Abstract
Magnetic interference compensation is critical for enhancing the accuracy of unmanned aerial vehicle (UAV) magnetic anomaly detection. To address the constrained compensation performance of the conventional Tolles-Lawson (T-L) model, which stems from insufficient parametric dimensionality, this study proposes a dynamic-enhanced extended compensation model. The novelly introduced attitude angle and attitude angular rate-coupled features expand the parameter set from 18 to 34 terms, significantly enhancing the characterization of the magnetic field. To overcome the limitations of linear regression in modeling the nonlinear relationships inherent in small aeromagnetic datasets, we developed a genetic algorithm-optimized shallow backpropagation neural network (GA-BP). This network establishes high-precision correlations between the extended parameters and magnetic interference noise. Experimental results demonstrated that the proposed model effectively captured the coupling characteristics between dynamic flight attitudes and the interference field, leading to significant gains in key performance metrics. This approach provides novel optimization pathways for anti-interference capabilities in airborne detection systems, offering substantial practical value for enhancing UAV aeromagnetic surveys.