Abstract
We investigated how monocular and binocular cues contribute grasping a slanted 3D object, and whether grasping performance was consistent with perception. Depth information is needed for multiple purposes: aligning the grip with the orientation of the object, scaling the grip aperture, and selecting grasp contact points for a stable grip. We measured these components when subjects reached to grasp objects that presented conflicting 3D cues, and measured perceived slant for comparison. Consistent cue stimuli were square planar objects presented at different slants and plane orientations. For conflict conditions, we created objects that, when viewed at a 45° slant, had the same projected contour as a rotated square object that is slanted by 55° (compressed) or 35° (elongated). For these objects, contour information specifies an incorrect 3D slant and orientation. Slant estimates revealed that perceived slant of cue conflict objects was between the slant specified by contour and binocular cues. When subjects' grasping these objects, the hand orientation was consistent with perceived slant: the grip was less slanted as the hand approached elongated objects, and vice versa. Grip aperture was similarly intermediate between the size specified by contour and binocular cues. However, the contact points for grasping objects were not influenced by binocular cues. Grasp points were consistent with a rectangular interpretation of the projected contours, rather than a skewed object that would be consistent with an intermediate slant. While this is not consistent with perceived slant, it would be consistent with perceived shape of the cue conflict stimuli, which appear to be elongated or compressed rectangular objects. We observed no dissociation between the use of 3D cues for perception and action. Grip control and grasp points were affected by 3D cues in the same way as perceived slant and shape.
Meeting abstract presented at VSS 2017