基于SVM的疲劳驾驶人脸图像特征检测算法  

Research on Feature Detection Algorithm of Fatigue Driving Face Image Based on SVM

作　　者：刘梦佳[1] 赵建国[1] LIU Mengjia;ZHAO Jianguo(College of Mechanical and Power Engineering,Zhengzhou University,Zhengzhou 450052,CHN)

机构地区：[1]郑州大学机械与动力工程学院,郑州450052

出　　处：《半导体光电》2020年第2期278-282,共5页Semiconductor Optoelectronics

摘　　要：针对传统图像识别算法对疲劳驾驶检测精度差、准确率低的缺陷,提出了一种利用人脸图像特征提取的疲劳驾驶检测方法。首先将实时采集到的车辆驾驶员面部图像进行预处理,借助Dlib检测出图像中的人脸区域并进行人脸图像特征点的标注,然后使用基于眼睛纵横比(Eye Aspect Ratio,EAR)的方法进行图像中人眼疲劳特征的识别,基于嘴唇纵横比(Mouth Aspect Ratio,MAR)的方法进行图像中嘴部疲劳特征的识别,最后利用支持向量机(SVM)的方法将两种特征融合起来进行疲劳驾驶检测。实验表明,该方法可以准确地定位出特征点,疲劳检测的识别率达84.29%,可以有效地识别出疲劳状态。Aiming at the defects of the traditional image recognition algorithm,which has poor precision and low accuracy in fatigue driving detection,an effective evaluation method of fatigue driving detection using face image data is proposed.Through real-time acquisition of the vehicle driver\s face image,the face image was preprocessed first,the face area in the image was detected with the help of Dlib and the feature points of the face image were marked,then the eyeaspect-ratio(EAR)-based method was used to recognize the fatigue feature of the human eyes in the image,the mouth-aspect-ratio(MAR)-based method was used to recognize the fatigue feature of the mouth in the image,and finally the support vector machine(SVM)was applied to combine the two features for fatigue driving detection.Experimental results show that the method can locate the feature points accurately,and the recognition rate of fatigue detection reaches 84.29%,which can effectively identify the fatigue state.

关 键 词：疲劳驾驶 人脸图像检测 人脸特征点定位 眼睛纵横比 支持向量机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]