Abstract:
The data collected by the NIITEK GPR, high range resolution (HRR) ground-penetrating
radar (GPR), results in a digitized raw video representing signals re
ected at the surface
of as well as internal to the landmine due to changes in impedances, materials dielectric
properties. This digitized signal has to undergo several stages of preprocessing in order to
produce a binary-valued- sequence. A part of this sequence contains a speci c length of
string that is a characteristic of a mine pattern. The mine pattern, to be recognized from
a longer string of processed GPR data, has to be presented to a landmine detector. The
landmine detector not only detects the mine but also classi es them and discriminates the
mines from clutter (noise).
Three pattern recognizers, one reset Finite State Machine (FSM) and two behaviorally
equivalent Parallel Correlators are designed to detect multiple landmines simultaneously.
Alternative implementations of these processing modules are compared with respect to chip
area in terms of number of slices (real-estate or chip area) and speed (processing time), as a function of the number of landmines to be simultaneously recognized. It is found that
the reset FSM is smallest in size of all the architectures but the slowest, whereas, the rst
Parallel Correlator is largest in size and the second Correlator, the fastest.
An alternative pattern recognizer, a non-reset Finite State Machine, sometimes known as
a Rabin-Scott Machine, is also analyzed in terms of chip area (slices) and speed(processing
time) but with respect to the relevant parameter, the maximum number of states in FSM.
It was apparent that both- the size as well as the speed of the FSM increases with the
increase in the number of states.
