On June 25, 1998, we performed a second set of tests at Fort AP Hill.  On this
trip, we were hosted by Tim Schweitzer and Ken Strittmatter from the Mine
Neutralization group at Fort Belvoir.  Prior to our trip, Tim and Ken prepared
a mine lane for our tests.  They added clutter to a set of anti-personnel mines
that had been buried earlier in a mine lane at AP Hill.  In all, the mine lane
contained 23 objects, the location of each marked with paint on the soil
directly above it.

We made our way through the mine lane twice, each time taking three to six
measurements on each object buried in the lane.  After completing the second
run-through, Tim and Ken revealed the identity of each of the 23 objects.

Significance of this Test

This test differed from previous ones in two principal ways.  First, this was
the first test we had performed that was truly blind, in which we knew neither
the nature nor the location of any of the test objects.  Second, this was the
first test we had performed on landmines located in reconsolidated soil.  In
other words, these mines had been buried for a significant amount of time prior
to our test, creating a more realistic test environment than our previous tests.

Contents of the Lane

In all, the lane contained:


As we proceeded through the lane, we used the measurements from the device and
our own senses to guess the identity of each object.  Our guesses as to the
identity of two of the objects differed on the two runs; for the other 21
objects, our guesses were the same on each run.  Our guesses fit into the broad
classifications of plastic mine, hard object, or soft object.  Here are the
results from the two runs:
First Run
mines detected:                           6/8
hard objects detected as mines:           2/8
soft object detected as mines:            5/7
Second Run
mines detected:                           8/8
hard objects detected as mines:           2/8
soft object detected as mines:            5/7

Performance Conclusions

The main problem with the device was the high number of false alarms when testing
"soft" objects such as wood sticks.  In its different stages of development, the
acoustic probe has always been more successful in distinguishing mines from harder
objects such as rocks and metal than from soft objects such as wood.  However, the
high rate of false alarm seen in these tests is a new phenomenon.  Given the
change in the severity of this problem between previous tests and this one, it is
possible that changes to the mechanical part of the probe prior to this test might
be responsible for the increased rate of false alarm.

General Conclusions

Due to its blind nature, this test gave us a better feel for the process of mine
discrimination in the real world.  We found that our own demining senses led us
to the same conclusion as the acoustic probe for the majority of the trials.  It
is hard in non-blind tests to understand the significance of the deminer's
experience in distinguishing a landmine from a clutter object.  But having
completed this test, the most realistic test we have performed, it seems that
personal experience is as powerful a tool as the acoustic probe.  If this is
true, then it does not make sense for deminers to equip themselves with an
acoustic probe.  Only a significant increase in information provided to the
deminer can offset the cost, complexity, fragility, training and additional
support required by an automated discrimination device.

Next Steps

  1. Revisit the physics issues behind the acoustic probe. Is there any way to pull
    out more features than the one we currently use? We will obtain additional technical

  2. Look into near-field RF method. We will make some lab measurements that investigate
    the dielectric properties of plastic AP mines versus ground clutter. Specifically, we
    will use a vector network analyzer driving a piece of open-ended hardline coax to observe
    complex reflection coefficients as a function of frequency, for different objects.