文献类型：期刊文献

中文题名：基于M估计的稳健配准

英文题名：Robust Registration Based on M Estimation

作者：刘晶[1];林大钧[1]

机构：[1]华东理工大学机械与动力工程学院,上海200237

年份：2013

卷号：39

期号：4

起止页码：472

中文期刊名：东华大学学报（自然科学版）

外文期刊名：Journal of Donghua University(Natural Science)

收录：CSTPCD;;北大核心:【北大核心2011】；CSCD:【CSCD2013_2014】；

语种：中文

中文关键词：M估计;配准;四元数法

外文关键词：M estimation; registration; quaternion method

摘要：在计算机视觉和逆向工程领域,数字化测量数据与其CAD模型间的配准是一个值得研究的问题.基于3个基准点的对齐方法实现数字化测量数据与CAD模型的粗配准,使得变换后的测量数据与CAD模型比较接近,再通过M估计来实现稳健配准,这可以去除原始测量数据中的不好数据,同时较快地获得模型变换并实现精确配准.试验结果表明,该算法能很好地实现点集与三维模型间的配准.
Registration between digitized measuring data and CAD model has been intensely studied in computer vision and reverse engineering. The registration method based on three reference points are utilized to obtain rough registration, this can make transformed digitized data be close with CAD model, then robust registration is implemented by M estimation. This method can remove outliers of initial digitized data and calculate the rotation matrix and translation matrix accurately. Experiment demonstrates that the approach can realize the registration between point set and three dimensional model.