基于面部特征与深度学习的疲劳驾驶状态检测研究

Fatigue Driving State Detection Based on Facial Features and Deep Learning

在驾驶机动车时,驾驶员的面部信息尤其是眼睛和嘴巴最能够反映驾驶员的疲劳状态。为了提高机动车驾驶的安全性,本文提出了一种基于面部特征和深度学习的疲劳驾驶状态检测研究模型。首先设计一种改进的三级级联卷积神经网络检测驾驶员人脸图像,再使用轻量级特征提最小单元结构定位人脸关键点,通过基于眼睛纵横比(Eye Aspect Ratio,EAR)和基于嘴唇纵横比(Mouth Aspect Ratio,MAR)的方法判定眼睛疲劳和嘴部疲劳状态,最后利用支持向量机(SVM)融合眼部和嘴部疲劳特征进行疲劳驾驶状态检测。通过实验表明,该算法可以准确地定位出人脸关键点,且具有较高的疲劳检测准确率和较好的鲁棒性。

When driving a motor vehicle,the driver’s facial information,especially the eyes and mouth,can best reflect the driver’s fatigue state.In order to improve the safety of motor vehicle driving,a fatigue driving state detection research model based on facial features and deep learning is proposed in this paper.First,design an improved three-level cascade convolution neural network to detect drivers face image.Then,use lightweight feature extraction minimum unit structure to locate face key points,determine the eye fatigue state and mouth fatigue state based on the Eye Aspect Ratio(EAR)and Lip Aspect Ratio(MAR)method。Finally,use support vector machine(SVM)fusing fatigue characteristics of the eyes and mouth to detect the state of driving fatigue.Experiment shows that this algorithm can accurately locate the key points of human face,and has high fatigue detection accuracy and good robustness.