[Soc-2013-dev] Weekly Report #7 Motion Tracking

[Joseph Mansfield]

=This week=

I was still a little unsure about how the multiview algorithm would work
and Keir was pretty busy, so we didn't have much chance to talk about. I
continued working on interface to multiview algorithms, setting up the data
structure for the output. Basically, the EuclideanReconstruction class
contains the reconstructed camera poses and 3D points. For the camera
poses, I now store a vector of EuclideanCameras for each view and provide
member functions to access and set them. The 3D points didn't need to
change, because they don't correspond to particular views. All points can
generally be seen from all views. I also made small changes to the other
work I had done.

I finally managed to get my clean-up commit into trunk and merge it into my
branch, so now I'm working with only the clean interface. I also spent some
time working out how the user interface for multiview reconstruction works
(and we might actually call it multicamera reconstruction instead).

=Next week=

I have spoken to Keir now about how the multiview algorithm should work
(and it's much simpler than I expected), so I plan to have multicamera
reconstruction implemented in libmv and reviewed before my holiday in
Latvia on Tuesday.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/soc-2013-dev/attachments/20130803/dc9f0ae4/attachment.htm