Soil pollution accurate positioning method based on geophysical exploration technology

Abstract

The invention discloses a soil pollution accurate positioning method based on a geophysical exploration technology. The method comprises the following steps: S1, analyzing geologic features of a research area: analyzing the geologic features of the research area; s2, constructing a geological model; s3, measuring line arrangement and data acquisition; s4, data processing: obtaining a two-dimensional resistivity and polarizability image and an internal structure image of ground penetrating radar data by adopting least square inversion, training an anomalous body recognition model by utilizing a convolutional neural network, automatically recognizing anomalous body features in ERT and GPR data, and automatically screening out potential pollution areas; s5, data explanation: performing comprehensive analysis on inversion results of ERT and GPR, constructing an intelligent explanation model, automatically adjusting explanation parameters by the intelligent explanation model according to data features, and providing a plurality of explanation schemes; and S6, establishing a three-dimensional electrical structure model: establishing the three-dimensional electrical structure model of the research area by adopting a Kriging interpolation method according to the ERT inversion result and the earth surface coordinate data.

Description

Technical Field

[0001] The present invention relates to the technical field of soil pollution control, and particularly relates to a precise soil pollution location method based on geophysical exploration technology. Background Art

[0002] In the past 30 years, with the rapid development of social economy and high-intensity human activities, the problem of soil environmental safety in China has become increasingly prominent, attracting wide social attention. The precise soil pollution location method based on geophysical exploration technology can achieve full coverage of the site and pollution detection of the underground profile, presenting the potential pollution situation of the site from the perspectives of "surface" and "line". The successful application of this method has introduced geophysical exploration technology into the fields of soil pollution investigation and prevention, expanding its application scope. At the same time, the continuous development and improvement of this method will also promote the further innovation and development of geophysical exploration technology in the field of environmental protection.

[0003] However, in geophysical exploration, a large amount of data needs to be collected, processed, and interpreted. The traditional manual interpretation method is not only time-consuming and laborious, but also easily affected by the experience and subjective judgment of interpreters, resulting in inconsistency and uncertainty of interpretation results. Therefore, a precise soil pollution location method based on geophysical exploration technology is proposed. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings existing in the prior art that the traditional manual interpretation method is not only time-consuming and laborious, but also easily affected by the experience and subjective judgment of interpreters, resulting in inconsistency and uncertainty of interpretation results, and to propose a precise soil pollution location method based on geophysical exploration technology.

[0005] In order to achieve the above purpose, the present invention adopts the following technical scheme: A precise soil pollution location method based on geophysical exploration technology, comprising the following steps: S1: Analysis of geological characteristics of the study area: Analyze the geological characteristics of the study area; S2: Construction of a geological model: On the basis of the analysis of geological characteristics, construct a geological model according to geological exploration data and previous research results; S3: Layout of survey lines and data collection: Layout survey lines of the high-density resistivity method (ERT) and the ground penetrating radar method (GPR) in the study area for data collection; S4: Data Processing: Preprocess the collected high-density resistivity data and ground penetrating radar data, and use least squares inversion to obtain two-dimensional resistivity and polarization rate images, as well as internal structure images of ground penetrating radar data. Combine the results of geological drilling and geological surveys, and use a convolutional neural network to train an anomaly recognition model to automatically identify the anomaly characteristics in ERT and GPR data and automatically screen out potential polluted areas; S5: Data Interpretation: Conduct a comprehensive analysis of the inversion results of ERT and GPR to determine the location, scale, and type of underground bad geological bodies. Combine multidisciplinary knowledge such as geology and geophysics to construct an intelligent interpretation model. The intelligent interpretation model automatically adjusts the interpretation parameters according to data characteristics, provides multiple interpretation schemes, and evaluates the rationality of each scheme to help interpreters make more accurate judgments; S6: Establishment of 3D Electrical Structure Model: Based on the ERT inversion results and surface coordinate data, use Kriging interpolation to establish a 3D electrical structure model of the study area. The 3D electrical structure model is used to display the spatial distribution characteristics of underground bad geological bodies.

