Distributed edge and contour line detection for environmental monitoring with wireless sensor networks

UnrestrictedSeveral signal processing problems associated with wireless sensor networks deployed for the monitoring of certain environmental physical phenomena are examined in this research. A couple of distributed algorithms are proposed to extract the essential information of the monitored physical phenomena such as edges and contour lines so that a remote base station may utilize the information for further analysis. These networks often span over a large geographical area and they cannot be adequately monitored or tracked using traditional localization techniques.; Due to different sensing devices of deployed sensor nodes, measured sensor data can take a binary or a scalar value. For a binary field (phenomenon observed through the sensing devices which provide binary outputs only), the edge (or the boundary region) often provides the important information of concern. A new statistical approach together with a false alarm control scheme to edge region detection is proposed to increase the discrimination capability of the true edge line location. Unlike previous work in the literature, no hierarchical network topology and geographical information about sensor nodes are needed. Simulation results demonstrate that the proposed approach has better performance when the noise interference is high and the probing range is small. For a scalar field (phenomenon observed through the sensing devices which are able to provide real-number measurements), the two-dimensional contour lines offer an efficient way to depict its density distribution. Three approaches to contour lines estimation are proposed to provide different levels of system performance for density observation based on the cost of system installment and its computation capability. Simulation results demonstrate that the proposed approaches can extract contour lines of the density distribution well in the presence of sensing and channel noise