Ship trajectory anomaly detection based on multi-feature fusion

The detection of ship trajectory anomaly based on AIS data is a challenge in marine and computer fields. Unfortunately, traditional AIS data lack the information of ship's size and environment, which will result in false alarms in a harsh environment. To address this challenge, we propose to detect trajectory anomaly concerning more information beyond traditional AIS data (i.e., information about ship's size and environment). To combine traditional features and new features efficiently and rationally, we define two sensors: hull pressure sensor and environmental change sensor. These sensors can combine different types of features well and take advantage of multi-feature fusion. Then, we propose a ship trajectory anomaly detection method based on multi-feature fusion. Experiments on real data show that our method is superior to traditional methods in efficiency and performance.