying ultramafic rocks by the support vector machine with improved geological factors in the embodiment of the present invention;

[0057] Figure 4 is a schematic structural diagram of the system for identifying ultramafic rocks by the support vector machine with improved geological factors in the embodiment of the present invention. Detailed Embodiments

[0058] The present invention will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only for explaining the present invention, rather than limiting the present invention. Additionally, it should be noted that for the sake of description, only the parts related to the present invention are shown in the drawings rather than all the structures.

[0059] It should be understood that in various embodiments herein, the magnitudes of the serial numbers of the above processes do not imply the order of execution, and the order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments herein.

[0060] Embodiment 1

[0061] An embodiment of the present invention provides a method for identifying ultrabasic rocks by a support vector machine improved by geological factors, as Figure 1 shown, the method includes:

[0062] Step 100, obtaining aeromagnetic data of the area to be measured; performing reduction to the pole calculation and potential field separation on the aeromagnetic data to respectively obtain reduced-to-the-pole magnetic anomaly data and residual magnetic anomaly data;

[0063] Step 200, respectively calculating the vertical first derivative and the analytical signal amplitude corresponding to the reduced-to-the-pole magnetic anomaly data and the residual magnetic anomaly data; by performing derivative calculations on the reduced-to-the-pole magnetic anomaly and the residual magnetic anomaly respectively, the vertical first derivative (VDR) of each can be obtained, and high-frequency signals in the magnetic anomaly can be extracted to reflect deep features and improve the resolution of the anomaly data at the same time.

[0064] Step 300, performing standardization calculations on the reduced-to-the-pole magnetic anomaly data, the residual magnetic anomaly data, the vertical first derivative and the analytical signal amplitude corresponding to the reduced-to-the-pole magnetic anomaly data, and the vertical first derivative and the analytical signal amplitude corresponding to the residual magnetic anomaly data respectively to obtain first standardized data;

[0065] Step 400, obtaining radioactive data of the area to be measured and performing standardization calculations to obtain second standardized data; wherein, the radioactive data includes the total channel, uranium element content data, thorium element content data and potassium element content data. The standardization calculation is a rising ridge-shaped distribution standardization calculation, and finally the data values are unified to the range of 0 to 1. By filtering and denoising the anomalies in the airborne radioactive data, the measurement errors and the influence of surface interference objects can be eliminated, shallow information can be reduced, and the accuracy of airborne radioactive anomalies can be improved.

[0066] Step 500, using the first standardized data and the second standardized data as the prediction set;

[0067] Step 600: Use the distribution of known ultramafic rocks in the geological map, along with the corresponding radioactive data and magnetic survey data in their respective regions, as a sample set and input it into the improved support vector machine with geological factor constraints in the genetic algorithm. Through learning and training, obtain the optimal model parameters of the vector machine to acquire the optimal vector machine model. The geological factors include an area factor, a fragmentation factor, and a shape factor, which respectively describe the distribution range, fragmentation degree, and strip aspect ratio characteristics of the ultramafic rocks. In a specific embodiment of the present invention, three geological factors are proposed, namely: an area factor, a fragmentation factor, and a shape factor, to measure the distribution and morphological characteristics of ultramafic rocks. According to the actual geological distribution characteristics of ultramafic rocks, which are small-scale, sporadic, and distributed along a certain trend, the geological factors can reflect the distribution characteristics of ultramafic rocks from three aspects: distribution area, sporadic degree, and aspect ratio along the strike.

[0068] Step 700: Input the prediction set into the optimal vector machine model to obtain the distribution of ultramafic rocks in the area to be measured. By using the same data processing method for ultramafic rocks in the known area and the area to be measured, it is ensured that the feature vectors of the sample set and the prediction set are consistent, and both undergo effective feature data extraction and data standardization processes, ensuring the prediction accuracy of the support vector machine for the area to be measured.

