Geomagnetic detection method based on artificial intelligence

Abstract

The invention discloses a geomagnetic detection method based on artificial intelligence. The detection method comprises the following steps of S1, data acquisition, S2, feature extraction, S3, pattern recognition, S4, interpretation and prediction, S5, visual display, and S6, early warning and safety. According to the invention, a worker needs to check a place needing to be detected and confirm the range, other interference factors and weather conditions before the work is started, so that the influence on the detection work is reduced, and then the sensor is calibrated and tested, so that the problem that the test result is influenced by the failure of the sensor is prevented; geomagnetic detection is carried out on the area by using the unmanned aerial vehicle and the satellite, so that the detection range is enlarged, then filtering processing is carried out on the collected data, and time sequence data and normal data are processed through a Kalman filter, so that subsequent data analysis work is facilitated; the influence of wrong data and invalid data on a test result is reduced, and the range and accuracy of geomagnetic detection are improved.

Description

Technical Field

[0001] This invention relates to the field of geomagnetic detection technology, specifically to a geomagnetic detection method based on artificial intelligence. Background Technology

[0002] Geomagnetic detection is a technology that uses changes in the Earth's magnetic field to detect underground structures and mineral resources. It obtains underground geological information by analyzing changes in the Earth's magnetic field at different locations on the ground. It is widely used in mineral resource exploration, oil exploration, and geological surveys. Traditional geomagnetic detection requires manual data filtering and analysis, which reduces the difficulty and complexity of the work. Based on the original geomagnetic detection method, an artificial intelligence-based method is proposed. However, in traditional geomagnetic detection, the fixed sensor installation results in a small detection range, and the raw data is often messy, making subsequent analysis difficult and reducing the scope and accuracy of the detection. Therefore, "the original small detection range and cumbersome analysis work inevitably affect the efficiency of geomagnetic detection," specifically in the following aspects:

[0003] (1) When conducting geomagnetic detection, the fixed installation of the sensor results in a small detection range and messy raw data, which makes it inconvenient to analyze the results later, thus reducing the scope and accuracy of geomagnetic detection; (2) When processing the data, the normal model construction is not conducive to the accuracy of the result prediction. There is no operation to optimize and adjust the model, thus reducing the optimization of the prediction effect; (3) Because geomagnetic detection involves a large amount of sensitive information and location data, the data cannot be viewed by staff at will, which easily reduces the risk and privacy of the data. Summary of the Invention

[0004] The purpose of this invention is to provide an artificial intelligence-based geomagnetic detection method to solve the problems of the current detection methods mentioned in the background art: (1) When conducting geomagnetic detection, the fixed installation of the sensor results in a small detection range and messy raw data, which is inconvenient for subsequent analysis of the results, thereby reducing the range and accuracy of geomagnetic detection; (2) When processing the data, the normal model construction is not conducive to the accuracy of the result prediction, and there is no operation to optimize and adjust the model, thereby reducing the optimization of the prediction effect; (3) Because geomagnetic detection involves a large amount of sensitive information and location data, the data cannot be viewed by staff at will, which easily reduces the risk and privacy of the data.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a geomagnetic detection method based on artificial intelligence, the detection method comprising the following steps:

[0006] S1, Data Acquisition

[0007] 1. Confirm the data collection location: ① Delineate the scope, ② Detect interference factors; 2. Calibrate the sensors: ① Use wireless sensors, drones, and satellites, ② Deploy and install the sensors; 3. Data preprocessing: ① Data integration, ② Data filtering and noise reduction, ③ Use a Kalman filter for time series data.

[0008] S2, Feature Extraction

[0009] 1. Raw data fusion: ① Raw data summarization, ② Raw data classification; 2. Feature extraction: ① Data dimensionality reduction using principal component analysis (PCA), ② Feature extraction using an autoencoder (AE).

[0010] S3, Pattern Recognition

[0011] 1. Multi-source information fusion: ① Fusion of data from different sensors, ② Fusion of different frequencies; 2. Data analysis: ① Deep learning, recurrent neural networks (RNN) and long short-term memory networks (LSTM), ② Data mining; 3. Online learning: ① Adaptive updates, ② Using ensemble learning frameworks (SGAE).

