一种信息融合的驾驶员疲劳检测系统设计与实现  

Design and implementation of a driver fatigue detection system based on information fusion

作　　者：孙岩 董恒 梁西博 周立新 SUN Yan;DONG Heng;LIANG Xibo;ZHOU Lixin(The Unit 63726 of PLA,Yinchuan 750000,China)

机构地区：[1]63726部队,宁夏银川750000

出　　处：《电子设计工程》2024年第11期71-75,81,共6页Electronic Design Engineering

摘　　要：疲劳驾驶是交通事故发生的主要原因之一,为减少事故发生概率,以树莓派作为服务器,搭建网站式检测系统,置于驾驶舱采集图像流后多线程处理,线程一采用Dlib(图像处理开源库)人脸关键点检测器,通过实时计算驾驶员眼部的纵横比和嘴部长宽,与设定阈值进行比较,得出眨眼和打哈欠的统计次数和频率,得出疲劳状态;线程二通过混合域CBMA注意力模块改进传统VGG16网络,优化了模型算法,实现疲劳状态自动判别;线程三采用头部姿态检测(HPE)算法,计算驾驶员头部姿态的欧拉角,统计点头次数和频率,得出疲劳状态;各线程处理结果,送至融合算法,综合判别驾驶员状态,若为困倦状态,语音提醒和向绑定邮箱发送抓拍图像及结果。远端用户可通过http网页实时查看驾驶员视频图像,同时接收服务器推送邮件。通过多次实验,验证了系统的有效性和判定结果准确性。Fatigue driving is one of the main causes of traffic accidents.In order to reduce the probability of accidents,a website⁃type detection system is built with Raspberry PI as the server,and multi⁃thread processing is carried out after image acquisition stream in the cockpit.In thread one,a Dlib(open source image processing library)face key point detector is adopted to calculate the aspect ratio of the driver’s eyes and the length and width of the mouth in real time.The statistical number and frequency of blinking and yawning were obtained by comparing with the set threshold.Thread 2 improved the traditional VGG16 network by using the hybrid domain CBMA attention module,optimized the model algorithm,and realized the automatic identification of fatigue state.The third thread used the head attitude detection(HPE)algorithm to calculate the Euler Angle of the driver’s head attitude,counted the number and frequency of nodding,and obtained the fatigue state.The processing results of each thread are sent to the fusion algorithm to comprehensively judge the driver’s state.If the driver is sleepy,the voice reminds and sends the captured image and results to the binding mailbox.Remote users can view the driver’s video images in real time through http web pages and receive emails pushed by the server.The effectiveness of the system and the accuracy of the decision result are verified by many experiments.

关 键 词：疲劳驾驶 树莓派 Dlib VGG16 HPE算法 视频图像 

分 类 号：TN911.7[电子电信—通信与信息系统]