Step 7Software
The images below show the two camera images as they look out towards a coke can and a DrPepper can. You can see the focal difference between the two images and also the vertical disparity between the two cameras despite being mounted very close to each other. This disparity can be reduced by using a prism to split a single view into two views for two cameras but we found the quick method of using two webcams close to each other to be sufficient.
Note on the left side of the image the close Coke can is out of focus and the far DrPepper can is in focus. In the right side image the situation is reverse. If you look at the edges of this image you can see the edge strengths reflect the focus of the object. The white lines signal a higher edge transition which means the object is more in focus. The bluer lines signal a weaker response.
Each image is broken into 3 vertical sections. Left, middle and right. We use these areas to determine if an obstacle exists in those areas and if so steer the robot away. These bands are highlighted back into one side of the original image so that we can verify their correctness. The lighter areas in these images signal that the object is close. This tells the robot to move away from that direction.
The downside to this technique is that objects need texture. From the next image we can see two red blocks which are placed in the same position as the cans but they do not respond to this technique. The issue is that the red blocks do not have any internal texture. This feature requirement is similar to that needed for stereo and optical flow techniques.
| « Previous Step | Download PDFView All Steps | Next Step » |
4
comments
|
Add Comment
|
CarlS
says:
![]() |
Add Comment
|


















































