【摘要】 在飞行座舱多屏交互场景中，飞行员往往需快速切换视线以完成态势感知与精确操作，高精度的跨屏视线追踪成为提升人机交互效能的关键技术挑战。然而，现有基于单视角的视线估计方法在头部大幅运动时易受图像遮挡影响，导致精度显著下降，难以满足复杂交互场景下的高精度操作需求。本文提出了一种面向飞行座舱的跨屏视线估计方法，结合双视角图像采集和深度学习技术，实现了跨屏环境中的高精度视线估计。利用双屏双相机系统获取的双视角人脸图像数据，有效克服了传统单视角视线估计中的视角盲区问题。针对飞行座舱中频繁的视线切换需求，构建了一套跨屏视线估计系统，能够高效进行注视屏幕选择与注视点估计。试验结果表明，相较于单视角传统方法，本方法的视线估计平均精度提升33.54%，系统鲁棒性与实时性得到同步优化，为飞行座舱多屏交互提供了一种可验证的解决方案。更多还原

【Abstract】 In multi-screen interaction scenarios in flight cockpits, pilots often need to switch their gaze quickly to achieve situation awareness and precise operations. High-precision cross-screen gaze tracking has become a key technical challenge for improving human-computer interaction efficiency. However, existing single-view-based gaze estimation methods are susceptible to image occlusion when the head moves significantly, leading to a significant drop in accuracy, which makes them difficult to meet the high-precision operation requirements in complex interaction scenarios. In this paper, we propose a cross-screen gaze estimation method for flight cockpits, which combines dualview image acquisition and deep learning technology to achieve high-precision gaze estimation in cross-screen environments. Using dual-view facial image data acquired by a dual-screen and dual-camera system, the method effectively overcomes the view blind spot issue in traditional single-view gaze estimation. To address the frequent gaze switching requirements in flight cockpits, a cross-screen gaze estimation system is constructed, which can efficiently perform gaze screen selection and gaze point estimation. Experimental results show that compared with traditional single-view methods, the average accuracy of gaze estimation of our method has improved by 33.54%, and the system robustness and real-time performance have been optimized simultaneously, providing a verifiable solution for multi-screen interaction in flight cockpits.更多还原