A study was conducted to estimate the pose of an object in a distributed camera network. The cameras were assumed to be internally and externally calibrated. They potentially saw a subset of the object feature points in the middle of clutter points from the background. The cameras individually recovered the object's pose by using their knowledge of the model where pose was defined to be the rotation matrix and the translation vector from a known origin that placed and oriented the object in space. These pose estimates were exchanged between neighbors, performing all updates locally to obtain a single estimate consistent across all cameras.
|Number of pages
|Johns Hopkins APL Technical Digest (Applied Physics Laboratory)
|Published - Mar 16 2010
ASJC Scopus subject areas
- General Engineering
- General Physics and Astronomy