[0012] S4. Explanation and Prediction

[0013] 1. Determine the model: ① Determine the geological structure, ② Determine the distribution of mineral resources, ③ Determine environmental changes; 2. Interpret the model and results; 3. Predict the results.

[0014] S5, Visual Display

[0015] 1. Results display: ① 3D visualization technology, ② interactive analysis, ③ multi-format display; 2. User interface: staff can operate and view the results; 3. Predictive results: intelligently provides decision-making solutions.

[0016] S6. Early Warning and Security

[0017] 1. Real-time monitoring: ① Detect abnormal terrain, ② Abnormal magnetic field changes, ③ Magnetic field anomaly early warning; 2. Access control: ① Data encryption, ② Identity verification, ③ Anomaly viewing alarm; 3. Identity recording: Record the identity of the person viewing the data.

[0018] Preferably, the detector uses a geomagnetic sensor, and the geomagnetic intensity is calculated using the formula H = B / μ0.

[0019] Preferably, in the geomagnetic intensity calculation formula, H is the magnetic field strength, in amperes per meter (A / m), B is the magnetic flux density, in tesla (T), and μ0 is the permeability constant (vacuum permeability), approximately 4π × 10⁻⁷ tesla·m / (amperes·m), in T·m / A. This formula shows that in a vacuum, the magnetic field strength H is proportional to the magnetic flux density BB.

[0020] Preferably, the formula used in the Kalman filter is X(k)=AX(k-1)+BU(k)+W(k), plus the system measurement value Z(k)=HX(k)+V(k).

[0021] Preferably, in the Kalman filter formula, X(k) is the system state at time k, U(k) is the control quantity of the system at time k, A and B are system parameters, which are matrices for multi-model systems, Z(k) is the measured value at time k, and H is the parameter of the measurement system, which is a matrix for multi-measurement systems.

[0022] Preferably, the recurrent neural network, i.e., RNN, uses the calculation formulas ht=σ(WxXt+WhHt-1+bh) and yt=σ(WyHt+by).

[0023] Preferably, in the recurrent neural network (RNN) calculation formula, Ht is the hidden state at the current time step, Wx is the weight matrix of the input layer, Xt is the input at the current time step, Wh is the weight matrix from hidden layer to hidden layer, Ht-1 is the hidden state at the previous time step, bh is the bias term of the hidden layer, σ is the activation function, typically Sigmoid, Tanh, and ReLU, Wy is the weight matrix from hidden layer to output layer, by is the bias term of the output layer, and yt is the output at the current time step.

[0024] Preferably, the forget gate calculation formula in the Long Short-Term Memory network, i.e., LSTM, is ft=σ(Wf*[Ht-1,Xt]+bf).

[0025] Preferably, in the Long Short-Term Memory (LSTM) network, i.e., the forget gate calculation formula, σ represents the Sigmoid activation function, Wf is the weight matrix, bf is the bias, Ht-1 is the hidden state of the previous time step, and Xt is the current input.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] This invention requires personnel to inspect the area to be detected before starting work, confirming the scope and other interfering factors, as well as weather conditions, to reduce the impact on the detection work. Then, the sensors are calibrated and tested to prevent sensor malfunction from affecting the test results. Next, drones and satellites are used to conduct geomagnetic detection of the area, thereby increasing the detection range. The collected data is then filtered, and a Kalman filter is used to separate time-series data from normal data, facilitating subsequent data analysis and reducing the impact of erroneous and invalid data on the test results, thus improving the range and accuracy of geomagnetic detection.

[0028] This invention improves the modeling and prediction of geomagnetic fields over large areas by fusing data collected from different sensors and integrating ground magnetic field data with satellite remote sensing data. This enhances the accuracy and reliability of geomagnetic field estimation. The data is then used to construct two different prediction models using recurrent neural networks and long short-term memory networks, allowing for analysis from different perspectives. Different neural network architectures and training strategies are employed to improve the model's accuracy and robustness. Finally, an ensemble learning framework is used to incorporate real-time data into the model training, enabling online learning and adaptive updates to maintain the model's real-time performance and accuracy.

