Research On Application Of Density-based Clustering Analysis In Ship Traffic Supervision

In recent years,with the increasing volume of maritime trade,the size of ships becomes larger and the speed becomes faster,and the number of ships sailing at sea has been increasing.The increasingly complex maritime traffic environment puts forward higher requirements for maritime traffic supervision.Data analysis and mining of massive AIS data can obtain effective navigation information of ships and provide strong support for maritime traffic supervision.Among them,density-based clustering analysis method plays an important role in research of maritime traffic.The paper focuses on the application of density-based clustering analysis in ship trajectory clustering and real-time monitoring of multi-ship collision risk in sea area.Regarding the application research of density-based clustering analysis in ship trajectory clustering,a distributed ship trajectory clustering method based on OPTICS using Spark platform is proposed.The ship's longitude,latitude and the course to the ground are taken into account to construct the similarity measurement model between ship trajectory.Taking the AIS data of Tianjin Port in January 2016 as experimental data,the cluster is deployed on the Spark platform and four threads are created.After completing AIS data preprocessing,ship motion point recognition and ship trajectory extraction and simplification using Douglas-Peucker algorithm under multi-threads,the rules of regional trajectory partition are set up and in accordance with the rules will all ship trajectories be divided into four parts,and OPTICS trajectory clustering is carried out independently in four threads.Finally,the clustering results are spliced and visualized.The result of trajectory clustering can reflect the characteristics of ship trajectories in the waterway of Tianjin Port,and the ship trajectories with different characteristics are distinguished successfully.In the application research of density-based clustering for real-time monitoring of Multi-ship collision risk in sea area,a real-time monitoring method of Multi-ship collision risk in sea area is proposed,which first uses DBSCAN clustering algorithm to cluster all ships in sea area,and then calculates the collision risk of each class of ships separately.A Multi-ship collision risk calculation model considering the effects of DCPA,TCPA and ship azimuth distribution is established.Based on the AIS data from 00:00:00 to 00:01:00 on January 1,2016,the dynamic navigation information of each ship is synchronized by using linear interpolation method.After DBSCAN clustering of ships in a certain time,about four fifths of ships without collision risk are filtered as noise points.The risk of Multi-ship collision among ship and other ships in the same cluster is was calculated and visualized.Experimental results show that this method can extract the ships that are close to each other in space and filter out the ships that are far away from other ships in the stage of non-collision risk,so as to achieve the purpose of reducing the amount of calculation.Theoretical verification of the multi-ship collision risk calculation model is carried out and the function image of the model is drawn.The verification results show that the method is feasible.