Networked Surveillance System Using Object-Centric Motion-Based Tracking and Classification

This paper presents a novel technology for networked surveillance systems. The system integrates an innovative method of object centric automatic detection, classification and tracking. It processes the video streams captured from multiple cameras installed in different parts of a building. The system after detecting a moving object turns on its classification and tracking algorithms. The classifier analyzes the movement of the object as normal or anomalous behavior using classifier masks. The main monitoring station is prompted only if an anomalous posture or movement is detected. The tracker keeps a track of that suspicious object and shows an object centric view of it. When an object is under the surveillance of more than one camera, the system displays only one video stream, which is independent of the camera. For every object a new window is opened and it stands same for multiple objects. The novelty of our approach is that the security operator does not need to switch between multiple camera views to track an object, thus making it suitable for effective surveillance applications.