Research on CNN-Based Remote Reference Denoising Methodology for Fully Airborne Natural-Field Electromagnetic Detection

The Airborne Natural-Field Tipper Detection System represents an advanced rapid detection methodology specifically engineered for geophysical prospecting in harsh environments characterized by extreme cold, high altitude, and challenging terrains. However, airborne electromagnetic interference induces significant distortion in acquired datasets, thereby impeding precise localization of subsurface anomalous bodies. To address this challenge, this study proposes a Remote Reference (RR) Denoising Method Integrated with Convolutional Neural Networks (CNN). The framework establishes a nonlinear mapping relationship between high signal-to-noise ratio (SNR) ground reference station data and co-located noisy aerial measurements, subsequently constructing a CNN-based denoising model. Experimental validation demonstrates that the proposed approach achieves an SNR improvement of 11.9 dB, effectively verifying its capacity for intelligent noise attenuation and subsurface feature identification.