Robust abandoned object detection and analysis based on online learning

Abstract

In this paper, we propose a novel approach based on online learning for accurate and effective detection of abandoned objects. Most existing methods for abandoned objects detection only detect abandoned objects without considering of the logic owner of the abandoned object. These methods need an advanced trained human detector to discriminate abandoned objects from still persons frequently. However, human detection is a challenge in robotic vision system, which always needs off-line training. The proposed framework without specific advanced trained human detector is able to detect abandoned objects and analyze their owners. The online framework is based on a valid assumption for objects and persons in natural scenes. Based on the assumptions that objects are moved by their logic owners and all the moving objects are humans in the scene, online classifiers are established with a certain moving objects just in the scene, which can assist us to detect abandoned objects and analyze their owner in true sense. Instead of a pixel based background model, a robust block based background model is established using online boosting method, which is able to adapt to a large variety of environment and complex changes. In the evaluation over the PETS 2006 and AVSS 2007 datasets, the proposed technique performs robustly and efficiently.