[0069] In the embodiment of the present invention, the original aeromagnetic anomaly reduction to the pole calculation is performed on the aeromagnetic survey data to obtain the reduced-to-pole magnetic anomaly data. The potential field separation is carried out on the reduced-to-pole magnetic anomaly to obtain the residual magnetic anomaly data, which can eliminate the influence of inclined magnetization of the non-reduced-to-pole magnetic data and the influence of large-scale regional fields, extract weak deep information, and improve the accuracy of data feature extraction. The prediction results conform to the actual geological distribution characteristics of ultramafic rocks, have high geological interpretability, and integrate the measured geoscience data on the surface - shallow subsurface - subsurface, and can effectively predict ultramafic rocks.

[0070] In the method for identifying ultramafic rocks using a support vector machine improved by geological factors in a specific embodiment of the present invention, preferably, obtaining the first standardized data and the second standardized data includes:

[0071] Convert the aeromagnetic data to obtain the converted aeromagnetic data; perform denoising processing on the radioactive data to obtain the denoised radioactive data;

[0072] Perform the following formula-based standardization processing on the converted aeromagnetic data and the denoised radioactive data respectively to obtain the first standardized data and the second standardized data:

[0073]

[0074] Wherein, X is the denoised radioactive data or the converted aeromagnetic data, Y is the first standardized data or the second standardized data, a is the preset data lower limit, and b is the preset data upper limit; wherein, when the denoised radioactive data or the converted aeromagnetic data is not higher than a, the first standardized data or the second standardized data is 0, and when the denoised radioactive data or the converted aeromagnetic data is not lower than b, the first standardized data or the second standardized data is 1.

[0075] In a method for identifying ultrabasic rocks by a support vector machine improved by geological factors according to a specific embodiment of the present invention, preferably, the preset data lower limit and the preset data upper limit are obtained by the following method:

[0076] Group the distribution range of the denoised radioactive data or the converted aeromagnetic data, and count the number of characteristic data points corresponding to the corresponding group to obtain a characteristic data distribution curve; for example, divide it into 100 groups in descending order, and count the number of characteristic data points distributed in the corresponding range of the corresponding group to obtain a characteristic data distribution curve.

[0077] Obtain the gradient curve of the characteristic data distribution curve by the difference quotient method, and obtain the maximum value and the minimum value of the gradient curve;

[0078] Take the denoised radioactive data or the converted aeromagnetic data corresponding to the maximum value as the preset data lower limit a, and take the denoised radioactive data or the analyzed aeromagnetic data corresponding to the minimum value as the preset data upper limit b. Through this step, the relevant characteristic data range can be unified, such as the range of 0 to 1, and compared with the standardized calculation formula of the ascending half-Cauchy distribution, the number of selected parameters is reduced; while eliminating the dimension, the convergence speed of machine learning can be improved, and the characteristics of the high-value and low-value distributions of the original data can be extracted, effectively enhancing the ability to identify and obtain weak information in the deep part.

[0079] In a method for identifying ultrabasic rocks by a support vector machine improved by geological factors according to a specific embodiment of the present invention, preferably, taking the distribution of known ultrabasic rocks in the geological map and the corresponding radioactive data and magnetic survey data in the corresponding area as a sample set includes:

[0080] Divide the distribution of the known ultramafic rocks into ultramafic rocks and non-ultramafic rocks, which are used as the labels of the sample set. Combine them with the magnetic measurement data and radioactive data of the corresponding regions to construct a sample set. For example, according to the geological map of the known region, set the label of the ultramafic rock location in the known region to 1, and the label of the non-ultramafic rock location to 0. Then combine the label with the standardized results of radioactive and magnetic measurement data to formulate the sample set. By using the distribution data characteristics of the ultramafic rocks exposed on the surface or inferred from the geological map in the known region as training samples, perform machine learning training on the initial model of the improved support vector machine to construct a method for identifying the distribution of ultramafic rocks from known regions to unknown regions.

[0081] Obtain the training set and the test set from the sample set. Among them, the training set and the test set of the sample set are obtained by the uniform random method, where the training set accounts for 80% and the test set accounts for 10%. By separating the training set and the test set, the purpose is to ensure that the test set for measuring the model performance has never participated in any training process, improving the prediction accuracy and the reliability of the measurement standard.

