The Magnetic Anomaly Inversion of Magnetic Targets Based on Convolutional Neural Network Using Undersampled Data

In magnetic target detection, the inversion of the target object can be achieved by observing and recording the magnetic anomaly values of magnetic bodies, thereby obtaining the position information of the target magnetic anomaly body. In order to identify the shape and location of the magnetic anomaly objects accurately, an under-sampled data magnetic anomaly inversion method based on convolutional neural networks was proposed. In this method, a large number of magnetic anomaly observation data were generated through the forward model, the observation data were undersampled, and then the undersampled data were used as input to train the network. Finally, the validation of the model data was achieved. The experiment proves that this method can accurately reflect the position information and magnetization strength values of magnetic anomaly bodies.