视线跟踪系统中眼睛睁闭检测算法研究

Eye state detection algorithm of eye tracking system

眼睛睁闭检测在视线跟踪系统中具有重要意义。为提高检测的准确性,提出了一种基于Harris算子的检测方法。该方法首先利用Harris算子计算图像的角点量,然后搜索图像中角点量最大的位置,以该位置为中心,设置一个区域,统计该区域内角点量总和占整幅图像角点量总和的比例,通过将该比例与阈值相比较来确定眼睛的状态。实验结果表明,本算法具有较高的鲁棒性和准确率,并能满足实时性要求。

Eye state detection is important in eye tracking system. In order to improve the correctness, a Harris operator based new algorithm is proposed. First, Harris operator is used to calculate the cornerness of eye image. Then, the position of maximum cornerness is set as center of a region. The proportion of total cornerness in the region and all images are used to detect eye state. Experimental results definitely verify the robustness and accuracy of the new algorithm. Especially, this new algorithm is suitable for real-time applications.