基于角膜折射与滑移补偿的凝视估计算法

打开文本图片集

中图分类号：TN27；TP391 文献标识码：A doi：10.37188/CJLCD.2025-0213 CSTR：32172.14.CJLCD.2025-0213

Gaze estimation algorithm based on corneal refraction and slip compensation

XIA Yuchen 1，2 ，WU Shiqian 1，2 *，WAN Zhonghua³，YIN Zihang 1，2 ， YANG Mingxuan 1，2

（1. School of Electronics and Information， Wuhan University ofScience and Technology， Wuhan 430081，China; 2. Institute of Aduanced Displays and Imaging， Henan Academy of Sciences， Zhengzhou 45OO46，China; 3.School ofMechanical Engineering，North University of China， Taiyuan O3oO5l，China）

Abstract： To improve the gaze estimation accuracy and slippage-robust in head-mounted eye tracking devices，a gaze estimation algorithm combining corneal refraction and lower eyelid slip compensation was proposed. First，an initial eye model was constructed based on the pupil back-projection method to obtain the virtual pupil position. The pupil positioning error caused by corneal refraction was corrected using the Snell refraction law to obtain the true pupil position. The optimized eye model was then iteratively obtained. Subsequently，vertical device slip was monitored in real time through lower eyelid region corner detection and density analysis to compensate for gaze estimation. Experimental results show that the proposed method can compensate for the slip eect when the device is worn，with an average gaze estimation error of 2.882∘ ，which is superior to two-dimensional and three-dimensional methods. The proposed gaze estimation algorithm combining corneal refraction and lower eyelid slip compensation can effectively improve gaze estimation accuracy and significantly enhance the system's robustness to device slip in real-world wearable scenarios，verifying the reliability and practicality of the proposed method.

Key words： head-mounted display；gaze estimation；corneal refraction；slippage

1引言

近年来，随着智能眼镜的快速兴起，头戴式眼动追踪技术在虚拟现实[1]、增强现实[2-3]及人机交互[4等领域受到广泛关注。（剩余13689字）