BioResources,
Journal Year:
2024,
Volume and Issue:
19(3), P. 4531 - 4546
Published: May 20, 2024
The
quality
detection
of
alfalfa
hay
is
crucial
for
the
development
animal
husbandry.
In
this
study,
a
method
based
on
fusion
multisource
information
including
near-infrared
spectroscopy,
image
processing
techniques,
and
electronic
nose
proposed.
After
SG
convolution
smoothing,
feature
wavelengths
were
extracted
using
Competitive
Adaptive
Re-weighting
Scheme
Successive
Projections
Algorithm
from
spectral
data.
data
denoised
adaptive
wavelet
thresholding,
color
texture
features
histograms
random
forest
algorithms,
respectively.
Electronic
principal
component
analysis
was
used
dimensionality
reduction.
Support
Vector
Machine,
Extreme
Learning
Multi-Layer
Perceptron
employed
to
establish
models
image,
gas
information,
their
combination,
Experimental
results
demonstrate
that
data,
effectively
enhances
classification
accuracy
model.
test
set
reaches
100%,
with
root
mean
square
error
determination
coefficient
values
0.1728
0.9239,
respectively,
surpassing
prediction
established
solely
individual
information.
This
study
provides
new
insights
into
detection.
Monitoring
chicken
freshness
is
vital
for
food
safety,
reducing
the
risk
of
foodborne
illnesses.
Electronic
noses
(E-nose)
have
emerged
as
valuable
tools
assessing
quality
by
detecting
changes
in
volatile
compounds
linked
to
freshness.
We
developed
a
portable
E-nose
assess
stored
at
room
temperature
and
refrigerator
(4
°C)
45
days.
tracked
bacterial
growth
through
Total
Plate
Count
(TPC)
identified
organic
(VOCs)
using
Gas
Chromatography-Mass
Spectrometry
(GC-MS)
over
storage
period.
utilized
Polynomial
Feature
Extractions
(poly0,
poly1,
poly2)
analyze
sensor
responses.
Principal
Component
Analysis
(PCA)
reduced
data
dimensionality
visualization,
followed
Linear
Discriminant
(LDA)
classification.
LDA,
coupled
with
poly2,
effectively
classified
based
on
days
temperatures.
To
predict
freshness,
we
employed
Support
Vector
Regression
(SVR)
our
device.
Results
showed
high
predictive
accuracy:
values
0.95
0.93
0.74
3.89
under
refrigerated
conditions,
respectively.
Our
study
demonstrated
that
VOCs
correlate
samples
varying
durations
conditions.
Notably,
poly2
feature
extraction
model
excelled
predicting
surpassing
other
polynomial
orders.
This
highlights
significance
an
appropriate
discerning
different
conditions
durations.
showcases
potential
technology
monitoring
offering
insights
into
dynamics
VOC
various
settings.
