Page 526 - Fundamentals of Radar Signal Processing

FIGURE 6.29 versus threshold scale factor α in order statistics CFAR.
OS
The selected order statistic is chosen as approximately 0.75N.

Figure 6.30 shows the performance of the OS CFAR on the same data set

considered previously using N = 20 again and choosing the k = 15th order
statistic to set the threshold. The use of the ordered statistic instead of a mean
estimate makes the detector almost completely insensitive to masking by closely
spaced targets so long as the number of cells contaminated by interfering targets
does not exceed N – k. In this example, both closely spaced targets are readily
detected. Although they were used in Fig. 6.30 for consistency, guard cells are
less important in an OS CFAR since the rank ordering process will not be
affected by targets spreading outside of the test cell. The effect of the window

length and choice of order statistic k on the behavior at clutter edges are
discussed in (Rohling, 1983). If k ≤ N/2, there will be extensive false alarms at
clutter edges. Thus, k is usually chosen to satisfy N/2 < k < N. Typically, k is on
the order of 0.75N (Nathanson, 1991).