驾驶人眼睛定位及特征提取算法研究

Research on Driver's Eyes Positioning and Feature Extraction Algorithm

为进一步提高驾驶人眼睛状态检测对环境、头部姿态的适应性和鲁棒性,提出一种基于主动红外成像和双目立体视觉技术的眼睛定位和特征提取的方法。对采集的3名驾驶人脸部红外图像平滑处理算法、阈值分割算法进行了对比研究,得出适于驾驶人眼睛状态检测的实时算法。在驾驶人脸部区域定位的基础上,采用连通区域标记法对眼睛区域进行精确定位,通过最小二乘椭圆拟合算法获取瞳孔位置,使用Harris角点检测获取普尔钦光斑位置。研究表明:该方法能有效对驾驶人眼睛进行定位,并准确提取相关特征信息。

In order to further improve the adaptability and robustness of drivers eye state detection for different head poses and environments, an eyes positioning and feature extraction algorithm based on active infrared imaging and binocular stereo vision technology was presented. Through the comparison for different infrared image smoothing algorithm and threshold segmentation algorithm for three drivers face infrared image, pointed out the suitable algorithm. On the basis of the driver face localization, the eye area can get the precision location by connected components labeling algorithm, and the pupil position can obtain by based on Least Square method. The Purkinje image position was used Harris corner detection algorithm. This study shows that presented algorithms is effective to get driver＇s eyes positioning and to extract feature.