19-04-2012, 01:55 PM
VIDEO SECURITY FOR AMBIENT INTELLIGENCE
paper on Video Security for Ambient Intelligence.doc (Size: 559 KB / Downloads: 44)
INTRODUCTION:
In this paper, a video security application for people detection, tracking, and counting in indoor environments is presented. It is based on computer vision techniques to analyze and process video streams acquired from multiple video cameras placed at different entrances of the building or for monitoring interesting areas. Image segmentation is performed to detect the people in the scene. Their movement is tracked and counted as entering or exiting the building.
System Architecture and Processing Steps
The proposed system is based on a network of smart sensors that cooperatively aim to understand the movements and activities inside a monitored building. Here, smart sensors are represented by subsystems able to detect and to track people, to extract important features for event detection purposes, and to send information for distributed data association. High level nodes get information generated by the sensors and, on the basis of rules in a knowledge base, associate different simple events to build more complex ones. In our case, simple events are represented by states of a finite automata, see fig1 corresponding to simple actions as walking, taking the stairs, etc. Complex events, on the other hand, do not represent only a sequence of events performed by a single person, but also have important Semantic value.
PEOPLE COUNTING
describes the general architecture of the people counting module. Input data are represented by a list of objects correctly detected by the tracking module and a geometric model of the monitored scene describing the position on the top-view map of the areas, where the counting operation is performed.
CONCLUSION
This paper presents an application for indoor person detection and counting in an AmI framework. The system processes color images acquired by several cameras placed at different entrances
of public buildings in order to estimate the number of accesses. The system works in real-time and it is able to detect moving people in the scene and calculate their number.