摘要
针对复杂海洋环境下,船舶行为识别与轨迹预测困难的问题,提出一种基于多任务学习模型。将卷积神经网络(convolutional neural network,CNN)与通过注意力机制优化过的双向长短时记忆网络(bidirectional long short-term memory,BiLSTM)并联作为训练网络,对船舶的行为识别与轨迹预测两个任务进行联合训练;选取船舶自动识别系统(Automatic Identification System,AIS)提供数据基础。研究结果表明:所提出的模型具有更高的预测和识别精度,能有效地辅助海事部门监管工作,同时多任务学习模型也为海上智能交通研究提供新思路。
To solve the difficulty of vessel behavior recognition and trajectory prediction in complex marine environment,a model based on multi-task learning was proposed.The convolutional neural network(CNN)was connected in parallel with the bidirectional long short-term memory(BiLSTM)optimized by the attention mechanism as a training network,to complete the missions of vessel behavior identification and trajectory prediction at the same time.Vessel automatic identification system(AIS)was selected to provide data base.The research results show that the proposed method has higher prediction accuracy and identification accuracy.It can effectively assist the supervision work of maritime supervision.Meanwhile,the idea of multi-task learning provides a new idea for the study of maritime intelligent transportation.
作者
杨红
韩鹏
刘畅
宫珊珊
YANG Hong;HAN Peng;LIU Chang;GONG Shanshan(School of Information Science&Technology,Dalian Maritime University,Dalian 116026,Liaoning,China)
出处
《重庆交通大学学报(自然科学版)》
CAS
CSCD
北大核心
2022年第4期1-7,共7页
Journal of Chongqing Jiaotong University(Natural Science)
关键词
交通工程
智能交通
行为识别与轨迹预测
多任务学习
船舶
船舶自动识别系统
traffic engineering
intelligent traffic
behavior recognition and trajectory prediction
multi-task learning
vessel
vessel automatic identification system(AIS)