A year and a half ago, when radar technologist Elliott Brown visited the
dolphins and beluga whales performing mine hunting exercises for the US Navy in
San Diego, he was impressed with their hypersensitive natural sonar. These
marine mammals could identify small objects lodged in the coastal seabed far
more easily than anything the Navy has developed itself
(see 鈥淔lipper鈥檚 secret鈥�, 快猫短视频, 28 June 1997).
The only thing that hampered the animals鈥� effectiveness was the difficulty they
had communicating with their human handlers.
Clearing mines from the cluttered, turbid waters near a coastline is a tricky
business. Conventional sonar systems quickly become confused because their sound
waves get distorted. Often the signals are reflected many times between the
seafloor and the surface. Turbulence and changes in temperature and salinity can
also bend sound waves, making matters worse. And even if a clear signal can be
obtained, the resolution of most sonar systems is too poor to identify many
modern underwater mines, which can be less than a metre across.
As a result, the world鈥檚 navies rely on divers to search for and disarm
underwater mines, and the limited visibility in coastal waters means that mine
clearance is painfully slow. Often it is easier to avoid mine-infested coastal
waters altogether. During the 1991 Gulf War, the Allied forces ruled out a
marine invasion of Kuwait because the coastal waters could not be adequately
cleared of mines. Because of this impotence, the US Navy has decided that
developing an effective way of spotting underwater mines is crucial.
Advertisement
Sound and vision
Brown is a programme manager at the Defense Advanced Projects Agency (DARPA)
in Arlington, Virginia, a research and development branch of the US military
which has an annual budget of over $2 billion. Inspired by the
performance of the marine mammals, Brown returned to his office determined to
develop an acoustic system that could rival the dolphins鈥� and whales鈥� sonar. His
goal was to build a small battery-operated underwater video camera that produced
images using sound rather than light. The camera would have to produce sharp
images of small objects in very shallow water at distances of between 3 and 10
metres. Most important of all, it would have to cope with the severe acoustic
conditions that exist in coastal waters. 鈥淭herein lies the challenge,鈥� he
says.
In principle, Brown鈥檚 acoustic video camera works just like a conventional
underwater camera. It has the acoustic equivalent of a spotlight to illuminate
the scene. The reflected sound waves are then focused by an acoustic lens onto
an array of sound-sensitive pixels. Each of these creates an electronic signal
and a computer uses this data to reconstruct an image of the scene which is then
displayed.
In practice, however, Brown鈥檚 sonoelectronics programme faces huge
challenges. Resolving objects only a few centimetres across requires sound with
a frequency of a few megahertz. This frequency is at least an order of magnitude
higher than frequencies used by conventional sonar systems and it is rapidly
absorbed as it travels though water. This limits the range of the camera to a
few metres. However, lower frequencies are less strongly absorbed, so Brown can
increase the range by reducing frequency giving the camera the ability to
鈥渮oom鈥�. Ultimately, the camera will operate at 3 megahertz and have a range of
between 1 and 5 metres.
Silicon carvings
Creating and detecting these high frequencies is not easy. Acoustic waves
must be generated by a vibrating structure that sets up pressure waves in the
water. The higher the frequency, the faster this structure must vibrate, and
conventional microphones simply do not work at this level. So Brown鈥檚 programme
is developing microphones carved out of silicon chips that can turn electrical
energy into high-frequency sound waves and vice versa
(see Diagram).
One of Brown鈥檚 most promising designs consists of a tiny silicon membrane
that makes up half of a tiny capacitor, an example of a microelectromechanical
device. By changing the electric field inside the capacitor, the membrane can be
made to vibrate, sending out pressure waves. The big advantage of this device is
that it can operate in reverse. A pressure wave hitting the membrane will set it
vibrating. And by monitoring the way this vibration changes the capacitance of
the device, it is possible to determine the frequency and amplitude of the
incoming wave. Brown is hoping that a 10-centimetre-square array of
approximately 1000 pixels will both illuminate the scene and pick up the
echoes.
The task is complicated by the fact that the sound sent out cannot be made up
from ordinary sine waves. These are coherent, laser-like acoustic emissions
where all the waves are in step, and they cause a phenomenon called speckle.
Coherent sound waves can be thought of as a series of flat wavefronts moving
towards an object. Ideally, the wavefront should reflect off a single point on
the object and return to the camera. But because objects鈥攑articularly
artificial ones鈥攈ave many flat surfaces, a single wave can be reflected at
several points along its front. These extra echoes appear as bright spots or
speckle on the image. 鈥淵ou get all kinds of unwanted reflections when you use
coherent illumination and this confuses the imaging system,鈥� says Brown.
