UNet 기반 수동 소나 신호분석을 통한 표적 선박의 프로펠러축 회전수 추정

초록

This study proposes a UNet-based deep learning algorithm to remove noise contained in underwater radiated signals, thereby improving target detection and tracking performance in underwater environments. Analysis results show that for the first and second harmonic frequency components, UNet-based method achieved power spectral density (PSD) values up to approximately 6.2 dB higher than the conventional method in low-detection environments where the signal-to-noise ratio (SNR) is below zero. In other low-detection conditions, an average improvement of more than 3 dB was also observed. Furthermore, by calculating the interval differences between the peaks of each harmonic component, the propeller shaft rotation (PSR) was estimated to be 19.53 Hz.

키워드