渔业信息与战略2025，Vol.40Issue(4)：304-315,12.DOI:10.13233/j.cnki.fishis.2025.04.008
基于流速场的自注意力LSTM养殖鱼水平空间游动轨迹预测
Prediction of horizontal swimming trajectory of farmed fish based on flow field using self-attention LSTM
摘要
Abstract
This study focuses on the prediction of the horizontal swimming trajectory of farmed fish,aiming to solve the problem that traditional methods are difficult to accurately capture the spatiotemporal characteristics of fish movement due to the neglect of environmental factors.The paper proposes an LSTM model(long short-term memory)that combines flow velocity information and a spatiotemporal enhanced LSTM model that integrates flow velocity field.The model takes water velocity and the current position of the fish as input features,and introduces a self-attention mechanism to more effectively capture the complex spatiotemporal dependencies in the movement trajectory of a single fish.The experimental results show that compared with the traditional model,the proposed model reduces the mean absolute error(MAE),mean square error(MSE)and relative error(RE)by 62.4%,81.8% and 52.0%,respectively.The study shows that by integrating environmental features(such as water velocity)and using deep learning technology,the accuracy of fish swimming position prediction can be significantly improved.In addition,the research provides a novel approach based on environmental features for animal behavior prediction.关键词
轨迹预测/LSTM/自注意力机制/水流速度/时空依赖关系Key words
trajectory prediction/LSTM/self-attention mechanism/water flow velocity/spatio-temporal dependency分类
农业科技引用本文复制引用
刘夕,朱子奇,涂家海,周斌,熊威..基于流速场的自注意力LSTM养殖鱼水平空间游动轨迹预测[J].渔业信息与战略,2025,40(4):304-315,12.基金项目
国家自然科学基金项目(61702382) （61702382）
湖北省教育厅科研计划项目:一种轻量化"双人双锁"危化品管控模块的模型研究(B2023278) （B2023278）
