Application of Face Recognition Based on CNN in Fatigue Driving Detection

Abstract

Fatigue driving detection technology based on the external characteristics of the driver has made some progress in many aspects, but the method of driver facial feature extraction needs to be further improved, and the driver's eye location takes a long time, which affects the system recognition rate. The paper uses the fatigue driving detection method to achieve better results than the traditional detection method. The authors applied the convolutional neural network to face recognition, and improve the pupil localization algorithm, effectively overcoming the problem of large calculation of the original algorithm. According to the characteristics of driver's eyes with different width and height ratios in different states, a simple and feasible method of eye state judgment is realized, and the driver's fatigue state is judged by PERCLOS algorithm. The convolutional neural network model is applied to ORL face database, and the face recognition rate is 85%. The improved Hough transform method has a positioning accuracy of 92% for the driver's eyes, respectively, and the recognition rate for the driver's eye state is 83.9%. The authors designed the prototype system of fatigue driving detection based on face recognition which realizes the functions of driver's face feature detection, eye location, eye state judgment and fatigue judgment. The experimental results show that the recognition rate of fatigue is 87.5%.