Online classification of frequent behaviours based on multidimensional trajectories

Abstract

In early warning surveillance domain, frequent behaviours could be mined from historical trajectories. Based on the mined frequent behaviours, online classification technology could be used to classify the continuously updated target trajectories. Online classification of frequent behaviours is of great significance for situation assessment, threat assessment and command decision. There are a lot of researches on trajectory classification. However, they cannot distinguish behaviours whose space position is similar but the moving speed and direction are quite different. The online classification performances of them are also not appropriate for early warning surveillance application. The authors propose a conformal multi-class classifier based on conformal predictor and propose a multi-factor non-conformity measure based on multidimensional trajectories firstly. Then, they present a sequential multi-factor Hausdorff nearest neighbour conformal multi-class classifier, which could online learn and classify frequent behaviours. Experiments in both simulated military scenario and realistic civilian scenario show the presented algorithm has a good performance to online classify frequent behaviours and would have a wide prospect in early warning surveillance systems.