基于ASM的驾驶员面部疲劳状态识别方法  

Method of driver's face fatigue state recognition based on ASM

作　　者：闫河[1] 杨晓龙 张杨 董莺艳 王鹏 YAN He;YANG Xiao-long;ZHANG Yang;DONG Ying-yan;WANG Peng(College of Computer Science and Technology,Chongqing University of Technology,Chongqing 400054,China)

机构地区：[1]重庆理工大学计算机科学与工程学院,重庆400054

出　　处：《计算机工程与设计》2018年第10期3240-3245,共6页Computer Engineering and Design

基　　金：国家自然科学基金面上基金项目(61173184)

摘　　要：针对驾驶员疲劳状态识别的需求,提出一种基于主动形状模型的驾驶员面部疲劳状态识别方法。利用Helen数据集训练得到具有194个特征点的人脸主动形状模型,结合haar级联检测得到精确的人脸和眼睛轮廓区域;通过光流法实现对人脸及上下眼睑特征点的有效跟踪,获得每帧图像的头部姿态及眼睛睁开度,实现对眼睛疲劳状态识别;结合正负图像训练得到的支持向量机判定驾驶员是否属于重度疲劳情况下的低头行为。实验结果表明,该方法可以在毫秒级别检测到眼睛状态和头部姿态,准确率达到92.5%,能有效识别驾驶员面部疲劳状态。Aiming at the requirement of driver’s fatigue state recognition,a method of driver’s face fatigue state recognition based on active shape model(ASM)was proposed.The face ASM with 194 feature points was obtained by Helen dataset trai-ning,and the accurate face and binocular contour area were obtained by combining haar cascade detection.Through the optical flow algorithm,the face and the upper and lower eyelids feature points were tracked effectively,the head posture and the eyes open rate were obtained from each frame image,and the eyes fatigue states were recognized.Support vector machine(SVM)trained by positive and negative image was combined to determine whether the driver belonged to severe fatigue in the case of head-dipping.Experimental results indicate the proposed method can detect the state of the eyes and the head posture in the millisecond level,and the accuracy rate reaches 92.5%,which can effectively recognize the driver’s face fatigue state.

关 键 词：面部疲劳状态识别 主动形状模型 光流法 眼睛睁开度 支持向量机 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]