Study On Ship’s Trajectory Clustering Model Based On AIS Data

Due to the rapid growth of maritime traffic and its increasingly complex environment, the trajectory of time and space generated in the production activities has been becoming more and more, hence, how to achieve effective supervision and management from any unusual trajectory and spot them has been the most significant part of realizing maritime intellectual transportation, which is based on obtaining its typical trajectory. Apparently, the traditional researching method is time-consuming and great efforts required, as well as low efficiency, while effective and potential information on ships can be shown by massive AIS data from characteristics maritime traffic, which is beneficial to be access to any data. This paper is based on the maritime traffic engineering theory and data mining technique, AIS data base, ways of combining theories with practices, emphasizing on ship’s trajectory clustering and its algorithm. The main work is as follows:(1) Builded the ship trajectory clustering model experimental data platform based on the IVISSEA system. By collecting, storing and preprocessing the AIS data, we have set up AIS data base so as to analyze the flow of maritime traffic and acquire the data of ship movement.(2) Established the ship trajectory partitioning model and clustering algorithm. During the research, we have made use of course speed rate and minimum description length method to describe the original trajectory accurately. Furthermore, we have combines the classical DBSCAN clustering algorithm to build the whole process of clustering eventually.(3) Optimized the method of trajectory similarity algorithm. Due to the ship’s trajectory composed of AIS connection, and each AIS point contains abundant information, such as location, course, speed, etc., therefore the ship trajectory also has its own structural characteristics. As a consequence, we have come up with a similar measuring method for ship trajectory in this paper, combing two distance measurement methods of the Hausdorff and the structure similarity between lines.(4) Using the sweep line method to get a typical trajectory algorithm. In order to have a better access to representative trajectory, this paper has used the sweep line algorithm to get the typical trajectory and proposed the coordinate conversion algorithm on how to improve the efficiency.(5) Realizing the ship’s trajectory clustering model with the AIS database, and verified by the related theory of marine traffic engineering and the actual vessel traffic flow situation. Take the main channel of Xiamen port and the direct ship “COSCOSTAR” between Fujian and Taiwan ship’s trajectory as an example, it combines with the historical AIS data to cluster the ship’s trajectory in this area, obtaining the typical ship trajectory. According to the different research objects, the accuracy by using the different algorithms, the area encounter and a specific ship’s encounter, and combining the actual ship traffic flow.The results show that the typical trajectory of ship can be obtained better by this trajectory clustering model, and can lay the premise condition for analysis and predict ship motion behavior, find and dispose ship’s unusual behavior in time, and lay solid foundation for realizing intelligent vessel traffic management system to monitor ship behavior.