Using Measured RCS in a Serial, Decentralized Fusion Approach to Radar-Target Classification

A decentralized fusion problem is a hypothesis-testing problem where the decision algorithm is not provided with raw measurements but rather is provided only with results from so-called local decision functions. In a serial decentralized fusion problem, these local decisions are fused into a global decision one-by-one until all local decisions have been included or until a stopping criterion is achieved. In this paper, we consider serial decision fusion with a single radar sensor using a local decision function based on the radar cross section (RCS) characteristics measured across multiple dwells. In a single-sensor situation, the serial topology is convenient because local decisions are naturally sequential in time. A serial topology also supports the calculation of system performance measures and adapts easily to a distributed or multi-sensor situation. From dwell to dwell, measured RCS fluctuates due to noise, errors, and target scintillation. By examining the results from multiple dwells, the measured RCS can be separated into fixed-amplitude and Rayleigh components. The relative magnitudes of these components vary depending on both target type and sensor-to-target geometry. Spreading measurements and local decisions in time allows the sensor to observe a moving target from different aspect angles and thus introduces changing geometry into the decision data. In this way local decisions at different times are based on observations of different fixed to Rayleigh-amplitude ratios. Results based on a high-fidelity radar simulation along with different fixed- versus-Rayleigh RCS profiles are used to investigate issues related to serial multi-hypothesis decentralized fusion. Simulation results and performance measures demonstrating the efficacy of the proposed method are presented and sensitivities with respect to RCS profiles and to local- decision rates are investigated.