A 3D Driver Head Pose Estimation Method Based on Depth Image
Yafei Wang, Bowen Zhang, Xianping Fu · 2014
The driver's head pose is an indicator of driver's attention and plays a significantly role in driving safety. Traditional 2D color image processing method has disadvantage due to its vulnerability under poor environmental lighting conditions in vehicle. In this paper, we propose a novel method with calibrated gaze zone to estimate the driver head pose combining RGB and depth data. The head rotation can be calculated by Iterative Closest Point algorithm in 3D point cloud. In order to reduce the amount of calculation of ICP for head pose estimation, particle filter is used to track and learn the status of head movement. The head templates of different gaze zones and their neighbors for ICP also make contribution to an accurate result. The experimental results show that our approach has a low error rate and is acceptable in real driving environment.