A neural network-based automatic semi-variogram modeling approach for geomagnetic map construction in multi-source...

Area of Science:

• Geophysics
• Geomagnetism
• Geostatistics

Background:

• High-precision geomagnetic maps are crucial for navigation.
• Traditional methods for creating these maps rely on subjective semi-variogram modeling, limiting accuracy.

Purpose of the Study:

• To develop an automated framework for geomagnetic map construction.
• To improve the accuracy and efficiency of geomagnetic mapping for navigation.

Main Methods:

• Introduced geomagnetic map via auto-semi-variogram kriging (GMAS-K) framework.
• Integrated geomagnetic map via auto-semi-variogram convolutional neural network (GMAS-CNN) for automatic semi-variogram parameter inference.
• Employed an encoder-decoder architecture with a multiple convolutional block attention module (M-CBAM) for enhanced feature representation and cross-scale consistency.

Main Results:

• GMAS-K produced smoother and more accurate geomagnetic maps compared to ordinary kriging.
• The automated approach streamlined the geomagnetic mapping workflow.
• Demonstrated superior performance in inferring semi-variogram parameters.

Conclusions:

• The proposed GMAS-K framework effectively automates geomagnetic map creation.
• Coupling deep learning with geostatistical interpolation advances geomagnetic mapping and navigation.
• The method shows significant potential for improving navigation accuracy.