基于旋转图像特征描述子改进的ICP算法  

Improved ICP algorithm based on spin-image feature descriptor

作　　者：郎萍[1] 高媛[1] 秦品乐[1] 王丽芳[1] 

机构地区：[1]中北大学计算机与控制工程学院,山西太原030051

出　　处：《计算机工程与设计》2016年第10期2750-2753,2774,共5页Computer Engineering and Design

基　　金：山西省自然基金项目(2015011045)

摘　　要：为获取更好的配准结果,在ICP算法的基础上,提出一种基于旋转图像(spin-image)特征描述子改进的ICP算法。获取两个待匹配点云的旋转图像特征描述子,通过比较两个描述子之间特征的相似程度,估算它们之间的对应关系并计算初始匹配参数,利用最小二乘法迭代进行ICP配准。实验结果表明,改进后的算法在配准精度和运行时间上都有较大改善。To get better results of registration,on the basis of ICP algorithm,an improved ICP algorithm based on spin-image feature descriptor was proposed.The spin-image feature descriptors were obtained from two matching point clouds,the corresponding relations between them were estimated,and the initial parameter was calculated by comparing the similarity between those feature descriptors.The least square method was used to complete ICP registration.Experimental results show that the improved algorithm improves the registration accuracy and the operation time.

关 键 词：迭代最近点算法 旋转图像特征描述子 配准 对应关系 最小二乘法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]