Magnetic Moment Estimation Algorithm Based on Convolutional Neural Network

Estimating the magnetic moment of a magnetic target is a key aspect of magnetic anomaly detection when accurately locating the target. Current methods primarily rely on vector magnetic field signals or gradient signals, which face challenges such as steering errors and the high costs associated with sensor arrays. This paper proposes a magnetic moment estimation algorithm that combines a scalar magnetic field sensor and the three components of the local geomagnetic field with a convolutional neural network (CNN). The simulation results demonstrate that the proposed algorithm performs well in noise environments with a signal-to-noise ratio (SNR) greater than −5 dB. At an SNR of −5 dB, the relative error in the magnetic moment magnitude is 4.92%, while the absolute errors in the magnetic moment declination and inclination are 4.73° and 1.54°, respectively. The vector angle difference is 2.7°. The algorithm achieves high estimation accuracy even with a scalar sensor and under poor noise conditions, offering a novel and effective approach for future magnetic moment estimation research.