Driver's Head Detection Model in Color Image for Driver's Status Monitoring

Abstract

In this paper, we proposed a simple driver's head detection model for driver's status monitoring, in order to implement interactive safety driving agent system. For intensifying the driver's head region, we consider normalized color feature combination method with skin color filter. And, the center-surround difference and normalization algorithms are applied for making saliency map of driver's head in order to nt only reduce the background noise, but also more reinforce the intensity of driver's head regions than non-head area. The driver's head regions are simply selected through searching local maximum energy with histogram projection method. And, we adopt the 2D+3D active appearance model (AAM) for tracking driver's head region and estimating driver's head pose. Experimental results show that proposed model can successfully detect driver's head region and estimate driver's head pose.