[0029] This invention enables timely detection and early warning of anomalies during geomagnetic exploration. It utilizes deep learning technology and sensors, with sensors installed on the ground to measure magnetic field changes, while deep learning technology is used to identify and predict anomalies. Furthermore, encryption and access control measures are implemented at multiple stages of data acquisition, storage, transmission, and processing to ensure data security and privacy. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a flowchart of the geomagnetic detection method of the present invention;

[0031] Figure 2 This is a partial operational logic code diagram of the geomagnetic detection method of the present invention;

[0032] Figure 3 This is a code diagram illustrating the operational logic of Principal Component Analysis (PCA) in this invention.

[0033] Figure 4 This is a code diagram illustrating the operational logic of the SGAE (Synthetic Learning Framework) in this invention. DETAILED DESCRIPTION

[0034] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0035] Example 1:

[0036] Please see Figure 1-Figure 4 An artificial intelligence-based geomagnetic detection method includes the following steps:

[0037] S1, Data Acquisition

[0038] 1. Confirm the data collection location: ① Delineate the area, ② Detect interference factors. 2. Calibrate the sensors: ① Use wireless sensors, drones, and satellites, ② Deploy and install the sensors. 3. Data preprocessing: ① Data integration, ② Data filtering and noise reduction, ③ Use a Kalman filter for time series data. Before starting the work, staff need to check the area to be detected, confirm the range and other interference factors, as well as the weather conditions, to minimize the impact on the detection work. Then, calibrate and test the sensors to prevent sensor malfunction from affecting the test results. Next, use drones and satellites to conduct geomagnetic detection of the area, thereby increasing the detection range. Then, filter the collected data, and use a Kalman filter to compare the time series data with normal data, thus facilitating subsequent data analysis and reducing the impact of erroneous and invalid data on the test results, improving the range and accuracy of geomagnetic detection.

[0039] S2, Feature Extraction

[0040] 1. Raw Data Fusion: ① Raw data summarization, ② Raw data classification; 2. Feature Extraction: ① Data dimensionality reduction using Principal Component Analysis (PCA), ② Feature extraction using an Autoencoder (AE). After preprocessing the raw data, the raw data is then fused and classified to integrate data from different sources, thereby reducing the manipulation of the raw data and improving the speed of data processing. Next, Principal Component Analysis is used to reduce the dimensionality of the data, facilitating subsequent processing of multiple datasets. Finally, an Autoencoder is used to extract data features, facilitating subsequent data modeling.

[0041] S3, Pattern Recognition

[0042] 1. Multi-source information fusion: ① Fusion of data from different sensors; ② Fusion of different frequencies. 2. Data analysis: ① Deep learning, recurrent neural networks (RNN) and long short-term memory networks (LSTM); ② Data mining. 3. Online learning: ① Adaptive updating; ② Using an ensemble learning framework (SGAE) to fuse data collected from different sensors, including ground magnetic field data and satellite remote sensing data, to improve geomagnetic field modeling and prediction over large areas, thereby improving the accuracy and reliability of geomagnetic field estimation. Then, the data is used to construct two different prediction models through recurrent neural networks and long short-term memory networks, facilitating analysis from different perspectives. Different neural network architectures and training strategies are used to improve model accuracy and robustness. Finally, the ensemble learning framework incorporates real-time collected data into model training, enabling online learning and adaptive updates to maintain model real-time performance and accuracy.

[0043] S4. Explanation and Prediction

[0044] 1. Determine the pattern: ① Determine the geological structure, ② Determine the distribution of mineral resources, ③ Determine environmental changes; 2. Interpret the model and results; 3. Predict the results. During the model training process, some highly interpretable features are retained to help engineers and managers understand the model's decision-making process and results. Then, through machine learning technology, a large amount of information is obtained to help technicians better understand and interpret the data and make predictions about the results.

[0045] S5, Visual Display

[0046] 1. Results Display: ① 3D visualization technology, ② interactive analysis, ③ multi-format display; 2. User Interface: Staff can operate and view the results; 3. Predictive Results: Intelligent decision-making solutions are provided. This information is spatially analyzed and visualized to better understand the underground structure and distribution. By converting the data into maps and charts, technicians can better understand and interpret the data. Then, visualization technology and interactive analysis tools are used to display and interpret the data. The data is presented in the form of maps and time series, and an interactive user interface is provided so that users can more intuitively understand and operate the detection results. The output results are visualized to help engineers and managers understand the model's decision-making process and results.

