摘要
提出一种基于眼睛状态识别的驾驶员疲劳状态监测算法。在首帧定位人眼并在眼睛区域内提取多个具有互补性的特征,使用粒子滤波算法进行直接跟踪;在后续图像帧中利用CAMShift算法对初始检测到的人脸区域进行实时跟踪,在人脸区域中同步检测眼睛,检测结果用于粒子滤波器的后验确认和修正。在跟踪眼睛的同时检测每帧中的眼睛状态,通过连续帧中的眼睛状态判断出驾驶员精神状态。实验表明:该文算法对人脸旋转和尺度变化、摄像机抖动,以及遮挡的影响均具有较高的鲁棒性。
A driver fatigue monitoring algorithm based on eye state recognition is proposed here.In the first frame,eye windows are detected by face detection followed with an accurate eye localization algorithm.For eye tracking,several complementary features are extracted from the detected eye window and fused in probability form in particle filter tracking framework.The CAMShift algorithm is also exploited to track the face region,and the eye detection algorithm is still used to detect reliable eye pairs,so as to verify and correct the tracking result of the particle filter.The spirit status of drivers can be detected by analyzing eye states in continuous image frames.Experiments show that the proposed algorithm can well handle pose changes,camera jitter and partial occlusion.
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2010年第4期448-453,共6页
Journal of Nanjing University of Science and Technology
基金
国家自然科学基金(60632050)
中国博士后基金
江苏省科技计划高技术研究项目(BG2005008)
