Abstract A new method is presented for reasoning with constraints on the three orientation parameters of an object which arise from the identification of features discovered in a single image. The method uses a geometrical model of the object and a tetrahedral representation of view-direction. The final angle parameter, the roll angle, is computed for each view direction. This method is subject to errors from the data and the assumptions inherent in the use of view directions. These errors are taken into account by considering upper and lower bounds on the roll angle. A method is presented for using constraints on the orientation parameters to prune the search for a consistent set of labelled image lines and to guide an iterative inversion of the view perspective.
[1]
Geoffrey D. Sullivan,et al.
Model Based Vision using a Planar Representation of the Viewsphere
,
1988,
Alvey Vision Conference.
[2]
Geoffrey D. Sullivan,et al.
Model based perspective inversion
,
1989,
Image and Vision Computing.
[3]
Geoffrey D. Sullivan,et al.
Model based perspective inversion
,
1989,
Image Vis. Comput..
[4]
Geoffrey D. Sullivan,et al.
Consistent labelling of image features using an assumption-based truth maintenance system
,
1989,
Image Vis. Comput..
[5]
Chris Goad,et al.
Special purpose automatic programming for 3D model-based vision
,
1987
.