CBAM-Based Residual Network for Magnetic Anomaly Target Detection

Magnetic anomaly detection techniques have been extensively utilized in civil and military applications. One of the most prevalent detection methods is known as the orthogonal basis functions, which is not satisfactory in colored background noise and low signal-to-noise ratios. In this paper, a residual network based on convolutional block attention module (CBAM) is proposed. The input to the network is the time-frequency spectrum obtained by wavelet transform of the raw magnetic data. The residual modules help the network to learn features better and improve performance, while the CBAM facilitates focusing on important features of the data. Experimental results show that the detection performance of the proposed network has better detection performance than some existing methods at low signal-to-noise ratios.