Magnetic Anomaly Detection Using a Reduced Set of Principal Components Based on 2-D Pattern Recognition

Magnetic anomaly detection (MAD) is used for detection of visually obscured ferromagnetic objects such as sunken ship wrecks, unexploded ordnance (UXO), archaeological artifacts and mineral prospection, to mention a few. In a typical scenario, a magnetic survey is carried-out using an aircraft equipped with a scalar magnetometer. Taking advantage of several survey lines for detection rather than a single line may improve the ability to distinguish between a target signal and interference. The survey lines are used to build a magnetic anomaly map where 2D dipole structure can be detected inside the map. In the present work, we investigate the 2D patterns of a dipole field in three different scenarios. Principal component analysis (PCA) enables us to calculate the 2D orthonormal basis functions (OBFs) required for construction of an effective detector for these scenarios. We show that superfluous OBFs reduce the effectiveness of the detector. We exemplify the results with real-world data obtained using an airborne magnetic survey. The good detection performance paves the way for the development of real-time detection system.