Finding prototypes to estimate trajectory development in outdoor scenarios

Abstract

The incorporation of qualitative information into tracking systems is a challenging topic of research in the computer vision area. Within the context of an outdoor scenario, it is usual to find some patterns of human motion and, in general there is a reduced set of typical entrance and exit points. In this work we present a method to analyze an existing set of trajectories obtained in a selected outdoor environment. Trajectories are clustered based on their entrance and exit points in order to model prototypes that are used to on–line estimate the development of a new trajectory. Those prototypes are modeled using spline curves to avoid the inaccuracy pulled from the vision system. Interesting applications comprise abnormal behavior detection, spatio–temporal event analysis and nd semantic interpretation of human behavior based on the context.