[0082] For a method for identifying ultramafic rocks by a support vector machine improved by geological factors according to a specific embodiment of the present invention, preferably, learning and training to obtain the best model parameters of the vector machine to obtain the best vector machine model includes:

[0083] Encode the kernel function parameters and penalty factors in the vector machine as the population gene strings in the genetic algorithm to obtain the encoded parameters;

[0084] Substitute the encoded parameters into the vector machine. After learning the training set to obtain the best vector machine model, calculate the test set data and count the confusion matrix parameters of the test set data. The confusion matrix parameters include: accuracy, recall rate, and F1 score; mainly measure the fitting ability of the support vector machine model for known labels and the generalization ability of the data that has not participated in training.

[0085] Take the weighted average of the F1 score and the area factor in the geological factors as the fitness function of the genetic algorithm;

[0086] Calculate the fitness function of each population in the population of the genetic algorithm, and copy, select, cross, and mutate the population to obtain a new population for cyclic iteration;

[0087] When the fitness function meets the cyclic iteration termination condition, the genetic algorithm optimization ends, and the best parameter combination of the kernel function parameters and penalty factors is obtained.

[0088] The combination of two important parameters in the vector machine, the kernel function parameter and the penalty factor, is optimized; the kernel function parameter and the penalty factor are used as the output of the genetic algorithm optimization. After setting the initial population and individuals, the population iterations such as replication, crossover, and mutation are performed, and the best parameter combination of the vector machine can be selected globally. The fitness function of the genetic algorithm is the weighted average of the F1 score of the confusion matrix of the test set and the area factor in the geological factor. It can combine the calculation accuracy and the distribution range of ultrabasic rocks, and the optimized vector machine model parameters can ensure the distribution characteristics of ultrabasic rocks while ensuring the calculation accuracy.

[0089] In some embodiments, the feature vector of the support vector machine model input in this embodiment is consistent with the number of feature vectors of the training set, for example, 8 in this example, including 4 groups of magnetic anomaly data and 4 groups of radioactive data, and the support vector machine model parameters are: the penalty factor ranges from 0 to 10, the kernel function is the RBF Gaussian kernel function, the gamma value ranges from 0 to 10, the category weight is 0.5:1.0, the input 1 label data is oversampled by 5.0 times, and the input 0 label data is undersampled by 0.5 times; the output result is a one-dimensional array, that is, the predicted ultrabasic rock classification labels at the corresponding points in the test area, 1 label is ultrabasic rock, and 0 label is non-ultrabasic rock.

[0090] In some embodiments, the selection of penalty factors, kernel functions, and gamma values ​​can effectively reduce the strictness of the vector machine segmentation calculation, thereby avoiding overfitting and improving calculation stability and efficiency. The combination of the F1 score of the test set confusion matrix and the area factor in the geological factor can effectively control the direction of the genetic algorithm's global search and guide it to select a model parameter combination with high accuracy and high consistency with the actual ultrabasic rock geological characteristics. Setting different weights and sampling methods for data categories can effectively alleviate the problem of serious imbalance in the distribution of actual geological data labels. By increasing the weight of category 1 participating in the calculation and oversampling it, reducing the weight of category 0 calculation and undersampling it, the influence of the predicted target ultrabasic rock on the prediction results is enhanced, and the accuracy of the prediction range is improved.

[0091] In some embodiments, when the vector machine model performs calculations on the prediction set, the distribution classification results of ultrabasic rocks at corresponding points in the test area are obtained, which are Class 0 and Class 1, and the ultrabasic rock results predicted by the vector machine in the study area are obtained.

[0092] In some instances, by combining the support vector machine prediction results with the magnetic susceptibility and lithology of the ground-collected samples in the area to be measured, it is possible to effectively screen out other types of rocks with a certain magnetic susceptibility that can cause a certain magnetic anomaly, such as granite, etc.; the comprehensive application of aeromagnetic anomalies, radioactive data, and surface physical property data reflects multi-source geoscience information from the deep - shallow subsurface to the surface, reduces the non-uniqueness in ultrabasic rock prediction, and improves the prediction accuracy and geological interpretability.

