视线追踪系统中注视点估计算法研究

针对非接触式视线跟踪系统中注视点估计算法鲁棒性差的问题,提出一种基于角度映射的注视点估计算法。首先,根据人眼特性及视觉成像原理,瞳孔中心相对于角膜反射光斑的位置与注视点相对于红外光源的位置之间具有一定的角度映射关系,据此估计出注视点近似位置。然后,在眼球模型结构的基础上分析了眼球角膜曲面对注视点造成的偏差,通过弧长对误差进行补偿。最后,利用非线性多项式模型对眼球视轴和光轴之间的偏差进行拟合,得到最终的视线落点。实验证明,该系统具有较高的精度和自由度,在水平和垂直方向上最大误差均小于1 cm。

The Research on Gaze Estimation for Eye Tracking System

Aiming at the problem of poor robustness of gaze point estimation algorithm of un-contact eye tracking system, one novel gaze point estimation algorithm was proposed in this paper based on angle mapping. First, according to human eyes characteristics and visual imaging theory, the gaze points are estimated with the angle mapping relationship between the correspond positions of pupil center and corneal reflection points and the correspond positions of gaze point and infrared lights. Next is the analysis of gaze point error of ocular cornea surface based on eyeball structure and compensation for it with arc length. Finally the nonlinear polynomial model was also used to fitting the error of eye visual axis and optical axis. In our experiment, the deviation of this algorithm is less than 1 centimeter both in horizontal and vertical directions. The result showed that the system we proposed has higher precision and degree of freedom.