Navigation Behavior Analysis And Trajectory Prediction Expert System Based On AIS

With the application and popularization of the Automatic Identification system(AIS),a large amount of AIS information data comes with it.In recent years,the prediction of ship trajectory using AIS data has become a hot issue,and the prediction of ship trajectory has important guiding significance for collision avoidance,traffic control,risk assessment and navigation behavior analysis.At present,the work of ship trajectory prediction mainly includes two parts:one is reconstruction of past trajectory and prediction the future position,the other is establishment of trajectory classification prediction model and the prediction of ship trajectory by classification algorithm.Route planning is the key technology of ship intelligent navigation system,so it is particularly important to provide better navigation guidance for ship operators by rational use of the ship's AIS historical data for predicting trajectory.In view of the limitations of the present ship navigation behavior analysis and trajectory prediction,this paper proposes an expert trajectory prediction system applicable to all ship trajectories.There are mainly the following innovations:(1)A density-based clustering method is used to detect and remove noise points in ship trajectories,through density-based clustering analysis of AIS data,noise points can be automatically identified and removed,and the processing process is simple and efficient.(2)In this paper,a new method of trajectory feature extraction E-vector is proposed.Based on the hierarchical clustering analysis of trajectory AIS data points,the Euclidean distance is used as the similarity measurement method,and the number of feature points of the feature trajectory is determined by using the Dynamic Time Warping algorithm,so as to realize E-vector trajectory feature extraction.(3)A trajectory prediction model is established by using the Long-Short Term Memory(LSTM)network,which takes the feature trajectory of the historical trajectory as the input and the ship's trajectory in the future as the output,it can reduce the complexity of experiments and save space and time.(4)The trajectory prediction expert system mainly consists of three modules.The first module is the AIS data preprocessing module,which is mainly responsible for the original AIS data storage and the removal of noise points.The second module is the trajectory feature extraction module,which is mainly responsible for extracting E-vector feature.The last module is trajectory prediction module,which is mainly responsible for making use of the existing historical trajectory data to realize the prediction of future trajectory.In this paper,a trajectory prediction expert system is established.Firstly,the E-vector trajectory is extracted,and the comparison experiments are made of new method with the Douglas-Peucker algorithm and the Least-squares Cubic Spline Curves Approximation algorithm.It is found that the trajectory feature E-vector algorithm has the minimum distortion rate,less time consumption and the best recovery degree.The model of trajectory prediction is established by using the LSTM network,and the experiments are made to verify least prediction error.The research results of this paper lay a foundation for further improving maritime shipping information services,and it also has important guiding significance and certain application value for maritime departments to realize maritime traffic monitoring,ship collision avoidance,traffic control,risk assessment and navigation behavior analysis.