[0093] In a method for identifying ultrabasic rocks using a support vector machine improved by geological factors according to a specific embodiment of the present invention, preferably, the method further includes:

[0094] The area factor is obtained by the following calculation formula:

[0095]

[0096] where F_area is the calculated area factor, Area_ulramafic is the total area of the identified ultrabasic rocks, and Area_all is the total area of the area to be measured;

[0097] The scatter factor is obtained by the following calculation formula:

[0098]

[0099] where F_scatter is the calculated scatter factor, and Num_ulramafic is the total number of the identified ultrabasic rocks;

[0100] The shape factor is obtained by the following calculation formula:

[0101]

[0102] where F_shape is the calculated shape factor, length i is the length of the i-th ultrabasic rock, and width i is the width of the i-th ultrabasic rock.

[0103] In a method for identifying ultrabasic rocks using a support vector machine improved by geological factors according to a specific embodiment of the present invention, preferably, the method further includes:

[0104] The fitness function of the genetic algorithm is obtained by the following calculation formula:

[0105] Fittness = F1 + F_area * weight,

[0106] Wherein, Fitness is the fitness function of the genetic algorithm, F1 is the F1 score, F_area is the area factor, and weight is the weight coefficient. The weight coefficient is generally selected to be 10-15.

[0107] In the method for identifying ultrabasic rocks by a geological factor-improved support vector machine described in a specific embodiment of the present invention, preferably, the optimal parameter combination of the kernel function parameter and the penalty factor is obtained, which includes:

[0108] The initial population size of the genetic algorithm is 50, and the maximum number of iterations is 200;

[0109] The vector machine kernel function is an RBF function, the penalty factor ranges from 0 to 10, and the kernel function parameter γ value ranges from 0 to 10;

[0110] The accuracy of the genetic algorithm in calculating the penalty factor and the kernel function parameter is 0.1.

[0111] The method for identifying ultrabasic rocks by a geological factor-improved support vector machine according to a specific embodiment of the present invention preferably inputs the prediction set into the optimal vector machine model to obtain the distribution of ultrabasic rocks in the test area, specifically including:

[0112] Inputting the prediction set into the optimal vector machine model to obtain the classification of the distribution of ultrabasic rocks at corresponding points in the area to be predicted;

[0113] The classification is specifically: ultrabasic rocks are classified as Class 1, and non-ultrabasic rocks are classified as Class 0. In a specific embodiment, when a sample set is prepared according to the distribution of ultrabasic rocks in a known area, the known area is divided into ultrabasic rocks and non-ultrabasic rocks. For example, based on a geological map of the known area, a label of 1 is set for the ultrabasic rock position in the known area, and a label of 0 is set for the non-ultrabasic rock position, and the label is combined with the standardized results of radioactive and magnetic survey data to prepare a sample set.

[0114] The method for identifying ultrabasic rocks by a support vector machine improved by geological factors described in a specific embodiment of the present invention is preferably further comprising: screening the ultrabasic rocks obtained after the prediction set is calculated, according to known surface samples, retaining the ultrabasic rock bodies that do not meet the conditions, and removing the ultrabasic rock bodies that meet the conditions; wherein the screening conditions are: the surface lithology is non-ultrabasic rock, and the magnetic susceptibility value is greater than 2000×10-5SI. For example, coarse-grained granite with a magnetic susceptibility value of 4000×10-5SI. Specifically, the magnetic susceptibility of ultrabasic rocks generally exceeds 3000×10-5SI, or even 5000×10-5SI, but some granites still have certain magnetism that can cause magnetic anomaly characteristics; comprehensive consideration of surface magnetic susceptibility and lithology, combined with the prediction of ultrabasic rocks, can further screen ultrabasic rocks, thereby removing interference from other types of rock bodies with magnetism.

