The invention discloses a real-time fatigue driving detection method. The method comprises the following steps: 1, acquiring image data from a video shot by a camera; 2, according to the obtained image data, performing state recognition, including face recognition, and completing eye and mouth positioning and classification; 3, calculating various parameters, and forming a multi-parameter composite criterion; 4, performing fatigue detection: determining whether fatigue exists or not according to a multi-parameter composite criterion, and if not, turning to the step 1; and if so, giving an alarm. According to the technology of the invention, an SSD (Single Shot MultiBox Detector) network is mainly utilized to identify faces in video images and position and classify eyes and mouths, and thena PERCLOS method, blinking frequency, yawn frequency and the like are utilized as fatigue judgment standards. In view of the detection effect of the SSD network, the technology of the invention has higher detection accuracy and real-time performance.