<s>
Structure	B-Algorithm
from	I-Algorithm
motion	I-Algorithm
(	O
SfM	O
)	O
is	O
a	O
photogrammetric	B-Application
range	B-Algorithm
imaging	I-Algorithm
technique	O
for	O
estimating	O
three-dimensional	O
structures	O
from	O
two-dimensional	O
image	O
sequences	O
that	O
may	O
be	O
coupled	O
with	O
local	O
motion	O
signals	O
.	O
</s>
<s>
It	O
is	O
studied	O
in	O
the	O
fields	O
of	O
computer	B-Application
vision	I-Application
and	O
visual	O
perception	O
.	O
</s>
<s>
Finding	O
structure	B-Algorithm
from	I-Algorithm
motion	I-Algorithm
presents	O
a	O
similar	O
problem	O
to	O
finding	O
structure	O
from	O
stereo	B-Algorithm
vision	I-Algorithm
.	O
</s>
<s>
In	O
both	O
instances	O
,	O
the	O
correspondence	B-Algorithm
between	O
images	O
and	O
the	O
reconstruction	B-Algorithm
of	O
3D	O
object	O
needs	O
to	O
be	O
found	O
.	O
</s>
<s>
To	O
find	O
correspondence	B-Algorithm
between	O
images	O
,	O
features	O
such	O
as	O
corner	O
points	O
(	O
edges	O
with	O
gradients	O
in	O
multiple	O
directions	O
)	O
are	O
tracked	O
from	O
one	O
image	O
to	O
the	O
next	O
.	O
</s>
<s>
One	O
of	O
the	O
most	O
widely	O
used	O
feature	O
detectors	O
is	O
the	O
scale-invariant	B-Algorithm
feature	I-Algorithm
transform	I-Algorithm
(	O
SIFT	O
)	O
.	O
</s>
<s>
It	O
uses	O
the	O
maxima	O
from	O
a	O
difference-of-Gaussians	B-Algorithm
(	O
DOG	O
)	O
pyramid	O
as	O
features	O
.	O
</s>
<s>
Another	O
common	O
feature	O
detector	O
is	O
the	O
SURF	B-Algorithm
(	O
speeded-up	O
robust	O
features	O
)	O
.	O
</s>
<s>
In	O
SURF	B-Algorithm
,	O
the	O
DOG	O
is	O
replaced	O
with	O
a	O
Hessian	O
matrix-based	O
blob	O
detector	O
.	O
</s>
<s>
Also	O
,	O
instead	O
of	O
evaluating	O
the	O
gradient	O
histograms	O
,	O
SURF	B-Algorithm
computes	O
for	O
the	O
sums	O
of	O
gradient	O
components	O
and	O
the	O
sums	O
of	O
their	O
absolute	O
values	O
.	O
</s>
<s>
Therefore	O
,	O
comparing	O
to	O
SIFT	O
,	O
SURF	B-Algorithm
is	O
a	O
faster	O
feature	O
detector	O
with	O
drawback	O
of	O
less	O
accuracy	O
in	O
feature	O
positions	O
.	O
</s>
<s>
Another	O
type	O
of	O
feature	O
recently	O
made	O
practical	O
for	O
structure	B-Algorithm
from	I-Algorithm
motion	I-Algorithm
are	O
general	O
curves	O
(	O
e.g.	O
,	O
locally	O
an	O
edge	O
with	O
gradients	O
in	O
one	O
direction	O
)	O
,	O
part	O
of	O
a	O
technology	O
known	O
as	O
pointless	O
SfM	O
,	O
useful	O
when	O
point	O
features	O
are	O
insufficient	O
,	O
common	O
in	O
man-made	O
environments	O
.	O
</s>
<s>
RANSAC	B-Algorithm
(	O
random	B-Algorithm
sample	I-Algorithm
consensus	I-Algorithm
)	O
is	O
the	O
algorithm	O
that	O
is	O
usually	O
used	O
to	O
remove	O
the	O
outlier	O
correspondences	O
.	O
</s>
<s>
In	O
the	O
paper	O
of	O
Fischler	O
and	O
Bolles	O
,	O
RANSAC	B-Algorithm
is	O
used	O
to	O
solve	O
the	O
location	O
determination	O
problem	O
(	O
LDP	O
)	O
,	O
where	O
the	O
objective	O
is	O
to	O
determine	O
the	O
points	O
in	O
space	O
that	O
project	O
onto	O
an	O
image	O
into	O
a	O
set	O
of	O
landmarks	O
with	O
known	O
locations	O
.	O
</s>
<s>
There	O
are	O
several	O
approaches	O
to	O
structure	B-Algorithm
from	I-Algorithm
motion	I-Algorithm
.	O
</s>
<s>
A	O
somewhat	O
intermediate	O
approach	O
is	O
out-of-core	B-Application
SfM	O
,	O
where	O
several	O
partial	O
reconstructions	O
are	O
computed	O
that	O
are	O
then	O
integrated	O
into	O
a	O
global	O
solution	O
.	O
</s>
<s>
Structure-from-motion	O
photogrammetry	B-Application
with	O
multi-view	O
stereo	O
provides	O
hyperscale	O
landform	O
models	O
using	O
images	O
acquired	O
from	O
a	O
range	O
of	O
digital	O
cameras	O
and	O
optionally	O
a	O
network	O
of	O
ground	O
control	O
points	O
.	O
</s>
<s>
Structure	B-Algorithm
from	I-Algorithm
motion	I-Algorithm
is	O
also	O
useful	O
in	O
remote	O
or	O
rugged	O
environments	O
where	O
terrestrial	O
laser	O
scanning	O
is	O
limited	O
by	O
equipment	O
portability	O
and	O
airborne	O
laser	O
scanning	O
is	O
limited	O
by	O
terrain	O
roughness	O
causing	O
loss	O
of	O
data	O
and	O
image	O
foreshortening	O
.	O
</s>
<s>
Structure	B-Algorithm
from	I-Algorithm
motion	I-Algorithm
can	O
be	O
placed	O
in	O
the	O
context	O
of	O
other	O
digital	O
surveying	O
methods	O
.	O
</s>
<s>
The	O
first	O
operational	O
phase	O
is	O
an	O
accurate	O
preparation	O
of	O
the	O
photogrammetric	B-Application
surveying	O
where	O
is	O
established	O
the	O
relation	O
between	O
best	O
distance	O
from	O
the	O
object	O
,	O
focal	O
length	O
,	O
the	O
ground	O
sampling	O
distance	O
(	O
GSD	O
)	O
and	O
the	O
sensor	O
’s	O
resolution	O
.	O
</s>
<s>
Furthermore	O
,	O
structure-from-motion	O
photogrammetry	B-Application
represents	O
a	O
non-invasive	O
,	O
highly	O
flexible	O
and	O
low-cost	O
methodology	O
to	O
digitalize	O
historical	O
documents	O
.	O
</s>
