Automatic target detection in heterogeneous and spatially correlated Lognormal clutter

In this paper, we address the problem of target detection in a spatially correlated Lognormal clutter when no prior information is available about the presence or not of any clutter edge and/or interferences in the reference window. We propose a detector based on censoring and detection algorithms with constant false censoring and alarm rates (CFCAR). The CFCAR property is guaranteed by use of linear bi-parametric adaptive thresholds based on the best linear unbiased estimators (BLUEs) of the Normal parameters obtained via a logarithmic transformation. The censoring algorithm starts up the heterogeneity test from the two most left ranked cells and proceeds forwardly. The detection algorithm uses the set of reference cells supposed homogeneous to estimate the unknown background level. Monte Carlo simulations show that the proposed detector and its corresponding fixed-point(s) censoring BLU-CFAR detector perform equally well.