Research On Mining Ship Navigation Rules Based On Clustering Algorithm

With the development of the economy,the shipping industry has become increasingly prosperous,and the number of ships has continued to increase,resulting in increased pressure on shipping channels and ports.In order to ensure the safety of ship navigation,the maritime department has developed the Automatic Identification System(AIS)of the ship.AIS data contains ship navigation information,which provides a data basis for mining ship navigation rules.By excavating ship navigation rules such as ship channels and frequent activity areas,it can provide technical support for maritime departments for ship route planning and abnormal detection.In order to deal with the problems existing in the current mining of ship channels and frequent activity areas through AIS data,this thesis has carried out the following two researches,the main contents are as follows:1.Aiming at the problem that the current trajectory clustering algorithm separates trajectory feature representation and clustering assignment tasks,resulting in poor clustering effect and inaccurate channel extraction,a depth-based approach that simultaneously performs trajectory feature representation and clustering assignment is proposed.Clustering method to extract ship fairways.First,preprocess the ship trajectory data;then,use the ship trajectory data to train the autoencoder,and extract the initial features of the ship trajectory through the encoder part;then,combine the custom clustering layer with the encoder part to build a deep clustering network to realize the trajectory The task of feature representation and cluster assignment is carried out at the same time to complete the cluster analysis of ship trajectories;finally,the typical trajectories in the trajectory cluster are used to represent the ship’s channel.2.Aiming at the problem of inaccurate extraction of frequent activity areas,which requires artificially setting density thresholds and only considers the spatial information of trajectory data,a grid density peak clustering method considering time and space information is proposed to extract Areas where ships frequently move.First,determine the space-time granularity,and combine the box plot and the elbow method in the grid density peak clustering method to automatically select the cluster center grid;then,according to the local density distribution of the grid in each cluster,the density threshold is automatically selected.,obtain the frequent movement area of ships with a single time-space granularity;finally,through the fusion of time and space,obtain the frequent movement area of ships with multiple time-space granularity.In the experiment in a specific sea area,clustering based on deep representation learning and spatial clustering based on density-based noise application extracted 0 and9 channels,while the method in this thesis extracted 17 channels,and the extraction effect was better;traditional grid clustering The method can only extract the frequent activity area with a single spatiotemporal granularity,while the method in this thesis can extract the frequent activity area with multiple spatiotemporal granularity,and the extracted frequent activity area can more accurately reflect the ship activity.