[0006] The above further includes: Furthermore, in S1, the geological features include formation lithology, geological structure, and karst fracture development. Understand the distribution, thickness, and lithology of the strata through geological surveys, conduct chemical composition analysis, mineral identification, and rock mechanical property tests on the collected rock samples through laboratory tests; Through field geological surveys, draw the geological structure map of the study area, use remote sensing image data to identify and analyze the geological structure characteristics of the study area, and use geophysical exploration to detect underground geological structures; Understand the karst landforms, karst forms, and fracture development through geological surveys, understand the occurrence conditions and migration laws of groundwater through hydrogeological exploration, and use geophysical exploration to detect the development of underground karst fractures.

[0007] Furthermore, in S3, the survey lines of the high-density resistivity method (ERT) and the ground penetrating radar method (GPR) are arranged in the study area. The specific steps are as follows: Analysis of the study area: It is necessary to comprehensively analyze the terrain undulation, geological features, and location of potential pollution sources in the study area, which includes consulting geological maps, topographic maps, and existing environmental monitoring data to understand the geological structure, formation lithology, soil type, and distribution of potential pollution sources in the study area; Survey line design: Based on the results of the study area analysis, design a survey line layout plan. The survey lines cover the entire study area, especially near potential pollution sources and in areas with complex geological features, and the survey lines are densified to improve the detection accuracy. The direction of the survey lines should be determined according to the strike of the geological structure and the migration direction of potential pollutants; On-site layout: When laying out the survey lines on-site, it is necessary to ensure that the positions of the survey lines are accurate and the markings are clear. At the same time, the influence of terrain undulations on the layout of the survey lines needs to be considered, and adjustments should be made if necessary to ensure the continuity and integrity of the survey lines.

[0008] Further, in S3, the specific steps of the data acquisition are as follows: Instrument selection: According to the detection purpose and geological characteristics, select geophysical exploration instruments. Commonly used instruments include electromagnetic induction instruments, high-density resistivity instruments, ground-penetrating radars, etc. These instruments can measure the physical properties of underground media, such as resistivity, dielectric constant, etc., so as to reflect the existence and distribution of soil pollution. Parameter setting: After instrument selection, set the detection parameters, including emission frequency, receiving sensitivity, and sampling interval. The parameter settings should ensure that the collected data has high quality and integrity. Data acquisition: According to the survey line layout plan, conduct data acquisition along the survey lines. During the acquisition process, it is necessary to ensure the stability of the instrument and standard operation, and avoid the influence of external interference on the data. At the same time, it is necessary to record the environmental conditions during acquisition, such as temperature, humidity, etc., for subsequent data processing and correction.

[0009] Data quality inspection: After the acquisition is completed, conduct quality inspection on the data, which includes checking the integrity, continuity, and consistency of the data, etc. For abnormal data or missing data, processing or re-measurement is required to ensure the reliability of the data.

[0010] Further, in S4, the steps of obtaining the two-dimensional resistivity and polarization rate images and the internal structure image of the ground-penetrating radar data by using the least squares inversion are as follows: Establish an inversion model: According to the geophysical exploration principle, establish an inversion model for high-density resistivity and ground-penetrating radar data. The inversion model reflects the relationship between the physical properties of underground media (such as resistivity, dielectric constant, etc.) and the observed data. Adopt the least squares inversion method: Using the least squares principle, conduct inversion calculations on the observed data to obtain the two-dimensional resistivity and polarization rate images and the internal structure image of the ground-penetrating radar data. The least squares inversion method solves the model parameters by minimizing the error between the observed data and the model prediction data. Let the observed data be , the model prediction data be , the model parameters be , then the objective function of the least squares inversion is expressed as: , by solving the that minimizes value, the inversion result can be obtained.

[0011] Further, in S4, by combining geological drilling and geological survey results, an anomaly recognition model is trained using a convolutional neural network to automatically identify anomaly features in ERT and GPR data and automatically screen out potential contaminated areas, including the following steps: Data preparation: Comprehensively analyze the geological drilling and geological survey results and the inversion results of ERT and GPR to determine the accurate location and type of underground bad geological bodies, label the known anomaly features in the ERT and GPR data to form a training dataset; Model training: Use a convolutional neural network to train the training dataset to obtain an anomaly recognition model; Model verification and optimization: Use the verification dataset to verify the model, evaluate the recognition accuracy and generalization ability of the model, and optimize and adjust the model according to the verification results; Automatically identify anomaly features: Use the trained anomaly recognition model to automatically identify anomaly features in the ERT and GPR data; Automatically screen potential contaminated areas: According to the recognition results of the anomaly features, automatically screen out potential contaminated areas.

