Matching omnidirectional images
from United States  [214 posts]
12 years
The images didn't appear to upload with my initial post, so here they are now.

Starting picture:
Displaced 12" to the right:
Displaced 12" directly back:
Displaced 12" to the left:
Displaced 12" directly forward:
Anonymous 12 years

You may actually have an easier time if you do not polar unwrap the image but instead use the radial distortion module to flatten out the image. If you take a look at that page's documention


you will see a soccer ball court where the same issue was raised. Basically with the above image you have a very complex movement structure to determine. With the radial distortion all you need is to do X,Y and Z movement which is essentially an alignment procedure between two images.

We think we have exactly the module you need (after the radial distortion is added) but would need to test it out on your original images from above (i.e. the non-unpolar warpped ones). Can you post those images here?


from United States  [214 posts] 12 years
Hi STeven,

Cool!  I never would have thought of using the Radial Distortion module.  Here are six new original images.  This time I used a ruler to make the displacements 15" from the starting location.  Note that the robot actually faces to the right in the images because of the way I had to mount the camera.

So the images are:

- starting position
- 15" to the right
- 15" directly back
- 15" to the left
- 15" directly forward

And this last one I displaced the robot 15 inches to the right and also rotated it 90 degrees clockwise to see if you can also pull out the rotation.

- 15" to the right and rotated 90 degrees clockwise


This forum thread has been closed due to inactivity (more than 4 months) or number of replies (more than 50 messages). Please start a New Post and enter a new forum thread with the appropriate title.

 New Post   Forum Index