According to statistics, the accident rate caused by fatigue driving and mental inconcentration in traffic accidents accounts for 93%. To solve this problem, 68 key points of face are introduced in this paper. Based on deep learning and machine vision algorithm concepts, combined with YOLOv8 model and related fatigue driving judgment mechanism, PERCLOS algorithm, MAR algorithm, EAR algorithm and HPE algorithm are used to improve the accuracy and reliability of the system. A set of algorithms for recognizing tired driving behavior is successfully constructed.