Case Studies on Generative Adversarial Networks in Precision Farming
Advances in geospatial technologies book series,
Год журнала:
2025,
Номер
unknown, С. 291 - 320
Опубликована: Апрель 30, 2025
The
chapter
reviews
the
applicability
of
Generative
Adversarial
Networks
in
precision
agriculture,
with
an
emphasis
on
its
role
enhancing
remote
sensing
technology.
This
ranges
from
resolution
augmentation
for
satellite
and
drone
images
using
GAN-based
models
like
SRGAN
CycleGAN
to
generating
synthetic
data
training
that
will
help
crop
health
monitoring,
soil
analysis,
yield
prediction.
case
study
demonstrates
tremendous
improvements
image
quality
decision-making,
further
reach
into
weather
simulation,
real-time
UAV
IoT
integration.
Язык: Английский
Single-Shot X-ray to Multi-View Projections for 3D Pork Shoulder Bone Analysis
Research Square (Research Square),
Год журнала:
2025,
Номер
unknown
Опубликована: Май 22, 2025
Abstract
Pork
is
an
important
meat
product
for
the
European
Union,
which
exported
over
4.2
million
tons
in
2023,
valued
at
€8.1
billion.
Automating
labor-intensive
deboning
process
of
significant
interest,
particularly
through
development
advanced
inline
inspection
systems
capable
analyzing
pork
shoulder
bone
structures.
While
computed
tomography
(CT)
provide
high-contrast
3D
reconstructions,
their
large
size
and
high-cost
present
substantial
barriers
to
adoption
industrial
processing.
This
study
addresses
these
challenges
by
introducing
a
novel
approach
that
uses
single
X-ray
projection
combination
with
deep
neural
networks
predict
segmentation
map
structures
using
conventional
reconstruction
algorithms.
To
this
end,
U-Net
network
variants
were
trained
on
high-resolution
CT
scans
90
shoulders.
These
augmented
synthetic
data
simulate
different
orientations
conveyor
belt,
ensuring
model’s
robustness.
The
minimum
number
projections
needed
accurate
was
determined
based
simulations,
60
evenly
spaced
between
0°
180°
found
optimal.
Feldkamp-Davis-Kress
(FDK)
algorithm
chosen
its
efficiency
cost-effectiveness
model
achieved
Dice
score
0.94
SSIM
0.96
test
data,
demonstrating
ability
59
missing
reconstruct
structure
accurately.
method
proposed
paper
has
potential
advance
processing
enhancing
precision,
reducing
waste,
streamlining
operations.
Язык: Английский
Improving Non-Line-of-Sight Identification in Cellular Positioning Systems Using a Deep Autoencoding and Generative Adversarial Network Model
Sensors,
Год журнала:
2024,
Номер
24(19), С. 6494 - 6494
Опубликована: Окт. 9, 2024
Positioning
service
is
a
critical
technology
that
bridges
the
physical
world
with
digital
information,
significantly
enhancing
efficiency
and
convenience
in
life
work.
The
evolution
of
5G
has
proven
positioning
services
are
integral
components
current
future
cellular
networks.
However,
accuracy
hindered
by
non-line-of-sight
(NLoS)
propagation,
which
severely
affects
measurements
angles
delays.
In
this
study,
we
introduced
deep
autoencoding
channel
transform-generative
adversarial
network
model
utilizes
line-of-sight
(LoS)
samples
as
singular
category
training
set
to
fully
extract
latent
features
LoS,
ultimately
employing
discriminator
an
NLoS
identifier.
We
validated
proposed
indoor
factory
(dense
clutter,
low
base
station)
scenarios
assessing
its
generalization
capability
across
different
scenarios.
results
indicate
that,
compared
state-of-the-art
method,
markedly
diminished
utilization
device
resources
achieved
2.15%
higher
area
under
curve
while
reducing
computing
time
12.6%.
This
approach
holds
promise
for
deployment
terminals
achieve
superior
localization
precision,
catering
commercial
industrial
Internet
Things
applications.
Язык: Английский
Unveiling the Urban Morphology of Small Towns in the Eastern Qinba Mountains: Integrating Earth Observation and Morphometric Analysis
Buildings,
Год журнала:
2024,
Номер
14(7), С. 2015 - 2015
Опубликована: Июль 2, 2024
In
the
context
of
current
information
age,
leveraging
Earth
observation
(EO)
technology
and
spatial
analysis
methods
enables
a
more
accurate
understanding
characteristics
small
towns.
This
study
conducted
an
in-depth
urban
morphology
towns
in
Qinba
Mountain
Area
Southern
Shaanxi
by
employing
large-scale
data
innovative
form
measurement
methods.
The
U-Net3+
model,
based
on
deep
learning
technology,
combined
with
concave
hull
algorithm,
was
used
to
extract
precisely
define
boundaries
31,799
buildings
morphological
town
core
were
measured,
areas
defined
using
calculated
tessellation
cells.
Hierarchical
clustering
applied
analyze
12
characteristic
indicators
89
towns,
various
metrics
determine
optimal
number
clusters.
identified
eight
distinct
clusters
towns’
differences.
Significant
differences
between
observed.
results
revealed
that
exhibited
diverse
shapes
distributions,
ranging
from
irregular
sparse
compact
dense
forms,
reflecting
layout
patterns
influenced
unique
each
town.
use
morphometric
method,
cellular
biological
morphometry,
provided
new
perspective
deepened
structure
micro
perspective.
These
findings
not
only
contribute
development
quantitative
for
planning
but
also
demonstrate
novel,
data-driven
approach
conventional
studies.
Язык: Английский