Robust variability index CFAR for non-homogeneous background

Radar signal detection using constant false alarm rate (CFAR) detectors encounters many non-ideal situations making it difficult to characterise the background. These include the presence of multiple targets, clutter edges and their combination in the reference window. Designing an efficient CFAR for these situations is a non-trivial problem. Algorithms based on ordered statistics (OS), outlier rejection using sorting and sample by sample hypothesis testing, variability index (VI), ordered data VI are proposed in the literature. These approaches require expensive sorting or prior information on the depth of censoring. In this study, the authors propose robust VI CFAR (RVI-CFAR) that obviates sorting. RVI-CFAR computes the threshold in multiple stages. The first stage uses VI-CFAR to determine an adaptive threshold. Outlier rejection in the computation of background, mean ratio (MR) and VI is carried out in subsequent stages. The updated MR and VI statistics are used to refine switching decisions at every stage of processing. RVI-CFAR exhibits low CFAR loss in homogeneous and multiple target scenarios, meanwhile achieving superior performance compared to other censored CFAR techniques. The proposed RVI-CFAR is evaluated and shown to be robust for all the cases of non-homogeneity compared to OS CFAR.