Method of detecting obstacle around vehicle

Patent metadata

Field	Value
Publication number	US-10354151-B2
Application number	US-201715706227-A
Country	US
Kind code	B2
Filing date	Sep 15, 2017
Priority date	Apr 20, 2017
Publication date	Jul 16, 2019
Grant date	Jul 16, 2019

Abstract

Disclosed is a method of detecting obstacle around a vehicle. The method of detecting an obstacle around a vehicle, includes: acquiring an image of the obstacle around the vehicle using a monocular camera; creating, by a controller, a distance based cost map, a color based cost map and an edge based cost map from the image; and integrating, by the controller, the distance based cost map, the color based cost map, and the edge based cost map to create a final cost map, and estimating, by the controller, a height of the obstacle from the final cost map.

First claim

What is claimed is:

1. A method of detecting an obstacle around a vehicle, comprising:
acquiring an image of the obstacle around the vehicle using a monocular camera;
creating, by a controller, a distance based cost map, a color based cost map and an edge based cost map from the image; and
integrating, by the controller, the distance based cost map, the color based cost map, and the edge based cost map to generate a final cost map, and estimating by the controller a height of the obstacle from the final cost map.

2. The method of claim 1 , wherein the creating of the distance based cost map comprises:
generating a plurality of Delaunay triangles by connecting feature points in the image and performing interpolation of distance information using a plane equation generated by three vertexes of each of the plurality of the Delaunay triangles; and
estimating a disparity of pixels included in each of the plurality of the Delaunay triangles from a result of the interpolation.

3. The method of claim 1 , wherein the creating of the color based cost map comprises:
setting an area where color similarities are measured in the image, measuring the color similarities between all feature points existing in the area, and selecting a greatest color similarity value as a final color similarity; and
calculating a difference between sums of the color similarities from the measured color similarities to create the color based cost map.

4. The method of claim 1 , wherein the creating of the edge based cost map comprises:
performing edge detection on the image; and
performing distance transformation based on edge detection such that pixels located closer to an edge have lower values.

5. The method of claim 1 , wherein the final cost map is calculated as:
c t =w d c d +w c c c +w e c e ,
where: w d is a weight of the distance based cost map, w c is a weight of the color based cost map, w e is a weight of the edge based cost map, c d is the distance based cost map, c c is the color based cost map, and c e is the edge based cost map.

6. A method of detecting an obstacle around a vehicle, comprising:
acquiring an image of the obstacle around the vehicle using a monocular camera;
reconstructing, by a controller, three-dimensional positions of corresponding points in the image;
integrating, by the controller, previously reconstructed three-dimensional corresponding points with the currently reconstructed three-dimensional corresponding points based on a relative positional relationship between the previously reconstructed three-dimensional corresponding points and the currently reconstructed three-dimensional corresponding points;
calculating, by the controller, a disparity value by applying a virtual baseline value formed by a movement of the monocular camera to a depth value obtained through three-dimensional reconstruction of the corresponding points; and
estimating, by the controller, a boundary of the obstacle based on the disparity value.

7. The method of claim 6 , wherein the disparity value is calculated as:

d
=

f
⁢

B
Z

,

where, the d is the disparity value to be obtained, the B is the virtual baseline value of the monocular camera, the Z is the depth value obtained through three-dimensional reconstruction, and the f is a focal length of the monocular camera.

8. The method of claim 6 , further comprises changing, when an angle of view of the monocular camera is a wide angle that is greater than or equal to a predetermined angle of view, a u-axis to an incident angular axis θu through the following equation:

θ
u

=

atan
(

u
-

o
x

f

)

,

where, the u is the value of the u-axis which is the horizontal axis of the image, the 0 x is the center point of the monocular camera, and the f is the focal length of the monocular camera.

9. A method of detecting an obstacle around a vehicle, comprising:
acquiring an image of the obstacle around the vehicle using a monocular camera;
reconstructing, by a controller, three-dimensional positions of corresponding points in the image, calculating, by the controller, a disparity value by applying a virtual baseline value formed by a movement of the monocular camera to a depth value obtained through three-dimensional reconstruction of the corresponding points, and estimating a boundary of the obstacle based on the disparity value by the controller; and
creating, by the controller, a distance based cost map, a color based cost map and an edge based cost map from the image, and estimating, by the controller, a height of the obstacle using the distance based cost map, the color based cost map, and the edge based cost map.

10. The method of claim 9 , wherein the disparity value is calculated as:

d
=

f
⁢

B
Z

,

where the ‘d’ is the disparity value to be obtained, the B is the virtual baseline value of the monocular camera, the Z is the depth value obtained through three-dimensional reconstruction, and the f is a focal length of the monocular camera.

11. The method of claim 9 , further comprising changing, when an angle of view of the monocular camera is a wide angle that is greater than or equal to a predetermined angle of view, a u-axis to an incident angular axis θu through the following equation:

θ
u

=

atan
(

u
-

o
x

f

)

,

where the u is the value of the u-axis which is the horizontal axis of the image, the 0 x is the center point of the monocular camera, and the f is the focal length of the monocular camera.

12. The method of claim 9 , wherein the creating of the distance based cost map comprises:
generating a plurality of Delaunay triangles by connecting feature points of the image and performing interpolation of distance information using a plane equation generated by three vertexes of each of the plurality of the Delaunay triangles; and
estimating a disparity of pixels included in each of th

Assignees

• Hyundai Motor Co Ltd
• Kia Motors Corp
• Korea National Univ Of Transportation Industry Academic Cooperation Foundation

Inventors

• Yoon Sung Won
• Kim Jeyeon
• Jo Hoon
• Jung Ho Gi
• Suhr Jae Kyu

Classifications

• G06T7/579 Primary

  from motion · CPC title

• G06V10/761

  Proximity, similarity or dissimilarity measures · CPC title

• G06V20/58

  Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads · CPC title

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title

• G06V10/56

  relating to colour · CPC title