Research On Magnetic Anomaly Detection Based On Multi-feature Ensemble Learning

Magnetic anomaly detection is a passive detection technology for detecting small changes in magnetic field.It has the advantages of simple structure and not easy to be exposed.It is widely used in earthquake prediction,submarine detection,mineral resources detection,positioning and tracking,weapon guidance and other fields.It has very high application value and research significance.Based on this background,the thesis conducts research on magnetic anomaly detection based on multi-feature ensemble learning.The main work of this thesis is as follows:(1)Aiming at the problem that the magnetic anomaly magnetic field signal is submerged in the geomagnetic background noise,the frequency characteristics of the magnetic anomalous magnetic field are studied by establishing a mathematical model of the magnetic anomalous magnetic field.Aiming at the problem that the feature vector extracted by a single method cannot fully characterize the magnetic anomaly features,a multi-feature extraction method can be used to obtain more complementary and complete feature vectors.(2)Aiming at the problem of redundant nonlinear relationship among multiple features in eigenvectors,a multi-feature fusion method of magnetic anomaly features is studied.The dimensionality reduction and fusion capabilities of the principal component analysis(PCA),the kernel principal component analysis(KPCA)and the weighted kernel principal component analysis(WKPCA)are compared and analyzed.The experimental results show that the weighted kernel principal component analysis multi-feature fusion method has better dimensionality reduction fusion ability.(3)Aiming at the problem of low accuracy of traditional magnetic anomaly detection algorithms,this thesis studies ensemble learning and its combination strategy,and compares and analyzes the ability of various combination strategy algorithms to improve the accuracy and rationality of detection results.The experimental results show that the improved evidence theory can have the highest accuracy under the condition that the decision results are reasonable.(4)Magnetic anomaly detection is studied based on the denoising method of magnetic anomaly signal,multi-feature extraction method,multi-feature fusion theory,ensemble learning and improved evidence theory.The experimental results show that the magnetic anomaly detection method based on multi-feature ensemble learning proposed in this thesis has higher magnetic anomaly detection accuracy and better reliability.