摘要
驾驶人的疲劳状态会大大增加车祸发生的概率,为有效避免这类事件的发生,以OpenCV机器视觉库为核心,基于树莓派构建疲劳驾驶检测系统。驾驶人在疲劳状态下眨眼、打哈欠、低头犯困的频率明显增加,因此在驾驶人上方安装摄像头实时采集驾驶人面部图像,根据视频流中Dlib开源库标识出的68个人脸特征关键点,计算出眨眼频率与打哈欠频率。提取头部3D特征精确计算出头部姿态欧拉角,多次实验设定最优阈值。最终,得到疲劳判定的准确率为87%。
The fatigue state of drivers will greatly increase the probability of traffic accidents.In order to effectively avoid the occur⁃rence of such events,the fatigue driving detection system is built based on Raspberry Pi with OpenCV machine vision library as the core.In the fatigue state,the frequency of drivers blinking,yawning and lowering their heads to fall asleep significantly increases.Therefore,a camera is installed above the driver to collect real-time facial images of drivers.According to the 68 face feature key points identified by the Dlib open source library in the video stream,the blinking frequency and yawning frequency are calculated.The head posture Euler Angle was accurately calculated by extracting 3D head features,and the optimal threshold was set by many experi⁃ments.Finally,the accuracy of fatigue determination is 87%.
作者
张伯辰
施鑫杰
霍梅梅
Zhang Bochen;Shi Xinjie;Huo Meimei(College of Computer and Computing Science,Zhejiang University City College,Hangzhou 310015)
出处
《现代计算机》
2021年第23期129-132,共4页
Modern Computer
基金
国家大学生科研创新项目(202013021002)。
