Segmentation Reconstruction and Prediction of AIS Trajectory Based on Broad Learning System

Baohua He, Yi Zuo, Weihong Wang, Licheng Zhao, Tieshan Li, C. L. Philip Chen

Abstract: With the widespread adoption of the automatic identification system (AIS), the collected AIS data has become vast volumes, and cause the analysis and processing of vessel trajectories with highly time consuming. Linear models (LMs) are simple and fast to be widely applied in trajectory reconstruction and prediction. However, existing studies generally pursue individual vessel trajectories independently, and the accuracy and stability of LMs are still unsuitable for generalized processing of a large number of trajectories. To address this limitation, we adopt broad learning system (BLS) to establish a prediction model for trajectories. This paper includes three parts of trajectory segmentation, feature extraction and model training. Firstly, K-means clustering is used to segment the trajectories, and divide them into small pieces based on navigation characteristics. Secondly, considering the time series analysis of trajectory data, the segmented trajectory is processed using first-order and second-order differencing to obtain training data. Finally, by training the time series data of different trajectories, a generalized trajectory prediction can be achieved.

Submission Number: 75