Information Technology Reference
In-Depth Information
with the car, when it runs through the road. Here, let
Pic Set
stand for the set
of the pictures, i.e.,
Pic Set
=
{
pic
1
,pic
2
, ..., pic
n
}
.
Pic Set
, it contains the information of the vehicle's
license plate. Let
LPD
indicate the real character string appearing on the vehi-
cle's license plate. Let
lpd
i
indicate the character string of the vehicle's license
plate extracted from the picture
pic
i
.Forthe
n
pictures contained in
Pic Set
,
n
character strings could be extracted from the pictures. Here, let
LP Set
indicate
the set of
n
character strings, i.e.,
LP Set
=
For a picture
pic
i
,
pic
i
∈
{
lpd
1
,lpd
2
, ..., lpd
n
}
.
With these assumptions, two cases will be discussed here.
-
For a
lpd
i
,if
lpd
i
=
LPD
, we believe that the camera
ca
i
captures the right
information appearing on the vehicle's license plate.
-
For a
lpd
i
,if
lpd
i
=
LPD
, we believe that the camera
ca
i
did not capture
the right information appearing on the car's license plate.
Furthermore, if we could not get the
LPD
in advance, an assumption as
defined by Def. 1 is presented here for get the right information of the car's
license plate.
Definition 1.
For a
lpd
i
,
lpd
i
∈
,if
there are
m
other elements contained in
Pic Set
, which have the same data
with
lpd
i
, the probability of
lpd
i
captured by the cameras along the road could be
computed by
(
m
+1)
/n
.Fora
lpd
i
, if it owns the maximum probability value, it
will be treated as the right information of the license plate.
LP Set
,
LP Set
=
{
lpd
1
,lpd
2
, ..., lpd
n
}
Def. 1 is enabled by an assumption that it is a small probability event for the
cameras along the road to make same mistake.
In this paper, we assume that all cars on roads are not stationary but floating.
Since two adjacent cameras stand only a few kilometers apart in the urban
areas, this assumption assures us that floating cars inevitably come across several
cameras and thereby get captured by them. As to cars which are captured only
once by a camera and then disappeared from the neighboring camera, we deem
it as an exception since this situation is rather rare.
Several formal definitions are given in the following to support our method in
detail.
Definition 2 (A Trustworthy License Plate,
TLP
).
TLPs
are those plates that are needless to correct. Given a set of cameras
CA
=
, if a vehicle goes through all the cameras in
CA
and
is captured as the same plate
LP
, as depicted in Fig. 1, we claim that
LP
is a
correctly recognized plate whose probability can be lower bounded by
{
ca
1
,ca
2
, ..., ca
n
}
n
Pr
(
LP
)
≥
1
−
p
i
(1
−
p
j
)
(1)
j
=1
Here,
p
i
indicates the precision of the camera
ca
i
∈
CA
.
