Fatigue driving state detection for tanker truck drivers based on multi-feature fusion

As the significance of truck transportation in the modern economy continues to grow, the issue of fatigue driving in tanker trucks has garnered significant attention. Therefore, this study proposes a multimodal fatigue driving detection method. It involves conducting driving experiments with tanker trucks, collecting driving operation data, electrocardiogram data, and eye-tracking data. After data preprocessing, a multimodal driving dataset is generated. Data mining techniques are used to extract 42 driving feature values, and then, through correlation analysis, 27 feature values are selected for fatigue state detection. Subsequently, the K-means method is employed to classify driving data into four fatigue levels, and a random forest algorithm is used for fatigue state recognition. Experimental results demonstrate that the proposed fatigue detection algorithm achieves a precision rate of 92.6%, effectively identifying different fatigue driving states. This approach provides insights and theoretical support for subsequent driver risk assessment and targeted traffic management.