摘要
为了减少因疲劳驾驶所引发的交通事故的发生,人眼定位在疲劳驾驶检测技术中起着重要的作用.对人眼定位的过程中,采用多摄像头获取图像并对这些图像进行筛选,将检测到的人脸面积最大,将能够完整检测到双眼状态的图像做为最佳输入图像,采用Adaboost(迭代级联分类器)算法对其进行人脸定位以减小检测区域,再对其进行灰度投影将检测范围缩小在眉眼区域,然后进行一种新的LMS(最小均方误差)模板匹配,精确定位眼睛区域.在人脸成功定位的基础上,该算法经过二次眼睛定位,较传统的模板匹配方法,平均模板匹配检测时间提高到了30.5ms,准确率提高到了97%.实验表明:将灰度投影法和改进的LMS(最小均方误差)模板匹配两种人眼定位的方法相结合来进行人眼的定位,与传统的模板匹配相比,提高了检测的效率和检测的准确率,使得疲劳驾驶检测系统能更准确地进行实时检测,该方法能适用于疲劳驾驶检测等需要快速人眼定位的场合.
Eye location plays an important role in the fatigue driving detection to reduce the traffic accident. During the process of the eye location, the images were got by multi-camera and chosen the best to input. At first, the face was located by Adaboost method. Second, the area was narrowed in facial features by the gray projection and applied a new LMS template matching. Then the area of eye was located accurately. Based on the face location accurately, the efficiency of this algorithm increased to 30.5ms and the accuracy of the eye location was improved to 97% through the twice eye location The experiment shows that the efficiency and the accuracy are improved greatly after the combination of the gray projection and the LMS template matching. So it is suitable for the fatigue driving detection which needs rapid eye location.
出处
《武汉工程大学学报》
CAS
2013年第6期67-72,共6页
Journal of Wuhan Institute of Technology
基金
智能机器人湖北省重点实验室提供的基金资助
