An Aeromagnetic Compensation Method Based on Attention Mechanism

• Published in: IEEE geoscience and remote sensing letters Vol. 22; pp. 1 - 5
• Main Authors: Ma, Xiaoyu, Zhang, Jinsheng, Liao, Shouyi, Li, Ting, Li, Zehao
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Aeromagnetic compensation; Aircraft; Algorithms; Atmospheric modeling; attention mechanism; Compensation; Data analysis; Data models; Data processing; geomagnetic navigation; Interference; linear regression; Machine learning; Magnetic field measurement; magnetic interference; Magnetometers; Magnetosphere; Methods; Navigation; Neural networks; Regression analysis; Regression models; Training; Vectors
• ISSN: 1545-598X; 1558-0571
• DOI: 10.1109/LGRS.2024.3508080

Aeromagnetic interference is one of the important factors limiting the application of aeromagnetic data on aircraft platforms. Therefore, magnetic compensation is necessary for aeromagnetic data processing, which is of great significance to improve the accuracy of geomagnetic navigation. In recent years, aeromagnetic compensation methods can be mainly divided into two categories: linear regression methods based on the Tolles-Lawson (T-L) model and data-driven methods based on machine learning. However, the accuracy of linear regression methods is subject to the complexity of the model and the problem of multicollinearity, while data-driven methods require the quantity and quality of aeromagnetic measurement data. To solve this problem, we proposed an aeromagnetic compensation method taking advantage of both the T-L model and neural network. The T-L model parameters are trained through our network, while the attention mechanism is applied in the hidden layer to enhance the feature extraction ability of the model for time series. We validate our method by applying it to an open-access dataset. The Experimental results demonstrate that our method has higher compensation accuracy and generalization performance than the classical algorithms.