단시간 수중음향 신호를 활용한 합성곱 신경망 기반의 선박 소음 탐지 = CNN-based Shipping Noise Detection using Short-time Underwater Acoustics Signal

다국어 초록 (Multilingual Abstract)

Most of the noise generated by humans in the ocean is ship radiated noise caused to fishing and commercial shipping. Recently, deep learning technology has been used to detect shipping noise. In this study, the convolutional neural network is trained by a shipping noise spectrogram divided into 1-minute units to detect a near distance ship. Inception-V3, ResNet-50, VGG-16 and the proposed model were used to learn and evaluate 1-minute shipping noise. As a result, the F1 scores were 97.42%, 98.42%, 98.16% and 97.88% for Inception-V3, ResNet-50, VGG-16 and the proposed model, respectively. These models showed satisfactory performance in detecting shipping noise. It was confirmed that the proposed model showed equivalent detection performance with about 1/8 parameters compared to ResNet-50. For future works, it is expected that it will be possible to detect long-distance shipping noise by using additional noise data and AIS(Automatic Identification System).

국문 초록 (Abstract)

해양에서 인간에 의해 발생하는 대부분의 소음은 어업 및 상업 운송과 관련된 선박 방사소음이 주요한 원인이다. 최근 선박소음을 자동으로 탐지하기 위한 방법으로 딥러닝 기술이 활용되고 있다. 본 연구에서는 1분 단위로 분할한 선박소음 신호 기반의 스펙트로그램 이미지를 합성곱 신경망 기반 학습을 수행하여 근거리 선박소음 및 배경소음을 자동으로 탐지하는 연구를 수행하였다. 현재까지 많이 사용되고 있는 합성곱 신경망 모델인 Inception-V3, ResNet-50, VGG-16와 본 연구에서 제안한 모델을 이용하여 1분 단위의 선박소음을 학습 및 평가를 수행하였다. 분석 결과 F1 점수는 모델별로 각각 Inception-V3 97.42%, ResNet-50 98.42%, VGG-16 98.16%, 제안된 모델은 97.88%로 나타나 선박소음을 탐지함에 있어 준수한 성능이 나타났다. 이 때, 제안된 모델은 F1 점수가 가장 높게 나타난 ResNet-50 모델에 비해 약 1/8의 적은 파라미터로 동등한 탐지 성능을 보이는 것을 확인할 수 있었다. 추후에는 다양한 선박소음 및 선박자동식별장치(AIS, Automatic Identification System) 자료를 동시에 활용하여 원거리 선박소음 또한 자동으로 탐지가 가능할 것으로 판단된다.

참고문헌 (Reference)

• 1 김선효 ; 정섬규 ; 강돈혁 ; 김미라 ; 조성호, "천해역 선박 소음 자동 탐지를 위한 인공지능 기법 적용" 한국음향학회 39 (39): 279-285, 2020
• 2 S. H. Cho, "Wind-dependent ambient noise level estimation in shallow water using wind speed data" 223 : 108653-, 2021
• 3 K. Simonyan, "Very deep convolutional networks for large-scale image recognition"
• 4 C. Li, "Underwater target classification using deep learning" 22-25, 2018
• 5 M. F. Mckenna, "Underwater radiated noise from modern commercial ships" 131 (131): 92-103, 2012
• 6 T. Guo, "Underwater Target Detection and Localization with Feature Map and CNN-Based Classification" IEEE 1-8, 2022
• 7 G. Song, "Underwater Noise Classification based on Support Vector Machine" 1048-1051, 2021
• 8 S. H. Cho, "Spatial mapping of underwater noise radiated from passing vessels using automatic identification system data" 57 (57): 07LG07-, 2018
• 9 H. Niu, "Ship localization in Santa Barbara Channel using machine learning classifiers" 142 : EL455-460, 2017
• 10 S. Shen, "Ship Type Classification by Convolutional Neural Networks with Auditory-Like Mechanisms" 20 (20): 253-, 2020
• 11 C. Szegedy, "Rethinking the Inception Architecture for Computer Vision" 2818-2826, 2016
• 12 R. J. Urick, "Principles of Underwater Sound" McGraw-Hill 202-233, 1983
• 13 J. S. Park, "Noise directionality estimated using ship tracking data in the Southern Sea of Korea" 57 (57): 07LG06-, 2018
• 14 D. Ross, "Mechanics of Underwater Noise" Pergamon 272-287, 1976
• 15 M. J. Wilmut, "Inversion of Lloyd Mirror Field for determining a source’s track" 32 (32): 940-947, 2007
• 16 A. Krizhenvsky, "ImageNet classification with deep convolutional neural networks" 60 (60): 1-9, 2017
• 17 International Marine Organization, "Guidelines for the reduction of underwater noise from commercial shipping to address adverse impacts on marine life"
• 18 C. Szegedy, "Going deeper with convolutions" 1-9, 2015
• 19 M. Irfan, "DeepShip : An underwater acoustic benchmark dataset and a separable convolution based autoencoder for classification" 183 : 115270-, 2021
• 20 K. He, "Deep Residual Learning for Image Recognition" 770-778, 2016
• 21 김지섭 ; 윤영글 ; 한동균 ; 나형술 ; 최지웅, "Convolutional neural network 기법을 이용한 턱수염물범 신호 판별" 한국음향학회 41 (41): 235-241, 2022
• 22 S. Shen, "Auditory Inspired Convolutional Neural Networks for Ship Type Classification with Raw Hydrophone Data" 20 (20): 990-, 2018
• 23 J. W. Choi, "Analysis of Characteristics of Underwater Ambient Noise Measured at the Coasts of Korea" 107-, 2016
• 24 X. Lurton, "An introduction to underwater acoustics" Springer 2002
• 25 S. C. Wales, "An ensemble source spectra model for merchant ship-radiated noise" 111 (111): 1211-1231, 2002
• 26 S. B. Kim, "A study on the sources of ambient sea noise in the coastal water of pusan" 26 (26): 180-183, 1990
• 27 S. M. Ha, "A Study on Analysis of Image Classification Accuracy Based on Xception Model" Namseoul University 2020