Using Virtual Reality to Investigate the Representation of 3D Space

9th June, 2011, 12.00-13.00

Pearson (North East Entrance) LT

Andrew Glennerster

School of Psychology and Clinical Language Sciences,
University of Reading, Reading RG6 6AL

Much of the work on human 3D vision has concentrated on static, binocular vision rather than the more general case of a moving observer. Experiments from our lab using virtual reality show that freely-moving observers ignore large changes in the scale of the room they are in (eg a four-fold change in size) and that the representation of depth can be intransitive (A > B > D while A < C < D). In navigation experiments, we have shown that errors in returning to a previously-viewed location are better predicted by a view-based than a 3D reconstruction model. This evidence in favour of view-based representation coincides with an increasing interest in view-based models of 3D structures in computer vision.

Andrew Glennerster is Reader and Unit of Assessment Leander in the School of Psychology and Clinical Language Sciences at the University of Reading. He studies human vision, particularly binocular stereopsis and the representation of 3D structure. His main interest is in how the visual system could represent the shape of surfaces and the layout of a scene when the observer is free to move. He trained as a doctor in Cambridge and studied stereo vision for his DPhil. in the Psychology department at Oxford. Together with a group of researchers in the Physiology, Psychology and Engineering departments at Oxford I have set up an immersive virtual reality laboratory to study spatial perception and motor control in freely-moving observers.