增强现实中室外场景光照一致性研究

项目名称： 增强现实中室外场景光照一致性研究

项目编号： No.61303089

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 张锐

作者单位： 山东财经大学

项目金额： 23万元

中文摘要： 增强现实技术将虚拟物体叠加到真实场景的视频图像中，而光照一致性是确保虚拟物体表面光照效果真实感、使虚实物体无缝融合的关键技术。由于室外场景的几何复杂性和光照复杂性，其光照估计是当前未解决的难点和热点问题。本项目以此为切入点，在场景三维模型未知的情况下，围绕雾霾天气情况下室外场景视频图像的光照估计问题进行研究。分析研究室外场景光照估计所采用的模型，构建准确的太阳光、天空光及地面反射光模型；将真实场景中雾霾的影响引入到光照估计的框架中，研究雾霾对视频图像质量的影响，构建雾天的室外场景光照明模型；根据雾天视频图像的本质特性增加约束条件，建立能量方程，求解光照参数，并剔除运动物体等因素的影响；实现一个包含这些关键技术的虚实融合原型系统。本项目的研究成果将为解决室外场景的光照一致性问题提供新思路、新理论和新方法，为光照估计在物体识别、视频跟踪、视频分割等领域的有效应用提供简单实用、稳定有效的新技术。

中文关键词： 光照估计；雾天室外图像；色度一致性；雾天整体光照明模型；虚实融合

英文摘要： Augmented reality can overlay the computer-generated virtual objects onto the real scene. Illumination consistency plays an important role in making the whole scene seem to be seamless and achieving high realism of augmented reality system. Due to the geometric complexity and the illumination complexity of the outdoor scene, its illumination estimation is a challenge at present. In order to resolve the illumination estimation when three-dimensional model is unknown, the project studies the video images of the outdoor scene during fog and haze. We firstly analyze the light illumination model of the outdoor scene and construct the accurate model of sunlight, skylight and ground reflected light. Then the effect of fog and haze is introduced into the framework of the illumination estimation and the illumination model in the fog weather is constructed. In the third, we add constraint conditions according to the nature of the fog video image, establish the energy equation, solve illumination parameters, and exclude the impact of factors such as moving objects. A prototype system for the seamless synthesis of virtual and reality is developed finally. The research results of this project will provide a new theory and method for solving the illumination consistency of outdoor scenes, and provide a simple and efficient t

英文关键词： Illumination estimation；Fog-degraded outdoor image；hue consistency；Global illumination model of the hazy weather；Virtual-real fusion