CAJ | 학술논문

疲劳驾驶是导致交通事故的主要原因之一，因此，以人眼检测为核心的疲劳驾驶检测受到了普遍地关注。目前传统的疲劳驾驶检测方法采用“人脸?人眼”模型，即先定位人脸，随后再进行人眼检测。在AdaBoost算法的基础上，使用变模板匹配方法进行人眼检测，并改进了“人脸?人眼模型”，采用多特征的分类器来有选择的跳过人脸检测步骤，从而快速准确地定位人眼。在DM642芯片上进行实验，人眼识别率达到了90%以上，且平均每秒能处理40帧图片，证明了该方法的高效性和实用性。
피로가사시도치교통사고적주요원인지일，인차，이인안검측위핵심적피로가사검측수도료보편지관주。목전전통적피로가사검측방법채용“인검?인안”모형，즉선정위인검，수후재진행인안검측。재AdaBoost산법적기출상，사용변모판필배방법진행인안검측，병개진료“인검?인안모형”，채용다특정적분류기래유선택적도과인검검측보취，종이쾌속준학지정위인안。재DM642심편상진행실험，인안식별솔체도료90%이상，차평균매초능처리40정도편，증명료해방법적고효성화실용성。
Since fatigue driving is one of the main causes of traffic accidents,the fatigue drivingdetection based on eye de?tection has been concerned generally. At present,“human face?eye”model is mainly used in the traditional fatigue driving de?tection,in which AdaBoost algorithm is used to locate human face,and then the human eye detection is performed. In this pa?per,a template?changed matching method is used to make human eye detection based on the AdaBoost algorithm,and im?prove the“human face?eye”model. A multi?feature classifier is adopted to skip the face detection step selectively,so as to lo?cate the human eyes more quickly and accurately. An experiment was carried out on DM642 chip. The result shows that the hu?man eye recognition rate is more than 90%,and the average processing speed is 40 images per second,which has proved the efficiency and practicality of the method.