Stereovision Node

[fnannizzi]

To research:

• [ ] How is a disparity map generated?
• [ ] How is a disparity map used?
• [ ] Why can't Francesca understand disparity maps?
• [ ] Will images be significantly distorted because the cameras are underwater?
• [ ] When/where can we get some better reference images?

Design features:

• [ ] Baseline calibration coefficient - want to easily adjust code for multiple baselines
• [ ] 2D correspondence search - 1D is too optimistic, most likely the cameras will be in different planes

Expected challenges:

• [ ] Fine-tuning the depth measurements could be hard.
• [ ] We're looking at the bottom of the pool - pretty far away, and pretty murky. Hopefully points and pixels will be distinct.

[uscauv]

Hey so what are you having trouble with as far as understanding disparity maps?

[uscauv]

Ugh I always forget that I'm signed in as uscauv and not canondetortugas

[uscauv]

Also, as far as naming of that milestone, is the goal to have the optical flow stuff done too or just to get this vision stuff pinned down well enough to know exactly what we need on the hardware side?

[canondetortugas]

I think this explains the disparity map generation part pretty well but the MRF part might go over your head a bit since you haven't done too much probability stuff: http://www.eecis.udel.edu/~grauerg/globalStereoPresCompVisionClass.pdf

[canondetortugas]

Also, thanks to openCV you don't actually need to understand how to generate one: http://glowingpython.blogspot.com/2011/11/computing-disparity-map-in-opencv.html

Making one yourself would be kinda cool though.