基于体积和纹理特征的深度像匹配  

A range image registration method base on the inter-space with texture

作　　者：刘星明[1] 刘则毅[2] 刘晓利[1] 李阿蒙[1] 

机构地区：[1]深圳大学光电工程学院,光电器件与系统教育部重点实验室,广东省光电子器件与系统重点实验室,深圳518060 [2]深圳大学数学与计算科学学院,深圳518060

出　　处：《深圳大学学报（理工版）》2012年第1期91-94,共4页Journal of Shenzhen University(Science and Engineering)

基　　金：国家自然科学基金资助项目(60775021)~~

摘　　要：提出一种深度像精匹配方法,利用深度像重叠区域的空间体积和纹理特征,定义深度像匹配的误差测度函数,构造三角网格近似物体表面.通过寻找不同坐标系下深度像之间的精确空间坐标变换关系,实现深度像的精匹配.实验结果表明,用体积和纹理特征能准确反映深度像之间的距离关系,且对准精度高,收敛速度快.A novel algorithm for precise registration was proposed. An error metric based on the inter-space with texture in the overlap region was defined, and an approximate surface was constructed. By finding the rigid transformation relationships between range images in the different coordinate systems, the proposed algorithm was able to achieve precise registration with respect to the metric function. The experiment results show that the inter- space with texture metric function can indicate the distance between range images with high accuracy and fast convergence.

关 键 词：计算机视觉 深度像匹配 多视角几何 三角面片 K-D树 最近点迭代算法 三维数字成像及造型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]