Detection and tracking over networks

Abstract

Knowing the location of the nodes in wireless sensor networks is essential for their operation. However, such localization in large scale networks is challenging. Additionally, the traditional localization techniques based on GPS and satellite navigation provide low accuracy and are useless indoors. Therefore, an alternative technology has to be proposed. This motivates this project's study of using wireless camera networks to track moving objects. In particular, first standard localization techniques used in a WSN are studied. Then it is explained how to detect and track objects using video cameras. Finally, an analysis is made on how video images and sensor measurements are combined in order to obtain better results. The good performance of these fusion techniques is illustrated with examples. It is demonstrated in a simulation that the localization techniques can perfectly track a moving object if there is no measurement noise. These results also demonstrate the trade- off between localization accuracy and measurement noise. Moreover, it is affirmed that increasing the number of anchor nodes, reduces the error in target localization. Finally, the detection and tracking of objects using a camera based tracking system are simulated. The general conclusion of this work is that even though video tracking systems are promising technology, for them to be a realistic option more research has to be done in signal processing, communications, computer vision and mathematical analysis.