Information Technology Reference
In-Depth Information
stands for all the number of character
char
i
occurance appeared on
LPs
un-
der weather
w
in the time period of
t
. Since RFID detector are accurate to
be convincible,
N
(
char
i
|
w,t
) would be generated according to RFID detector
results. Accordingly,
N
(
char
i
ₒ
w,t
) stands for the number of that char-
acter
char
i
transforms into
char
j
occurance recognized by ALPR under weather
w
in the time period of
t
.
Then given character
char
i
and
char
j
, respectively stand for real character
and recognized character at the same position in a
LP
,
char
i
,char
j
∈
char
j
|
Char Set
.
And
w
∈
Weather Set,t
∈
Time Set
,let
f
(
char
i
,char
j
,w,t
)=
N
(
char
i
ₒ
char
j
|
w,t
)
(4)
N
(
char
i
)
Here,
N
(
char
i
) stands for all the number of character
char
i
occurance ap-
peared on the
LP
taken by
ca
i
. Formula (4) means that the character transition
probability of
char
i
transforms into
char
j
under weather
w
in the time period
of
t
.
CTPT is a table which has five columns including real character, recognized
character, weather, time period and transition probability. By Formula (4), we
can generate a CTPT for every
ca
i
,
ca
i
∈ CA
for all weather conditions and
time periods. CTPT would be updated in a certain period
˃
, and in this paper,
we set
˃
as one week temporarily. In the further research,
˃
would be discussed
in detail.
SLPs
Correction.
After Algorithm 1 is finished, most of correctly recognized
plates have been identified. For those plates that are not marked as
TLP
in
Algorithm 1, we deem them as
SLPs
.A
SLP
is not definitely an incorrectly
recognized plate, just because the high probability it has. However, if this
SLP
is indeed incorrectly recognized, its corresponding correct plate should occur in
its
k
-degree neighbors multiple times for some
k
.
Now, we define the similarity comparing
LP
to other
LPs
. Suppose there
are
n
characters in
LP
and
char
i
(
LP
) stands for the
i
th
character of
LP
.The
similarity of
LP
and
LP
recognized by camera
ca
is described as follows:
sim
(
LP,LP
)=
n
P
(
char
i
(
LP
)
ₒ char
i
(
LP
)
|w,t
)
(5)
i
=1
w,t
) stands for the probability of the
i
th
character of LP transforms into the
i
th
character of
LP
under the weather
w
in
the time period
t
. According to Nave Bayes theory [18],
char
i
(
LP
)
Here,
P
(
char
i
(
LP
)
ₒ
|
sim
(
LP, LP
)=
i
=1
CTPT
ca
(
char
i
(
LP
)
,char
i
(
LP
)
,w,t
)
∗ N
(
char
i
(
LP
)
|w,t
)
N
(
char
i
(
LP
)
|w,t
)
(6)