A Reliable and Data-Efficient Magnetic Field Prediction Method for Seafloor Exploration Platforms via Prior-Constrained Boundary Integrals

The static magnetic field from large seafloor exploration platforms severely interferes with weak geological signals. Accurately predicting and compensating for this interference is critical for deep-sea surveys. However, traditional inversion methods using limited spatial measurements have severely ill-posed coefficient matrices, amplifying near-field noise and causing massive divergence during far-field extrapolation. To address this, we propose a reliable and data-efficient magnetic field prediction method utilizing prior-constrained boundary integrals. First, a virtual plane is constructed between the platform and the measurement plane. A differential recursive algorithm extracts the local magnetic field on this plane from limited measurements to serve as physical prior information. Incorporating this knowledge to structurally constrain the boundary integral inversion fundamentally mitigates the ill-posed problem. Simulations and scaled physical experiments demonstrate that this method prevents near-field noise overfitting, achieving enhanced far-field reliability. By maximizing the utility of limited spatial data, the maximum relative error on the far-field prediction plane is reduced from 10.5% to 8.3% in simulations, and from 13.2% to 9.8% in physical experiments. This provides a highly reliable approach for marine magnetic interference compensation.