当前位置: X-MOL 学术Mach. Vis. Appl. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Head and camera rotation invariant eye tracking algorithm based on segmented group method of data handling
Machine Vision and Applications ( IF 3.3 ) Pub Date : 2020-08-31 , DOI: 10.1007/s00138-020-01112-2
Mohammad Reza Mohebbian , Javad Rasti

Eye-gaze tracking through camera is commonly used in a number of areas, such as computer user interface systems, sports science, psychology, and biometrics. The robustness of the head and camera rotation tracking algorithm has been a critical problem in recent years. In this paper, Haar-like features and a modified version of the group method of data handling, as well as segmented regression, are used together to find the base points of the eyes in a facial image. Then, a geometric transformation is applied to detect precise eye-gaze direction. The proposed algorithm is tested on GI4E and Columbia Gaze datasets and compared to other algorithms. The results show adequate accuracy, especially when the head/camera is rotated.

中文翻译:

基于分段分组数据处理方法的头部和相机旋转不变眼跟踪算法

通过摄像机进行的视线跟踪通常用于许多领域,例如计算机用户界面系统,体育科学,心理学和生物识别技术。头部和摄像机旋转跟踪算法的鲁棒性是近年来的关键问题。在本文中,将类似Haar的特征和数据处理分组方法的改进版本以及分段回归一起用于查找面部图像中眼睛的基点。然后,进行几何变换以检测精确的视线方向。该算法在GI4E和Columbia Gaze数据集上进行了测试,并与其他算法进行了比较。结果显示出足够的精度,尤其是在旋转头部/摄像机时。
更新日期:2020-08-31
down
wechat
bug