Research On Magnetic Anomaly Detection Method Based On Machine Learning

The magnetic field generated by the magnetic target is superimposed on the geomagnetic field,causing the distribution of the magnetic field in the space around it to change,thus forming magnetic anomalies.The detection and location of underwater targets can be realized by detecting magnetic anomalies.However,in the process of underwater magnetic detection,the weak magnetic anomaly signal generated by the target is usually deeply buried in the background magnetic noise,and the detection method of weak magnetic anomaly largely determines the actual detection ability of underwater magnetic field.In this paper,the problem of magnetic anomaly detection is converted into a binary classification problem of magnetic field signals,that is,to identify whether magnetic anomaly signals exist in the binary classification problem of measured magnetic signals.Therefore,a magnetic anomaly detection method based on machine learning is proposed.The main work of this paper is as follows:Construct the original data set of magnetic anomaly detection model.The magnetic anomaly signal is modeled by magnetic dipole model,and the scalar signal expression of the magnetic anomaly target signal is obtained in the measuring coordinate system.The simulation parameters conforming to the actual situation are selected and the magnetic anomaly signal is simulated by the simulation tool.In order to simulate the real situation of the magnetic target submerged in geomagnetic noise,background noise is replaced by Gaussian white noise,which is superimposed on the magnetic anomaly signal,and the original sample set of magnetic anomaly signal is constructed in the generated mixed signal.A magnetic anomaly detection model is constructed.The magnetic anomaly detection model is used to classify the magnetic anomaly signal and judge whether there is a magnetic target,so as to detect the magnetic anomaly.Firstly,the feature extraction of the original magnetic anomaly data set is carried out by principal component analysis,and the noise and redundant features are eliminated,which improves the classification accuracy of the magnetic anomaly detection model.Then,a magnetic anomaly detection model is constructed according to three machine learning algorithms: backpropagation neural network,radial basis function neural network and support vector machine.In order to obtain the optimal parameters of support vector machines magnetic anomaly detection model,the benchmark test function is introduced to compare the performance of ant colony optimization,genetic algorithm and particle swarm algorithms from three evaluation points of accuracy,robustness and convergence speed.Finally,the optimal particle swarm algorithm is selected as the parameter optimization algorithm.Finally,three kinds of magnetic anomaly detection models,namely principal component analysis-particle swarm analysis-support vector machine,principal component analysisbackpropagation neural network and principal component analysis-radial basis function neural network,are obtained by using the feature extracted magnetic anomaly feature information into the model for training.Simulation experiment verification.By comparing and analyzing the accuracy of different classifiers,the average accuracy of the magnetic anomaly detection model of principal component analysis-particle swarm optimization-support vector machines is 4% higher than that of principal component analysis-ant colony optimization-support vector machines model and 4% higher than that of principal component analysis-genetic algorithm-support vector machines model.The average accuracy is 10% higher than that of the principal component analysis-backpropagation neural network model and 10% higher than that of the principal component analysis-radial basis function neural network model,which verifies the superiority of the principal component analysis-particle swarm algorithm-support vector machine magnetic anomaly detection model constructed based on machine learning and improves the underwater magnetic anomaly detection capability.