Collision Detection in Cluttered Driving Scenes

Abstract

Summary:The purpose of the present experiment was to examine whether drivers' detection of collisions was altered when the driving scene was cluttered with scene objects. In this experiment stationary scene objects were manipulated by positioning them behind an approaching object and driver motion induced. We found that observers' collision detection performance (d') decreased with the presence of scene objects. These results indicate that the ability to detect a collision is altered by the presence of scene objects. In addition, performance was dependent on display duration, with greater sensitivity at increased durations. Moreover, the results showed a significant criterion shift between scene objects present and scene objects absent, with a decrease in identifying a collision object (hit rate) when scene objects were present but no difference in identification of a collision event when scene objects were absent. This suggests that the decreased performance was due to the inability to accurately determine a collision event because of apparent motion of background scene objects due to driver motion. Because the displays used in this experiment are akin to driving in a cluttered environment, the results of this research have important implications regarding driving safety and crash rates particularly in urban environments with complex scenes. Specifically, the results suggest that one factor in cluttered driving scenes is the apparent motion of background scene objects due to driver motion.

Effects of Adjacent Vehicles on Judgments of a Lead Car During Car Following

“…Participants exhibited greater thresholds for detection of a target car’s approach when an additional foveally located car (late arriving) was approaching rather than stationary, supporting the notion that adjacent vehicles may have differential effects on the detection of approach motion (or deceleration) and the judgment of TTC. Furthermore, our finding that non-decelerating adjacent vehicles decreased sensitivity ( d ′) is consistent with those of Lemon and Andersen (2015), who reported lower d ′ values when additional background scene objects were present (rather than absent) in a collision detection task that involved simulated self-motion.…”