基于深度学习的面部特征检测与疲劳驾驶预警  

Facial Feature Test and Fatigue DrivingWarning Based on Deep Learning

作　　者：张苗苗 柴国强[1] 于海乐 徐昊璇 ZHANG Miao-miao;CHAI Guo-qiang;YU Hai-le;XU Hao-xuan(School of Physics and Information Engineering,Shanxi Normal University,Taiyuan 030000,Shanxi)

机构地区：[1]山西师范大学物理与信息工程学院,山西太原030000

出　　处：《电脑与电信》2022年第12期1-5,共5页Computer & Telecommunication

基　　金：国家自然科学基金,项目编号:62201333;山西省基础研究计划,项目编号:20210302124647;山西省高等学校科技创新项目,项目编号:2021L269。

摘　　要：疲劳驾驶是道路交通安全的无形杀手。为保证检测的高效性、准确性,设计出一种基于深度学习的疲劳驾驶检测系统。首先,运用OpenCV对实时采集的图像进行灰度化预处理;然后,通过方向梯度直方图特征检测并提取人脸区域,利用Dlib模型标定68个人脸特征点;最后,利用改进的眨眼、打哈欠和点头检测算法分别计算出眼睛、嘴巴的长宽比以及头部姿态的欧拉角,并与其对应的阈值作对比,进而判定驾驶员是否处于疲劳状态并做出相应的预警措施。实验表明,该系统进行检测时,准确率高达97%,验证了其有效性。Fatigue driving is an invisible killer of road traffic safety.A fatigue driving detection system based on deep learning is proposed in this paper to ensure the high efficiency and accuracy of detection.First,OpenCV is applied to gray preprocess the collected image.Second,directional gradient histogram is used to extract feature and the pre-trained Dlib model is applied to calibrate 68 face feature points.Finally,the improved detected algorithms of blink,yawn and nod are used to calculate the length width ratio of eyes and mouth and head posture Euler angle,respectively,and compare with its corresponding threshold to determine whether the driver is in fatigue state and take early warning measures.Experiments show that the proposed system has 97%accuracy,verifying its effectiveness.

关 键 词：深度学习 人脸特征点 疲劳驾驶预警 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]