[0115] In a preferred embodiment, the measurement and evaluation of the prediction results include three parts: various parameters of the confusion matrix calculated by the test set, the values ​​of geological factors, and the recognition accuracy of known ultrabasic rocks.

[0116] In a preferred embodiment, the accuracy and distribution characteristics of the ultrabasic rock are evaluated and measured, specifically including:

[0117] The confusion matrix parameters of the test set calculated by the vector machine model are used to describe the model's fitting ability and generalization ability for known labels;

[0118] The prediction set is the value of the geological factor calculated by the vector machine model, which is used to describe the distribution characteristics of ultrabasic rocks in a certain distribution range and sporadically in a certain direction;

[0119] The accuracy of the ultrabasic rock conditions calculated in the test area and the known ultrabasic rocks is used to describe the degree of consistency between the predicted results and the actual situation;

[0120] The correct rate of the number of known ultrabasic rocks, Rate_num, is obtained by the following calculation formula:

[0121]

[0122] Among them, Num_known_all is the number of all known ultrabasic rocks, and Num_identification is the number of known ultrabasic rocks correctly identified;

[0123] The area accuracy rate Rate_area of ​​the known ultrabasic rocks is obtained by the following calculation formula:

[0124]

[0125] Among them, Area_known_all is the area of all known ultramafic rocks, and Area_identification is the area of correctly identified known ultramafic rocks.

[0126] This embodiment presents the distribution result of ultramafic rocks in a certain area identified according to the method described in Embodiment 1.

[0127] Figure 2 It is a schematic diagram of the distribution result of ultramafic rocks identified by the method of identifying ultramafic rocks with a support vector machine improved by geological factors provided by an embodiment of the present invention. As Figure 2 shown, the abscissa is the position in the X direction, and the ordinate is the position in the Y direction, with both units being m. The aeromagnetic data of the study area is processed and transformed, the radioactive data is denoised, and after ascending ridge-shaped distribution standardization, the labels are delimited in combination with the ultramafic rock distribution in the geological map and sorted as the training set; the training set is input to learn to obtain a support vector machine model improved by a genetic algorithm with appropriate geological factor constraints; the aeromagnetic and radioactive data of the entire study area are similarly processed, transformed, and standardized, and the prediction set is sorted out; the improved support vector machine model obtained by training calculates the probability distribution result of ultramafic rocks in the study area for the prediction set data; according to the magnetic susceptibility and lithology data of the samples collected on the surface of the area to be predicted, the ultramafic rock results predicted by the improved support vector machine are screened one by one to remove the interference of granite causing magnetic anomaly interference, and the final delineation result of ultramafic rocks in the area to be predicted is obtained. Figure 2 The black solid-line delineated polygon in

[0128] The present invention can simultaneously use aeromagnetic and airborne radioactive data, as well as the magnetic susceptibility and lithology of surface samples to constrain the distribution of ultramafic rocks from different physical property perspectives, reflect the comprehensive geological information from deep to shallow surface to ground, and greatly improve the accuracy of prediction and delineation; the data processing, transformation, and standardization process effectively completes the extraction of abnormal characteristics of weak information in the deep part, solves the problems of insufficient resolution ability of the original data and lack of deep information, and at the same time improves the stability and anti-interference ability of machine learning calculations; the calculation process is based on the objective data of aerial surveys, reduces the influence of human experience differences, and improves the efficiency; in the selection of the best parameter combination of the support vector machine model, the parameters of the confusion matrix and geological factor data are combined to comprehensively ensure the accuracy of the calculation and the degree of coincidence of the ultramafic rock distribution characteristics, and improve the geological interpretability of predicting ultramafic rocks.

[0129] Embodiment 3

[0130] Figure 3 It is a statistical table of the evaluation indexes of the calculation results of the method for identifying ultramafic rocks with a support vector machine improved by geological factors provided by an embodiment of the present invention. AsFigure 3 As shown in the figure, the evaluation indexes for predicting the distribution effect of ultrabasic rocks include three parts: the parameters of the confusion matrix calculated from the test set, the values of three geological factors, and the recognition accuracy rate of the actually known ultrabasic rocks.