[0047] S6. Early Warning and Security

[0048] 1. Real-time monitoring: ① Detecting abnormal terrain, ② Abnormal magnetic field changes, ③ Magnetic field anomaly early warning; 2. Access control: ① Data encryption, ② Identity verification, ③ Anomaly viewing alarm; 3. Identity recording: Recording the identity of personnel viewing the data. During geomagnetic exploration, we can promptly detect anomalies and issue early warnings. This is achieved through the use of deep learning technology and sensors. Sensors are installed on the ground to measure magnetic field changes, while deep learning technology is used to identify and predict anomalies. Encryption and access control measures are then implemented in multiple stages of data acquisition, storage, transmission, and processing to ensure data security and privacy.

[0049] In this embodiment: the detector uses a geomagnetic sensor, and the geomagnetic intensity is calculated using the formula H = B / μ0.

[0050] In this embodiment: In the geomagnetic intensity calculation formula, H is the magnetic field strength, in amperes per meter (A / m), B is the magnetic flux density, in tesla (T), and μ0 is the permeability constant (vacuum permeability), which is approximately 4π×10⁻⁷ tesla·m / (amperes·m), in T·m / A. This formula shows that in a vacuum, the magnetic field strength H is proportional to the magnetic flux density BB.

[0051] In this embodiment, the formula used in the Kalman filter is X(k)=AX(k-1)+BU(k)+W(k), plus the system measurement value Z(k)=HX(k)+V(k).

[0052] In this embodiment: X(k) in the Kalman filter formula is the system state at time k, U(k) is the control quantity of the system at time k, A and B are system parameters, which are matrices for multi-model systems, Z(k) is the measured value at time k, and H is the parameter of the measurement system, which is a matrix for multi-measurement systems.

[0053] In this embodiment: the recurrent neural network, i.e., RNN, uses the calculation formulas ht=σ(WxXt+WhHt-1+bh) and yt=σ(WyHt+by).

[0054] In this embodiment: In the Recurrent Neural Network (RNN) calculation formula, Ht is the hidden state at the current time step, Wx is the weight matrix of the input layer, Xt is the input at the current time step, Wh is the weight matrix from hidden layer to hidden layer, Ht-1 is the hidden state at the previous time step, bh is the bias term of the hidden layer, σ is the activation function, usually Sigmoid, Tanh, and ReLU, Wy is the weight matrix from hidden layer to output layer, by is the bias term of the output layer, and yt is the output at the current time step.

[0055] In this embodiment: the forget gate calculation formula in the Long Short-Term Memory network, i.e., LSTM, is ft=σ(Wf*[Ht-1,Xt]+bf).

[0056] In this embodiment: In the Long Short-Term Memory (LSTM) network, σ represents the Sigmoid activation function, Wf is the weight matrix, bf is the bias, Ht-1 is the hidden state of the previous time step, and Xt is the current input.

[0057] Example 2:

[0058] As shown in the figure, unlike Embodiment 1, this embodiment presents an artificial intelligence-based geomagnetic detection method, which includes the following steps:

[0059] Step 1: Establish an intelligent inversion estimation architecture for radar detection environment based on prior knowledge base.

[0060] Step 2: Establish a decision model for the correlation between radar detection environment and sea clutter; calculate the correlation factors and correlation weights between the two using grey relational analysis and improved entropy method; conduct correlation verification; and formulate conflict resolution strategies to achieve correlation mining between radar detection environment and sea clutter.

[0061] Step 3: Establish a prior knowledge base for radar detection environment, which is based on both simulation data and measured data.

[0062] Step 4: Using the K-means algorithm, based on actual radar echo data and a prior knowledge base, we estimate the presence, type, and characteristic parameters of atmospheric ducts to achieve accurate estimation of the radar detection environment. Furthermore, we improve accuracy by establishing an analysis of the impact of sea state and sea conditions on atmospheric duct characteristics.

[0063] In summary, compared with Example 2, Example 1 utilizes drones and satellites to conduct geomagnetic detection of the region, thereby increasing the detection range. Furthermore, the collected data is filtered, and a Kalman filter is used to separate the time-series data from the normal data, facilitating subsequent data analysis and reducing the impact of erroneous and invalid data on the test results. This improves the range and accuracy of geomagnetic detection.

