Multistep Information Fusion for Target Detection Using UWB Radar Sensor Network

In this paper, we propose a multistep information fusion scheme for target detection through foliage and wall, using ultrawideband (UWB) radar sensor networks. We apply an information theory to detect target with poor signal quality in dynamic forest-environment. This method is motivated by the fact that echoes from the stationary target that is obscured by foliage has strong random characteristics. This is resolved by three steps of information fusion. For the first step of information fusion, we use Kullback-Leibler divergence-based weighting and generated a modified histogram. In the second step, we use entropy- and mutual information-based information fusion. Finally, we use three different fusion methods: 1) Dempster and Shafer theory of evidence; 2) proportional conflict redistribution rule 5; and 3) Bayesian network for decision fusion. Results show that when echoes are in poor quality, accurate detection can be achieved by applying our method. To demonstrate that our algorithm could be applied to other scenarios, we apply it to sense-through-wall human detection using different UWB radars, and simulation results show that our approach works well.