Development & Testing of a Camera-based Driver Monitoring System

Distracted or drowsy driving is one of the major causes of road accidents. Often, long and monotonous road journeys lead to distracted or drowsy driving. Therefore, there is a need for a system which alerts a distracted or drowsy driver. Moreover, as the levels of autonomy move beyond SAE Level 2, the system assumes a larger share of the dynamic driving task. Under challenging circumstances, the system might ask the driver to take back vehicle control. To ensure safety, it is therefore important to monitor the driver’s state to gauge their ability to take back the vehicle’s control. A driver-monitoring system (DMS), also referred to as driver state sensing (DSS) system, is an advanced safety feature to track driver drowsiness or distraction, and to issue a warning or alert to get the driver’s attention back to the task of driving. This paper presents a novel camera-based driver drowsiness system developed using a standard webcam and an embedded hardware. The system leverages deep learning for face detection, face landmark extraction and head pose estimation. Key facial landmarks are used to determine blink rate, track eye gaze and detect yawn. To improve robustness against brightness changes, the software also implements an auto calibration feature. The system uses multiprocessing capabilities of the embedded hardware for improved performance. The system was tested for functionality and performance inside the lab and on a real vehicle.