This research project addresses the problem of detecting, classifying, and tracking intruders. It combines the sensing potential of self-configuring and instantly-deployable wireless sensor networks (WSN’s) with the reasoning capabilities of artificial neural networks (ANN’s). A system has been designed to be able to detect intruders of many sizes and types: from reptiles to human beings to large vehicles, in a variety of terrains especially those that are vast and hard to reach.
We investigated the implementation of a smart, economical solution for the above surveillance problem. Smart as it utilizes artificial neural networks to classify the different types of intruders. Economical as it utilizes small, cheap wireless sensor nodes that are self-powered and self-configuring. The target terrains to be protected by the implemented system are those that are usually too vast to be patrolled by human guards. Such terrains include oil pipelines, the national borders, and large-scale farms. Traditional solutions to such surveillance problem are expensive, complex and hard to maintain.
The goal of the project is to implement a reliable, smart, and feasible surveillance system that is able to detect, classify, and track various types of intruders over large, deserted areas. To reach the above goal, the project was implemented by a mesh network of wireless sensor nodes as part of a multitier network. The data collected by the sensors is sent to a central server. An artificial neural network system runs on the server to detect, classify and track the intruders in real-time. The data captured from the WSN showed the need for the ANN. The results show that ANN can accurately distinguish between large-size and small-size intruders.