ICP方法匹配深度图像的实现  

Implementation of Registering Range Images by ICP

作　　者：张宗华[1] 彭翔[1] 胡小唐[1] 

机构地区：[1]天津大学精密测试技术及仪器国家重点实验室,天津300072

出　　处：《天津大学学报（自然科学与工程技术版）》2002年第5期571-576,共6页Journal of Tianjin University：Science and Technology

基　　金：国家自然科学基金资助项目 (698750 1 3);中国博士后基金资助项目

摘　　要：针对ICP(迭代最近点 )方法中寻找深度像之间点对应关系这一难点问题 ,首先介绍对ICP方法匹配深度图像的原理并给出单位四元数和旋转角度间的关系 ,然后提出一种验证两个深度图像中对应点对有效性的方法 ,给出了判断待匹配深度图像之间的点对应条件的准则 ,同时给出了实现这种准则的数据结构以及程序设计思想 .最后 ,用真实深度图像证明此方法的有效性 ,并用“主次缝合线”Iterative Closest Points(ICP)algorithm is a popular method for the registration of multiple range images.However,the main difficulty using ICP algorithm is to look for the correspondence between a pair of range images to the registered.This article briefly describes the principle of ICP algorithm and provides a relationship between unit-quaternion and rotating angles.Then we propose a rule including five criteria in order to identify accurately the correspondent points of the range images.We also present an idea concerning the data structure and pseudo-code for the program design.Practical range images are applied for the implementation of proposed rule to show the efficiency of our approach.In addition,a stitching-based integration method is employed to merge multiple range images.

关 键 词：迭代最近点方法 图像匹配 曲面造型 表面重构 三维数字图像造型 ICP方法 深度图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]