<s>
Geometric	B-Algorithm
feature	I-Algorithm
learning	I-Algorithm
is	O
a	O
technique	O
combining	O
machine	O
learning	O
and	O
computer	B-Application
vision	I-Application
to	O
solve	O
visual	O
tasks	O
.	O
</s>
<s>
Researchers	O
simulate	O
humans	O
 '	O
ability	O
of	O
recognizing	O
objects	O
to	O
solve	O
computer	B-Application
vision	I-Application
problems	O
.	O
</s>
<s>
(	O
2002	O
)	O
applied	O
feature	O
learning	O
techniques	O
to	O
the	O
mobile	B-General_Concept
robot	I-General_Concept
navigation	I-General_Concept
tasks	O
in	O
order	O
to	O
avoid	O
obstacles	O
.	O
</s>
<s>
They	O
used	O
genetic	B-Algorithm
algorithms	I-Algorithm
for	O
learning	O
features	O
and	O
recognizing	O
objects	O
(	O
figures	O
)	O
.	O
</s>
<s>
Geometric	B-Algorithm
feature	I-Algorithm
learning	I-Algorithm
methods	O
can	O
not	O
only	O
solve	O
recognition	O
problems	O
but	O
also	O
predict	O
subsequent	O
actions	O
by	O
analyzing	O
a	O
set	O
of	O
sequential	O
input	O
sensory	O
images	O
,	O
usually	O
some	O
extracting	O
features	O
of	O
images	O
.	O
</s>
<s>
This	O
technique	O
is	O
widely	O
used	O
in	O
the	O
area	O
of	O
artificial	B-Application
intelligence	I-Application
.	I-Application
</s>
<s>
Geometric	B-Algorithm
feature	I-Algorithm
learning	I-Algorithm
methods	O
extract	O
distinctive	O
geometric	O
features	O
from	O
images	O
.	O
</s>
<s>
Corners	O
of	O
an	O
object	O
can	O
be	O
extracted	O
through	O
Corner	B-Algorithm
detection	I-Algorithm
.	O
</s>
<s>
The	O
outline	O
of	O
an	O
object	O
can	O
be	O
easily	O
detected	O
by	O
finding	O
the	O
edge	O
using	O
the	O
technique	O
of	O
edge	B-Algorithm
detection	I-Algorithm
.	O
</s>
<s>
Blobs	O
:	O
Blobs	O
represent	O
regions	O
of	O
images	O
,	O
which	O
can	O
be	O
detected	O
using	O
blob	B-Algorithm
detection	I-Algorithm
method	O
.	O
</s>
<s>
Extracting	O
geometric	O
feature	B-Algorithm
vector	I-Algorithm
at	O
location	O
x	O
can	O
be	O
computed	O
according	O
to	O
the	O
reference	O
point	O
,	O
which	O
is	O
shown	O
below	O
:	O
</s>
<s>
Feature	O
space	O
was	O
firstly	O
considered	O
in	O
computer	B-Application
vision	I-Application
area	O
by	O
Segen	O
.	O
</s>
<s>
The	O
probably	O
approximately	O
correct	O
(	O
PAC	O
)	O
model	O
was	O
applied	O
by	O
D	O
.	O
Roth	O
(	O
2002	O
)	O
to	O
solve	O
computer	B-Application
vision	I-Application
problem	O
by	O
developing	O
a	O
distribution-free	O
learning	O
theory	O
based	O
on	O
this	O
model	O
.	O
</s>
<s>
The	O
input	O
is	O
a	O
feature	B-Algorithm
vector	I-Algorithm
and	O
the	O
output	O
is	O
1	O
which	O
means	O
successfully	O
detect	O
the	O
object	O
or	O
0	O
otherwise	O
.	O
</s>
<s>
The	O
main	O
purpose	O
of	O
SVM	B-Algorithm
is	O
to	O
find	O
a	O
hyperplane	O
to	O
separate	O
the	O
set	O
of	O
samples	O
where	O
is	O
an	O
input	O
vector	O
which	O
is	O
a	O
selection	O
of	O
features	O
and	O
is	O
the	O
label	O
of	O
.	O
</s>
