基于多幅图像的三维重建  

Multiple-image-based 3D reconstruction

作　　者：朱庆生[1] 罗大江[1] 葛亮[1] 刘金凤[1] 

机构地区：[1]重庆大学计算机学院,重庆400044

出　　处：《计算机工程与设计》2010年第10期2351-2353,2357,共4页Computer Engineering and Design

基　　金：国家863高技术研究发展计划基金项目(2006AA10Z233);国家自然科学基金项目(60773082)

摘　　要：针对当前三维重建系统中大量的交互和频繁的失真,根据运动进行物体三维重建的原理,设计了一个简便易操作的三维重建系统。首先经过相机标定,立体匹配点,进而运用双视点三维重建原理即三角测量得到三维坐标;然后运用运动的连续性,综合多视点下的三维信息结合图像特征点的二维局部属性进行优化,最后经过三角剖分和纹理渲染得到物体真实的形状。与传统的三维重建系统相比,该系统能够在用户交互很少的情况下进行三维重建,且能重建出完整的低失真的三维模型。Since there are so many interactions and distortions in current systems,a simple three-dimensional reconstruction system based on the theories of motion are designed.At first through the camera calibration and stereo point matching,the three-dimensional coordinates by employing double-viewpoint method（triangulation principles） are gotten.Then by means of motion continuity,multi-viewpoint three-dimensional information with local properties of feature points in image is integrated.Finally,through triangle meshing and texture rendering,the model of the object can be obtained.Compared with the traditional three-dimensional reconstruction system,this one can reconstruct the whole three-dimensional model with low distortion even in the case of litter user interaction.

关 键 词：相机标定 立体匹配 三角测量 三角剖分 纹理渲染 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]