The Automatic Identification System (AIS), as a shipboard broadcast system, is widely used for maritime supervision. Based on the analysis of AIS data, the behavioral characteristics of ships can be obtained and the characteristics of the current waterway traffic flow can be extracted, which provide strong support for tasks such as traffic flow prediction and traffic safety risk analysis. Currently, ship trajectory clustering based on AIS data is a popular research area, and often used to analyze the behavior patterns of ships. However, traditional trajectory clustering methods have drawbacks such as long running time, high computational cost, and low accuracy. This directly affects the accuracy of further work such as ship trajectory prediction and ship collision avoidance. Besides, the performance of low accuracy and high time consumption is difficult to meet the needs of the shipping industry development today.In view of the current situation, this paper attempts to introduce the trajectory clustering method based on the virtual shipping network. By combining computer vision technology with the AIS data density visualization method, the AIS data density grid heat map is formed. Based on this, a dataset of node target detection in the marine traffic network is constructed, and the nodes of the marine traffic network and their locations are extracted using deep learning-based target detection model and node extracting method. A marine traffic network is constructed, and the ship trajectory clustering under the network is realized.
Visualized Node Detection-Driven Shipping Network Construction and Ship Trajectory Clustering
2024-04-26
4369163 byte
Conference paper
Electronic Resource
English
Ship route prediction method based on ship trajectory clustering
European Patent Office | 2021
|A novel clustering method for ship trajectory prediction
IEEE | 2024
|