Road region detection
A road region detection method is provided. The method includes: obtaining a first image captured by a camera at a first time point and a second image captured by the camera at a second time point (S 101 ), converting the first and second images into a first top view and a second top view, respectively (S 103 ), obtaining a movement vector matrix which substantially represents movement of a road region relative to the camera between the first and second time points (S 105 ), and determining whether a candidate point belongs to the road region by determining whether a position change of the candidate point between the first and second top views conforms to the movement vector matrix. The accuracy and efficiency may be improved.
1. A method for detecting road regions, comprising:
obtaining a first image captured by a camera at a first time point and a second image captured by the camera at a second time point;
converting the first image and the second image into a first top view and a second top view, respectively;
obtaining a movement vector matrix which substantially represents movement of a road region relative to the camera between the first time point and the second time point; and
determining whether a candidate point belongs to the road region by determining whether a position change of the candidate point between the first top view and the second top view conforms to the movement vector matrix.
2. The method according to claim 1 , wherein a scale of the first top view and the second top view is substantially similar to a real-world scale.
3. The method according to claim 1 , wherein the movement vector matrix is obtained by:
obtaining a rotation matrix R 1 and a translation matrix T 1 which substantially represent movement of the camera between the first time point and the second time point; and
obtaining the movement vector matrix, comprising a rotation matrix R 2 and a translation matrix T 2 , based on R 1 , T 1 , and one or more extrinsic parameters of the camera.
4. The method according to claim 3 , wherein, if the camera's pitch angle α equals to zero, then R 2 equals −R 1 , and T 2 equals −T 1 .
5. The method according to claim 3 , wherein, if the camera's pitch angle α does not equal to zero, then R 2 equals −R 1 , and T 2 equals
-
(
cos
α
0
0
0
1
0
0
0
sin
α
)
*
T
1
.
6. The method according to claim 1 , wherein the movement vector matrix is obtained by:
Identifying a group of feature points on the first top view;
tracking the feature points on the second top view; and
calculating R 2 and T 2 by solving an objective function:
arg
min
R
2
,
T
2
∑
{
I
2
(
X
T
2
,
Y
T
2
)
-
I
1
[
f
(
X
T
1
,
Y
T
1
)
]
}
2
,
where argmin defines a group of feature points of an argument for which the function Σ{I 2 (X T2 ,Y T2 )−I 1 [f(X T1 ,Y T1 )]} 2 attains a minimum value,
where I 2 (X T2 ,Y T2 ) defines a set of coordinates indicating the position of a feature point on the second top view,
where I 1 [f(X T1 ,Y T1 )] defines a set of coordinates calculated based on:
f
(
X
T
1
,
Y
T
1
)
=
(
R
2
T
2
0
T
1
)
(
X
T
1
Y
T
1
)
,
where
(
X
T
1
Y
T
1
)
defines a set of coordinates indicating the position of the feature point on the first top view.
7. The method according to claim 1 , wherein determining whether a candidate point belongs to the road region comprises:
obtaining a first set of coordinates of the candidate point on the first top view;
obtaining a second set of coordinates of the candidate point on the second top view;
calculating a third set of coordinates using the first set of coordinates and the movement vector matrix;
calculating a distance between the second set of coordinates and the third set of coordinates; and
determining whether the candidate point belongs to the road region by determining whether the distance is less than a predetermined threshold value.
8. A system for detecting road regions, comprising:
a processing device configured to:
obtain a first image captured by a camera at a first time point and a second image captured by the camera at a second time point;
convert the first image and the second image into a first top view and a second top view, respectively;
obtain a movement vector matrix which substantially represents movement of a road region relative to the camera between the first time point and the second time point; and
determine whether a candidate point belongs to the road region by determining whether a position change of the candidate point between the first top view and the second top view conforms to the movement vector matrix.
9. The system according to claim 8 , wherein a scale of the first top view and the second top view is substantially similar to a real-world scale.
10. The system according to claim 8 , wherein the processing device is further configured to:
obtain a rotation matrix R 1 and a translation matrix T 1 which substantially represent movement of the camera between the first time point and the second time point; and
obtain the movement vector matrix, comprising a rotation matrix R 2 and a translation matrix T 2 , based on R 1 , T 1 , and one or more extrinsic parameters of the camera.
11. The system according to claim 10 , wherein, if the camera's pitch angle α equals to zero, then R 2 equals −R 1 , and T 2 equals −T 1 .
12. The system according to claim 10 , wherein, if the camera's pitch angle α does not equal to zero, then R 2 equals −R 1 , and T 2 equals
-
(
cos
α
0
0
0
1
0
0
0
sin
α
)
*
T
1
.
13. The system according to claim 8 , wherein the processing device is further configured to:
Identify a group of feature points on the first top view;
track the feature points on the second top view; and
calculate R 2 and T 2 by solving an objective function:
arg
min
R
2
,
T
2
∑
{
I
2
(
X
T
2
,
Y
T
2
)
-
I
1
[
f
(
X
T
1
,
Y
T
1
)
]
}
2
,
where argmin defines a group of feature points of an argument for which the function Σ{I 2 (X T2 ,Y T2 )−I 1 [f(X T1 ,Y T1 )]} 2 attains a minimum value,
where I 2 (X T2 ,Y T2 ) defines a set of coordinates indicating the position of a feature point on the second top view,
where I 1 [f(X T1 ,Y T1 )] defines a set of coordinates calculated based on:
f
(
X
T
1
,
Y
T
1
)
=
(
R
2
T
2
0
T
1
)
(
X
T
1
Y
T
1
)
,
where
(
X
T
1
Y
T
1
)
defines a set of coordinates indicating the position of the feature point on the first top view.
14. The system according to claim 8 , wherein the processing device is further configured to:
obtain a first set of coordinates of the candidate point on the first top view;
obtain a second set of coordinates of the candidate point on the second top view;
calculate a third set of coordinates using the first set of coordinates and the movement vector matrix;
calculate a distance between the second set of coordinates and the third set of coordinates; and
determine whether the candidate point belongs to the road region by determining whether the distance is less than a predetermined threshold value.
15. A system for detecting road regions, comprising:
means for obtaining a first image captured by a camera at a first time point and a second image captured by the camera at a second time point;
means for converting the first image and the second image into a first top view and a second top view, respectively;
means for obtaining a movement vector matrix which substantially represents movement of a road region relative to the camera between the first time point and the second time point; and
means for determining whether a candidate point belongs to the road region by determining whether a position change of the candidate point between the first top view and the second top view conforms to the movement vector matrix.