Minimising speckle is an important issue. Another way to illuminate the scene
is with an incoherent sound source鈥攖he acoustic equivalent of a household
lightbulb. But unless the illumination is hugely powerful, noise-like signals
are difficult to distinguish from the natural noise of surf and turbulence. A
diver could not carry enough batteries to power such a strong signal.
So the sonoelectronics programme is testing a technique that creates sound
with both coherent and noise-like properties. The idea is to use a random number
generator to determine the shape of the acoustic wavefronts. This gives them a
noise-like quality that is ideal for illumination. But because the shape of this
wavefront is known, these signals can be distinguished from natural noise. This
type of signal is known as quasi-coherent noise and telecommunications companies
use a similar technique to transmit low-power radio signals through the
atmosphere. The military also uses the technique to camouflage their signals in
a background of noise.
But generating quasi-coherent noise and then distinguishing it from real
noise requires heavy-duty computer processing power. Squeezing this into a
camera the size of a shoebox is not easy. 鈥淚t鈥檚 one of the major challenges we
face,鈥� says Brown.
Building a sound-sensitive array that is also the source of illumination
creates other problems. The connections to each pixel must be capable of
carrying signals both to and from the processing unit. This is not
straightforward since many electronic components鈥攄iodes, for
example鈥攐nly work when signals flow in one direction. Alternatively, each
pixel could have two connections, one for the incoming signal and one for the
outgoing one. But this vastly increases the complexity of any circuitry that has
to be designed. And compared to arrays that are optimised for one task or the
other, such an arrangement is also power hungry. Brown has set himself a design
goal specifying that the camera must work for two hours on batteries. 鈥淭his
places severe constraints on the amount of power we can use.鈥�
Acoustic lens
By comparison, designing the lens for an acoustic camera is straightforward.
鈥淭here is a very strong analogy between acoustics and optics,鈥� explains Brown.
It turns out that many of the laws of optics translate well into the acoustic
world. In fact, an acoustic lens is simply a lens-shaped disc of plastic. The
sound waves are refracted by the material and focused by its shape in the same
way that an optical lens focuses light.
A more challenging problem is turning the electronic signals from the pixel
array into a meaningful image. An optical video camera displays the pattern of
light that hits the pixel array. These pixels produce electronic signals that
are proportional to the amount of light that hits them and can then be used to
control the brightness of pixels in the display.
Life is not so simple in the acoustic world. As well as recording the pattern
of sound waves that hit the pixel array, the acoustic video camera records the
time delay between the illuminating flash and the echo. This gives the object鈥檚
range. So instead of recording a two-dimensional scene, the camera records
information about a three-dimensional volume.
鈥淚mage processing is difficult,鈥� says Brown. Simply rendering an image takes
the kind of processing power found only in a workstation, 鈥渁nd we need to do it
on a single processor鈥�. For the moment he is hoping to find a way of displaying
raw data that a diver could make sense of intuitively. 鈥淭hat could save a lot of
processing power.鈥�
For the display itself, Brown is hoping to rely on other research groups
within DARPA. 鈥淭here are a lot of people working on displays and we hope to
piggyback on that.鈥� The idea is to use a heads-up display in which the image is
projected inside the diver鈥檚 face mask.
Brown hopes to have all the pieces working together within approximately two
years. To carry out this integration, he has employed the services of two
contractors in Massachusetts鈥擫ockheed-Martin IR Imaging Systems in
Lexington and Teratech Corporation in Burlington.
If all goes to plan, the acoustic video camera should find applications
outside the military. For example, Brown would like to modify it so that it can
work in air as well as water. Such a camera would be able to peer through thick
smoke and fog, which could be a boon for fire fighters.
There are other spin-offs too. The acoustic illumination system has to throw
very narrow beams of sound onto the subject area. The techniques Brown is
developing for this could be used to develop highly directional microphones and
speakers that would be aimed electronically rather than mechanically.
Brown may also modify the camera to peer into solid materials. This would
offer a way of spotting defects inside solid objects and could provide high-quality
ultrasound images for surgeons carrying out certain kinds of non-invasive surgery.
DARPA also anticipates other military uses for the technology such as better
acoustic communications links with submarines or for divers. But whether the
navy will ever be able to outdo the dolphins and the beluga whales remains to be
seen.