Knowledge-based situational analysis of unusual events in public places

Combining appropriate methods from computer vision and artificial intelligence enables further progress in smart video surveillance. In this work, an Interacting Multiple Model (IMM) filter is used for person tracking due to the fact that a single motion may not capture the complex dynamics from persons. In addition, context information from the IMM is used for controlling the background model to detect left luggage. The combination of this processing chain serves as input for the situation recognition in addition to person detection and tracking. The computer vision components are integrated in the distributed Cognitive Vision System (dCVS) architecture, which is applied up to now to Traffic, Robotics, Smart Homes, and Video Surveillance. For this work, we cope with situations dealing with unusual events in public places