摘要： 针对疲劳驾驶检测中头部运动多变情况下的人眼跟踪问题，提出一种改进的Mean-shift算法，通过对像素值分布特征及目标梯度方向的密度分布特征的迭代公式进行交替迭代运算，实现对运动目标的平移跟踪和旋转跟踪，以改善头部深度旋转运动下的人眼跟踪性能。实验结果表明，该方法在驾驶员头部运动姿态多变的情况下，能快速有效地对人眼进行跟踪和定位。

关键词: 人眼跟踪, 疲劳驾驶, 改进的Mean-shift算法, 梯度方向, 多姿态

Abstract: Aiming at the problem of eye tracking under the head movement varied in the process of drowsy driving detection, this paper proposes an improved Mean-shift algorithm. By using pixel gray value distribution feature and the density distribution feature of target gradient direction, it performs an alternate iterated operation for the two characteristics iteration formula, and carries the rotation and translation movement tracking better for the moving target, so the method can improve the eye tracking under the head rotation in depth. Experimental results show that the method can effectively track and position the human eye under multi-pose head movement.

Key words: human eyes tracking, drowsy driving, improved Mean-shift algorithm, gradient direction, multi-pose

中图分类号: 

• TP301.6