A discriminant analysis-based automatic ordered statistics scheme for radar systems

The ordered statistics (OS) scheme is an effective constant false alarm rate (CFAR) technique deployed in many radar systems. It is widely deployed because of its simplicity and effectiveness under conditions of both homogeneous and non-homogeneous radar returns. However, the problem of inaccurate censoring typically degrades its performance since it is often difficult to accurately determine the actual number of interfering targets and clutter edges in the reference window per time. In this article, we address this problem based on the principle of discriminant analysis (DA) towards automatically and effectively estimating the kth rank ordered element of the OS scheme. Our scheme, termed the DA-OS scheme, works without requiring a priori knowledge about the statistical characteristics of the input radar returns. The results obtained via Monte Carlo simulation indicate that the DA-OS scheme achieves a small CFAR loss of about 0.392 dB relative to the cell averaging (CA) scheme under conditions of homogeneous radar returns at a probability of detection of 0.5. It outperforms other notable traditional schemes, including the CA, smallest-of CA, greatest-of CA, and the fixed OS schemes under conditions of non-homogeneous radar returns. Finally, it provides a number of desirable qualitative characteristics as against other existing censoring techniques. •A new censoring method is proposed for the ordered statistics (OS) CFAR scheme.•The proposed scheme does not depend on prior information about the radar return.•The BCV metric is used to determine the homogeneity/non-homogeneity of radar returns.•Our scheme outperforms other CFAR schemes such as the CA, SOCA, GOCA, and OS schemes.•Our scheme provides improved characteristics against other censoring schemes.