基于卷积神经网络的安全驾驶智能监测系统  

Intelligent Monitoring System for Safe Driving Based on Convolutional Neural Network

作　　者：王振辉 陈虹烨 张志华[1] WANG Zhenhui;CHEN Hongye;ZHANG Zhihua(School of Electronic and Information Engineering,University of Science and Technology Liaoning,Anshan,Liaoning 114051,China)

机构地区：[1]辽宁科技大学电子与信息工程学院,辽宁鞍山114051

出　　处：《自动化应用》2025年第6期36-39,共4页Automation Application

基　　金：辽宁科技大学大学生创新创业训练计划项目(S202410146097)。

摘　　要：随着汽车保有量的持续增长,司机安全驾驶问题导致的事故在道路安全事故中占比越来越高,因此司机驾驶监测识别技术对道路交通安全系统非常重要。为实现汽车行驶过程中的安全监管,基于卷积神经网络对司机驾驶状态进行智能监测识别。采用基于卷积神经网络的YOLOv7算法,对司机驾驶过程中的非安全目标(如手机、烟头、水瓶等)进行分心驾驶识别;通过Dlib实现人脸检测,并利用卷积神经网络提取人脸关键点特征,最终通过计算疲劳驾驶程度对司机的打哈欠、眯眼等行为进行疲劳监测识别。结果表明,该监测系统对分心驾驶的识别准确率为97.1%,对疲劳驾驶的识别准确率为96.2%,证明该监测系统可为司机道路安全驾驶提供良好的保障。As the number of vehicles continues to grow steadily,accidents caused by unsafe driving behaviors by drivers account for an increasingly high proportion of road safety incidents,rendering driver monitoring and recognition technologies particularly crucial for road safety systems.To achieve safety oversight during vehicle operation,an intelligent monitoring and recognition system for drivers'driving states is implemented based on Convolutional Neural Network.The YOLOv7 algorithm,which is grounded in Convolutional Neural Network,is employed to recognize distracted driving by detecting non-safety targets(such as mobile phones,cigarettes,water bottles,etc.)during the driving process.Face detection is achieved through Dlib,and Convolutional Neural Network are utilized to extract facial keypoint features.Ultimately,behaviors indicative of tired driving,such as yawning and eye squinting,are monitored and recognized by calculating the degree of fatigue.The results indicate that the monitoring system achieves an accuracy rate of 97.1%for distracted driving recognition and 96.2%for tired driving recognition.This demonstrates that the monitoring system can provide robust safeguards for drivers'road safety.

关 键 词：YOLOv7 卷积神经网络 目标识别 疲劳驾驶 分心驾驶 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]