<![CDATA[Traffic accidents caused by drowsiness continue to pose a serious threat to road safety. Many of these accidents can be prevented by alerting drivers when they begin to feel sleepy. This research introduces a non-invasive system for detecting driver drowsiness based on visual features extracted from videos captured by a dashboard-mounted camera. The proposed system utilizes facial landmark points and a facial mesh detector to identify key areas where the mouth aspect ratio, eye aspect ratio, and head pose are analyzed. These features are then fed into three different classification models: 1D-CNN, LSTM, and BiLSTM. The system’s performance was evaluated by comparing the use of these features as indicators of driver drowsiness. The results show that combining all three facial features is more effective in detecting drowsiness than using one or two features alone. The detection accuracy reached 0.99 across all tested models.]]>
Copyrights © 2025
