TY - GEN
T1 - Creating Geopositioned 3D Areas of Interest from Fleet Gaze Data
AU - Bickerdt, Jan
AU - Gollnick, Christian
AU - Sonnenberg, Jan
AU - Kasneci, Enkelejda
N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2022
Y1 - 2022
N2 - In order to observe driver’s attention levels, different approaches are followed. They include simple methods counting driver input changes [6], machine learning based approaches based on driver input [17], and methods considering additional inputs such as environmental data and eye tracking data [3–5, 7, 12, 16]. Recent studies have proposed geopositioned 3D AOIs as a tool for driver intention observation. Geopositioned 3D AOIs are three dimensional Areas (boxes), with fix geopositiones (e.g. GPS) which have to be observed for a safe completion of driving maneuvers. Examples are pedestrian waiting areas, crosswalks, and traffic light. Creating these AOIs by hand is a tedious task with ample room for potential errors, as the created AOIs might differ from the real AOIs drivers look at. We therefore propose a pipeline to generate real 3D AOIs from gaze clouds. To generate relevant gaze clouds we use the points of closest encounter in fleet gaze data collected in a driving simulator setup. The results show that the generation of 3D AOIs from fleet data is possible and the created AOIs are mostly consistent with the expected AOIs.
AB - In order to observe driver’s attention levels, different approaches are followed. They include simple methods counting driver input changes [6], machine learning based approaches based on driver input [17], and methods considering additional inputs such as environmental data and eye tracking data [3–5, 7, 12, 16]. Recent studies have proposed geopositioned 3D AOIs as a tool for driver intention observation. Geopositioned 3D AOIs are three dimensional Areas (boxes), with fix geopositiones (e.g. GPS) which have to be observed for a safe completion of driving maneuvers. Examples are pedestrian waiting areas, crosswalks, and traffic light. Creating these AOIs by hand is a tedious task with ample room for potential errors, as the created AOIs might differ from the real AOIs drivers look at. We therefore propose a pipeline to generate real 3D AOIs from gaze clouds. To generate relevant gaze clouds we use the points of closest encounter in fleet gaze data collected in a driving simulator setup. The results show that the generation of 3D AOIs from fleet data is possible and the created AOIs are mostly consistent with the expected AOIs.
KW - Areas of interest
KW - Automotive
KW - Eye tracking
KW - Fleet data
UR - https://www.scopus.com/pages/publications/85132982865
U2 - 10.1007/978-3-031-04987-3_2
DO - 10.1007/978-3-031-04987-3_2
M3 - Conference contribution
AN - SCOPUS:85132982865
SN - 9783031049866
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 21
EP - 34
BT - HCI in Mobility, Transport, and Automotive Systems - 4th International Conference, MobiTAS 2022, Held as Part of the 24th HCI International Conference, HCII 2022, Proceedings
A2 - Krömker, Heidi
PB - Springer Science and Business Media Deutschland GmbH
T2 - 4th International Conference on HCI in Mobility, Transport, and Automotive Systems, MobiTAS 2022 Held as Part of the 24th HCI International Conference, HCII 2022
Y2 - 26 June 2022 through 1 July 2022
ER -