Multiple state-based video tracking for surveillance applications

The objective of this paper is to present a set of techniques for video analysis aiming a visual surveillance system for monitoring scenarios, detecting and tracking moving objects. The proposed method integrates motion, spatial position and color information to track object blobs which are segmented using Eigenbackground subtraction. To deal with occlusions and splits, detected blobs are organized in a multi-state transition management. Experiments confirming the efficiency of our method were conducted on a range of image sequences from PETS and CAVIAR datasets. Keywords-Surveillance, activity modelling, tracking, Eigenspace models, segmentation.