TABLE OF CONTENTS The amount of data required for adaptation is of the order of magnitude of 2 N M for the optimum processor (Chapter 4) and 2 LK for the space-time FIR filter processor after Chapter 7(see REED et al. [421]).
Basically the adaptation of the clutter filter (e.g., updating the space-time covariance matrix) [7] can be done by use of clutter echo data either gathered form the time or range dimensions or both. In the case of sidelooking array radar the range dimension is the preferable choice because the clutter Doppler is independent of range.
In the case or a forward looking radar the clutter Doppler depends strongly on range which requires that the clutter filter has to be updated while running through the data along the range dimension. This causes additional computational load for the adaptation.
However, we are faced with another, even more stringent dilemma. On the one hand the amount of data associated with a sufficient narrow Doppler band may not be sufficient for adaptation, so that losses in SNIR may occur. On the other hand, if data are taken form a large number of range elements, all of them being associated with different Doppler frequencies, we can expect a broadening of the clutter notch according to the Doppler bandwidth in the data which results again in degradation of slow target detection. As a compromise the incoming data may be subdivided into...
