Information Technology Reference
In-Depth Information
Fig. 2.
Patches displacement between two images
Algorithm 1.
Yaw angle estimation
function
yaw estimation(
i
t
,i
t−
1
)
Obtain patches
p
it
and
p
it−
1
from
i
t
and
i
t−
1
for all
{p
it
,p
it−
1
}
do
ʔ
d
i
ₐ
findDisplacement(
p
it
,p
it−
1
)
m
it
ₐ
m
it−
1
+
ʔ
d
i
end for
ˈ ₐ
R
z
ₐ
H
ₐ
findHomography(
m
it
,
m
it−
1
)
return
ˈ
end function
Algorithm 2.
Displacement determination between patches
function
findDisplacement(
p
it
,p
it−
1
)
P
it
ₐ FFT
(
p
it
)
P
it−
1
ₐ FFT
(
p
it−
1
)
C ₐ CPS
(
P
it
,P
it−
1
)
r ₐ IFFT
(
c
)
ʔ
d
i
ₐ
max(
r
)
return
ʔ
d
i
end function
3 A Multilevel Decomposition Algorithm
This work proposes a parallel algorithm to calculate the homography based yaw
angle determination equations described in the previous section, over a parallel
architecture. This algorithm take advantage of the natural division of the images
in patches using a domain decomposition method, following the data-parallel
approach. The parallel algorithm is implemented in a ground station which is
composed by oce desktop computers connected with a pre-existing LAN to
form a distributed system. From now on, we refer to the computers as
nodes
,
and each node has multiple processors running one thread each.