[0012] Further, in S5, by combining multidisciplinary knowledge such as geology and geophysics, an intelligent interpretation model is constructed, including the following steps; Data preprocessing: Preprocess the original data of ERT and GPR, including noise elimination, terrain correction, etc., to improve the data quality; Feature extraction: Extract feature information from the preprocessed data, such as resistivity values, radar wave reflection intensity, frequency, etc., and the feature information will be used to construct the intelligent interpretation model; Model construction: Combine multidisciplinary knowledge such as geology and geophysics, and use a support vector machine to construct an intelligent interpretation model. The intelligent interpretation model uses SMO to find the optimal separating hyperplane, and the form of the separating hyperplane is , where, is the weight vector, is the input feature vector, is the bias term. The SMO is an iterative algorithm for solving the dual problem of the support vector machine. In each iteration, the SMO selects two Lagrange multipliers for update to maintain satisfaction of the KKT conditions, which is expressed as: Calculate the gradient of L(α): Calculate the gradients of the Lagrangian function L(α) with respect to and , and the Lagrangian function L(α) is defined as where, is the kernel function, which is used to calculate the similarity between two samples. The gradients with respect to and are expressed as , where λ is the reciprocal of the regularization parameter C, i.e., λ = 1 / C; Solve the quadratic programming problem: Simplify the problem to a quadratic programming problem to find the new and values that will maximize L(α) subject to the constraints. The quadratic programming problem is expressed as , where the elements of the B matrix, are related to the kernel function and the labels, c is a constant. Solving this quadratic programming problem, we can obtain the new and values; Update and : Use the new values found by the solver to update and . The updated values must satisfy the constraints of the SVM, and the constraints are expressed as ; Calculate the new value of b: After updating and , select a support vector (i.e., or sample with a value greater than 0) to calculate the new value of b, , where w is the weight vector, expressed as . Select a support vector such that >0, then we have . Substitute the expression of w into the above formula to obtain the new value of b; Check the KKT conditions to check whether the updated , and b satisfy the KKT conditions. If not, continue the iteration; If satisfied, the algorithm converges and the optimal separation hyperplane is found. The intelligent interpretation model automatically adjusts the interpretation parameters according to the data characteristics and provides multiple interpretation schemes. The interpretation schemes include different combinations of the location, shape, and rock type of the geological body;

[0013] Scheme evaluation: Evaluate the multiple interpretation schemes provided by the intelligent interpretation model; Optimization and iteration: According to the evaluation results, optimize and iterate the intelligent interpretation model. By continuously optimizing the model parameters and algorithms, improve the interpretation ability and accuracy of the model.

[0014] Furthermore, in S6, the establishment of the three-dimensional electrical structure model of the study area by using the Kriging interpolation method includes the following steps: Data collection and processing: Collect the ERT inversion results and surface coordinate data, and preprocess the collected data, including noise removal, outlier detection and processing, data format conversion, etc., to ensure the accuracy and consistency of the data; Interpolation method selection: The Kriging interpolation method is adopted. The key formulas in the Kriging interpolation method include the semivariogram function and the Kriging interpolation formula. The semivariogram function is used to describe the spatial autocorrelation between data points, and the formula is , where is the semivariance value, is the distance between two points, is the number of point pairs with a distance of , and are the observed values at positions and respectively. The Kriging interpolation formula is used to estimate the value of an unknown data point based on known data points, and the formula is , where is the estimated value at the unknown data point , are the weight coefficients, is the observed value at the known data point . The weight coefficients are determined by minimizing the variance of the prediction error;

[0015] 3D model construction: Using the ERT inversion results and surface coordinate data, combined with the Kriging interpolation method, a 3D electrical property structure model of the study area is constructed, expanding the 2D data into 3D space to form a 3D electrical property structure image; Model verification and optimization: The constructed 3D model is verified to ensure that it can accurately reflect the spatial distribution characteristics of underground bad geological bodies, which is achieved by comparing with the actual geological exploration results. According to the verification results, the model is optimized, such as adjusting interpolation parameters, adding data points, etc., to improve the accuracy and reliability of the model; Result presentation: The optimized 3D model is presented in an intuitive way, such as 3D visualization images, geological maps, etc.

