arXiv (Cornell University),
Journal Year:
2022,
Volume and Issue:
unknown
Published: Jan. 1, 2022
Lane
marking
detection
is
fundamental
for
both
advanced
driving
assistance
systems.
However,
detecting
lane
highly
challenging
when
the
visibility
of
a
road
low
due
to
real-life
environment
and
adverse
weather.
Most
methods
suffer
from
four
types
challenges:
(i)
light
effects
i.e.,
shadow,
glare
light,
reflection
etc.;
(ii)
Obscured
eroded,
blurred,
colored
cracked
caused
by
natural
disasters
weather;
(iii)
occlusion
different
objects
surroundings
(wiper,
vehicles
etc.);
(iv)
presence
confusing
like
lines
inside
view
e.g.,
guardrails,
pavement
marking,
divider
etc.
Here,
we
propose
robust
tracking
method
with
three
key
technologies.
First,
introduce
comprehensive
intensity
threshold
range
(CITR)
improve
performance
canny
operator
in
edges.
Second,
two-step
verification
technique,
angle
based
geometric
constraint
(AGC)
length-based
(LGC)
followed
Hough
Transform,
verify
characteristics
prevent
incorrect
detection.
Finally,
novel
defining
horizontal
position
(RHLP)
along
x
axis
which
will
be
updating
respect
previous
frame.
It
can
keep
track
either
left
or
right
markings
are
partially
fully
invisible.
To
evaluate
proposed
used
DSDLDE
[1]
SLD
[2]
dataset
1080x1920
480x720
resolutions
at
24
25
frames/sec
respectively.
Experimental
results
show
that
average
rate
97.55%,
processing
time
22.33
msec/frame,
outperform
state
of-the-art
method.
Authorea (Authorea),
Journal Year:
2024,
Volume and Issue:
unknown
Published: Aug. 1, 2024
Lane
detection
is
a
critical
component
of
autonomous
driving
systems,
enabling
vehicles
to
identify
and
navigate
within
lanes
accurately.
This
paper
presents
novel
approach
enhancing
lane
ac
curacy
using
the
Mask
R-CNN
algorithm.
By
leveraging
capabilities
R-CNN,
proposed
algorithm
demonstrates
efficient
precise
road
lanes,
including
classification
types
angle
evaluation
for
steer
ing
purposes.
The
algorithm's
functionality
encompasses
determining
bounding
boxes
through
image
cropping,
classification,
data
configuration
schematic
environ
mental
surveillance.
Through
extensive
testing,
has
shown
superior
performance
in
scenarios
with
challenging
conditions
such
as
insufficient
lighting
line
degradation.
results
indicate
significant
improvement
accuracy,
making
it
promising
solution
advancing
systems.
This
paper
describes
the
implementation
of
a
learning-based
lane
detection
algorithm
on
an
Autonomous
Mobile
Robot.
It
aims
to
implement
Ultra
Fast
Lane
Detection
for
real-time
application
SEATER
P2MC-BRIN
prototype
using
camera
and
optimize
its
performance
Jetson
Nano
platform.
Preliminary
experiments
were
conducted
evaluate
algorithm's
in
terms
data
processing
speed
accuracy
two
types
datasets:
outdoor
public
dataset
indoor
internal
from
area
BRIN
Workshop
Building
Bandung.
The
revealed
that
runs
more
optimally
platform
after
conversion
TensorRT
compared
ONNX
model,
achieving
speeds
approximately
101
ms
CULane
105
TuSimple,
which
is
about
22
times
faster
than
previous
model.
While
demonstrates
good
dataset,
falls
short
dataset.
Future
work
should
focus
transfer
learning
fine-tuning
enhance
accuracy.
2021 5th CAA International Conference on Vehicular Control and Intelligence (CVCI),
Journal Year:
2022,
Volume and Issue:
unknown, P. 1 - 4
Published: Oct. 28, 2022
Lane
line
detection
is
one
of
the
important
tasks
environment
perception
system
autonomous
vehicles,
which
must
be
very
time
sensitive
and
robust.
To
this
end,
paper
proposes
a
lane
implementation
method
based
on
OpenCV
platform,
can
applied
to
smart
cars
in
specific
places,
mainly
including
image
preprocessing
fitting.
Applying
morphological
operations
stage
effectively
fill
wear
information
lines,
least
squares
used
adjust
lines
after
Hough
transformation.
