Left-object detection through background modeling

Video surveillance systems are becoming extensively deployed in many environments due to the increasing needs of public security and crime prevention. In this paper, we propose a comprehensive solution for managing abandoned objects, which means that the system can deal with objects that are abandoned, removed, or partially occluded. The system contains two adaptive abandoned object detection (AOD) methods that are both based on the proposed texture modeling method associated with a mixture of Gaussians for a real environment. The first method is more efficient than the second one, but the latter is more robust than the former. The proposed methods have been proved to be characterized with prominent efficiency and robustness according to mathematic analyses and experimental results. The designed automatic detection system helps human operators not only to ease tedious monitoring work but also to focus only on suspicious abnormal events.