基于机器视觉多信息融合的疲劳状态检测

发布时间：2018-01-24 02:47

本文关键词： 非接触式心率测量摄像头相位误差系统时钟抖动误差疲劳状态检测多信息融合　出处：《天津大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着生活节奏的加快和娱乐性活动的增多,人们每天用于休息的时间越来越短,从而很难能够以较充足的精力投入到每天的工作和学习中,这不但直接影响自己的工作能力而且在特殊的行业很有可能造成极大的经济损失和人员伤亡。特别是在航空航天、长途运输、大型服务器监管等行业,一旦操作员在精神疲劳状态下工作,将会使整个系统处于极其危险的境地。而目前,精神疲劳的检测设备往往都存在很多缺点,难以应用于这些场合,因此,急需一种新型的疲劳状态检测系统。本文提出了一种基于机器视觉的多信息融合的疲劳状态检测系统,首先,提出一种基于机器视觉的心率检测方法,并针对该检测系统中存在的CMOS摄像头引起的相位误差和图像采集系统引起的时钟抖动误差,提出了基于幅频叠加算法的相位误差消除方法和基于时间表的三次样条插值方法消除采集系统延时引起的随机误差,从而提高了该检测系统的测量精度。其次,将这种心率检测算法通过插值重构后做心率变异性分析,并结合实验发现基于机器视觉的心率变异性分析中的时域均值和频域高频分量与人的疲劳状态具有一定的相关性,因此可以采用时域均值和频域高频分量作为疲劳状态研究的特征量。最后,将这种心率变异性分析的疲劳状态检测方法和P80疲劳判定标准相结合,并通过实验验证了融合后的系统能够判断出受试者不同的疲劳状态和深度疲劳状态,同时也提出了融合模型,可以根据不同的应用场景设置不同的疲劳检测深度,从而达到对疲劳状态预警的作用。因此,这种基于一个摄像头的多信息融合的疲劳状态检测方法可以用于不同深度的疲劳检测和预警,并且具有操作简单、成本低廉、无侵入性等优点,具有极大的应用价值。
[Abstract]:With the quickening pace of life and the increase of recreational activities, people spend less and less time to rest every day, so it is difficult to devote more energy to daily work and study. This not only directly affects their own working ability but also may cause great economic losses and casualties in special industries, especially in aerospace, long-distance transportation, large-scale server supervision and other industries. Once the operator works in the state of mental fatigue, the whole system will be in a very dangerous situation. At present, the mental fatigue detection equipment often has a lot of shortcomings, so it is difficult to be used in these situations. A new fatigue state detection system is urgently needed. In this paper, a new fatigue state detection system based on multi-information fusion of machine vision is proposed. Firstly, a heart rate detection method based on machine vision is proposed. The phase error caused by the CMOS camera and the clock jitter error caused by the image acquisition system are analyzed. The phase error elimination method based on amplitude-frequency superposition algorithm and cubic spline interpolation method based on timesheet are proposed to eliminate the random error caused by the delay of the acquisition system, thus improving the measurement accuracy of the detection system. The heart rate variability is analyzed by interpolation reconstruction. Combined with experiments, it is found that the time-domain mean and frequency-domain high-frequency components of HRV analysis based on machine vision have a certain correlation with human fatigue state. Therefore, the time-domain mean and frequency domain high-frequency component can be used as the characteristic value of fatigue state research. Finally, the fatigue state detection method of heart rate variability analysis and P80 fatigue criterion can be combined. The experimental results show that the fusion system can judge the different fatigue state and the deep fatigue state of the subjects, and the fusion model is also proposed. Different fatigue detection depth can be set according to different application scenarios, so as to achieve the function of early warning of fatigue state. This multi-information fusion fatigue state detection method based on a single camera can be used for fatigue detection and early warning at different depths, and has the advantages of simple operation, low cost, no invasion and so on. It has great application value.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【参考文献】