[0016] The present invention has the following beneficial effects: In the present invention, two geophysical exploration techniques, namely the high-density resistivity method (ERT) and the ground penetrating radar method (GPR), are combined. These two techniques have their own advantages and complement each other, jointly improving the accuracy of soil pollution location. By combining multidisciplinary knowledge such as geology and geophysics to construct an intelligent interpretation model and using the results of geological drilling and geological surveys to train an anomaly recognition model with a convolutional neural network, the interpretation parameters can be automatically adjusted, multiple possible interpretation schemes can be provided, and the rationality of each scheme can be evaluated, thus significantly improving the efficiency and accuracy of data interpretation. Description of the drawings

[0017] Figure 1It is a step diagram of a method for precise positioning of soil pollution based on geophysical exploration technology proposed by the present invention. Specific embodiments

[0018] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0019] Please refer to Figure 1 As shown, the present invention is a method for precise positioning of soil pollution based on geophysical exploration technology, including the following steps: S1: Analysis of geological characteristics of the study area: Analyze the geological characteristics of the study area; S2: Construction of a geological model: Based on the analysis of geological characteristics, construct a geological model according to geological exploration data and previous research results; S3: Layout of survey lines and data collection: Layout survey lines of the electrical resistivity tomography (ERT) and ground penetrating radar (GPR) methods in the study area for data collection; S4: Data processing: Preprocess the collected high-density resistivity data and ground penetrating radar data, obtain two-dimensional resistivity and polarization rate images by least squares inversion, as well as internal structure images of the ground penetrating radar data. Combining the results of geological drilling and geological surveys, use a convolutional neural network to train an anomaly recognition model to automatically identify the anomaly characteristics in the ERT and GPR data, and automatically screen out potential pollution areas; S5: Data interpretation: Conduct a comprehensive analysis of the inversion results of ERT and GPR to determine the location, scale, and type of underground bad geological bodies. Combining multidisciplinary knowledge such as geology and geophysics, construct an intelligent interpretation model. The intelligent interpretation model automatically adjusts the interpretation parameters according to data characteristics, provides multiple interpretation schemes, and evaluates the rationality of each scheme to help interpreters make more accurate judgments; S6: Establishment of a three-dimensional electrical structure model: According to the ERT inversion results and surface coordinate data, use the Kriging interpolation method to establish a three-dimensional electrical structure model of the study area, and the three-dimensional electrical structure model is used to display the spatial distribution characteristics of underground bad geological bodies.

[0020] In one embodiment, for the above-mentioned S1, in S1, the geological features include formation lithology, geological structure, and the development of karst fissures. Through geological surveys, the distribution, thickness, and lithology of the strata are understood. Chemical composition analysis, mineral identification, and rock mechanical property tests are carried out on the collected rock samples through laboratory tests; through field geological surveys, geological structure maps of the study area are drawn. Using remote sensing image data, the geological structure features of the study area are identified and analyzed, and geophysical exploration is used to detect underground geological structures; through geological surveys, karst landforms, karst forms, and the development of fissures are understood, and through hydrogeological exploration, the occurrence conditions and migration laws of groundwater are understood, and geophysical exploration is used to detect the development of underground karst fissures.

