全封闭式高端塔台模拟机中投影显示图像高质量重现方法研究

Although the research on the visual system of tower simulator has been going on for a long time，degradation of projection display image caused by various reflection effects is still a big problem to be solved. Unfortunately, however, it is often ignored and has a serious impact on its practical application. This project is committed to the study of high quality image reconstruction algorithm, evaluation criteria and algorithm applicability. Firstly, analyzing the complex reflection phenomenon in large scale high immersion environment based on the physical display process of projection image. After that, a reflective model is built to describe the transmission process of light energy in the display system. And on this basis, we design a reflection compensation algorithm to achieve the reproduction of high quality projection display image. Secondly, an image quality evaluation criteria according with human visual perception is built to varify the effectiveness of the compensation algorithm. Lastly, an evaluation of the applicability of reflection compensation method is carried out by studying the effect of different factors on the projection display system. The research content of this project is not only different from the traditional image correction problem in the tower simulator, but also an important extension to it. It is a new attempt, with strong innovation and challenge. Meanwhile, it also has high practical value.

尽管对塔台模拟机视景系统的研究由来已久，但一大亟待解决的问题—全封闭式高端塔台模拟机中由各种反射现象引起的投影显示图像降质退化，常常被忽略，严重影响了其实际应用。课题致力于复杂空间中投影显示图像高质量重现算法、评价标准及算法适用性的研究：首先，结合投影图像物理显示过程，分析大规模高沉浸环境中复杂反射现象，建立描述投影显示系统中光能传输过程的反射模型，并在此基础上设计反射补偿算法，实现投影显示图像高质量重现；其次，建立满足人眼视觉感知的图像质量评价标准，验证补偿算法的有效性；最后，研究不同环境因素下投影显示图像重现效果，评估反射补偿方法的适用性。本项目的研究内容既区别于塔台模拟机中传统图像校正问题，又是对其的重要扩展，是一种全新的尝试，具有很强的创新性和挑战性，同时也具有很高的实用价值。

复杂投影显示系统中由各种反射现象引起的投影显示图像降质退化，常常被忽略，严重影响了其实际应用。课题致力于复杂空间中投影显示图像高质量重现算法、评价标准及算法适用性的研究，主要研究内容包括：（1）研究投影图像物理显示过程，搭建投影系统，采集相关数据集；（2）研究基于深度学习的投影显示图像高质量重现方法；（3）满足人眼视觉感知的图像质量评价方法；（4）研究投影补偿网络的适用性。取得的重要成果包括：（1）建立了规模化的数据集，用于训练和测试；（2）提出并实现了基于冗余反射光学习的曲面投影系统光污染补偿、基于注意力机制的曲面投影系统光污染高质量补偿、基于卷积神经网络的单投影图像反射补偿模型和基于多投影图像反射补偿模型优化等多种方法，并运用多种客观和主观评价方法验证了算法的有效性；（3）开展了算法适用性研究，设计并实现了投影系统互反射模拟软件。项目提出的创新性方法不仅以多篇学术论文形式在本领域重要期及会议上发表，亦在实际系统中进行了有效性验证。