Cite as:
Allison R, Jenkin M, 1999, "Depth perception with a stereoscopic robot head" Perception 28 ECVP Abstract Supplement

Depth perception with a stereoscopic robot head

R Allison, M Jenkin

TRISH-2 is a stereoscopic robot head that arose from the TRISH-1 platform. The robot consists of two computer-controlled CCD cameras acting as eyes. The cameras are mounted on motorised bases and have two extrinsic degrees of freedom. They can be independently panned (azimuth) under computer control. Torsion about the optic axis of each eye is achieved in software. The entire head can also be panned (azimuth) or tilted (elevation). Each camera provides additional optical degrees of freedom under computer control, with independent settings for focus, zoom, aperture, exposure, shutter speed, and video gain. Using TRISH-2, we investigated optimising the optical and rotational parameters for specific stereoscopic visual tasks. These techniques are often analogous to mechanisms proposed for biological vision. For example, Howard and Kaneko (1994 Vision Research 34 2505 - 2517) proposed a modified version of the deformation theory of inclination perception. Vertical shear disparity is averaged over the binocular field and used as the vertical disparity term in computing inclination from deformation disparity. To implement this theory, we use global cyclodisparity to set the torsional position of the eyes and then use deformation disparity to compute inclination. Torsional control of the stereo head improved efficiency of stereoscopic processing and enhanced performance for computing surface structure on inclined surfaces. Other analogies to biological stereoscopic mechanisms were considered as well as algorithms with no biological counterparts.

These web-based abstracts are provided for ease of seaching and access, but certain aspects (such as as mathematics) may not appear in their optimum form. For the final published version of this abstract, please see
ECVP 1999 Abstract Supplement (complete) size: 1478 Kb