Driver fatigue detection based on multi-feature fusion facial features

Fatigue driving is the main factor causing traffic accidents. To reduce the traffic risk, we explores a driver fatigue detection method based on OpenCV combined with PHOG algorithm. By extracting facial feature points from the eyes and mouth of drivers, this study employs an adaptive eyelid closure threshold for the eyes and a feature correction algorithm based on Euler angles. This approach enables real-time computation of fatigue indicators such as PERCLOS, blink frequency, and yawn frequency. By weighting and fusing these characteristic indicators, a fatigue score for the driver is derived. Based on predefined thresholds, the driver’s fatigue level is assessed, and corresponding warning messages are issued. Experimental results demonstrate the effectiveness of the proposed method in identifying driver fatigue, providing a reliable auxiliary tool for enhancing driving safety.