Moving Object Detection from Mobile Platforms Using Stereo Data Registration - Computational Intelligence Paradigms in Advanced Pattern Classification - page 32

Fig. 3 Synthesized views representing frames ( n ) (from Fig. 1 ( le ft ) ) in the coordinate sys-
tems of frames ( n + 1 ) , by using their corresponding rigid displacements: [ R q | t q ]

where P n + 1
i ( x , y , z )

denotes the mapping of a given point from frame n into the next frame.
Note that for static 3D points we ideally have P n + 1 P n + 1

i ( x , y , z ) i ( x , y , z ) =

.

Once the whole set of points P n has been mapped, we can also synthesize the

corresponding 2D view ( I n + 1
(

) as follows:

u , v )

u 0 +

x n + 1
i

f u n + 1

i

=( round ) , (2)

z n + 1
i

v 0 +

y n + 1
i

f v n + 1

i

=( round ) ,

z n + 1
i

where f denotes the focal length in pixels, ( u 0 , v 0 ) represents the coordinates of the

x n + 1
i

y n + 1
i

z n + 1
i

camera principal point, and ( , , ) correspond to the 3D coordinates of

Fig. 4 map of moving regions, from frames ( n ) and ( n + 1 ) presented in

Fig. 1 ( top ) . ( right ) Image difference between these consecutive frames: ( | I ( n ) − I ( n + 1 ) | )to
illustrate their relative displacement.

( le ft ) D

( u , v )