Gravity and magnetic field data integration using autoencoder neural networks

To address the differences in location and accuracy of gravity and magnetic field data from different periods, we used a structured similarity-constrained autoencoder neural network to normalize the data obtained at different scales to a common standard. Based on this approach, we propose a method suitable for integrating multi-scale gravity and magnetic field data.Through examples of multi-scale integration of gravity and magnetic data from theoretical models and measured data, it is demonstrated that our proposed integration method performs well in the integration of multi-scale gravity and magnetic data. The method has small computational errors and is consistent with the characteristics of gravity and magnetic field data, thus having significant practical value and potential for widespread application.