This paper presented a multi-mode fatigue detection system. It included visual detection, multi-point pressure detection and physiological signal detection. Visual detection mainly carried out on eyes and mouth feature extraction and the states detection. First, skin color detection was used into face image extraction,and then one-dimensional wavelet transform and gray integral projection were used to detect the eye and mouth in human face image. Finally, the status of the eyes and mouth were detected to judge fatigue. The principle of multi-point pressure detection was that during fatigue state period, the body and the back will be still in a long time. So we can use sensors to detect pressure points to judge fatigue. Physiological signal detection was detected by the speed of the human wrist pulse signal. The paper mainly presented visual detection that used the camera to capture face images. The algorithm has advantages of low computation complexity, high speed and better detection effect.