Magnetic Anomaly Detection Based on Full Connected Neural Network

Magnetic anomaly detection (MAD) has been widely used for detecting some hidden ferromagnetic objects. Orthonormal basis function (OBFs) detector is one of the most popular methods of MAD. The OBFs detector works effectively under white Gaussian noise. However, the practical geomagnetic noise is colored noise with a power spectral density of 1/f^α (f is frequency and α is noise exponent), and the signal-to-noise ratio (SNR) is usually very low. In order to improve magnetic anomaly detection performance in the case of colored noise and low SNR, a novel detection method by using full connected neural network (FCN) is proposed in the paper. Firstly, the detector based on FCN is designed and two kinds of features that include the signal's statistical property and the magnetic moment's characteristics of the target are extracted and used as the input of neural network; Then, the optimal network structure with proper number of layers and nodes is obtained; Finally, the detection performance of the detector under different SNRs and orientations of target's magnetic moment is evaluated. Simulation results show that the proposed method has better performance and achieves an incremental detection probability of about 5% to 40% under colored Gaussian noise with different noise exponent than traditional method. In the end, experiments under real geomagnetic noise also verify the effectiveness of the proposed method.