My project is to determine how a video camera was moving from
the video frames.This is achieved by first determining which way the camera was
facing in each frame by drawing perspective lines, as described in a paper my
mentor, Prof. Košecká, coauthored --- the 2002 video compass paper. This past
week, I've spent my time getting up at 9 AM, going to the lab I'm working in,
and working there for eight hours. I've accomplished a number of things, three
of which come to mind: I've arranged the main loop of the code I've been
working on so that it is more readable; now there are several blocks of code
that each work as a unit and can be seen in one screen, instead of 200 lines
that must be understood all at once. I found a bug causing incorrect rotation
and position estimates --- I corrected it after determining that I had been
describing how the room was rotated relative to the camera, instead of
describing how the camera was rotated relative to the room. I also sped up
processing for the rest of my project by saving data from image processing to a
file.
