Improved distributed automatic target recognition performance via spatial diversity and data fusion

Abstract

Radar target classification is examined from the viewpoint of improving classification performance through the use of spatial diversity. Improved radar target classification has been demonstrated previously by using at least one additional perspective in a generic environment but the impact of sensor placement has been less studied. In this paper, we examine the use of multiple high range resolution (HRR) profiles to demonstrate how selection of sensor locations can improve classification rates. Specifically, performance improvements are demonstrated after identifying the optimal set of perspectives and employing a simple decision fusion network (DFN) algorithm for defined signal-to-noise (SNR) levels. We show percentages of correct classification (PCC) can be maintained in scenarios where SNR has been reduced by up to 9 dB on a single sensor basis.