Biomedical Engineering Reference
In-Depth Information
preferred direction and a large negative response
in the null direction.
In a physical sensor system based on this
algorithm, the motion detector may be tuned to
specific speeds of interest by varying the time
delay and the physical displacement between a
pair of sensing elements. An array of such detec-
tors could be developed to detect a wide range
of motion velocities and directions.
In addition to motion processing, other vision
processes of interest include light adaptation
and lateral inhibition.
1.3.3.4 Lateral Inhibition
Another important vision process is
lateral inhi-
bition
. Lateral inhibition allows an excited cell to
mediate the response of its neighbors. This pro-
vides for an overall improvement of response.
The concept of lateral inhibition was investi-
gated by Hartline
et al.
in the horseshoe crab
(
Limulus polyphemus
)
[62]
.
The horseshoe crab is equipped with com-
pound eyes with approximately 800 ommatidia
per eye. The visual axis of the ommatidia
diverge, with a slight overlap between adjacent
pairs. Hartline
et al.
reported that each omma-
tidium functions as a single receptor unit. That
is, each ommatidia has its own single nerve
fiber. There is also an extensive system of cross-
connecting strands of nerve fibers
[62]
.
In a series of experiments, Hartline
et al.
illu-
minated a single ommatidium to elicit a response
in its corresponding nerve fiber. The nerve fiber
provided an output of approximately 65 impulses
per second. Ommatidia near the original were
then illuminated to determine their effect on the
response. Hartline
et al.
found that when the sur-
rounding light was activated, it inhibited the
response and decreased the number of pulses.
When the surrounding light was removed, the
original ommatidial response returned to its pre-
vious value
[62]
. For this investigative work in
lateral inhibition, Ragnar Granit, Haldan Hart-
line, and George Wald received the 1967 Nobel
Prize for Physiology or Medicine. Recently,
Strube has modeled lateral inhibition in an array
of fly-inspired biomimetic ommatidia and
showed a significant improvement in edge
detection capability
[63]
. Also, Petkov
et al.
have
used the response of Gabor filters to model late-
ral inhibition
[64]
.
1.3.3.3 Light Adaptation
Many vision systems must be able to view
scenes with illumination levels that span mul-
tiple orders of magnitude. That is, the system
must be able to operate from very low-light con-
ditions to very high levels of illumination. This
requires some mechanism for adaptation over a
wide range of illuminance levels. Different spe-
cies use a variety of
light adaption
mechanisms,
including photomechanical mechanisms such
as pupil changes, intensity-dependent summa-
tion mechanisms in both space and time, pho-
tochemical processes, and neuronal responses
[58]
.
In
M. domestica
, two different processes are
employed: chemical mediation of the photore-
ceptor response and membrance modification of
the rhabdomeres to adjust the amount of light
reaching the photoreceptors
[58]
. The fly further
conditions the light via a
log transform-subtraction-
multiplication
cellular-based algorithm. The
signal is first logarithmically compressed. The
average of surrounding ommatidia is then sub-
tracted from the log-compressed signal, which
removes the mean background illumination.
The final multiplication step provides a gain to
match the signal to the dynamic range of the
monopolar cells
[58]
. Dean has used a special
implementation of this algorithm to develop a
sensor system that operates over a wide dynamic
illumination range
[59-61]
.
1.3.3.5 Navigation
Srinivasan has studied the honeybee in great
detail and developed a hypothesis sup-
ported by experimental observations that have
