This paper mathematically analyzes and proposes new solutions for the
problem of estimating the camera 3D location and orientation (pose det
ermination) from a matched set of 3D model and 2D image landmark featu
res. Least-squares techniques for line tokens, which minimize both rot
ation and translation simultaneously, are developed and shown to be fa
r superior to the earlier techniques which solved for rotation first a
nd then translation. However, least-squares techniques fail catastroph
ically when outliers (or gross errors) are present in the match data.
Outliers arise frequently due to incorrect correspondences or gross er
rors in the 3D model. Robust techniques for pose determination are dev
eloped to handle data contaminated by fewer than. 50.0% outliers. Fina
lly, the sensitivity of pose determination to incorrect estimates of c
amera parameters is analyzed. It is shown that for small held of view
systems, offsets in the image center do not significantly affect the l
ocation of the camera in a world coordinate system. Errors in the foca
l length significantly affect only the component of translation along
the optical aids in the pose computation. (C) 1994 Academic Press, Inc
.