The
results
show
that
proposed
improve
operation
speed
without
affecting
accuracy
algorithm,
has
certain
practicality.
Sensors,
Journal Year:
2023,
Volume and Issue:
23(15), P. 6758 - 6758
Published: July 28, 2023
The
section
detection
of
the
pavement
is
data
basis
for
measuring
road
smoothness,
rutting,
lateral
slope,
and
structural
depth.
Pavement-Section
includes
longitudinal-section
inspection
cross-section
inspection.
In
this
paper,
based
on
multiple
laser
displacement
sensors,
fused
accelerometers
attitude
using
vehicle-mounted
high-speed
detection,
we
design
a
sensor-fused
acquisition
method,
establish
relevant
mathematical
model,
realize
automatic
longitudinal
transverse
sections.
acceleration
sensor
filtered
to
improve
accuracy
acquisition,
error
system
calculated
analyzed.
Through
actual
measurement,
profile
method
adopted
in
paper
can
not
only
accurately
detect
profile,
but
also
efficiency,
providing
cost-effective
mode
surface
detection.
BOHR International Journal of Smart Computing and Information Technology,
Journal Year:
2023,
Volume and Issue:
4(1), P. 86 - 94
Published: Jan. 1, 2023
Traffic
safety
is
enhanced
by
immediate
lane-line
monitoring
and
recognition
in
advanced
driving
assistance
systems.
A
new
method
of
recognizing
continuous
lane
lines
using
the
Hough
transform
proposed
this
study.
vehicle
equipped
with
a
camera
that
takes
pictures
road,
which
are
then
processed
to
enhance
visibility
lines.
transforms
applied
preprocessed
images
allow
system
recognize
In
order
ensure
lines,
Kalman
filter
has
been
used
comprehensive
set
real-time
scenarios
assess
performance
Python
OpenCV.
The
results
trial
demonstrate
system’s
viability
efficacy.
BOHR International Journal of Smart Computing and Information Technology,
Journal Year:
2023,
Volume and Issue:
3(1), P. 64 - 72
Published: Jan. 1, 2023
Traffic
safety
is
enhanced
by
immediate
lane-line
monitoring
and
recognition
in
advanced
driving
assistance
systems.
A
new
method
of
recognizing
continuous
lane
lines
using
the
Hough
transform
proposed
this
study.
vehicle
equipped
with
a
camera
that
takes
pictures
road,
which
are
then
processed
to
enhance
visibility
lines.
transforms
applied
preprocessed
images
allow
system
recognize
In
order
ensure
lines,
Kalman
filter
has
been
used
comprehensive
set
real-time
scenarios
assess
performance
Python
OpenCV.
The
results
trial
demonstrate
system’s
viability
efficacy.
arXiv (Cornell University),
Journal Year:
2022,
Volume and Issue:
unknown
Published: Jan. 1, 2022
Lane
marking
detection
is
fundamental
for
both
advanced
driving
assistance
systems.
However,
detecting
lane
highly
challenging
when
the
visibility
of
a
road
low
due
to
real-life
environment
and
adverse
weather.
Most
methods
suffer
from
four
types
challenges:
(i)
light
effects
i.e.,
shadow,
glare
light,
reflection
etc.;
(ii)
Obscured
eroded,
blurred,
colored
cracked
caused
by
natural
disasters
weather;
(iii)
occlusion
different
objects
surroundings
(wiper,
vehicles
etc.);
(iv)
presence
confusing
like
lines
inside
view
e.g.,
guardrails,
pavement
marking,
divider
etc.
Here,
we
propose
robust
tracking
method
with
three
key
technologies.
First,
introduce
comprehensive
intensity
threshold
range
(CITR)
improve
performance
canny
operator
in
edges.
Second,
two-step
verification
technique,
angle
based
geometric
constraint
(AGC)
length-based
(LGC)
followed
Hough
Transform,
verify
characteristics
prevent
incorrect
detection.
Finally,
novel
defining
horizontal
position
(RHLP)
along
x
axis
which
will
be
updating
respect
previous
frame.
It
can
keep
track
either
left
or
right
markings
are
partially
fully
invisible.
To
evaluate
proposed
used
DSDLDE
[1]
SLD
[2]
dataset
1080x1920
480x720
resolutions
at
24
25
frames/sec
respectively.
Experimental
results
show
that
average
rate
97.55%,
processing
time
22.33
msec/frame,
outperform
state
of-the-art
method.
