随着视觉计算技术的快速发展，人们期望计算机能够更精准、全面地理解日益复杂的视觉场景和生成新场景。这就要求在传统的几何形状建模之外，进一步突破材质属性（光、触、热等多模态属性）的视觉解析计算，从而满足航空航天、智能制造、公共安全等领域的重要需求。本项目围绕“多模态材质属性的视觉表征与关联计算”，针对多模态材质属性在获取、解析和关联等方面的难点，重点开展光反射属性的获取解析、非光反射属性的视觉表征、多模态材质属性的关联计算等方面的基础研究，力争取得高动态范围光反射数据单相机获取、光反射对称性约束的光度立体求解、基于微观几何解析的触感属性表征、基于光度演化解析的热感属性表征、语义学习引导的材质属性关联计算等方面的算法与方法创新，为支持更加精准高效的视觉场景解析计算提供理论创新和关键技术，推动视觉计算技术的发展和应用。

With the rapid development of vision computing, computers are expected to understand and produce visual scenes more accurately and comprehensively. As a result, besides the traditional geometry modeling, the key is to make breakthroughs in visual analysis and computing for material properties, such as light reflectance, haptics, and heat sensation, in order to meet the requirements from aerospace, intelligent manufacturing, and public safety. This project focuses on “visual representation and joint computing for multimodal material properties”, and its technical challenges related to capture, analysis and correlation. The proposed researches include capture and analysis for material reflectance, visual representation for non-optical properties, and joint computing for multimodal material properties. Expected achievements include monocular capture of HDR reflectance, reflectance-symmetry-based photometric stereo, haptic property from surface micro-structure, heat property from IR image analysis, and semantic-learning-guided joint computing for material properties. The proposed methods and techniques will lead to more accurate and effective visual scene analysis and computing, and promote the development and application of visual computing.