Ground plane obstacle detection for mobile robot navigation based on optical flow field
The main objective of this work is to develop a vision-based obstacle avoidance capability for autonomous mobile robot. The basic assumption is that the robot is moving on a planar pavement and any objects not lying on this plane are considered to be obstacle. The robot will not be able to detect ov...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2006
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/4905/1/WaiTaiKuanMFKE2006.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The main objective of this work is to develop a vision-based obstacle avoidance capability for autonomous mobile robot. The basic assumption is that the robot is moving on a planar pavement and any objects not lying on this plane are considered to be obstacle. The robot will not be able to detect overhanging object as obstacle. An important feature is that the knowledge of the camera parameters and vehicle motion is not required. The method used for obstacles detection is Ground Plane Obstacle Detection where the robot is moving with the camera is facing the ground plane, hence observing a planar surface in motion. With this assumption, there is a globally valid parameterisation for the corresponding optical flow field. The detection mechanism simply relies on the fact that the optical flow should be globally constant no matter what the motion direction and speed of the robot might be. Thus, detecting the incoherent area in the optical flow field is equivalent to detecting the obstacle. Information from colour images is introduced to solve the problem of under constraint in optical flow computation using gradient-based method. By treating a colour image as three individual monochrome images, produced from extracting red, green and blue colour component of a colour image, the system will become over constraint. Thus, neighbouring sampling approach is introduced to solve the problem. Since the interest area for obstacle avoidance can be minimize to a small area in front of the robot, therefore optical flow is only calculated on the interest area. This method of calculating optical flow from subsampled images has greatly speed up the process of optical flow computation. Besides, this system has proved that gradient-based optical flow computation can also be sufficient with only using two subsequent images. The experiments conducted have given evidences about the good performed of the proposed method for mobile robot navigation utilizing only information from optical flow field. |
|---|