摘要
海船异常行为检测对于海上航行安全、海域的智能监管及港口的发展规划都具有非常重要的意义。整理并总结了现阶段海船异常行为检测领域的相关方法,主要包括聚类分析、统计建模、神经网络以及基于预测等异常检测方法,在船舶位置异常、状态异常和潜在的行为异常检测等方面取得了较为显著的进展。同时梳理了海船异常行为检测领域在数据的时变性、模型的复杂性以及异常行为的语义分析等方面存在的异常检测准确性低、语义分析不明确等问题。并对海船异常检测领域未来在数据的增量学习、实时在线检测、情景语义分析和异常行为可视化等方面所存在的突破点做了展望和介绍。
As the abnormal behavior detection of marine vessels is of great significance to the safety of maritime navigation,intelligent supervision of sea areas and development plan of harbors,its present methods and achievements,challenges as well as future research direction were defined and summarized.It is urgent that a robust algorithm with high accuracy and high real-time performance be proposed to detect and study their abnormal behavior for the improvement of maritime navigation safety.
作者
唐皇
尹勇
神和龙
TANG Huang;YIN Yong;SHEN Helong(Laboratory of Marine Simulation and Control,Dalian Maritime University,Dalian 116026,Liaoning,P.R.China)
出处
《重庆交通大学学报（自然科学版）》
CAS
CSCD
北大核心
2019年第9期109-115,共7页
Journal of Chongqing Jiaotong University(Natural Science)
基金
国家“863”课题项目(2015AA016404)
海洋公益性行业科研专项经费项目(201505017-4)
云南内河航运船舶操纵模拟(851333)
关键词
交通工程
海船异常行为检测
轨迹挖掘
智能监管
数据挖掘
traffic engineering
abnormal behavior detection
trajectory mining
intelligent supervision
data mining
