About MonoSLAM

[motters]

I originally wrote this software in 2006 as a C# translation of Andrew
Davison's monocular slam demo, SceneLib 1.0. At that time I was
mainly using a Windows operating system, and so needed to carry out
some kind of port of the original code in order to really be able to
test out the system live with a webcam.

At present I'm not actively working on this, but setting it up as a
Google Code project provides a more organized way for anyone
interested in developing this further to submit patches or report
bugs. If enough developers are working on this project the discussion
group provides a way for like-minded people to collaborate and
exchange ideas.

Just in case people do actually start using this discussion group I
should give a heads up warning that I'm a totally ruthless moderator,
and anything even vaguely impolite or off topic will be deleted
without warning.

It should be noted that MonoSLAM as it exists at present is not a
completely practical system for robotics use. The main limitation is
the requirement to perform a scale calibration using the black square
target each time the system is initially run. Davison et al have made
various improvements since releasing SceneLib 1.0, such as multiple
motion models, inverse depth parameterisation and surface normal
estimation. These all nudge monocular SLAM towards greater
practicality, although they have not been implemented in the current
C# version.