Advances in finance, accounting, and economics book series,
Год журнала:
2024,
Номер
unknown, С. 457 - 482
Опубликована: Дек. 27, 2024
This
chapter
pioneers
the
utilization
of
YOLOv8,
an
advanced
object
detection
algorithm,
as
a
transformative
tool
to
address
pressing
issues
faced
by
farmers
when
wild
animals
encroach
upon
their
lands.
The
comprehensive
pipeline,
spanning
from
custom
dataset
processing
YOLOv8
model
deployment,
establishes
robust
framework
for
effective
integration
deep
learning
algorithms,
enabling
real-
time
analysis
imagery
and
sensor
data
in
farmlands.
power
lies
its
streamlined
architecture,
eliminating
traditional
reliance
on
Convolutional
Neural
Networks
(CNNs)
Recurrent
(RNNs),
thereby
enhancing
computational
efficiency.
groundbreaking
technology
offers
scalable
solution,
ushering
new
era
sustainable
coexistence
between
agriculture
wildlife
management.
Drones,
Год журнала:
2023,
Номер
7(6), С. 398 - 398
Опубликована: Июнь 15, 2023
In
recent
years,
UAV
remote
sensing
has
gradually
attracted
the
attention
of
scientific
researchers
and
industry,
due
to
its
broad
application
prospects.
It
been
widely
used
in
agriculture,
forestry,
mining,
other
industries.
UAVs
can
be
flexibly
equipped
with
various
sensors,
such
as
optical,
infrared,
LIDAR,
become
an
essential
observation
platform.
Based
on
sensing,
obtain
many
high-resolution
images,
each
pixel
being
a
centimeter
or
millimeter.
The
purpose
this
paper
is
investigate
current
applications
well
aircraft
platforms,
data
types,
elements
category;
processing
methods,
etc.;
study
advantages
technology,
limitations,
promising
directions
that
still
lack
applications.
By
reviewing
papers
published
field
we
found
research
classified
into
four
categories
according
field:
(1)
Precision
including
crop
disease
observation,
yield
estimation,
environmental
observation;
(2)
Forestry
forest
identification,
disaster
(3)
Remote
power
systems;
(4)
Artificial
facilities
natural
environment.
We
image
(RGB,
multi-spectral,
hyper-spectral)
mainly
neural
network
methods;
monitoring,
multi-spectral
are
most
studied
type
data;
for
LIDAR
data,
end-to-end
method;
review
examines
development
process
certain
fields
implementation
some
predictions
made
about
possible
future
directions.
Real-time
crowd
monitoring
plays
a
pivotal
role
in
effectively
managing
public
spaces
and
ensuring
safety.
This
study
investigates
the
fusion
of
IoT
devices
YOLO
object
detection
model
to
accurately
count
crowds.
facilitate
instantaneous
collection
data
from
cameras,
while
adeptly
identifies
individuals
within
recorded
video
frames.
The
rigorously
assesses
performance
three
variants:
V5,
V8
NAS.
Findings
reveal
that
NAS
surpasses
V5
mean
average
precision
(mAP),
achieving
an
exceptional
mAP
95.1%.
heightened
is
attributed
integration
Neural
Architecture
Search
(NAS)
into
model,
fine-tuning
its
architecture
specifically
for
counting
tasks.
It
analyzes
various
networking
models
proposed
earlier
studies
analyzing
crowded
scenes
spaces.
emphasizes
potential
hybrid
involving
IP
camera
module
Deep
Network
effective
sensing.
In
this
setup,
captures
footage,
DNN
detects
density
based
on
people
recognized.
approach
presents
encouraging
solution
real-time
management
environments.
AgriEngineering,
Год журнала:
2024,
Номер
6(2), С. 1235 - 1251
Опубликована: Май 2, 2024
Accurate
feeding
management
in
aquaculture
relies
on
assessing
the
average
weight
of
aquatic
animals
during
their
growth
stages.
The
traditional
method
involves
using
a
labor-intensive
approach
and
may
impact
well-being
fish.
current
research
focuses
unique
way
estimating
red
tilapia’s
cage
culture
via
river,
which
employs
unmanned
aerial
vehicle
(UAV)
deep
learning
techniques.
described
includes
taking
pictures
by
means
UAV
then
applying
machine
algorithms
to
them,
such
as
convolutional
neural
networks
(CNNs),
extreme
gradient
boosting
(XGBoost),
Hybrid
CNN-XGBoost
model.
results
showed
that
CNN
model
achieved
its
accuracy
peak
after
60
epochs,
showing
accuracy,
precision,
recall,
F1
score
values
0.748
±
0.019,
0.750
0.740
0.014,
respectively.
