Clustering And Visualization Of Ship Typical Trajectories Based On Improved DBSCAN

With the wide application of data mining in various fields,ship trajectory data has become an important area of concern for Marine traffic information.As a huge basic industry,shipping industry constantly generates massive data and stores it in the Automatic Identification System(AIS).It covers ship temporal and spatial information,involves ship communication and navigation,and provides rich data sources for ship trajectory prediction.At the same time,there exists the problem of data fault redundancy.In view of the AIS original data,the original data should be preprocessed to obtain effective data,and then typical trajectory clustering should be carried out to mine ship movement patterns,plan routes,and observe ship movement rules.Typical trajectory research provides intuitive visualization for Marine traffic flow analysis and vessel density analysis.It can lay a theoretical and technical foundation for establishing a scientific,intelligent and efficient maritime supervision system.This paper studies the AIS data acquired,selects the appropriate algorithm for clustering,improves DBSCAN algorithm,reduces query times and improves operational efficiency,presents the data in the form of images through data and image processing technology,visually displays the ship’s movement trajectory,and reveals the spatiotemporal evolution law of the ship’s entry movement trajectory.It is of great significance to provide support for maritime safety supervision and decision-making.The main research contents include:(1)Data preprocessing.AIS data is collected,cleaned,and the missing value is preprocessed to make up the difference.In view of the missing track data set,interpolation is carried out according to the time interval to ensure the integrity of the original track data.(2)Typical trajectory clustering algorithm theory.The classical trajectory clustering algorithm is described,and combined with the temporal and spatial characteristics of ship trajectories,the similarity of different ship trajectories is analyzed to determine the stagnation point of each ship.Then,the algorithm is used to cluster multiple latitude and longitude grid points,and the algorithm characteristics are summarized.Combining the algorithm analysis characteristics and ship data attributes,the clustering algorithm suitable for ship trajectories in port is determined.(3)Research on trajectory clustering and visualization based on DBSCAN algorithm.In view of DBSCAN clustering algorithm’s repeated area query of data,the improved DBSCAN clustering algorithm tries to reduce and shorten the process of repeated query of sample neighborhood,and avoids repeated query by querying sample neighborhood.On this basis,the results of reducing algorithm time consumption and improving accuracy are obtained.(4)Experimental simulation.The validity of the algorithm was verified by the water area of Tianjin port section,and the clustering convergence process was evaluated by the similarity measurement method.The refined cluster improves the data analysis efficiency,reduces the operation time cost,and generates visual information visualization images through the generated track points and ship sailing track lines.