Electronics,
Journal Year:
2024,
Volume and Issue:
13(19), P. 3826 - 3826
Published: Sept. 27, 2024
Existing
dehazing
methods
deal
with
real-world
haze
images
difficulty,
especially
scenes
thick
haze.
One
of
the
main
reasons
is
lacking
pair
data
and
robust
priors.
To
improve
ability
in
scenes,
we
propose
a
semi-supervised
codebook
learning
method.
The
used
as
strong
prior
to
guide
hazy
image
recovery
process.
However,
following
two
issues
arise
when
applied
task:
(1)
Latent
space
features
obtained
from
coding
degraded
suffer
matching
errors
nearest-neighbour
performed.
(2)
Maintaining
good
balance
quality
fidelity
for
heavily
dense
difficult.
reduce
nearest-neighbor
error
rate
vector
quantization
stage
VQGAN,
designed
unit
dual-attention
residual
transformer
module
(UDART)
correct
latent
features.
UDART
can
make
encoding
closer
those
corresponding
clear
image.
result,
design
density
guided
weight
adaptive
(HDGWA),
which
adaptively
adjust
multi-scale
skip
connection
weights
according
density.
In
addition,
use
mean
teacher,
strategy,
bridge
domain
gap
between
synthetic
enhance
model
generalization
scenes.
Comparative
experiments
show
that
our
method
achieves
improvements
0.003,
2.646,
0.019
over
second-best
no-reference
metrics
FADE,
MUSIQ,
DBCNN,
respectively,
on
dataset
URHI.
Journal of Location Based Services,
Journal Year:
2024,
Volume and Issue:
18(4), P. 381 - 407
Published: Jan. 24, 2024
To
ensure
good
usability,
Location
Based
Services
(LBS)
should
be
context-aware,
i.e.
adapting
the
information
and
services
according
to
context
of
their
user,
such
as
his/her
location,
tasks,
preferences,
underlying
geo-social
environment.
This
article
reviews
main
challenges
related
modelling
processing
in
LBS,
proposes
a
list
essential
research
opportunities
that
can
pursued
overcome
challenges.
These
are
classified
into
four
groups:
'modelling
environment',
mobile
user',
'context-aware
adaptation',
'ethical
data
processing'.
Sufficiently
addressing
these
issues
will
enable
LBS
provide
'5R',
'right'
information,
way,
at
time,
place,
person.
ACM Journal on Computing and Sustainable Societies,
Journal Year:
2024,
Volume and Issue:
2(2), P. 1 - 18
Published: April 27, 2024
We
propose
an
automated
lowest
floor
elevation
(LFE)
estimation
algorithm
based
on
computer
vision
techniques
to
leverage
the
latent
information
in
street
view
images.
Flood
depth-damage
models
use
a
combination
of
LFE
and
flood
depth
for
determining
risk
extent
damage
properties.
used
image
segmentation
detecting
door
bottoms
roadside
edges
from
Google
Street
View
The
characteristic
equirectangular
projection
with
constant
spacing
representation
horizontal
vertical
angles
allows
extraction
pitch
angle
camera
bottom.
bottom
was
obtained
depthmap
paired
image.
LFEs
were
calculated
depth.
testbed
application
proposed
method
is
Meyerland
(Harris
County,
Texas).
results
show
that
achieved
mean
absolute
error
0.190
m
(1.18
%)
estimating
LFE.
height
difference
between
(HDSL)
estimated
provide
estimation.
automatic
using
images
provides
rapid
cost-effective
compared
surveys
total
station
theodolite
unmanned
aerial
systems.
By
obtaining
more
accurate
up-to-date
data
method,
city
planners,
emergency
insurance
companies
could
make
precise
damage.
ISPRS International Journal of Geo-Information,
Journal Year:
2024,
Volume and Issue:
13(5), P. 153 - 153
Published: May 5, 2024
This
study
investigates
the
efficacy
of
an
open
vocabulary,
multi-modal,
foundation
model
for
semantic
segmentation
images
from
complex
urban
street
scenes.
Unlike
traditional
models
reliant
on
predefined
category
sets,
Grounded
SAM
uses
arbitrary
textual
inputs
definition,
offering
enhanced
flexibility
and
adaptability.
The
model’s
performance
was
evaluated
across
single
multiple
tasks
using
benchmark
datasets
Cityscapes,
BDD100K,
GTA5,
KITTI.
focused
impact
input
refinement
challenges
classifying
visually
similar
categories.
