基于面部倒立摆模型与信息熵的驾驶员疲劳检测

Driver Fatigue Detection Based on Facial Inverted Pendulum Model and Information Entropy

驾驶员疲劳检测的研究有助于降低交通事故的发生。本文提出一种基于面部倒立摆模型与信息熵的疲劳检测方法,首先,采用PFLD(Practical Facial Landmark Detector)模型检测驾驶员面部关键点坐标,并估计用于表示头部姿态信息的Pitch、Yaw以及Roll角度值;然后,以关键点坐标为输入建立面部倒立摆模型,计算模型中连杆系统在驾驶员驾驶过程中的动能和势能;之后,以倒立摆模型的动能、势能以及头部姿态数据作为衡量驾驶员疲劳状态变化的指示特征,基于滑动窗口计算各疲劳特征的信息熵值;通过CNN(Convolutional Neural Networks)处理疲劳特征信息熵值,建立信息熵与驾驶员疲劳状态之间的联系;最后,将CNN在各个时间点上的输出作为LSTM(Long Hort-term Memory)网络的输入特征,通过CNN-LSTM模型实现疲劳特征信息熵的分类预测。实验结果表明,所提模型的预测结果达到95.04%,验证了本文方法的有效性。

Driver fatigue detection is helpful for reducing the traffic accidents related to driver fatigue.This paper proposes a fatigue detection method based on the facial inverted pendulum model and information entropy.First,the Practical Facial Landmark Detector(PFLD)model is used to detect the coordinates of key points on the driver's face and estimate the Pitch,Yaw and Roll angles used to represent the head posture.Then,a facial inverted pendulum model is developed with key point coordinates as input.The kinetic energy and potential energy of the linkage system in the model are calculated during the driver's driving process.The kinetic energy,potential energy and head attitude data of the inverted pendulum model are used as the indicators of driver fatigue state changes,and the information entropy of each fatigue feature is calculated based on the sliding window.The information entropy values are concatenated on the temporal axis by a Convolutional Neural Networks(CNN)to extract the effect of the fatigue state on the information entropy over time.Then,the output of the CNN at each time point is used as the input feature of the Long hort-term memory(LSTM),and the fatigue feature information entropy is used to classify and predict the CNN-LSTM model.The experimental results show that the predicted result of this method reaches 95.04%,which indicates the effectiveness of the proposed method.