Information Technology Reference
In-Depth Information
θ
and
a
are measured as
⎧
⎨
18
π
arctan
S
E
y
−S
S
y
,
S
E
x
−S
S
x
if
S
E
y
−
S
S
y
≥
0and
S
E
x
−
S
S
x
>
0;
,
if
S
E
y
−S
S
y
≥
0and
S
E
x
−S
S
x
<
0;
180 +
18
π
arctan
S
E
y
−S
S
y
−
18
π
arctan
S
E
y
−S
S
y
180
S
E
x
−S
S
x
θ
=
(3)
,
⎩
S
E
x
−S
S
x
if
S
E
y
−
S
S
y
<
0and
S
E
x
−
S
S
x
≥
0;
−
18
π
arctan
S
E
y
−S
S
y
,
360
S
E
x
−
S
S
x
if
S
E
y
−
S
S
y
<
0and
S
E
x
−
S
S
x
<
0
.
v
=
v
x
+
v
y
(4)
a
=
S
E
v
−
S
S
v
(5)
S
E
t
−
S
S
t
A moving object trajectory is a finite sequence of segments:
S
m
=
{
S
1
,S
2
,...,S
m
}
and points
Pt
=
{
Pt
1
,Pt
2
,...,Pt
m
+1
}
.
Extracting.
In order to extract the important representative points and seg-
ments, we use angulation
θ
and acceleration
a
as constrains to detach those
unimportant ones.
Assume that
S
a
is a main segment and
S
b
is a subsequent segment of
S
a
.If
|
>Δ
,
S
b
can be another important segment.
The more specific algorithm is shown in Algorithm 1.
Then we can get main segments sequence
S
n
=
(
S
a
)
θ
−
(
S
b
)
θ
|
>ε
or
|
(
S
a
)
a
−
(
S
b
)
a
|
S
1
,S
2
,...,S
n
}
{
.
2.2 Trajectory Clustering
Raw Classification.
Trajectory direction histogram (TDH) expresses an ap-
proximative segments' direction distribution. It describes directional characters
of trajectories. It's an effective method of pre-process classification for large
amount of data.
First, in order to figure out trajectory direction histogram, let (
S
i
)
θ
∈
[0
,
360),
i
=1
,
2
,
3
,...,n
. And we divide [0
,
360) into
N
subintervals, each length of which
is
aver
=
36
N
.If
javer
N
,(
S
i
)
θ
belongs to the
jth
subinterval.
N
j
is the number of segments whose angles belong to the
jth
subinterval. So, TDH of the trajectory can be calculated as
(
S
i
)
θ
<
(
j
+1)
aver
,1
≤
≤
j
≤
H
=(
h
1
,h
2
,...,h
N
)
,h
j
=
N
j
N
(6)
Search WWH ::
Custom Search