Abstract
Recent work on perceiving the gist of a static image, whether a natural scene or a radiological scan, has shown that the visual system extracts details quickly, allowing for rapid discrimination and identification. While these findings are well-established in vision research, they have yet to penetrate to driving, where drivers are thought to take seconds to process the road scene. To test whether subjects could perceive the gist of a road scene, we performed two experiments where they were asked to discriminate between normal driving video and video showing pre-collision cues. In our first experiment, subjects viewed short clips (100 - 1000 ms duration) and were asked to discriminate pre-collision from normal driving scenes. Subjects were able to discriminate the scenes above chance at all durations, and performance plateaued beyond 300-400 ms. Using these results, we also investigated variability in rapid discrimination of driving environment as a function of position in the visual field. Subjects fixated one of five locations (ranging 20-30° from road center) and performed the same task as before, with a fixed video duration of 300 ms. Subjects remain capable of this discrimination when the roadway is in peripheral, rather than central vision, although we find a small decrease in sensitivity. Subjects' ability to perform this task suggests that the visual system is not only processing the moving scene in the time available, but successfully detecting a wide variety of potentially hazardous deviations from what would be normal in the scene. Understanding what drivers can and cannot perceive quickly is particularly timely given the advent of automated vehicles; drivers may not need to attend to the road all of the time, but must return attention to the road and rapidly acquire gist at a moment's notice.
Meeting abstract presented at VSS 2018