A camera vision system is implemented for detecting the driver's gaze and head orientations and detecting the momentary attention targets by using a mapping algorithm. In the developed rule-base classification algorithm, the cockpit view is divided into four clusters of interest: road ahead, windscreen, and left and right exterior mirrors. The early field experiments prognosticate up to 80 % performance for visual distraction detection. Cognitive distraction may occur rapidly, e.g. the mobile phone rings and the conversation steals the driver's attention. Moreover, the assumption is that one parameter alone does not reveal the distraction, but rather by fusing multiple sensor data, the robustness of the detection can be improved. For this reason the work carried out promotes use of the multi-dimensional support vector machine classification method to retrieve possible cognitive workload. According to the laboratory experiments the performance is varying between 68 - 86 % depending on the actual environmental factors.
|Publication status||Published - 2008|
|MoE publication type||Not Eligible|
|Event||XXXIX Annual Conference of Italian Operational Research Society, AIRO2008: Optimisation and logistics in transportation and communication networks - Ischia, Italy|
Duration: 8 Sep 2008 → 11 Sep 2008
|Conference||XXXIX Annual Conference of Italian Operational Research Society, AIRO2008|
|Period||8/09/08 → 11/09/08|
- visual workload
- cognitive workload
Kutila, M., Biamino, G., Markkula, G., & Parra, N. (2008). Driver Activity Assessment with a Machine Vision System. 71-72. Paper presented at XXXIX Annual Conference of Italian Operational Research Society, AIRO2008, Ischia, Italy.