A method for processing fluxgate aeromagnetic data based on machine learning
Through machine learning, processing flux gate avionics data is solved, and the problems of poor steering and high cost of sensors are achieved, achieving high-precision avionics exploration effect.
Patent Information
- Application Number
- CN202210250834.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-15
- Publication Date
- 2025-08-12
- Estimated Expiration
- 2042-03-15
AI Technical Summary
The existing flux gate sensors have poor steering in aerial magnetic survey, resulting in poor data quality. Although the noise level of domestic sensors can meet the requirements, the steering error is large, the imported equipment is high, the price is easy to damage, and the maintenance cost is high.
The machine learning method is used to process fluxgate aerial magnetic data, and the machine learning model is determined through data preprocessing, training data, and polynomial regression are carried out to reduce steering errors and improve data quality.
The use of machine learning technology reduces the steering error of flux gate aerial magnetic data, improves data quality, meets aerial magnetic exploration requirements, and reduces costs.
Smart Images
Figure CN114675335B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of geophysical magnetic exploration, and in particular to a method for processing fluxgate aeromagnetic data based on machine learning. Background Art
[0002] Fluxgate sensors are highly effective for measuring magnetic fields (such as the Earth's magnetic field), with applications in magnetic field monitoring, electromagnetic parameter detection, engineering inspection, and vehicle orientation and attitude measurement and control. Currently, the noise level of fluxgate sensors has reached the picoT level, meeting the requirements of aeromagnetic exploration. However, the orthogonality of the three fluxgate sensor elements is generally 0.1-1°, resulting in a conversion error of 50-400nT, making them unsuitable for aeromagnetic surveys.
[0003] Magnetic prospecting is an important method for mineral exploration, effectively supporting the development of the mining industry. Aeromagnetic technology is becoming an increasingly efficient and high-precision detection technology. Aeromagnetic exploration can be divided into manned and unmanned aerial vehicle (UAV) aeromagnetic methods based on the carrier. Manned aircraft aeromagnetic methods often use fluxgate sensors for their magnetic compensation systems, which cost hundreds of thousands of yuan to import and have large steering errors. UAV aeromagnetic methods often use imported optically pumped magnetic sensors or fluxgate sensors. Optical pumped magnetic sensors are expensive and easily damaged, requiring repairs at the factory, which is both economically and time-consuming. Domestic fluxgate sensors have reached a noise level of pT, which can meet the requirements of aeromagnetic measurements. However, they suffer from steering errors, resulting in poor application results.
[0004] Nowadays, in the era of big data, machine learning technology has gradually been applied to various fields, achieved good results, and promoted the progress of science and society. To this end, the applicant proposed to combine the technical advantages of machine learning and the accuracy of optical pump data, establish and train a model, and use the trained model to predict the fluxgate data, which should be able to reduce steering errors and improve data quality to meet the requirements of aeromagnetic surveys.
[0005] Therefore, it is particularly urgent to develop a method for processing fluxgate aeromagnetic data based on machine learning. Summary of the Invention
[0006] The task of the present invention is to overcome the shortcomings of the existing technology and provide a method for processing fluxgate aeromagnetic data based on machine learning, which can not only improve the accuracy of conventional aeromagnetic compensation, but also be applied to fluxgate UAV aeromagnetic survey and three-component aeromagnetic survey.
[0007] The task of the present invention is accomplished by the following technical solutions:
[0008] A method for processing fluxgate aeromagnetic data based on machine learning, which includes but is not limited to the following process steps:
[0009] 101. Get data preprocessing;
[0010] 102. Determine training data;
[0011] 103.Build machine learning models;
[0012] 104. Train machine learning models based on training data;
[0013] 105.Prediction of three-component fluxgate aeromagnetic data based on trained machine learning model.
[0014] The machine learning method in the present invention is a linear regression method. Since there is nonlinearity between each eigenvalue and the target data, in order to obtain better prediction, polynomial regression is required for the training and validation data, that is, the power of each feature is added as a new feature, and then the feature training model is expanded.
