Robust Edge Detection in Wireless Sensor Networks

The ability to geometrically represent sensed phenomena within a wireless sensor network can provide a more concise view than enumeration of all nodes identifying a phenomenon. A more concise view of sensed data can reduce the communication and energy costs of data analysis and extend the lifetime of the network. We propose a distributed edge detection technique that identifies connected perimeters for sensed phenomena within wireless sensor networks. The technique operates in arbitrarily deployed wireless sensor networks, such as those containing connectivity holes, and is capable of correctly identifying the perimeters of irregularly shaped phenomena. We implemented our technique and conducted extensive experiments; results show that our technique provides accurate perimeters for all sensed phenomena within a wireless sensor network even in the presence of irregularly shaped phenomena.