Electronic Thesis and Dissertation Repository


Doctor of Philosophy


Computer Science


Dr. Steven Beauchemin


The primary goal of this thesis is to provide processed experimental data needed to determine whether driver intentionality and driving-related actions can be predicted from quantitative and qualitative analysis of driver behaviour. Towards this end, an instrumented experimental vehicle capable of recording several synchronized streams of data from the surroundings of the vehicle, the driver gaze with head pose and the vehicle state in a naturalistic driving environment was designed and developed. Several driving data sequences in both urban and rural environments were recorded with the instrumented vehicle. These sequences were automatically annotated for relevant artifacts such as lanes, vehicles and safely driveable areas within road lanes. A framework and associated algorithms required for cross-calibrating the gaze tracking system with the world coordinate system mounted on the outdoor stereo system was also designed and implemented, allowing the mapping of the driver gaze with the surrounding environment. This instrumentation is currently being used for the study of driver intent, geared towards the development of driver maneuver prediction models.