Directional analysis of trajectories based on trajectory smoothing

Abstract

In this article we propose a framework to discover interesting directional patterns in trajectory data sets. The proposed framework has five stages; trajectory smoothing, directional segmentation, directional classification, filtering and finally clustering. The main contributions are in the stages for smoothing, directional classification and filtering. Trajectory smoothing is an important step in the analysis of complex, non-smooth trajectories data sets, such as animal movement data. In directional classification stage, different sub-trajectories are assigned to the classes corresponding to their directional orientation. In the filtration stage the outlier trajectories are removed from the respective classes using a novel convex hull based approach. We used animal movement data in this work.