[0131] The parameters statistically calculated from the confusion matrix of the test set, including precision, recall rate, and F1 score, represent the training and calculation effects of the improved support vector machine model. The higher the values of each index, the stronger the fitting degree and generalization ability of the improved support vector machine for the known ultrabasic rock labels. As can be seen from Figure 3 rows 2-4 of the table in the figure, the improved support vector machine model of the present invention has a relatively high accuracy rate calculated in the area to be predicted. The geological factors include area factor, fragmentation factor, and morphological factor, which measure the morphological characteristics of the predicted ultrabasic rock results and reflect the degree of coincidence between the prediction results and the actual ultrabasic rock distribution, and judge whether it conforms to the characteristics of small-scale, fragmented, and extending along a certain direction of the actual ultrabasic rocks. As can be seen from Figure 3 rows 5-7 of the figure, the distribution characteristics of the ultrabasic rocks calculated by the improved support vector machine model of the present invention in the area to be predicted are relatively consistent. The accuracy rate compared with the actually known ultrabasic rocks represents the effect of the prediction results in practical applications, including the number accuracy rate and the area accuracy rate, which respectively reflect the number and area ratios of the correctly identified predicted ultrabasic rocks. As can be seen from Figure 3 the statistical values in rows 8-9 of the figure, the improved support vector machine model of the present invention has a good practical application effect and a high recognition accuracy rate in the area to be predicted. The above three types of evaluation indexes are carried out from three stages of the calculation process, and evaluate the calculation effect of the method of the present invention in the main stages of machine learning training and calculation, geological interpretation, and practical application.

[0132] Example 4

[0133] Figure 4 As shown in the figure, it is a schematic structural diagram of a system for identifying ultrabasic rocks by an improved support vector machine with geological factors provided by an embodiment of the present invention. Figure 4 As shown in the figure, the system includes a sample construction module 410, a model construction module 420, a prediction and screening module 430, and an evaluation and measurement module 440, where:

[0134] The sample construction module 410 is used to obtain the sample set including aeromagnetic survey and radioactive data and the known ultrabasic rock labels on the geological map, and obtain the prediction set of the area to be predicted including aeromagnetic survey data and radioactive data; the data labels are derived from the actually known lithology on the geological map, where ultrabasic rocks are of type 1 and non-ultrabasic rocks are of type 0; perform denoising calculation on the radioactive data, and then perform ascending ridge-shaped distribution standardization calculation respectively; perform processing and conversion such as polarization, potential field separation, and derivative calculation on the magnetic survey data, and then perform the ascending ridge-shaped distribution standardization calculation.

[0135] The model construction module 420 is used to introduce a genetic algorithm into the support vector machine model to perform optimal selection of support vector machine parameter combinations. The fitness function of the genetic algorithm includes the confusion matrix parameters calculated from the test set and the area factor in the geological factors to balance the prediction accuracy and the distribution characteristics of ultramafic rocks. The training set in the sample set is input into the support vector machine improved by the genetic algorithm constrained by the geological factors for learning and training to obtain the final support vector machine model.

[0136] The prediction and screening module 430 is used to calculate the prediction set data using the support vector machine improved by the genetic algorithm constrained by the geological factors to obtain the classification results of ultramafic rocks in the area to be predicted. The predicted ultramafic rocks are combined with the physical properties and lithology data collected on the ground for screening to remove other rocks that can cause certain magnetic anomalies to obtain the final distribution of ultramafic rocks in the area to be predicted.

[0137] The evaluation and measurement module 440 is used to evaluate and measure the distribution of ultramafic rocks in the prediction stage, including aspects such as the calculation accuracy of the support vector machine model, the distribution characteristics of the predicted ultramafic rocks, and the degree of coincidence with known ultramafic rocks. The various parameters statistically calculated from the confusion matrix of the test set are used to measure the fitting ability and calculation accuracy of the support vector machine model for the labels. The geological factors are used to measure and evaluate whether the predicted ultramafic rocks conform to the true morphology of ultramafic rocks in terms of geological distribution characteristics. The degree of coincidence between the predicted ultramafic rocks and the known ultramafic rocks is measured by the area and number accuracy rates.

