Driver Sense

Abstract

Drowsiness has been a major cause of terrible accidents that have resulted in deaths and injuries all over the world. Globally, the number of fatal injuries is increasing day by day. Based on a 3D-deep convolutional neural network, we propose a condition-adaptive representation learning framework for driver drowsiness detection. Spatial-temporal representation learning, scene condition understanding, feature fusion, and drowsiness detection are the four models that make up the proposed framework. Learning spatial-temporal representations extracts features that can describe motions and appearances in the video at the same time. Scene condition understanding categorizes scene conditions relating to various aspects of drivers and driving situations, such as the status of wearing glasses, driving illumination, and the motion of facial elements such as the head, eye, and mouth. Automatic emergency braking (AEB) is an intelligent vehicle active safety system that helps drivers avoid certain types of collisions. Automatic emergency braking technology has become more widely used as automotive active safety technology has progressed, playing a key role in avoiding rear-end collisions as well as collisions with pedestrians and other road users. The technical characteristics of the automatic emergency braking system are examined, a subjective evaluation index is proposed, vehicle-to-vehicle, vehicle-to-pedestrian, and other typical scenes for automatic emergency braking subjective evaluation and actual vehicle verification are selected, and a subjective evaluation system for automatic emergency braking of passenger cars is constructed.