Advanced driver assistance systems provide a significant improvement for road safety and effective driving. Integrating driver observation in such systems is a crucial step in striving to reducing traffic accidents. For improved robustness in the designed system, we have chosen to approach the driver head pose estimation problem through regression methods based on images from a stereo camera. Therefore, the system operates solely on the 3D head information and is independent of facial feature detection, color and texture information. The proposed system contemplates real driving situations along with the design dictated position of the camera inside the car. The proposed regression algorithms for this work are support vector regression (SVR), random regression forest (RRF) and extremely randomized trees (ERT). Carried experimental studies show high accuracies for the proposed methods. Their algorithmic simplicity and measured time-costs further indicate their suitability for embedded real-time applications.
Driver Head Pose Estimation by Regression
Lect.Notes Mobility
2015-06-30
13 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Head Pose Estimation for Driver Monitoring
British Library Conference Proceedings | 2004
|Head pose estimation for driver monitoring
IEEE | 2004
|STEREO VISION-BASED DRIVER HEAD POSE ESTIMATION
British Library Conference Proceedings | 2014
|Improving Driver Pose Estimation
British Library Conference Proceedings | 2003
|Improving driver pose estimation
IEEE | 2002
|