海洋测绘2024，Vol.44Issue(1)：21-25,5.DOI:10.3969/j.issn.1671-3044.2024.01.005
基于Encoder-Decoder LSTM的船舶轨迹预测方法
A prediction method of vessel trajectory based on Encoder-Decoder LSTM
摘要
Abstract
Accurately predicting vessel trajectory is crucial for early warning and safe navigation,yet accuracy and stability remain major need to be solved at present.To remedy this,a vessel trajectory prediction method based on an Encoder-Decoder LSTM neural network is proposed.Firstly,vessel AIS trajectory data is preprocessed using methods such as denoising,segmentation,interpolation,stay point detection,and normalization to extract vessel sailing trajectories.Next,a vessel trajectory prediction model based on the Encoder-Decoder LSTM architecture is constructed,and the model parameters are initialized.Finally,the proposed model is trained and validated using real AIS data of ferries in the Tianshenggang waters in the Jiangsu section of the Yangtze River and compared with other widely-used trajectory prediction models.The results shows that this method can achieve accurate prediction of trajectories,and the predicted trajectories have a significant reference value.关键词
水路运输/船舶自动识别系统/船舶轨迹预测/编码器-解码器/长短期记忆网络Key words
waterway transportation/automatic identification system/vessel trajectory prediction/encoder-decoder/long short-term memory分类
天文与地球科学引用本文复制引用
李业,任鸿翔,张政..基于Encoder-Decoder LSTM的船舶轨迹预测方法[J].海洋测绘,2024,44(1):21-25,5.基金项目
国家自然科学基金(52071312) （52071312）
交通运输行业重点科技项目(2022-ZD3-035) （2022-ZD3-035）
大连市科技创新基金(2021JJ12GY031). （2021JJ12GY031）
