An information theoretical approach to sensor placement in a multi-sensor automatic target recognition environment

In this paper, we use a probabilistic divergence measure to identify radar sensor placements that yield high target classification rates. The derived divergence measure uses a lower bound of the Kullback-Leibler divergence to recognize significant differences in aspect-dependent target class probability distributions. Monte Carlo simulations are performed at various noise levels to demonstrate the similarity between the divergence measure and probabilities of correct classification (PCC). High range resolution (HRR) profiles are used as inputs to a multi-sensor classifier to identify the most probable target classification. Synthetic targets with dominant scatterers are employed to show the benefits of exploiting spatial diversity from prominent target features.