基于SVM的驾驶员面部朝向检测  

Driver′s face orientation recognition based on SVM

作　　者：张明恒[1,2] 郑雨[2] 赵一兵[1,2] 郭烈[1,2] 

机构地区：[1]大连理工大学工业装备结构分析国家重点实验室,辽宁大连116024 [2]大连理工大学汽车工程学院,辽宁大连116024

出　　处：《大连理工大学学报》2013年第6期879-884,共6页Journal of Dalian University of Technology

基　　金：教育部人文社会科学研究青年基金资助项目(12YJCZH280);国家自然科学基金青年基金资助项目(51205038)

摘　　要：驾驶员精神状态监测对于交通事故控制具有重要作用.基于支持向量机(SVM)建立信息融合的面部朝向检测模型,在驾驶员面部图像视觉特征(眼睛、嘴部及面部轮廓)检测基础上,通过分析头部运动过程中各特征的空间位置变化规律,利用SVM对驾驶员面部朝向进行多信息融合估计.结果表明,特征检测方法可实现面部各视觉特征的可靠检测,分析得出的各特征运动变化规律可用于对面部朝向的有效估计,从而为驾驶员驾车精神状态监测提供技术支持.Driver＇s spiritual state monitoring is important for reducing traffic accidents. Based on support vector machine （SVM）, a face orientation recognition model with information fusion is presented. Firstly, based on the detected multiple visual features such as eyes, mouth and facial contours, the features＇ position variations during the head moving are analyzed. Then, the face orientation is estimated with these characteristics through SVM method. The experimental results show that the proposed method for features detection has good reliability and the face orientation can be inferred according to the relevant characteristics. It is clear that the proposed method has strong potential effectiveness to be applied to the driver＇s spiritual state monitoring.

关 键 词：机器视觉 面部朝向 支持向量机(SVM) 安全辅助驾驶 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]