[0021] In one embodiment, for the above-mentioned S3, in S3, survey lines of the electrical resistivity tomography (ERT) and ground penetrating radar (GPR) are arranged in the study area. The specific steps are as follows: Analysis of the study area: It is necessary to comprehensively analyze the terrain undulation, geological features, and the locations of potential pollution sources in the study area. This includes consulting geological maps, topographic maps, and existing environmental monitoring data to understand the geological structure, formation lithology, soil type, and the distribution of potential pollution sources in the study area; Survey line design: Based on the results of the analysis of the study area, a survey line layout plan is designed. The survey lines cover the entire study area. Especially near potential pollution sources and in areas with complex geological features, the survey lines are densified to improve the detection accuracy. The direction of the survey lines should be determined according to the strike of the geological structure and the migration direction of potential pollutants; On-site layout: The survey lines are laid out on the ground. It should be ensured that the positions of the survey lines are accurate and the markings are clear. At the same time, the influence of terrain undulation on the layout of the survey lines needs to be considered, and adjustments are made if necessary to ensure the continuity and integrity of the survey lines.

[0022] In one embodiment, for the above-mentioned S3, in S3, the specific steps of data collection are as follows: Instrument selection: According to the detection purpose and geological features, geophysical detection instruments are selected. Commonly used instruments include electromagnetic induction instruments, high-density resistivity instruments, ground penetrating radars, etc. These instruments can measure the physical properties of underground media, such as resistivity, dielectric constant, etc., so as to reflect the existence and distribution of soil pollution; Parameter setting: After the instrument is selected, detection parameters are set, including emission frequency, receiving sensitivity, and sampling interval. The parameter settings should ensure that the collected data has high quality and integrity; Data collection: According to the survey line layout plan, data is collected along the survey lines. During the collection process, it should be ensured that the instrument is stable and the operation is standardized to avoid the influence of external interference on the data. At the same time, the environmental conditions during collection, such as temperature, humidity, etc., need to be recorded for subsequent data processing and correction.

[0023] Data quality inspection: After data collection, data quality inspection is carried out, which includes checking data integrity, continuity, consistency, etc. For abnormal or missing data, processing or re-survey is required to ensure data reliability.

[0024] In one embodiment, for S4, in S4, obtaining two-dimensional resistivity and polarizability images and internal structure images of ground penetrating radar data by using least squares inversion includes the following steps: Establish an inversion model: According to geophysical exploration principles, establish an inversion model for high-density resistivity and ground penetrating radar data. The inversion model reflects the relationship between the physical properties of underground media (such as resistivity, dielectric constant, etc.) and the observed data; Adopt the least squares inversion method: Using the least squares principle, perform inversion calculations on the observed data to obtain two-dimensional resistivity and polarizability images and internal structure images of ground penetrating radar data. The least squares inversion method solves for model parameters by minimizing the error between the observed data and the model predicted data. Let the observed data be , the model predicted data be , and the model parameters be , then the objective function of the least squares inversion is expressed as: , by solving for the value that minimizes , the inversion result can be obtained.

[0025] In one embodiment, for the above S4, in S4, combining geological drilling and geological survey results, training an anomaly recognition model using a convolutional neural network, and automatically identifying anomaly features in ERT and GPR data and automatically screening out potential contaminated areas includes the following steps: Data preparation: Comprehensively analyze the results of geological drilling and geological survey with the inversion results of ERT and GPR to determine the accurate location and type of underground bad geological bodies, and label the known anomaly features in ERT and GPR data to form a training dataset; Model training: Use a convolutional neural network to train the training dataset to obtain an anomaly recognition model; Model verification and optimization: Use the verification dataset to verify the model, evaluate the recognition accuracy and generalization ability of the model, and optimize and adjust the model according to the verification results; Automatically identify anomaly features: Use the trained anomaly recognition model to automatically identify anomaly features in ERT and GPR data; Automatically screen potential contaminated areas: According to the recognition results of anomaly features, automatically screen out potential contaminated areas.

