针对在轨卫星对船舶跟踪观测时存在的观测时间短、无法满足持续跟踪需求的问题，提出了基于CNN(Convolutional Neural Networks)、BiLSTM(Bidirectional Long Short-Term Memory)和自注意力机制的混合模型进行船舶航迹预测，以确保多颗卫星能够协同工作，实现对船舶的持续、有效跟踪观测。实验结果表明，该模型预测的轨迹很接近真实的轨迹，平均绝对误差比BiLSTM模型和CNN-BiLSTM模型分别降低了 40。7%、13。8%。因此，该模型对于轨迹预测具有较高的精度，能够为选取合适的卫星进行监视提供依据。
Intelligent Trajectory Prediction Technology for Moving Target
To address the problems of short observation time and inability to meet the continuous tracking demand when tracking and observing ships by in-orbit satellites,a hybrid model based on CNN,BiLSTM,and Self-Attention Mechanism for ship trajectory prediction to ensure that multiple satellites can work together to realize continuous and effective tracking and observation of ship.The experimental results show that the trajectories predicted by the model are very close to the real trajectories,and the Mean Absolute Error of the model is reduced by 40.7%and 13.8%compared with the BiLSTM model and CNN-BiLSTM model,respectively.Therefore,the model has high accuracy for trajectory prediction and can provide a basis for selecting appropriate satellites for monitoring.
国家科技期刊平台
NSTL
万方数据
