【作者】 王章野；

【导师】 彭群生；

【作者基本信息】 浙江大学 ， 计算机图形学， 2002， 博士

【摘要】 红外成像仿真和多光谱信息融合是现代高科技的关键技术之一，在红外制导与对抗、反恐怖活动、遥感探测、交通安全等众多领域有着广泛的应用。本文主要研究地面目标的红外成像仿真和多光谱成像的真实感融合。内容涉及以下几个方面：城市建筑物红外成像模型、地面目标与背景的真实感红外合成、动态红外场景的实时生成、基于量子光学原理的红外成像光照模型和多光谱融合真实感成像等。 本文第一章介绍了红外技术和红外成像仿真技术的基本知识、研究内容、应用领域和国内外发展现况。指出利用计算机图形学特别是虚拟现实的最新技术将是今后的发展方向。并介绍了多光谱信息融合技术的一些基本概念、研究内容和应用前景，并回顾了这一领域的国内外研究现况。 细致的城市场景红外成像建模以前很少有人涉及。本文第二章从气象学、传热学和红外物理学原理出发，分析了影响城市建筑物红外特性的各种主要因素，采用多层有限差分等方法求解建筑物及周围物体的表面温度，建立了城市建筑物红外成像模型；采用分类建模和特殊建模相结合的方法构建了大规模复杂城市场景的红外模型，绘制出了不同时刻城市复杂场景红外成像图。 目标与背景的真实感红外合成对红外仿真系统有着重要的意义。本文第三章从传热学原理出发，建立起坦克等地面目标的红外热传导模型，计算出坦克和地面接触后达热平衡时的温度场分布，绘制出目标与地面背景真实感合成的红外图像，并生成了地面目标驶离后不同时刻在地面留下的“红外阴影”效果图，这为反推某时刻之前红外目标的存在提供了依据。 动态红外场景的实时生成和交互式漫游的研究尚未见报道。本文第四章采用了场景几何造型和红外纹理图像混合的简化方法来进行红外动态场景的建模。提出了由可见图像转化生成场景目标红外图像的方法，采用了大气辐射传输计算软件和考虑热像仪噪声扰动机制的方法来预生成红外深度衰减纹理图像的模型，实现了动态红外场景的实时生成和虚拟漫游。浙江大学博士学位论文 量子光学的辐射现象是计算机图形学很少研究的课题.本文第五章从量子光学的理论出发，提出了一个新的基于量子光学的光照模型，该模型在传统光照模型的基础上增加了一项物体自身量子辐射项，可同时适用于红外波段和可见光波段的成像.依照该模型就可遇真地绘制出钢铁融化过程的光照效果;还可绘制出红外场景的镜面高光效果和红外波段穿透云雾成像的效果图. 现有的多光谱图像融合方法的结果经常丢失原光谱图像的光谱信息;且对不同的场景采用同一种融合方法，其效果有时并不好.本文第六章提出了多光谱图像真实感融合生成算法，算法强调在融合图像中保留原多光谱图像的光谱特征.提出了对相关性较弱的图像对实行基于HvS原理融合的新方法;对于相关性较强的图像对，根据场景目标距离的远近，分别采用小波变换和主成份分析进行融合.并进行了RGB多通道增强型方法的彩色融合.本算法对不同类型的场景得到的融合结果优于用单种方法进行融合，有利于对目标的探测和识别. 在本文的最后，作者对本文的研究工作进行了总结，并提出进一步研究的一些关健技术问题及本课题的发展方向.更多还原

【Abstract】 Infrared (IR) Image Synthesis and Multi-spectral Information Fusion are the key techniques in current high scientific and technological countermeasure environment and have been found wide application in many domains such as Infrared Guidance & Countermeasure, Anti-terrorism, Remote Sensor detection, traffic safety, etc. In this dissertation, the infrared image synthesis for ground targets and the realistic fusion of multi-spectral images are studied. The content of this dissertation includes following subjects: the IR imaging model of city buildings, the realist image synthesis between IR target and the ground background, the real time walkthrough for dynamic IR scene, a Quantity Optics based illumination model for IR image synthesis and the realistic fusion of multi-spectral images, etc.In Chapter 1, we first introduce some primary knowledge, research contents, application domains and related research works about IR Technology and Infrared Image Synthesis. And we point out that using the latest techniques of Computer Graphics especially that of Virtual Reality (VR) will be the developing trend in this field. Then we describe some basic concepts, research contents, application trends and related research works about Multi-spectral Information Fusion.Little works have been done about fine IR imaging modeling for urban scene. In Chapter 2, a new IR imaging model for city buildings has been established based on Meteorology, Heat Transfer and Infrared Physics. We first analysis all kinds of main factors affecting the IR characteristic of city buildings, then by using the methods such as multi-layer finite difference the surface temperatures of building and other surrounding objects are acquired. Combining the methods of classified modeling and special modeling we constitute the large-scale complex urban IR scene and render the IR images of urban scene at various time.The realistic synthesis between IR targets and the background is very important for IR image synthesis system. In Chapter 3 we establish the IR thermal model of ground targets such as tanks based on the principle of heat transfer and calculate the re-distribution of temperature field after the touch between targets and ground. Then we render the realistic IR images between the targets and the ground background. We also render different IR images of the ground at the different intervals after the targets have been driven away. Based on these IR images, we can determine the existence ofthe IR targets some certain interval ago.Reports about the real-time generation of dynamic IR scene and its interactive virtual walkthrough have never been seen yet. In Chapter 4 we constitute the dynamic IR scene by combining scene geometry modeling with IR texture images. In this way the complexity of IR scene is reduced. We also propose a method generating IR target images from visible images in the same scene. By pre-generation IR depth texture images using the calculation software of atmosphere radiation propagation and considering the mechanism of the noise disturbance of thermal imager, we successfully realize the real-time generation of dynamic IR scene and its virtual walkthrough.Displaying the radiation phenomena in the quantity optics is also an area less explored in Computer Graphics. In Chapter 5 we develop a new Quantity Optics based illumination model by incorporating a self-radiation term of the object in traditional illumination model. Our new model is suitable both for IR waveband imaging and for visible waveband imaging. The model successfully simulates the natural radiation phenomena such as the color of iron during the different fused phases. Our model can also simulates highlight (specular reflection) phenomena in IR scene and the special IR scene image through thick fog. This study lays a foundation to establish the unified illumination model for multi-spectral fusion imaging.The results of current algorithms for multi-spectral image fusion often lose spectral information of original spectral images and the fused images are not更多还原