Results
indicate
strong
in
single-category
but
highlighted
difficulties
multi-category
scenarios,
particularly
with
categories
bearing
close
or
visual
resemblances.
Adjustments
prompts
significantly
improved
detection
accuracy,
though
persisted
distinguishing
between
objects
such
as
buses
trains.
Comparative
analysis
state-of-the-art
revealed
SAM’s
competitive
performance,
notable
given
its
direct
inference
capability
without
extensive
dataset-specific
training.
feature
is
advantageous
resource-limited
applications.
concludes
that
while
vocabulary
mark
a
significant
advancement
segmentation,
further
improvements
integrating
image
text
processing
are
essential
better
scenarios.
Sustainability,
Journal Year:
2024,
Volume and Issue:
16(10), P. 3934 - 3934
Published: May 8, 2024
The
incidence
of
floods
is
rapidly
increasing
globally,
causing
significant
property
damage
and
human
losses.
Moreover,
Vietnam
ranks
as
one
the
top
five
countries
most
severely
affected
by
climate
change,
with
1/3
residents
facing
flood
risks.
This
study
presents
a
model
to
identify
susceptibility
using
analytic
hierarchy
process
(AHP)
in
GIS
environment
for
Hanoi,
Vietnam.
Nine
flood-conditioning
factors
were
selected
used
initial
data.
AHP
analysis
was
utilized
determine
priority
levels
these
concerning
assess
consistency
obtained
results
develop
flood-susceptibility
map.
performance
found
be
based
on
AUC
value
receiver
operating
characteristic
(ROC)
curve.
map
has
susceptibility:
area
very
high
flooding
accounts
less
than
1%
map,
high-
areas
nearly
11%,
moderate-susceptibility
more
65%,
low-
about
22%,
low-susceptibility
2%.
Most
Hanoi
moderate
level
susceptibility,
which
expected
increase
urban
expansion
due
impacts
urbanization.
Our
findings
will
valuable
future
research
involving
planners,
disaster
management
authorities
enable
them
make
informed
decisions
aimed
at
reducing
impact
enhancing
resilience
communities.
Cartography and Geographic Information Science,
Journal Year:
2024,
Volume and Issue:
51(3), P. 445 - 461
Published: Jan. 22, 2024
People
create
route
descriptions
based
on
their
mental
maps
to
provide
guidance,
which
represents
knowledge
of
the
environment.
Recent
studies
have
attempted
model
navigation
from
human
facilitate
communication.
However,
they
mainly
focus
outdoor
environments
and
do
not
address
representation
indoors
in
form
schematic
through
automatic
extraction
spatial
knowledge.
Schematic
been
commonly
applied
for
public
transportation
by
utilizing
abstract
representations
reduce
cognitive
load.
Compared
descriptions,
can
easy-to-understand
guidance.
In
this
paper,
we
present
a
novel
NLP-based
pipeline
automatically
generate
indoor
navigation.
The
experimental
data
consists
set
crowdsourced
that
follow
common
template
test
building
Soleway
web
service.
generated
were
presented
participants
an
online
survey,
it
was
found
92%
matched
well
with
corresponding
descriptions.
Thus,
proposed
method
is
effective
reliable
approach
modeling
ISPRS International Journal of Geo-Information,
Journal Year:
2024,
Volume and Issue:
13(6), P. 190 - 190
Published: June 7, 2024
Understanding
solar
radiation
in
urban
street
spaces
is
crucial
for
comprehending
residents’
environmental
experiences
and
enhancing
their
quality
of
life.
However,
existing
studies
rarely
focus
on
the
patterns
over
time
across
different
suburban
areas.
In
this
study,
view
images
from
summers
2013
2019
Shanghai
were
used
to
calculate
spaces.
The
results
show
a
general
decrease
compared
2013,
with
an
average
drop
12.34%.
was
most
significant
October
(13.47%)
least
May
(11.71%).
terms
data
gathered
sampling
points,
76.57%
showed
decrease,
while
23.43%
increase.
Spatially,
decreased
by
79.66%
every
additional
1.5
km
city
centre.
summary,
generally
shows
decreasing
trend,
variations
between
These
findings
are
vitally
important
guiding
planning,
optimising
green
infrastructure,
ecological
environment,
further
promoting
sustainable
development
improving
