An Embedded Solution of Gaze Estimation for Driver Assistance using Computer Vision

Drowsy driving is a severe public health issue that has to be addressed. According to recent studies, sleepy drivers are responsible for almost 20% of all car accidents. Trustworthy sleepiness detection is now one of the key goals in developing new advanced driver assistance systems. In this research work, a vision based approach has proposed that identify the drowsy driver by monitoring the eye closer and sight movement. These two methods have been combined to reduce the false-positive results and make the system more robust. Proposed system is not only have performance robustness, but also not required high computional power, and a light embedded system, such as Raspberry Pi, can be used to operationalize the proposed drowsness detection system. The proposed ialgorithms could be helpful vehicle safety solutions for alerting drivers about tiredness. This system was tested on a Raspberry pi-4 with 8GB RAM using a Logitech HD720 Webcam. The outcomes of the experiment appear to be quite positive and promising. For face and eye-tracking, the system could achieve a frame rate of more than 09 frames per second, the average accuracy rate for eye position and tracking reach 97.3%. As a result, the proposed strategy can be inferred to be a lowcost and effective solution for detecting driver drowsiness in real time.