摘要： 提出一种快速实时的组合型人眼识别方法，该方法由人眼检测和人眼跟踪2个部分组成。在检测过程中，采用级联AdaBoost分类器检测出人眼位置；在跟踪过程中，先利用卡尔曼滤波器追踪瞳孔，若瞳孔追踪失败，则使用平均位移追踪。该方法已在DM6446嵌入式系统中实现，实验结果证明该方法能快速识别人眼的位置。

关键词: 级联AdaBoost分类器, 卡尔曼滤波器, 平均位移跟踪器, 嵌入式系统

Abstract: This paper proposes a real-time robust integrated method for eye recognition which consists of two parts: eye detection and eye tracking. Eye detection is accomplished by cascade AdaBoost classifiers to find out the location of eyes. Eye tracking is a conventional Kalman filtering tracker based on the bright pupil. In case Kalman eye tracker fails, eye tracking based on the mean shift tracker to continue tracking the eyes. This method is applied to the embedded system DM6446. Experimental results show that this method can quickly recognize eyes position.

Key words: cascade AdaBoost classifier, Kalman filter, mean shift tracker, embedded system

中图分类号: 

• TP18