海洋渔业2026，Vol.48Issue(2)：163-176,14.
基于深度学习的南极磷虾泵吸桁杆拖网渔船船位状态和渔场信息识别
Deep learning based identification of ship position status and fishery information for Antarctic krill pump suction beam trawl fishing vessels
摘要
Abstract
Based on the AIS position data and fishing logs of the Chinese Antarctic krill continuous pump suction beam trawl fishing vessel Deep Blue from December 25th 2022,to January 15th 2023,an algorithm model(CNN-LSTM-attention model)for extracting operational status parameters of pump suction beam trawlers based on ship position data was constructed.This model classified fishing vessels into four states:sailing,preparing to fish,fishing,and drifting.Based on this model,the operational statuses of three Norwegian krill fishing vessels of the same type in 2023 were identified.Combined with ship position data,the fishing points of pump suction beam trawlers and those recorded in the fishing logs of traditional mid-water variable depth trawlers were identified,and a comparative analysis was conducted on the locations of operational hotspots,operational duration,and operational days of the two fishing methods.The results showed that the accuracy of the CNN-LSTM-attention model for extracting operational status parameters of Antarctic krill pump suction beam trawlers reached as high as 99.23%.The proportions of time spent in the four operational states—sailing,preparing to fish,fishing,and drifting,were 12.9%,2.2%,78.9%,and 6.0%respectively.During the main fishing season(January-May),the number of operational days of pump suction beam trawlers was similar to that of traditional mid-water variable depth trawlers;while in the non-main fishing season(June-September),there was a significant difference in the number of operational days between the two types of vessels.The average monthly operational days of the former was 23.4 d,compared with only 5.6 d for the latter.However,the daily operational duration of pump suction beam trawlers was significantly longer than that of traditional mid-water variable depth trawlers,with a gap of approximately 8.5 h.Meanwhile,the fishing grounds of pump suction beam trawlers and traditional mid-water variable depth trawlers had a high degree of overlap,which provided an important reference for the fishing ground prediction of the two fishing methods.The research results provide information reference for the operational behavior management,fishing ground dynamic prediction,and fishery supervision of China's pump suction beam trawlers.The accurate identification of fishing vessel operational status through deep learning methods can provide more precise monitoring and guidance for future fishery resource management.关键词
南极磷虾/深度学习/船位数据/作业状态/渔场热点Key words
Antarctic krill,/deep learning/ship position data/operation status/fishery hot spots分类
农业科技引用本文复制引用
李阳,韩海斌,苏冰,王雨涵,相德龙,孙煜琰,张衡,张巧芬..基于深度学习的南极磷虾泵吸桁杆拖网渔船船位状态和渔场信息识别[J].海洋渔业,2026,48(2):163-176,14.基金项目
青岛海洋科技中心山东省专项经费(2022QNLM030002-1) （2022QNLM030002-1）
国家重点研发计划(2022YFC2807504) （2022YFC2807504）
中国水产科学研究院东海水产研究所中央级公益性科研院所基本科研业务费专项资金(2021M06) （2021M06）
