Robust and continuous estimation of driver gaze zone by dynamic analysis of multiple face videos

Abstract

Abstract-Analysis of driver's head behavior is an integral part of driver monitoring system. Driver's coarse gaze direction or gaze zone is a very important cue in understanding driverstate. Many existing gaze zone estimators are, however, limited to single camera perspectives, which are vulnerable to occlusions of facial features from spatially large head movements away from the frontal pose. Non-frontal glances away from the driving direction, though, are of special interest as interesting events, critical to driver safety, occur during those times. In this paper, we present a distributed camera framework for gaze zone estimation using head pose dynamics to operate robustly and continuously even during large head movements. For experimental evaluations, we collected a dataset from naturalistic on-road driving in urban streets and freeways. A human expert provided the gaze zone ground truth using all vision information including eyes and surround context. Our emphasis is to understand the efficacy of the head pose dynamic information in predicting eye-gaze-based zone ground truth. We conducted several experiments in designing the dynamic features and compared the performance against static head pose based approach. Analyses show that dynamic information significantly improves the results.

Fuzzy-System-Based Detection of Pupil Center and Corneal Specular Reflection for a Driver-Gaze Tracking System Based on the Symmetrical Characteristics of Face and Facial Feature Points

“…The initial calibration for each driver demands as long as 5-15 min. Another research [36] derived the locations of a driver's forward gaze from facial feature points by using two cameras. The cameras were installed in the A-piller and near the rear-view mirror, respectively, to follow the movements of the driver's head.…”

“…[35], and the external illumination puts influence. Gaze tracking is based not on the pupil, but on head movements, which constrains the improvement of accuracy [36]. Figure 1 shows the overall flowchart of our fuzzy-system-based pupil and corneal reflection detection and gaze-tracking system, and we show only the rough procedure of proposed method in this Section.…”

Where is the driver looking: Analysis of head, eye and iris for robust gaze zone estimation

“…Vision based systems are commonly used for this purpose as they provide non-contact and non-invasive solution. In recent years, a number of efforts [10], [11] have been made towards robust and continuous monitoring of the driver. Such systems are often focused on head pose alone due to robustness requirement.…”

“…It is argued that the drivers tend to use both head and eye combination to direct their gaze to the target [5]. It is shown [10] that incorporating dynamic information of the head can provide better zone estimation accuracy. Perhaps, because head and eye co-ordination during driving [12] is exploited by head pose dynamics, revealing eye movement information indirectly.…”

“…A distraction detection study [5] suggests that even though precise eye gaze is a better predictor, head pose still provides a good estimate of the driver distraction. Tawari and Trivedi [10] showed that head pose dynamics provide better cues than static head pose alone to estimate coarse gaze directions or gaze zones such as front, left/right side mirrors, speedometer, center console, etc. The system, however, fails to distinguish between adjacent zones separated by subtle eye movement e.g.…”