基于U-Net网络与卡尔曼滤波的瞳孔检测跟踪算法  

Pupil detection and tracking algorithm based on U⁃Net network and Kalman filter

作　　者：张国静 王桂祥 ZHANG Guojing;WANG Guixiang(School of Arts and Sciences,Hangzhou Dianzi University Information Engineering College,Hangzhou 311305,China)

机构地区：[1]杭州电子科技大学信息工程学院文理学院,浙江杭州311305

出　　处：《现代电子技术》2025年第9期137-142,共6页Modern Electronics Technique

基　　金：杭州电子科技大学信息工程学院科研培育基金项目:基于深度学习的视频眼震采集系统的应用研究(KYP0323015)。

摘　　要：眼动跟踪是人机交互(HCI)及其应用的重要技术之一,越来越受到人们的重视。然而,在实际情况下往往受到大量非高斯噪声的影响,如不可控的光照、人眼的遮挡及连续的人眼移动等,这会导致瞳孔检测实时性和准确性的下降。因此,文中设计了一种基于U-Net语义分割网络的瞳孔检测方法。首先,利用该方法对瞳孔区域进行分割;然后对分割的瞳孔区域处理,确定其质心位置,达到瞳孔中心定位的目的;最后,又提出了一种改进的卡尔曼滤波器的稳态增益,通过在卡尔曼增益上引入分数阶反馈环路来实现,并利用改进的卡尔曼滤波器对瞳孔位置进行跟踪,消除非高斯噪声,可以大大提高瞳孔在线稳定检测的准确性。实验结果表明,所提方法能够实时跟踪人眼,具有较高的精确度和鲁棒性,且最佳均方根误差(RMSE)可达到0.78。Eye tracking is one of the important technologies in human-computer interaction(HCI)and its applications,which has been gaining increasing attention.However,in practice,it is often affected by a large amount of non-Gaussian noise,such as uncontrollable lighting,human eye being occluded,and continuous eye movements,which leads to a decrease in real-time performance and accuracy of pupil detection.Therefore,a pupil detection method based on the U-Net semantic segmentation network is proposed.In this method,the pupil region is segmented,and then the segmented pupil region is processed to determine its centroid position,so as to achieve the purpose of pupil center localization.Finally,a steady-state gain for improved Kalman filter is proposed by incorporating a fractional order feedback loop into the Kalman gain.The improved Kalman filter is used to track the pupil position and eliminate non-Gaussian noise,which can greatly improve the accuracy of online stable detection of pupils.Experimental results demonstrate that the method can track human eyes in real time with high precision and strong robustness,and its optimal root mean squared error(RMSE)can reach 0.78.

关 键 词：眼动跟踪 人机交互 非高斯噪声 U-Net网络 语义分割 瞳孔检测 分数阶 卡尔曼滤波 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP394.1[电子电信—信息与通信工程]