Identification of the Shaft-Rate Electromagnetic Field Induced by a Moving Ship Using Improved Learning-Based and Spectral-Direction Methods

The use of the shaft-rate electromagnetic fields generated by moving ships for detection and sensing purposes has several advantages, including effective target recognition and excellent concealment. It offers a solution to the challenges faced in detecting underwater targets. In this study, we propose a method to identify and analyze the shaft-rate electromagnetic field signals using an improved deep learning algorithm and a spectral-direction analysis technique. Initially, we apply variational mode decomposition (VMD) to identify the multifrequency characteristics of both synthesized and real extremely low-frequency (ELF) electromagnetic signals, creating a reliable sample library for deep learning. Next, we develop an improved deep learning model that combines the residual network (ResNet) with the aforementioned sample library to accurately detect the weak narrowband electromagnetic field signals hidden within the noise. Additionally, we use the spectral-direction analysis method to estimate the direction of the ship's movement. Finally, we validate our proposed method through a synthetic model and a field experiment. The results demonstrate the effectiveness of our approach in identifying the shaft-rate electromagnetic field signals and accurately estimating the direction of moving ships. The developed method shows the potential for accurate sensing and localization of moving ships.