基于面部特征的疲劳驾驶检测方法研究  

Research on Fatigue Driving Detection Method Based on Facial Features

作　　者：汪洪涛 谢牡丹 潘昊[2] WANG Hongtao;XIE Mudan;PAN Hao(Network Information Center, Wuhan University of Technology, Wuhan 430070, China;School of Computer Science and Artificial Intelligence, Wuhan University of Technology, Wuhan 430070, China)

机构地区：[1]武汉理工大学网络信息中心,武汉430070 [2]武汉理工大学计算机与人工智能学院,武汉430070

出　　处：《武汉理工大学学报（交通科学与工程版）》2021年第5期851-856,862,共7页Journal of Wuhan University of Technology（Transportation Science & Engineering）

基　　金：国家自然科学基金(21376185)。

摘　　要：疲劳驾驶是造成道路交通事故的主要因素之一.文中提出一个基于生成对抗网络的弱光增强算法,对驾驶员图像时行预处理,再采用一个改进的多任务级联卷积神经网络检测面部,定位关键点,最后通过深度学习进行面部与特征位置的预测,进而做出多指标的疲劳驾驶判断.Driving fatigue is one of the main factors causing road traffic accidents.In this paper,a weak light enhancement algorithm based on generated countermeasure network was proposed to preprocess the driver image.Then,the improved multi-task cascade convolutional neural network was used to detect the face and locate the key points.Through deep learning,face and feature positions were predicted,and then multi-index fatigue driving judgment was made.This study further improved the detection accuracy of fatigue driving.

关 键 词：疲劳驾驶 弱光增强 面部特征 卷积神经网络 

分 类 号：U463.6[机械工程—车辆工程]