Improved Principal Component Analysis for Magnetic Gradient Signal Detection Using Dual Three-Axis Magnetometers

Abstract

Magnetic anomaly detection (MAD) is an effective technique for detecting ferromagnetic materials in the presence of a geomagnetic field. In this article, by integrating and enhancing the orthogonal basis function (OBF) and principal component analysis (PCA), a high-sensitivity PCA (HSPCA) method is proposed, achieving both low computational complexity and high detection accuracy in MAD tasks. We first employ the PCA method to decompose the target signal used in the full magnetic gradient OBF (FMG-OBF) method for detection, referred to as the PCA baseline method, which achieves rapid and automated decomposition. It achieves the detection accuracy of FMG-OBF while reducing the real-time processing time by 61.1%. Furthermore, we propose the HSPCA method as an improvement by modifying the form of target signal to enhance detection sensitivity, establishing a mapping to eliminate the influence of redundant sampling parameters, and applying weights to the basis functions to balance their relative contributions. Ultimately, the computational time of this improved method is only 1.3% of that of the PCA baseline method, and its detection rate is increased by 53.1% compared to FMG-OBF at a noise level of −20 dB with a false alarm rate of 4%. Field experiments are conducted to validate the convenience and practicality of the proposed methods.