Application of the variability index (VI) statistic to radar CFAR processing

Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)


Electrical Engineering and Computer Science


Pramod K. Varshney


constant false alarm rate (CFAR target detection

Subject Categories

Electrical and Computer Engineering


The goal of constant false alarm rate (CFAR) processing is to perform target detection in an environment of unknown, time-varying noise/clutter power such that CFAR performance is maintained when no target is present. A typical CFAR processor dynamically determines a detection threshold based upon a scaled estimate of the local noise/clutter power. Existing CFAR algorithms are often effective for specific conditions, but seldom provide robust behavior in the variety of environments encountered in practice.

A new CFAR processor, known as the VI-CFAR, performs adaptive threshold target detection using a composite approach based on the well known CA-CFAR, SO-CFAR, and GO-CFAR algorithms. After envelope detection, radar samples are stored in a tapped delay line such that a test cell is surrounded on either side by a set of reference cells. The VI-CFAR dynamically chooses either the leading, lagging, or combined set of reference cells for background power estimation. Selection of the reference cells and the background estimation algorithm are based on the ratio of the means of the two half reference windows and on the "variability index (VI)" values calculated for the leading and lagging cells. Hypothesis tests using the variability indices and the mean ratio determine if the environment is homogeneous, contains multiple targets, or contains an extended clutter edge. Based on the decision, the VI-CFAR intelligently tailors the adaptive threshold calculation.

The performance of the VI-CFAR processor is determined for the square-law detector in homogeneous, multiple target, and clutter edge environments. The results are extended to the case of a linear-law detector and a two-channel linear detector with frequency diversity. The mathematical basis for the VI statistic is examined. In addition, the application of the VI statistic to the detection of homogeneity/non-homogeneity in non-Gaussian environments is investigated.

The VI-CFAR is shown to provide performance in a homogeneous environment comparable to the CA-CFAR. Multiple target detection performance is similar to the OS-CFAR, but with better false alarm performance in a clutter edge. Application to non-Gaussian environments is also feasible.


Surface provides description only. Full text is available to ProQuest subscribers. Ask your Librarian for assistance.