疲劳驾驶识别算法与技术  

The Recognition Algorithms of Fatigued Driver and Its Techniques

作　　者：邱桢[1] 罗露瑾[1] 应俊豪[1] 曼苏乐[1] 张秀彬[1] 

机构地区：[1]上海交通大学电子信息与电气工程学院

出　　处：《微型电脑应用》2012年第12期52-56,共5页Microcomputer Applications

摘　　要：针对驾驶员长途驾驶汽车时,会因难以克服的生理疲劳时常发生交通事故的残酷现状,提出一种疲劳驾驶的识别算法,能够有效阻止驾驶员在疲劳状态下,持续行驶的不良行为,以避免事故的发生。该算法通过对驾驶员脸部图像的采集与特征空间建立,采用与疲劳训练样本的类比计算,实时确认当前驾驶状态的类别属性。依据驾驶状态的类别属性即可确定是否需要对驾驶员实施干预。系统识别运算周期小于50ms,识别准确率高于96%。When a driver is driving a car for long-distance, he will be difficult to overcome physical fatigue, so the brutal traffic accidents often occur. In this paper, a fatigued driver recognition algorithm can effectively prevent the bad behavior of driver fatigue state to avoid accidents. Through the collection on the driver＇s face image and the establishment of its feature space, using the analo- gy of the fatigue training samples and it, the algorithm can confirm the category attributes of the current driving status in real time. According to the category attribute of the driving state, the system can determine whether interventions need to be implemented on the driver. The recognition period is less than 50ms; the recognition accuracy is greater than 96%.

关 键 词：疲劳驾驶 特征空间 状态识别 识别周期 识别准确率 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]