Motion Estimation of Driver's Head from Nostrils Detection

We propose a method of estimating driver's head movement by tracking a pair of nostrils. Nostrils clearly appear as a pair of dark areas in a facial image. S hape of nostrils have little influence of deformation compared to eyes. We use a concentric circular mask or density difference mask to detect difference of image density. We represent nostril position by a center of gravity of extracted nostril region, and apply moving average to a sequence of the detected nostrils positions. We decide reference positions of nostrils through a procedure that an interval of moving average is fixed into a standard cycle of model driver’s head motion. We estimate driver’s head movement from a difference between detected nostrils position and reference nostrils position. Cycle of driver’s head movement is available for a measure of monitoring driving condition. Experimental results show the effectiveness of our proposed method.