Depth perception in tele-presence systems.

Tele-presence refers to technologies enabling the remote presence of an observer or operator of robotics machines - through the use of monitoring and display devices. This involves the facilitation of 3D space perception on the basis of 2D pictures, a problem which is of interest to engineers, and psychologists who study space and picture perception. From a functional perspective the issue requires the specification of the necessary characteristics of a tele-presence system for effective task performance by a human observer. Since the central problem in picture perception is the conflict between the 3D re-presentation of the scene and the 2D surface of the picture, one possibility for tele-presence systems to reduce this conflict consists in the use of a camera which is slaved to observer movement. Thus the video picture is yoked to the head movement of the observer: changes in the video picture viewed by the observer emulate the changes that would have occurred in the visual field if the observer was viewing the scene directly. The explanation for reduced cue conflict and improved depth perception in pictures lies in the availability of motion parallax information. The main aim of this research was to see whether tele-presence which provides motion parallax information on a video picture improves depth perception compared to static tele-presence. While theoretical claims concerning the usefulness of motion parallax have a long history, the empirical findings are more equivocal. The basic design compared depth perception of a moving observer with that of a stationary observer. Two initial experiments showed that the movement condition leads to more accurate depth perception than the stationary condition, both under tele-presence and direct viewing conditions. Experiments 4 to 7 showed that active observation leads only to non- significantly better accuracy than passive observation. Interrupting the natural link between action and perception by reversing the picture tends to reduce the difference between the movement and the stationary condition. However, combining the analysis of the active, passive and reverse picture conditions did not lead to significant differences. A further experiment using an adjustment task supported the finding that reverse viewing does not reduce accuracy. In general the differences between the movement and the static condition while significant were not very strong which suggested that other sources of information such as visual angle information may have specified depth to a considerable extent. Simulation of fully remote tele-presence was expected to provide stronger differences. However, the differences were small and explainable in terms of short term learning processes resulting in perceptual fixity, i.e. an inability to take advantage of the information available. It was concluded that motion parallax is probably only a weak cue to depth under practical circumstances, and that learning effects in tele-presence systems require further attention. Future attention should be directed at learning processes and at the complexity of the stimulus displays. The study of learning processes may help to understand the consistent finding of large individual differences in using motion parallax information. And the study of more complex stimulus displays would enable a more adequate assessment of the ecological emphasis on the role of motion parallax