[0138] A specific embodiment of the present invention further provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the method described in any one of the above embodiments.

[0139] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0140] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0141] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means that implement the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0142] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are performed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0143] The foregoing description of specific exemplary embodiments of the invention has been presented for purposes of illustration and example. These descriptions are not intended to limit the invention to the precise forms disclosed, and it is apparent that many modifications and variations are possible in light of the above teaching. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the invention and its practical application, thereby enabling those skilled in the art to implement and utilize the invention in its various different exemplary embodiments, as well as various different selections and modifications. The scope of the invention is intended to be defined by the claims and their equivalents.

Claims

1. A method for identifying ultrabasic rocks by using a support vector machine improved by geological factors, characterized in that: The method comprises: Acquire aeromagnetic data of the area to be measured; perform polarization calculation and potential field separation on the aeromagnetic data to obtain polarization magnetic anomaly data and residual magnetic anomaly data respectively; Respectively calculating the vertical first-order derivative and the analytical signal amplitude corresponding to the polarization magnetic anomaly data and the residual magnetic anomaly data; Performing standardization calculation on the polarization magnetic anomaly data, the residual magnetic anomaly data, the vertical first-order derivative and the analytical signal amplitude corresponding to the polarization magnetic anomaly data, and the vertical first-order derivative and the analytical signal amplitude corresponding to the residual magnetic anomaly data, respectively, and obtaining first standardized data; Acquiring radioactivity data of the area to be tested and performing standardized calculation to obtain second standardized data; Using the first standardized data and the second standardized data as a prediction set; The distribution of known ultrabasic rocks in the geological map and the radioactive data and magnetic survey data of the corresponding areas are used as sample sets and input into the genetic algorithm improved support vector machine constrained by geological factors. The optimal model parameters of the vector machine are obtained through learning and training to obtain the optimal vector machine model. The geological factors include area factor, scatter factor and morphological factor, which respectively describe the distribution range, scatter degree and strip-like aspect ratio characteristics of ultrabasic rocks. The prediction set is input into the optimal vector machine model to obtain the distribution of ultrabasic rocks in the area to be tested.

2. The method for identifying ultrabasic rocks by using a geological factor-improved support vector machine according to claim 1, characterized in that: Acquiring the first standardized data and the second standardized data includes: Converting the aeromagnetic data to obtain converted aeromagnetic data; performing denoising on the radioactive data to obtain denoised radioactive data; The converted aeromagnetic data and the denoised radioactive data are respectively subjected to standardization processing according to the following formula to obtain the first standardized data and the second standardized data: Wherein, X is the denoised radioactivity data or the converted aeromagnetic data, Y is the first standardized data or the second standardized data, a is the preset data lower limit, and b is the preset data upper limit; wherein, when the denoised radioactivity data or the converted aeromagnetic data is not higher than a, the first standardized data or the second standardized data is 0, and when the denoised radioactivity data or the converted aeromagnetic data is not lower than b, the first standardized data or the second standardized data is 1.

3. The method for identifying ultrabasic rocks by using a geological factor-improved support vector machine according to claim 2, characterized in that: The preset data lower limit and the preset data upper limit are obtained in the following manner: Grouping the distribution range of the denoised radioactive data or the converted aeromagnetic data, counting the number of characteristic data points corresponding to the corresponding groups, and obtaining a characteristic data distribution curve; Obtaining a gradient curve of the characteristic data distribution curve, and obtaining a maximum value and a minimum value of the gradient curve; The denoised radioactivity data or converted aeromagnetic data corresponding to the maximum value is used as the preset data lower limit, and the denoised radioactivity data or analyzed aeromagnetic data corresponding to the minimum value is used as the preset data upper limit.

