Analysis and Modeling of Driver's Gaze Trajectories in Curves Driving

Abstract

To explore the relation between drivers' eye movements and vehicle steering performance, 25 participants were employed to drive a test vehicle on appointed tortuous routes. The drivers' visual behavior was measured while they were driving, and experiment data collected on motorway on-ramps and interchanges were used to analyze the drivers' visual characteristics. The distribution density of the fixation points showed that the drivers paid attention to the inner road edge when turning left and the outer road edge when turning right. Meanwhile, the heights of the gaze targets hardly changed. Therefore, a further investigation discovered that the gaze trajectory was parallel to the vehicle's driving trajectory on one-way curves. Then, six types of gaze trajectories were derived, and models for each type were established. Combining these models with the fact that gaze points are always distributed along the nearby road edge, it was deduced that the starting point of the gaze trajectory when driving on curves has a salient influence on the driver's gaze trajectory.

Deep Learning-Based Gaze Detection System for Automobile Drivers Using a NIR Camera Sensor

“…The research by Lee et al [ 27 ] used 18 gaze zones by using an additional zone (the upper position of region 6 of Figure 3 ) compared to 17 zones in our research. However, the case of gazing at this additional zone does not frequently occur while driving [ 48 , 49 , 50 , 51 ]. Therefore, in previous studies, they did not use this additional zone for experiments, either [ 24 , 36 , 38 , 39 , 40 ].…”

“…In previous research [ 48 , 49 , 50 , 51 ], they investigated the drivers’ visual strategies, the distribution of fixation points, driving performance, and gaze behavior by on-road experiment or driving simulator. Their research was focused on the analyses of the driver’s visual characteristics while driving instead of proposing new gaze detection methods.…”