Viewpoint invariance in the recognition of 3-D depth-rotated figures
Full text is not in this repository.
The aim of our experiments was to investigate the central mechanisms of visual perception involved in the identification of 3-D objects. We have made psychophysical measurements of the recognition thresholds of incomplete 2-D images of 3-D objects, varying their perspective during a participant's learning and testing, using the Gollin test. The results we obtained showed a dependence of recognition thresholds of a 3-D object upon (i) the number of times it had been previously presented and (ii) the viewing angle of the 3-D object. The threshold was about 20% of contour length at first presentation. Recognition was improved at the second presentation, irrespective of whether a different angle of viewing was used, after which recognition was constant. It appears that a template forms immediately after the first presentation of a 2-D projection of an object. This template has invariance properties since its use for object identification does not depend on the object's orientation when subsequently presented, at least up to an angle of 60 degrees.
|Additional Information:||European Conference on Visual Perception. Utrecht, The Netherlands. 24-28 Aug, 2008.|
|Research Areas:||A. > School of Science and Technology > Computer Science
A. > School of Science and Technology > Computer Science > Artificial Intelligence group
|Depositing User:||Devika Mohan|
|Date Deposited:||03 Dec 2009 07:22|
|Last Modified:||05 Oct 2015 15:42|
Actions (login required)