Start date: 07/01/15
End date: 12/31/16
- Senem Velipasalar | email@example.com | Syracuse University
- Akhan Almagambetov | firstname.lastname@example.org | Norwich University
- Yaw Adu-Gyamfi
About the research
A key to the development of effective crash countermeasures is an understanding of pre-crash causal and contributing factors. Accurately determining the cause or behaviors leading to traffic crashes is a very challenging process. Traditionally, traffic conflict observations by manual or automated means have been used to determine pre-crash causal factors. In recent years, naturalistic driving studies have been used to provide detailed and more accurate pre-crash causal information.
Naturalistic driving databases contain large datasets of low-resolution video streams (due to compression) of highly variable intersections, multiple flows, and turning movements of vehicles under very complex lighting conditions (e.g., constantly varying shadows). This makes automated traffic conflict detection and analysis very challenging. Also, existing vision-based traffic conflict detection systems are not designed to work in naturalistic real world settings; as a result, they fail to extract relevant information from these databases.
To maximize the use of this valuable safety dataset, vision systems with a high-level of understanding of scene dynamics around the naturalistic driver must be developed. The project aims at developing a vision system for understanding pre-crash causal factors through the detection and analysis of traffic conflicts in a naturalistic real world setting.
- Iowa State University
- Midwest Transportation Center