基于多特征经验融合的驾驶员疲劳状态检测方法  

Driver fatigue detection method based on multi-feature empirical fusion

作　　者：丁福生 秦彦彬 张岚祥 吕红明[1] Ding Fusheng;Qin Yanbin;Zhang Lanxiang;Lyu Hongming(School of Automotive Engineering,Yancheng Institute of Technology,Yancheng 224051,China)

机构地区：[1]盐城工学院汽车工程学院,盐城224051

出　　处：《国外电子测量技术》2024年第9期190-196,共7页Foreign Electronic Measurement Technology

基　　金：国家自然科学基金面上项目(51875494);江苏省研究生实践创新计划项目(KYCX24_XZ041)资助。

摘　　要：为解决驾驶员疲劳导致交通事故的问题,提出一种基于多特征经验融合的驾驶员疲劳状态视觉检测方法。首先,通过实时捕捉和提取驾驶员面部状态特征,包括眼部特征和嘴部特征,并采用经验融合模型对这些特征进行分析。接着,将多维度面部行为信息映射到卡罗林斯卡睡眠量表(KSS)分值,以此评估驾驶员的疲劳状态。最后,搭建实验验证了该检测方法的准确性、可靠性和有效性。实验结果表明,对于不同程度的疲劳状态,该方法的准确率分别为清醒状态90.34%,轻度疲劳90.17%,中度疲劳90.46%,重度疲劳97.67%。该检测方法能够准确评估驾驶员的疲劳程度,为提高行车安全性提供了有效的技术支持。To address the problem of traffic accidents caused by driver fatigue,this paper proposes a visual driver fatigue detection method based on multi-feature experience fusion.First,the driver's facial state features,including eye features and mouth features,are captured and extracted in real-time,and an experience fusion model is used to analyze these features.Then,the multi-dimensional facial behavior information is mapped to the Karolinska sleepiness scale(KSS)score to assess the driver's fatigue level.Finally,an experiment is conducted to verify the accuracy,reliability,and effectiveness of the detection method.The experimental results show that the accuracy of the detection method for different fatigue levels is 90.34%for the alert state,90.17%for mild fatigue,90.46%for moderate fatigue,and 97.67%for severe fatigue.This detection method can accurately assess the driver's fatigue level and provide effective technical support for improving driving safety.

关 键 词：驾驶疲劳 疲劳检测 面部特征 特征融合 机器视觉 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]