Driver Fatigue Monitoring Based on Facial Multifeature Fusion

In response to the problems of low efficiency and inaccuracy of existing fatigue driving recognition and detection algorithms, this paper proposes a machine learning-based driver fatigue detection algorithm. First, an adaptive median filtering algorithm is used to denoise the collected images, and then the Summed-area table is used to calculate its Har Like eigenvalue. Second, the Adaboost face detection and recognition algorithm and the tracking algorithm that predicts the relocation of the target position are used to quickly extract the features of the tracked face, and the improved cascade regression Tree model is used to locate the face feature points. The support vector machine algorithm is used to classify the collected eye feature values, setting a threshold to determine whether the eyes are in a closed state, and fusing the feature values of the mouth and head. Finally, through a fatigue detection algorithm based on multifeature weighted summation, repeated experimental simulations were conducted on the trained dataset to determine the weight values of the eye parameters as 0.55, the mouth parameters as 0.25, and the head parameters as 0.20. Based on the differences in the weight values of the feature values obtained from the simulation experiments, the fatigue level was divided into awake, mild fatigue, moderate fatigue, and severe fatigue. Repeated experimental data testing shows that the method has greatly improved the accuracy and recognition speed of sleep-deprived driving recognition and detection. Tests on a dataset of simulated driving have shown that our model has an average accuracy of approximately 95.1%.