[0026] In one embodiment, for the above S5, in S5, combining multidisciplinary knowledge such as geology and geophysics to construct an intelligent interpretation model includes the following steps; Data preprocessing: Preprocess the original data of ERT and GPR, including noise elimination, terrain correction, etc., to improve data quality; Feature extraction: Extract feature information from the preprocessed data, such as resistivity values, radar wave reflection intensity, frequency, etc., and the feature information will be used to construct an intelligent interpretation model; Model construction: Combining multidisciplinary knowledge such as geology and geophysics, use a support vector machine to construct an intelligent interpretation model. The intelligent interpretation model uses SMO to find the optimal separating hyperplane, and the form of the separating hyperplane is , where, is the weight vector, is the input feature vector, is the bias term. The SMO is an iterative algorithm for solving the dual problem of the support vector machine. In each iteration, the SMO selects two Lagrange multipliers for update, keeping the KKT conditions satisfied, which is expressed as: Calculate the gradient of L(α): Calculate the gradients of the Lagrangian function L(α) with respect to and , and the Lagrangian function L(α) is defined as where, is the kernel function, which is used to calculate the similarity between two samples. The gradients with respect to and are expressed as , where λ is the reciprocal of the regularization parameter C, i.e., λ = 1 / C; Solve the quadratic programming problem: Simplify the problem into a quadratic programming problem to find the new and values, which will make L(α) maximize under the constraint conditions. The quadratic programming problem is expressed as , where the elements of the B matrix, are related to the kernel function and the labels, and c is a constant. Solving this quadratic programming problem can obtain the new and values; Update and : Use the new values found by the solver to update and , and the updated values must satisfy the constraint conditions of the SVM, and the constraint conditions are expressed as ; Calculate the new value of b: After updating and , select a support vector (i.e., or Calculate the new value of b using samples greater than 0. , where w is the weight vector, expressed as , select a support vector such that >0, then there is Substitute the expression of w into the above formula to obtain the new value of b; check the KKT conditions and check whether the updated , and b satisfy the KKT conditions. If not, continue the iteration; if satisfied, the algorithm converges, find the optimal separation hyperplane, and the intelligent interpretation model automatically adjusts the interpretation parameters according to the data characteristics and provides multiple interpretation schemes, and the interpretation schemes include different combinations of the location, shape, and rock type of the geological body;

[0027] Scheme evaluation: Evaluate the multiple interpretation schemes provided by the intelligent interpretation model; Optimization and iteration: According to the evaluation results, optimize and iterate the intelligent interpretation model, and improve the interpretation ability and accuracy of the model by continuously optimizing the model parameters and algorithms.

[0028] In one embodiment, for the above S6, in S6, the method of establishing a three-dimensional electrical structure model of the study area using Kriging interpolation method includes the following steps: Data collection and processing: Collect ERT inversion results and surface coordinate data, and preprocess the collected data, including removing noise, detecting and processing outliers, data format conversion, etc., to ensure the accuracy and consistency of the data; Interpolation method selection: Use the Kriging interpolation method. The key formulas in the Kriging interpolation method include the semi-variance function and the Kriging interpolation formula. The semi-variance function is used to describe the spatial autocorrelation between data points, and the formula is , where is the semi-variance value, is the distance between two points, is the number of point pairs with a distance of , and are the observed values at positions and respectively. The Kriging interpolation formula is used to infer the value of an unknown data point based on known data points, and the formula is , where is the inferred value at the unknown data point , is the weight coefficient, is the observed value at the known data point . The weight coefficient is determined by minimizing the variance of the prediction error; ​

[0029] 3D model construction: Using the ERT inversion results and surface coordinate data, combined with the Kriging interpolation method, construct a 3D electrical structure model of the study area, extend the 2D data to the 3D space, and form a 3D electrical structure image; Model verification and optimization: Verify the constructed 3D model to ensure that it can accurately reflect the spatial distribution characteristics of underground bad geological bodies, which is achieved by comparing with the actual geological exploration results. Optimize the model according to the verification results, adjust the interpolation parameters, increase data points, etc., to improve the accuracy and reliability of the model; Result presentation: Present the optimized 3D model in an intuitive way, such as 3D visualization images, geological maps, etc.

