The long times-to-collision which can occur in traffic accidents require drivers to make judgements of distance and speed to supplement times-to-collision based directly on optic flow. The resulting errors due to overestimating familiar size for child pedestrians may cause over half of their casualties in Britain. For imminent collision, however, when optic flow is an effective source of time-to-collision, the algorithm which provides time-to-collision is generally assumed. By removing the need to judge angular position relative to a datum, the new algorithm helps to explain our ability to respond accurately to moving objects, and would also facilitate the rapid updating of time-to-collision necessary for motion which is accelerating or non-linear. The developments reported in this paper serve more to highlight the amazing capability of visual perception than to explain it. But if they are as fundamental as they appear to be, then not only is there great potential for preventing accidents and for new research and development into the perception of collision, but previous research may also merit reassessment.
Abstract