The method of detecting AIS isolated information based on clustering and distance

Abstract

For the growing number of Automatic Identification System (AIS) data, people have begun to apply data mining and machine learning techniques to extract valuable information from AIS data. About the AIS information, the analysis of isolated information is of great significance. They represent some position points apparently deviation from the traffic flow, which may represent the abnormal reception of AIS data or the abnormal behavior of the ships. In order to detect AIS isolated information, this paper uses the DBSCAN algorithm to cluster AIS data which can identify arbitrarily shapes firstly, the main traffic flow of data acquisition area is extracted; then provide a method to judge whether other clusters belonging to the isolated cluster, next propose an isolated point judgment method, apply the method for the noise points identified by the process of clustering, and get eventually isolated points. Finally, through the analysis of these isolated information by overlap chart and database query, the information has certain value to the analysis of the maritime traffic safety by Vessel Traffic Service (VTS)and the analysis of the abnormal behavior of ships.