Ship Trajectory Prediction Under Marine Lane Constraints Based On Maritime Big Data

In recent years,due to the popularity of mobile devices,global positioning systems and the development of location-based services,more and more trajectory data are accumulated.The research work of data mining and data analysis based on massive trajectory data has been paid more and more attention.In the field of marine traffic,ship trajectory prediction is of great significance in path planning,collision avoidance and other application scenarios.For example,through the trajectory prediction technology,we can predict that the ship will pass through the congested sea area,and then we can adjust the route in advance to avoid the congested road.Trajectory prediction can be divided into prediction under non-road network constraints and road network constraints.There are some limitations in the prediction under the condition of non-road network constraints,which can not effectively use the road network information for more accurate prediction.There is no man-made road boundary on the sea,but the ship also sails according to a certain lane.The lane can be extracted from the trajectories of many ships and form a lane network.Therefore,this paper studies the ship trajectory prediction based on the constraints of ocean waterway network.The main work of this paper is as follows:First of all,the characteristics and existing problems of the original ship trajectory data are analyzed,and a preprocessing method based on marine space-time big data is proposed.This method is based on the Map Reduce parallel computing framework,sampling,denoising,segmenting and interpolating the original data,and improves the quality of the data.Secondly,a method of constructing ocean channel network based on Delaunay triangulation is proposed.This method can extract the ocean channel network,and then express the connection relationship between the path nodes in the network structure through the adjacency matrix.The waterway polygon in Bohai region is selected in the experiment.The waterway network consists of 120 edges and 102 nodes,and the total length of the edges is 3597.154 km.This method can effectivelyextract and construct the ocean channel network and is superior to other related methods in terms of smoothness and integrity.Then,a ship trajectory prediction method based on maritime network is proposed.The factors affecting the navigation of the ship are analyzed,and the LSTM(Long Short-Term Memory,long-term and short-term memory network is selected as the prediction model to predict the next section that the ship will pass through in the voyage.In the experiment,33037 samples of length 5 are selected as the data set,the first four road sequence elements of each sequence are the predicted sequence,and the fifth sequence element is used as the sample label.the data set is divided into training set,verification set and test set at the proportion of 6:2:2.The prediction accuracy of the trained model on the test set is 88.25%.The precision of the trained model on the test set can reach 0.902,the recall rate is 0.895,and the F1-score is 0.89.Finally,the experimental verification prototype system of ocean channel network construction and ship trajectory prediction is realized,which provides data preprocessing,channel network construction,trajectory prediction operation and display pages,and supports the parameter adjustment of different functions.