A wireless sensor network for visual detection and classification of intrusions

  • Authors:

  • Andrzej Sluzek;Palaniappan Annamalai;Md. Saiful Islam

  • Affiliations:

  • School of Computer Engineering, Nanyang Technological University, Singapore and SWPS, Warszawa, Poland;IntelliSys Centre, Nanyang Technological University, Singapore;School of Computer Engineering, Nanyang Technological University, Singapore

  • Venue:
  • ICECS'05 Proceedings of the 4th WSEAS international conference on Electronics, control and signal processing
  • Year:
  • 2005

Quantified Score

Hi-index 0.03

Visualization

Abstract

The paper reports works on a wireless sensor network supporting a human operator monitoring a wide area of an unstructured environment. The main objective is to develop mechanisms allowing intelligent preliminary analysis of data (captured by a large number of sensing devices, including cameras) so that the human intervention is needed only for cases identified as "potentially dangerous intrusions". The developed feasibility-study platform contains two types of nodes: relatively simple warning nodes with a variety of sensors detecting a potential presence of intruders (proximity, vibration, magnetic activity, noise, etc.) and more sophisticated nodes equipped with cameras and capable to perform complex image analysis tasks so that the intrusion can be visually verified and (possibly) classified before involving a human operator. The nodes can wirelessly communicate using a standard protocol. Several mechanisms of visual detection have been proposed and implemented. Thus, the actual details of a network for a given application can be optimized, taking into account the current conditions, hardware constraints and user-defined requirements. Generally, minimization of the wirelessly transmitted data is of high priority. The paper briefly overviews the developed platform and discusses methodologies of the algorithms used in the system. Some implementation details are not disclosed.