Magnetic Anomaly Target Detection Using Neural Network With 2-D Multihead Attention

The magnetic anomaly target detection is a widely used technique to locate obscured ferromagnetic objects for both civilian and military purposes. To improve the detection performance at low signal-to-noise ratios (SNRs), we propose a deep learning network based on 2-D multihead attention in this article. First, three types of samples are prepared, including pure magnetic noise, noise mixed with complete signals, and noise mixed with truncated signals of different percentages. Then, the 1-D time sequences are converted as 2-D time-frequency spectra using wavelet synchrosqueezed transform (WSST) for better feature extractions. Afterward, a network is trained by using the time-frequency spectra for classification of the samples. The network mainly consists of two 2-D multihead attention modules and several convolutional layers. The first convolutional layer uses the time-frequency spectra of three orthogonal basis functions (OBFs) as the convolutional kernels to extract the rough signal features. The parameters of the 2-D multihead attention are obtained by convolutional operation for fine feature extractions. Experimental results demonstrate that the proposed scheme has higher detection performances than existing approaches to some extent, especially at low SNRs. It also shows that training the network using both complete and truncated signal samples is essential to raise the detection rate (DR) and reduce the false alarms.