4. The method for identifying ultrabasic rocks by using a geological factor-improved support vector machine according to claim 1, characterized in that: The distribution of known ultrabasic rocks in geological maps and the radioactive data and magnetic survey data of the corresponding areas are included as sample sets: The distribution of the known ultrabasic rocks is divided into ultrabasic rocks and non-ultrabasic rocks, which are used as labels for the sample set, and combined with the magnetic survey data and radioactive data of the corresponding area to construct a sample set; A training set and a test set are obtained from the sample set.

5. The method for identifying ultrabasic rocks by using a geological factor-improved support vector machine according to claim 4, characterized in that: Learning and training to obtain the best model parameters of the vector machine to obtain the best vector machine model includes: Encoding the kernel function parameters and penalty factors in the vector machine as population gene strings in a genetic algorithm to obtain encoding parameters; Bringing the encoding parameters into the vector machine, and after learning the training set to obtain the best vector machine model, calculating the test set data, and counting the confusion matrix parameters of the test set data, the confusion matrix parameters include: precision, recall rate and F1 score; Taking the weighted average of the F1 score and the area factor in the geological factor as the fitness function of the genetic algorithm; Calculating the fitness function of each population in the genetic algorithm population, and replicating, selecting, crossing, and mutating the population to obtain a new population for cyclic iteration; When the fitness function meets the loop iteration termination condition, the genetic algorithm optimization ends and the optimal parameter combination of the kernel function parameter and the penalty factor is obtained.

6. The method for identifying ultrabasic rocks by using a geological factor-improved support vector machine according to claim 5, characterized in that: The method further comprises: The area factor is obtained by the following calculation formula: Among them, F_area is the calculated area factor, Area_ulramafic is the total area of ​​the identified ultrabasic rocks, and Area_all is the total area of ​​the area to be tested; The scatter factor is obtained by the following calculation formula: Among them, F_scatter is the calculated scatter factor, and Num_ulramafic is the total number of ultrabasic rocks identified; The morphological factor is obtained by the following calculation formula: Among them, F_shape is the calculated shape factor, length i is the length of the ith ultrabasic rock, width i is the width of the ith ultrabasic rock.

7. The method for identifying ultrabasic rocks by using a geological factor-improved support vector machine according to claim 5, characterized in that: The method further comprises: The fitness function of the genetic algorithm is obtained by the following calculation formula: Fittness=F1+F_area*weight, Among them, Fitness is the fitness function of the genetic algorithm, F1 is the F1 score, F_area is the area factor, and weight is the weight coefficient.

8. The method for identifying ultrabasic rocks by using a geological factor-improved support vector machine according to claim 5, characterized in that: The best parameter combinations for kernel function parameters and penalty factors include: The initial population size of the genetic algorithm is 50, and the maximum number of iterations is 200; The vector machine kernel function is an RBF function, the penalty factor ranges from 0 to 10, and the kernel function parameter γ value ranges from 0 to 10; The accuracy of the penalty factor and kernel function parameter calculated by the genetic algorithm is 0.

1.

9. The method for identifying ultrabasic rocks by using a geological factor-improved support vector machine according to claim 1, characterized in that: Inputting the prediction set into the optimal vector machine model to obtain the distribution of ultrabasic rocks in the test area specifically includes: Inputting the prediction set into the optimal vector machine model to obtain the classification of the distribution of ultrabasic rocks at corresponding points in the area to be predicted; The classification is specifically: ultrabasic rocks are Class 1, and non-ultrabasic rocks are Class 0.

10. The method for identifying ultrabasic rocks by using a geological factor-improved support vector machine according to claim 1, characterized in that: The method further comprises: screening the ultrabasic rocks obtained after calculating the prediction set one by one according to known surface samples, retaining the ultrabasic rock bodies that do not meet the conditions, and removing the ultrabasic rock bodies that meet the conditions; wherein the screening conditions are: the surface lithology is non-ultrabasic rock, and the magnetic susceptibility value is greater than 2000×10 -5 SI.

Images