基于PERCLOS的机动车驾驶员驾驶疲劳的识别算法.docVIP

中国农业大学学报　　2002, 7(2): 104～ 109 　 Journal of China A griculturalU niversity 　 基于 PERCLOS 的机动车驾驶员驾驶疲劳的识别算法 郑　培 ①　宋正河　周一鸣 (中国农业大学车辆与交通工程学院) 摘　要　阐述了 PERCLO S 测评驾驶疲劳的机理, 对测试驾驶疲劳的几种方法的测试精度进行了比较, 认为 P80 是最好的。应用二维高斯模型、灰度直方图、灰度模式匹配等图像分析和识别手段定位和追踪眼睛睁开、 闭合的变化过程, 统计出眼睛闭合时间。利用概率和数理统计方法给出了一种行之有效的、基于 PERCLO S 的 机动车驾驶员疲劳程度测评的新算法。 关键词　PERCLO S; 驾驶疲劳; 图像识别; 灰度模式匹配; 测评算法 中图分类号　U 491131; TP 391141 PERCLOS-Based Recogn ition A lgor ithm s of M otor D r iver Fatigue Zheng Pei, Song Zhenghe, Zhou Yim ing (College of V ehicle and T raffic Engineering, China A gricultural U niversity, Beijing 100083, China) Abstract　A brief review of detecting and evaluating technique of mo to r driver fatigue is p rovided. A cco rding to the comparison betw een several commonly used detecting and evaluating techniques of mo to r driver fatigue, the P80 of PERCLO S (the p ropo rtion of tim e the eyes w ere clo sed at least 80 percent) is mo st significantly co rrelated w ith driver fatigue. A convincing case has been m ade that slow eyelid droop PERCLO S has the best po tential to detect fatigue. A feasible m ethod based on the m athem aticalmodel and symm etry analysis is p resented to detect and locate the driver’s eye in an im age and the gray scale mode m atching technique is used to determ ine the driver fatigue degree. A n examp le fo r detecting the eyelid clo sure over the pup il over tim e is given. The com binations of m athem atical model and symm etry analysis increases the robustness of the perfo rm ance w hile the target is defo rm ed. The schem e is suitable fo r hum an face location in intelligent hum an2m achine interface and p rovides a groundwo rk to p ractical app lication of face recognition techniques. The PERCLO S2based recognition algo rithm s w ill be considered as a mo st effective m ethod used to accomp lish real2tim e m easure of alertness fo r in2vehicle, drow siness2detection system s. Key words 　 PERCLO S; driver fatigue; im age recognition; gray scale mode m atching; detec