[0015] Compared with the prior art, the present invention has the following advantages or effects:
[0016] By leveraging the technical advantages of machine learning and the accuracy of optical pump data, a model is established and trained to predict fluxgate data, thereby reducing steering errors and improving data quality to achieve the desired aeromagnetic survey results. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 The present invention provides a process flow chart of a method for processing fluxgate aeromagnetic data based on machine learning.
[0018] Figure 2 yes Figure 1 Schematic diagram of the track before line preprocessing using the method shown.
[0019] Figure 3 yes Figure 1 Schematic diagram of the track after line preprocessing using the method shown.
[0020] Figure 4 yes Figure 1 Schematic diagram of fluxgate aeromagnetic data before machine learning training of the method shown.
[0021] Figure 5 yes Figure 1 Schematic diagram of fluxgate aeromagnetic data after machine learning training of the method shown.
[0022] The following is a further detailed description of the invention in conjunction with the accompanying drawings. DETAILED DESCRIPTION
[0023] like Figure 1-5 As shown, a method for processing fluxgate aeromagnetic data based on machine learning is characterized by including but not limited to the following process steps:
[0024] 101. Get data preprocessing;
[0025] 102. Determine training data;
[0026] 103.Build machine learning models;
[0027] 104. Train machine learning models based on training data;
[0028] 105.Prediction of three-component fluxgate aeromagnetic data based on trained machine learning model.
[0029] The method of the present invention may further be:
[0030] The step 101, obtaining data preprocessing, specifically includes:
[0031] (1) Open the file and obtain data including but not limited to optical pump data, three-component fluxgate data (magx, magy, magz), and positioning data (X, Y);
[0032] (2) Input the takeoff point positioning data (X1, Y1) and the altitude (H1), and determine and delete the data that is less than 30m from the takeoff point and less than (H1 + 30m) in altitude (this type of data is invalid during takeoff and landing);
[0033] (3) Input the starting coordinates, ending coordinates, line distance, maximum deviation, and number of lines of the first survey line, calculate the line direction angle (α), rotate the coordinate system according to the line direction angle (α), and generate a new coordinate system in the north-south direction (or east-west direction). After the coordinate system conversion, the X-direction (or Y-direction) values of the survey points on the same survey line are similar. Determine whether the difference between the X-direction (or Y-direction) and the preset coordinates is greater than the maximum deviation. Delete the data exceeding the value (invalid data such as those that do not meet the specification requirements and flight turns). Renumber each survey line according to the X-direction (or Y-direction) value and line distance parameters;
[0034] (4) The new coordinate system is transformed in the opposite direction according to the survey line direction angle (α) to obtain the original coordinate system;
[0035] (5) Delete the optical pump jump point data in the measurement line (there may be unlocked data during the optical pump flight, which needs to be deleted), and perform linear interpolation on the deleted data based on the previous and next point data.
[0036] The step 102, determining the training data, specifically includes:
[0037] (1) The optical pump aeromagnetic data (T) in the preprocessed data is separated as the target data (i.e., accurate value), and the other data are used as predicted feature values;
[0038] (2) Divide the data into training data (train) and verification data (test). The specific implementation process is as follows:
[0039] a. Call the sklearn.model_selection function library in scikit-learn and call the train_test_split function;
[0040] b.train_size is set to 0.7, and a random seed is set to make the model results reproducible;
[0041] c. The function randomly selects 70% of the data as the train data, which serves as the data basis for training the model;
[0042] d. Use the remaining 30% of the data as test data to verify the prediction effect of the training model. Step 103, establishing a machine learning model, specifically includes:
[0043] (1) Install the Python programming language, install and call the LinearRegression function library in scikit-learn;
[0044] (2) Call the PolynomialFeatures function to perform polynomial regression on the training data, and select the power parameter (degree) as 2;
[0045] (3) Call the LinearRegression function and use the regressed training data as input parameters to perform model training to obtain the training model RL.
