Sensors,
Journal Year:
2024,
Volume and Issue:
24(16), P. 5206 - 5206
Published: Aug. 11, 2024
Impact
craters
are
crucial
for
our
understanding
of
planetary
resources,
geological
ages,
and
the
history
evolution.
We
designed
a
novel
pseudo-spectral
spatial
feature
extraction
enhanced
fusion
(PSEF)
method
with
YOLO
network
to
address
problems
encountered
during
detection
numerous
densely
distributed
meter-sized
impact
on
lunar
surface.
The
illumination
incidence
edge
features,
isotropic
eigen
frequency
features
extracted
by
Sobel
filtering,
LoG
domain
bandpass
respectively.
Then,
PSEF
images
created
techniques
preserve
additional
details
from
original
DOM
data.
Moreover,
we
conducted
experiments
using
DES
optimize
post-processing
parameters
models,
thereby
determining
parameter
ranges
practical
deployment.
Compared
Basal
model,
model
exhibited
superior
performance,
as
indicated
multiple
measurement
metrics,
including
precision,
recall,
F1-score,
mAP,
robustness,
etc.
Additionally,
statistical
analysis
error
metrics
predicted
bounding
boxes
shows
that
performance
is
excellent
in
predicting
size,
shape,
location
craters.
These
advancements
offer
more
accurate
consistent
detect
surfaces,
providing
support
exploration
study
celestial
bodies
solar
system.
