Enhancing UAV aeromagnetic data denoising via multitype noise augmentation and deep learning
Chaohong Yan · 2025
In recent years, Unmanned Aerial Vehicle (UAV) aeromagnetic technology has faced challenges in practical applications due to multi-source complex noise arising from electromagnetic interference, dynamic flight trajectories, and sensor instability. Most existing methods rely on single-noise simulations, such as Gaussian noise, to approximate real-world noise for training deep learning models. However, these approaches struggle to effectively capture the diversity and complexity of noise in real-world environments. To address this limitation, this paper proposes a multi-type noise data augmentation method. By integrating Gaussian noise, impulse noise, and strip noise models, our approach expands the noise distribution coverage of training data, thereby significantly enhancing the robustness of denoising models. Based on this framework, we construct a large-scale hybrid dataset and evaluate the performance of several representative deep learning models for aeromagnetic data denoising. Experimental results demonstrate that, compared to conventional single-noise training frameworks, the proposed method markedly improves model generalization capabilities. It achieves superior denoising performance in both single-noise and complex noise scenarios, offering a reliable solution for UAV aeromagnetic data processing in challenging real-world conditions.