Research And Application Of The Trajectory Analysis Based On AIS

In the global environment of global economic globalization,with the rapid increase in the volume of marine traffic,this has increased the burden on the waters and has congested navigation channels.In this case,the problem of the ship itself and the accidents caused by human factors increase,causing huge economic losses.It is imperative to improve the decision-making level of the relevant maritime departments and vessel traffic service systems.One of the key issues is to achieve ship track forecasting and abnormal warning.At present,most ship trajectory prediction models still remain in the simple machine learning methods and models based on specific kinematic equations.Such models are highly constrained and have been difficult to cope with today’s complex maritime traffic situations.High-precision model training requires extensive data support.Compared with traditional radar equipment,AIS equipment has the advantages of small terrain influence,high positioning accuracy,etc.,and can provide more abundant ship trajectory feature data.Based on AIS data samples,this paper combines neural network and deep learning to study the algorithm model for ship trajectory prediction.Based on the parsed AIS data,a training simulation experiment for ship trajectory prediction model was designed and implemented.Finally,the application of the track prediction model in abnormal early warning and route planning was studied.The main research work of this article is as follows:1.Analyze and summarize the AIS related decoding and processing technology.Based on the introduction of AIS message types,the data structure of AIS information is presented,and the AIS information is decoded and processed to obtain a data source,which provides a data basis for track prediction based on AIS data.2.A shallow ship trajectory prediction model using BP neural network is deeply studied.According to the multi-dimensional characteristics of ship navigation trajectory,based on genetic algorithm,the model is improved on the ship prediction problem,and the influence of different network parameters on efficiency and performance is analyzed.3.Using the characteristics of deep learning and time series,a model based on the Recurrent Neural Networks-Long Short-Term Memory(RNN-LSTM)model was proposed to train the model,and the influence of the parameters was analyzed.The network’s prediction model is compared to reflect the LSTM model’s superior ability to process sequence data.4.The LSTM model is applied to trajectory detection,security early warning,route planning and other fields.The model can be combined with the electronic chart experiment platform to predict vessel trajectories and provide abnormal track warning functions.The results show that the RNN-LSTM ship trajectory prediction model using deep learning can achieve very good prediction results under the requirements of accuracy and performance,and provide technical support for intelligent transportation at sea.