[0064] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0065] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A geomagnetic detection method based on artificial intelligence, characterized in that: The detection method includes the following steps: S1. Data Collection 1. Confirm the collection location, ① define the range, ② detect interference factors, 2. Calibrate sensors, ① use wireless sensors, drones and satellites, ② place and install sensors, 3. Data preprocessing, ① data integration, ② data filtering and denoising, ③ use Kalman filter for time series data; S2. Feature extraction 1. Raw data fusion, ① raw data aggregation, ② raw data classification, 2. Feature extraction, ① data dimension reduction, using principal component analysis, i.e. PCA, ② feature extraction, using autoencoder, i.e. AE; S3. Pattern Recognition 1. Multi-source information fusion, ① Fusion of different sensor data, ② Fusion of different frequencies, 2. Data analysis, ① Deep learning, recurrent neural network, i.e. RNN and long short-term memory network, i.e. LSTM, ② Data mining, 3. Online learning, ① Adaptive update, ② Use integrated learning framework, i.e. SGAE; S4. Explanation and prediction 1. Determine the model, ① determine the geological structure, ② determine the distribution of mineral resources, ③ determine environmental changes, 2. Interpret the model and results, 3. Predict the results; S5. Visualization 1. Result display, ① 3D visualization technology, ② interactive analysis, ③ multi-format display, 2. Operation interface, staff can operate and view, 3. Prediction results, intelligent decision-making solutions; S6. Early warning and safety 1. Real-time monitoring, ① detection of abnormal terrain, ② abnormal magnetic field changes, ③ abnormal magnetic field warning, 2. Access control, ① data encryption, ② identity authentication, ③ abnormal viewing alarm, 3. Identity record, record the identity of the person viewing the data.

2. The method for geomagnetic detection based on artificial intelligence according to claim 1, characterized in that: The detector adopts a geomagnetic sensor, and the geomagnetic sensor calculates the geomagnetic intensity according to the formula H=B / μ0.

3. The method for geomagnetic detection based on artificial intelligence according to claim 2, characterized in that: In the geomagnetic intensity calculation formula, H is the magnetic field intensity, unit: ampere / meter, A / m, B is the magnetic induction intensity, unit: Tesla, T, μ0 is the magnetic permeability constant (vacuum magnetic permeability), which is approximately 4π×10 to the negative seventh power Tesla·meter / (ampere·meter), unit: T·m / A. This formula shows that in a vacuum, the magnetic field intensity H is proportional to the magnetic induction intensity BB.

4. The method for geomagnetic detection based on artificial intelligence according to claim 1, characterized in that: The formula used in the Kalman filter is X(k)=AX(k-1)+BU(k)+W(k), plus the system's measurement value Z(k)=HX(k)+V(k).

5. The method for geomagnetic detection based on artificial intelligence according to claim 4, characterized in that: In the Kalman filter formula, X(k) is the system state at time k, U(k) is the control quantity of the system at time k, A and B are system parameters. For multi-model systems, they are matrices, Z(k) is the measurement value at time k, and H is the parameter of the measurement system. For multi-measurement systems, H is a matrix.

6. The method for geomagnetic detection based on artificial intelligence according to claim 1, characterized in that: The recurrent neural network, i.e., RNN, uses the calculation formulas ht=σ(WxXt+WhHt-1+bh), yt=σ(WyHt+by).

7. The method for geomagnetic detection based on artificial intelligence according to claim 6, characterized in that: In the RNN calculation formula, Ht is the hidden state at the current moment, Wx is the weight matrix of the input layer, Xt is the input at the current moment, Wh is the weight matrix from hidden layer to hidden layer, Ht-1 is the hidden state at the previous moment, bh is the bias term of the hidden layer, σ is the activation function, usually Sigmoid, Tanh and ReLU are used, Wy is the weight matrix from hidden layer to output layer, by is the bias term of the output layer, and yt is the output at the current moment.

8. The method for geomagnetic detection based on artificial intelligence according to claim 1, characterized in that: The long short-term memory network, i.e., the forget gate calculation formula in LSTM is ft=σ(Wf*[Ht-1, Xt]+bf).

9. The method for geomagnetic detection based on artificial intelligence according to claim 8, characterized in that: In the long short-term memory network, i.e., the LSTM forget gate calculation formula, σ represents the Sigmoid activation function, Wf is the weight matrix, bf is the bias, Ht-1 is the hidden state of the previous time step, and Xt is the current input.

Images