基于人眼特征信息的驾驶人眼视线估计

提出了一种新的基于人眼特征信息的驾驶人眼视线估计的算法,且研究使用普通的摄像头。采用Harris角点检测算法对驾驶人上、下眼睑外轮廓进行角点检测,拟合人眼轮廓曲线;再对人眼区域图像进行色彩空间转换提取灰阶值分量,图像亚像素下进行Hough边缘检测,设定相应的虹膜边缘曲率阈值,准确识别人眼虹膜边缘信息。算法结合虹膜和人眼轮廓信息对驾驶人眼视线进行估计。将提出的算法应用到实车试验中,采用facelab5眼动仪对驾驶人视线估计角度结果进行验证。试验结果表明,所采用的人眼视线角度估计算法在实际的驾驶环境中准确率较高。

Gaze Estimation of Driver"s Eye Based on Eye"s Characteristic Information

A newSalgorithm based on eye"s characteristicSinformation for the gaze estimation of driver"s eye is proposed and using the common camera. First of all, the eye model is discussed and the new model is put forward according to the samples" analysis. As the components of eye model, the upper eyelid, (inner and outer) eye corners and iris are to be detected. Using HarrisScorner algorithm for detection of the driver"sSupper eyelid and lower eyelidScontourScorner, and fitting the human eye"s contour curve; Then, converting image color space and extracting component of gray-scale value of the human eye"s image, and adopting Hough edge detection algorithm in sub-pixel image. The corresponding threshold of the curvature of the iris"s edge is set, and the human iris edge information is identified accurately. The algorithm for the sake of the gaze estimation of driver"s eye which combined with the iris and the human eye contour information. According to our defined gaze model and the analysis of eye features, the gaze of human eye can be estimated. The proposed algorithm is applied to the real vehicle experiment, In order to verify the driver"s gaze estimation, and facelab5 eye movement instrument is used for the verification. The markers are detected by computer vision method, and the relationship between scene image and computer screen is constructed with point correspondences in two views. The point of gaze in the scene image is translated to computer screen coordinate. The experimental results show that the proposed algorithm for the gaze estimation of driver"s eye has a high accuracy in the actual driving conditions and the low - resolution images.