Automatic context-dependent driver attention monitoring using eye tracking and digital maps

Sammanfattning: Driver distraction is a contributing factor to car accidents here a warning system could potentially prevent some of these accidents by making the driver aware of distraction and redirecting the focus back to driving-related activities. The purpose of this work is to better understand where drivers direct their attention while driving and where the attention should be directed for the driver to be considered attentive. For this, zones determining where the driver is looking were created. This project aimed to analyze how drivers direct their gaze when approaching different types of intersections with different types of features, and after that use this data to create zones for where a driver should have looked before crossing each type of intersection to be considered attentive to driving. The data used in this project was received from Smart Eye AB in Gothenburg. Data were collected from 38 subjects driving a route in the city of Gothenburg while eye tracking and GPS data were sampled. These data sets were preprocessed and then combined into one set of data. This combined data set was used to analyze drivers' glance behavior in different types of intersections (three-way intersections, four-way intersections, roundabouts, and motorway junctions) with different features, such as pedestrian crossings, traffic lights, yield signs, etc. The most rewarding results were obtained from the three-way and four-way intersections where the driver makes a turn in the intersection. It was established that when approaching an intersection the proportion of glances in the forward zone decreases. In contrast, the ratio of glances in the left, right, outside left, and outside right zones increases. This could be interpreted as a need for attention to be directed here. The conclusion was that the work done in this project has great potential to become useful in the further development of a driver distraction algorithm for implementation in future car models to prevent accidents from occurring because of driver inattention.