XGBoost
reached
with
45
n_estimators,
recording
approximately
0.560
0.000
for
0.550
F1.
Regarding
model,
it
demonstrated
prediction
both
epochs
n_estimators.
value
was
around
0.760
precision
0.762
recall
0.754
0.752
0.019.
highest
compared
standalone
models
could
reduce
time
required
estimation
11.81%
CNN.
Although
testing
be
lower
than
those
from
previous
laboratory
studies,
this
discrepancy
is
attributed
real-world
conditions
settings,
involve
uncontrollable
factors.
To
enhance
we
recommend
increasing
sample
size
images
extending
data
collection
period
cover
one
year.
This
allows
comprehensive
understanding
seasonal
effects
evaluation
outcomes.
Applied Sciences,
Год журнала:
2024,
Номер
14(11), С. 4809 - 4809
Опубликована: Июнь 2, 2024
Maintaining
a
harmonious
balance
between
grassland
ecology
and
local
economic
development
necessitates
effective
management
of
livestock
resources.
Traditional
approaches
have
proven
inefficient,
highlighting
an
urgent
need
for
intelligent
solutions.
Accurate
identification
targets
is
pivotal
precise
farming
management.
However,
the
You
Only
Look
Once
version
8
(YOLOv8)
model
exhibits
limitations
in
accuracy
when
confronted
with
complex
backgrounds
densely
clustered
targets.
To
address
these
challenges,
this
study
proposes
optimized
CCS-YOLOv8
(Comprehensive
Contextual
Sensing
YOLOv8)
model.
First,
we
curated
comprehensive
detection
dataset
encompassing
Qinghai
region.
Second,
YOLOv8n
underwent
three
key
enhancements:
(1)
incorporating
Convolutional
Block
Attention
Module
(CBAM)
to
accentuate
salient
image
information,
thereby
boosting
feature
representational
power;
(2)
integrating
Content-Aware
ReAssembly
FEatures
(CARAFE)
operator
mitigate
irrelevant
interference,
improving
integrity
extraction;
(3)
introducing
dedicated
small
object
layer
capture
finer
details,
enhancing
recognition
smaller
Experimental
results
on
our
demonstrate
model’s
superior
performance,
achieving
84.1%
precision,
82.2%
recall,
84.4%
[email protected],
60.3%
[email protected],
53.6%
[email protected]:0.95,
83.1%
F1-score.
These
metrics
reflect
substantial
improvements
1.1%,
7.9%,
5.8%,
6.6%,
4.8%,
4.7%,
respectively,
over
baseline
Compared
mainstream
models,
strikes
optimal
real-time
processing
capability.
Its
robustness
further
validated
VisDrone2019
dataset.
The
enables
rapid
accurate
age
groups
species,
effectively
overcoming
challenges
posed
by
It
offers
novel
strategy
population
overgrazing
prevention,
aligning
seamlessly
demands
modern
precision
farming.
Moreover,
it
promotes
environmental
conservation
fosters
sustainable
within
industry.
SAE technical papers on CD-ROM/SAE technical paper series,
Год журнала:
2025,
Номер
1
Опубликована: Фев. 7, 2025
<div
class="section
abstract"><div
class="htmlview
paragraph">Human-wildlife
conflicts
pose
significant
challenges
to
both
conservation
efforts
and
community
well-being.
As
these
escalate
globally,
innovative
technologies
become
imperative
for
effective
humane
management
strategies.
This
paper
presents
an
integrated
autonomous
drone
solution
designed
mitigate
human-wildlife
by
leveraging
in
surveillance
artificial
intelligence.
The
proposed
system
consists
of
stationary
IR
cameras
that
are
setup
within
the
conflict
prone
areas,
which
utilizes
machine
learning
identify
presence
wild
animals
send
corresponding
location
a
docking
station.
An
equipped
with
high-resolution
sensors
is
deployed
from
station
provided
location.
camera
object
detection
technology
scan
specified
zone
detect
animal
emit
repelling
ultrasonic
sound
device
achieve
non-invasive
deterrence
provides
approaches
develop
algorithms,
optimize
strategies,
adapt
evolving
dynamics
wildlife
behavior.
promising
avenue
addressing
conflicts,
promoting
coexistence,
contributing
broader
field
technology-driven
ecological
management.</div></div>
