基于低信噪比AIS数据的船舶航迹精准预测

Accurate Prediction of Ship Tracks Based on Low SNR AIS Data

针对船舶自动识别系统(AIS)数据存在较大的实时系统噪声和测量噪声,且易丢包和产生错误的问题,提出了滑动卡尔曼循环网络。该方法采用加窗处理的方式,通过在各窗口中建立状态方程,并使用实时计算的噪声对其进行更新,将非线性高斯系统转化为线性高斯系统。将该高信噪比状态参数输入相应时刻的循环单元中,训练网络参数,从而拟合其非线性关系,实时预测船舶的航迹进行。试验对“大连—烟台”航段的AIS数据进行分析,并将所得结果与传统方法相对比,证明该方法具有较高的精度和较快的速度,可实时对船舶航行轨迹进行精准预测。

Automatic ship identification system(AIS)is widely adopted,however,its data usually contain lots of real-time system noises and measurement noises,it’s easy to loss package,and it contains error data.Aiming at this problem,a sliding Kalman cyclic network is proposed in this paper.This method adopts a windowing method,and converts the nonlinear Gaussian system into a linear Gaussian system by establishing a state equation in each window and updating it with real-time calculated noise.The high signal-to-noise ratio state parameters were input into the cyclic unit at the corresponding time.Thus,the network parameters are trained to fit the nonlinear relationship and to predict the ship's trajectory in real time.To validate its effectiveness,analyzes of AIS data collected on"Dalian-Yantai"route segment were performed.Compared with traditional methods,its higher accuracy and processing speed were proved,providing accurate trajectory prediction in real time.