[0030] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A soil pollution accurate positioning method based on geophysical detection technology, characterized in that: The following steps are involved: S1: Analysis of geological characteristics of the study area: Analyze the geological characteristics of the study area; S2: Constructing geological model: Based on the analysis of geological characteristics, geological exploration data and previous research results, constructing geological model; S3: Survey line layout and data collection: Arrange survey lines of high-density resistivity method (ERT) and ground penetrating radar method (GPR) in the study area and collect data; S4: Data processing: Preprocess the collected high-density resistivity data and ground penetrating radar data, use least squares inversion to obtain two-dimensional resistivity and polarizability images, as well as internal structure images of ground penetrating radar data, combine geological drilling and geological survey results, use convolutional neural network to train anomaly recognition model, automatically identify anomaly features in ERT and GPR data, and automatically screen out potential contaminated areas; S5: Data interpretation: Comprehensively analyze the inversion results of ERT and GPR to determine the location, scale and type of underground adverse geological bodies, and build an intelligent interpretation model based on multidisciplinary knowledge. The intelligent interpretation model automatically adjusts the interpretation parameters according to data characteristics, provides multiple interpretation schemes, and evaluates the rationality of each scheme; S6: Establishment of three-dimensional electrical structure model: Based on the ERT inversion results and surface coordinate data, the Kriging interpolation method is used to establish a three-dimensional electrical structure model of the study area. The three-dimensional electrical structure model is used to display the spatial distribution characteristics of underground adverse geological bodies.

2. According to claim 1, a soil pollution accurate positioning method based on geophysical detection technology is characterized in that: In S1, the geological characteristics include stratum lithology, geological structure and karst fissure development. Through geological surveys, the distribution, thickness and lithology of the strata are understood. Through laboratory tests, the collected rock samples are subjected to chemical composition analysis, mineral identification and rock mechanical property testing. Through field geological surveys, the geological structure map of the study area is drawn. Using remote sensing image data, the geological structure characteristics of the study area are identified and analyzed. Geophysical exploration is used to detect underground geological structures. Through geological surveys, the karst landforms, karst morphology and fissure development are understood. Through hydrogeological exploration, the occurrence conditions and migration laws of groundwater are understood. Geophysical exploration is used to detect the development of underground karst fissures.

3. The method for accurately locating soil pollution based on geophysical detection technology according to claim 1 is characterized in that: In S3, the survey lines of the high-density resistivity method (ERT) and the ground penetrating radar method (GPR) are arranged in the study area, and the specific steps are: Study area analysis: A comprehensive analysis of the topographic relief, geological characteristics and location of potential pollution sources in the study area is required; Survey line design: Based on the results of the study area analysis, a survey line layout plan is designed. The survey line covers the entire study area, especially near potential pollution sources and areas with complex geological features, and the survey line is intensified. The direction of the survey line should be determined according to the trend of the geological structure and the migration direction of potential pollutants; On-site layout: Layout of survey lines on site.

4. The method for accurately locating soil pollution based on geophysical detection technology according to claim 3 is characterized in that: In S3, the specific steps of data collection are: Instrument selection: Select geophysical exploration instruments according to the exploration purpose and geological characteristics; Parameter setting: After selecting the instrument, set the detection parameters, including the transmission frequency, receiving sensitivity and sampling interval; Data collection: Data collection is carried out along the survey line according to the survey line layout plan; Data quality check: After the collection is completed, the data quality is checked.

5. The method for accurately locating soil pollution based on geophysical detection technology according to claim 1 is characterized in that: In S4, the least squares inversion is used to obtain a two-dimensional resistivity and polarizability image, and an internal structure image of the ground penetrating radar data, including the following steps: Establishing an inversion model: Based on the principle of geophysical detection, an inversion model of high-density resistivity and ground penetrating radar data is established, wherein the inversion model reflects the relationship between the physical properties of the underground medium and the observed data; The least squares inversion method is used: the least squares principle is used to invert the observed data to obtain two-dimensional resistivity and polarizability images, as well as the internal structure images of the ground penetrating radar data. The least squares inversion method solves the model parameters by minimizing the error between the observed data and the model prediction data. Assume that the observed data is , the model predicts the data to be , the model parameters are , then the objective function of the least squares inversion is expressed as: , by solving Smallest value, and you can get the inversion result.

