Chaotic invariants modeling of motion patterns in multiple cameras

Abstract

Extracting the motion patterns from videos is a basic task in video surveillance and has become an active research area. In this paper, we propose a novel approach for discovering motion patterns in a scene observed by one or two cameras. The chaos theory is employed to compute the chaotic invariant features (CIFs) after obtaining all the trajectories. The CIFs and other features are combined to a feature vector and the trajectory is represented by the feature vector. Based on the CIFs, the motion patterns are grouped via mean shift clustering algorithm. Five challenging datasets are used for validation. Experimental results show the proposed method is effective in single camera and multiple cameras situation.