Fatigue driving is one of the main causes of traffic accidents in freight vehicles. Existing fatigue driving studies mostly use the vehicle running data or simulation data in the experiments, with certain defects in the validity and reliability of the model. The work collected a large number of naturalistic driving data of vehicle and drivers’ facial video to extract the sample data under different fatigue levels of drivers. Besides, the indicators of vehicle running states were selected based on significant driver fatigue levels. The BP neural network was used to establish the detection model of fatigue driving behaviors, considering the influence of the number of model training samples and other parameters on the accuracy of fatigue driving behavior detection. The status data of 50 freight vehicles for nearly one month was used to test the models of fatigue driving behavior detection. The analysis results showed that the accuracy of the model can reach more than 80%, which provides a new scheme for the detection of fatigue driving behavior and a theoretical basis for the warning of driver fatigue state.

CICTP