• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2006, Vol. 28 ›› Issue (6): 52-54.

• 论文 • 上一篇    下一篇

一种基于视觉的PERCLOS特征提取方法

王磊 吴晓娟 巴本冬 董文会   

  • 出版日期:2006-06-01 发布日期:2010-05-20

  • Online:2006-06-01 Published:2010-05-20

摘要:

PERCLOS是有效地检测驾驶员瞌睡的特征.本文在前人研究的基础上,提出了一种快速、有效的计算眼睛睁开程度的检测算法.针对车内光线会有变化的特点,对肤色滤波之后的人脸灰度图像通过累计直方图阈值法进行二值化,很好地将面部五官从肤色中分离出来.最后利用连通域搜索算法得到双眼睁开的大小,并最终提取到PERCLOS.

关键词: 驾驶疲劳/瞌睡 PERCLOS 肤色分割 累计直方图

Abstract:

PERCLOS is an effective cue for driver fatigue detection. In this paper, the authors present a fast and effective method to calculate the open degree  of a driver's eyes. Since the light inside the car is always changing, after skin segmentation, accumulative histogram threshold is used to binarize th   he face image, and then the eyes are separated from the skin color perfectly. A 4-connected component algorithm is used to calculate the area of two eye s, and then the open degree of eyes is got. Finally PERCLOS is obtained.

Key words: driver fatigue/drossiness, PERCLOS, skin segmentation;accumulative histogram