Combining feature engineering and random forest to improve the aeromagnetic anomaly detection performance

Aeromagnetic survey data is affected by various interferences, including aircraft-induced interference and complex background noise. Accurately detecting weak magnetic anomaly signals amidst complex interference has become a key research focus in magnetic detection. Traditional detection methods, which rely on a single detection metric to identify magnetic anomaly signals, face a critical challenge: elevated false alarm rates (FAR) when attempting to improve detection accuracy. To address this challenge, a feature fusion-based approach is proposed to reduce FAR in complex scenarios. Firstly, a comparative analysis across multiple dimensions of discriminative features is conducted. Secondly, feature fusion is implemented using the Random Forest algorithm to detect magnetic anomaly signals.