Aiming at the problem that the accuracy of the existing anti-fatigue early warning systems is less than 95% and they are susceptible to interference in the harsh driving environment of ultra-high altitude mines, a fatigue detection system based on visual and electroencephalogram (EEG) features was developed. The FastICA algorithm was used to remove ocular artifacts, and the EEG energy ratio and sample entropy were extracted. YOLOX was used for face detection, combined with an improved PFLD algorithm to extract eye state, yawn and head posture features. A multi-source information fusion decision algorithm based on D-S evidence theory and SVM was designed, which effectively integrates the probabilistic classification results of visual and EEG features. The application in the ultra-high altitude mine in Las Bambas, Peru, shows that the detection accuracy of the system based on multi-source information fusion is significantly improved to 97.6%, and the key fatigue features can be reliably captured. The fusion of visual and EEG features can significantly improve the accuracy and robustness of fatigue detection in ultra-high altitude environments. The fusion algorithm of D-S evidence theory and SVM is the key to achieving 97.6% high accuracy. The system provides reliable technical support for the transportation safety of high-altitude mines, which can reduce accident risk by about 15%, and also provides a new approach for anti-fatigue research in extreme environments.

High-altitude mine  /  Fatigue driving detection  /  EEG feature  /  Visual feature  /  D-S theory  /  SVM algorithm