6. The soil pollution accurate positioning method based on geophysical detection technology according to claim 1 is characterized in that: In S4, the method combines the geological drilling and geological survey results, uses a convolutional neural network to train an anomaly recognition model, automatically identifies anomaly features in ERT and GPR data, and automatically screens out potential contaminated areas, including the following steps: Data preparation: Comprehensively analyze the geological drilling and geological survey results with the inversion results of ERT and GPR to determine the exact location and type of underground adverse geological bodies, annotate the known abnormal body features in ERT and GPR data, and form a training data set; Model training: Use convolutional neural network to train the training data set to obtain an abnormal body recognition model; Model verification and optimization: Use the verification data set to verify the model, evaluate the recognition accuracy and generalization ability of the model, and optimize the model according to the verification results; Automatically identify abnormal body features: Use the trained abnormal body recognition model to automatically identify abnormal body features in ERT and GPR data; Automatic screening of potential contaminated areas: Automatically screen out potential contaminated areas based on the recognition results of abnormal body features.

7. The method for accurately locating soil pollution based on geophysical detection technology according to claim 1 is characterized in that: In S5, the intelligent interpretation model is constructed by combining multidisciplinary knowledge such as geology and geophysics, including the following steps; Data preprocessing: preprocess the raw data of ERT and GPR; Feature extraction: extract feature information from preprocessed data; Model construction: Combining multidisciplinary knowledge, using support vector machine to build an intelligent interpretation model, the intelligent interpretation model uses SMO to find the best separation hyperplane, the separation hyperplane is in the form of ,in, is the weight vector, is the input feature vector, is a bias term. The SMO is an iterative algorithm for solving the dual problem of support vector machines. In each iteration, the SMO selects two Lagrange multipliers for update to keep the KKT condition satisfied, which is expressed as: Calculate the gradient of L(α): Calculate the Lagrangian function L(α) about and The gradient of , the Lagrangian function L(α) is defined as in, is the kernel function, which is used to calculate the similarity between two samples. and The gradient of , where λ is the inverse of the regularization parameter C, i.e. λ = 1 / C; Solve the quadratic programming problem: Simplify the problem into a quadratic programming problem to find a new and values, which will maximize L(α) under the constraints. The quadratic programming problem is expressed as , where the elements of the B matrix are It is related to the kernel function and the label. c is a constant. Solving this quadratic programming problem can get the new and The value of; Update and : Updates with the new value found by the solver and , the updated value must satisfy the constraints of SVM, which are expressed as ; Calculate the new value of b: in the update and After that, a support vector is selected (i.e. or samples greater than 0) to calculate the new value of b, , where w is the weight vector, expressed as , choose a support vector , so that > 0, then Substitute the expression of w into the above formula to get the new value of b; check the KKT condition and check the updated , and b whether they meet the KKT condition, if not, the iteration continues; if they do, the algorithm converges and finds the best separation hyperplane, the intelligent interpretation model automatically adjusts the interpretation parameters according to the data characteristics, and provides a variety of interpretation schemes, the interpretation schemes including different combinations of the location, morphology and rock type of the geological body; Solution evaluation: Evaluate the various explanation solutions provided by the intelligent explanation model; Optimization and iteration: Based on the evaluation results, the intelligent explanation model is optimized and iterated.

8. The method for accurately locating soil pollution based on geophysical detection technology according to claim 1 is characterized in that: In S6, the three-dimensional electrical structure model of the study area is established by using the Kriging interpolation method, including the following steps: Data collection and processing: Collect ERT inversion results and surface coordinate data, and pre-process the collected data; Interpolation method selection: Kriging interpolation method is used. The key formulas in the Kriging interpolation method include the semivariance function and the Kriging interpolation formula. The semivariance function is used to describe the spatial autocorrelation between data points. The formula is: ,in, is the semivariance value, is the distance between two points, The distance is The number of point pairs, and The location and The Kriging interpolation formula is used to infer the value of the unknown data point based on the known data point. The formula is ,in, It is an unknown data point The estimated value at is the weight coefficient, is a known data point The observation value at Determined by minimizing the variance of the prediction error; Three-dimensional model construction: Using ERT inversion results and surface coordinate data, combined with Kriging interpolation method, a three-dimensional electrical structure model of the study area is constructed, and the two-dimensional data is expanded to three-dimensional space to form a three-dimensional electrical structure image; Model verification and optimization: Verify the constructed 3D model and optimize the model based on the verification results; Result presentation: The optimized three-dimensional model is presented in an intuitive way.

Images