An autonomous monitoring system based on object shape detection

In traditional monitoring systems, stationary cameras are supervised only by a human operator, who may easily miss some events recorded by a camera. Because it is imperative for a surveillance system to be reliable, its autonomy can be extended by applying computer vision algorithms to a video signal and also by the use of mobile robots capable of monitoring tight and occluded areas. In this paper, we present an overview of the concept of an autonomous monitoring system based on object shape detection. Our goal is to develop a real-time system which robustly and efficiently identifies objects on the basis of their approximate shape. For monitoring the environment we use active and smart cameras capable of remote position control, as well as mobots equipped with video sensors. After performing the object extraction from individual video frames, each new detected object is decomposed into simple graphical primitives like lines, circles, rectangles etc. and then identified in a database using the Query by Shape (QS) method.