[0046] The step 104, training the machine learning model based on the training data, specifically includes:
[0047] (1) Call the PolynomialFeatures function to perform polynomial regression on the validation data, and select the power parameter (degree) as 2;
[0048] (2) Call the predict function to predict the regression validation data and obtain the predicted value;
[0049] (3) Compare the predicted value with the target data to evaluate the prediction accuracy of the model.
[0050] The step 105 of processing the three-component fluxgate aeromagnetic data based on the trained machine learning model specifically includes:
[0051] (1) Process the three-component fluxgate aeromagnetic data to be processed according to the aforementioned preprocessing method (without the need for optical pump data);
[0052] (2) Substitute the processed data into the model RL to obtain the processed fluxgate aeromagnetic data.
[0053] As described above, the present invention can be better implemented. The above embodiments are only the best implementation methods of the present invention, but the implementation methods of the present invention are not limited to the above embodiments. Other changes, modifications, replacements, combinations, and simplifications made without departing from the spirit and principles of the present invention should be considered equivalent replacement methods and are included in the scope of protection of the present invention.
Claims
1. A method for processing fluxgate aeromagnetic data based on machine learning, characterized in that The specific process steps include:
101. Acquire data preprocessing, which includes optical pump aeromagnetic data, three-component fluxgate data (magx, magy, magz), and positioning data (X, Y); 102. Determine the training data, including: (1) Separate the optical pump aeromagnetic data (T) from the preprocessed data as the target data, and use the other data as the predicted feature values; (2) Divide the data into training data (train) and verification data (test). The specific implementation process is as follows: a. Call the sklearn.model_selection function library in scikit-learn and call the train_test_split function; b.train_size is set to 0.7, and a random seed is set to make the model results reproducible; c. The function randomly selects 70% of the data as the train data, which serves as the data basis for training the model; d. Use the remaining 30% of the data as test data to verify the prediction effect of the training model; 103.Build machine learning models; 104. Train machine learning models based on training data; 105. Predict three-component fluxgate aeromagnetic data based on trained machine learning models, including: (1) The three-component fluxgate aeromagnetic data to be processed is processed according to the aforementioned preprocessing method without the need for optical pump data; (2) Substitute the processed data into the aforementioned machine learning model to obtain the processed fluxgate aeromagnetic data.
2. The method according to claim 1, wherein The step 101, obtaining data preprocessing, specifically includes: (1) Open the file and obtain the optical pump data, three-component fluxgate data (magx, magy, magz), and positioning data (X, Y); (2) Input the takeoff point positioning data (X1, Y1) and the altitude H1, and determine and delete invalid data that is less than 30m from the takeoff point and less than H1+30m in altitude; (3) Input the starting coordinates, ending coordinates, line distance, and number of lines of the first survey line, calculate the line direction angle α, rotate the coordinate system according to the line direction angle α, and generate a new coordinate system in the north-south or east-west direction. After the coordinate system conversion, the X-direction or Y-direction values of the measuring points on the same survey line are similar. Based on the X-direction or Y-direction value and the survey line distance parameter, determine whether the difference between the X-direction or Y-direction and the preset coordinate is greater than the maximum deviation distance; (4) The new coordinate system is transformed in the opposite direction according to the survey line direction angle α to obtain the original coordinate system; (5) Delete the optical pump jump point in the measurement line and perform linear interpolation on the deleted data based on the data of the previous and next points.
3. The method according to claim 1, wherein The step 103 of establishing a machine learning model specifically includes: (1) Install the Python programming language, install and call the LinearRegression function library in scikit-learn; (2) Call the PolynomialFeatures function to perform polynomial regression on the training data, and select the power parameter degree as 2; (3) Call the LinearRegression function and use the regressed training data as input parameters to perform model training to obtain the training model RL.
4. The method according to claim 1, wherein The step 104, training the machine learning model based on the training data, specifically includes: (1) Call the PolynomialFeatures function to perform polynomial regression on the validation data, and select the power parameter degree as 2; (2) Call the predict function to predict the regression validation data and obtain the predicted value; (3) Compare the predicted value with the target data to evaluate the prediction accuracy of the model.