融合高斯混合模型和点到面距离的点云配准  

Point Cloud Registration Algorithm Combined Gaussian Mixture Model and Point-to-Plane Metric

作　　者：林桂潮[1,2] 唐昀超 邹湘军[1] 张青[2] 时晓杰[2] 冯文贤[3] Lin Guichao;Tang Yunchao;Zou Xiangjun;Zhang Qing;Shi Xiaojie;Feng Wenxian(Key Laboratory of Key Technology on South Agricultural Machine and Equipment Ministry of Education,South China Agricultural University,Guangzhou 510642;College of Mechanical and Automotive Engineering,Chuzhou University,Chuzhou 239000;School of Civil and Transportation Engineering,Guangdong University of Technology,Guangzhou 510006)

机构地区：[1]华南农业大学南方农业机械与装备关键技术教育部重点实验室,广州510642 [2]滁州学院机械与汽车工程学院,滁州239000 [3]广东工业大学土木与交通工程学院,广州510006

出　　处：《计算机辅助设计与图形学学报》2018年第4期642-650,共9页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(51578162);国家重点科技计划资助项目(2017YFD0700100);广东省科技计划项目(2014A010104011;2015A020209111);滁州学院校级规划项目(2016GH10;2016GH11);安徽省高校自然科学研究项目(KJ2015B20)

摘　　要：针对非线性光照变化、杂乱或遮挡等环境下目标定位精度低的问题,提出一种基于高斯混合模型和点到面距离的点云配准算法,以实现目标的精确定位.首先设模板点云元素服从高斯混合分布,根据点到面距离大小分配高斯混合模型中各组成部分的概率值,构建负对数似然函数;然后应用EM算法对点云优化,并推导了最大化步阶段Q函数的封闭解,提高算法实时性.以合成数据和实际的法兰零件点云为对象进行实验,结果表明,该算法配准精度和鲁棒性明显优于传统配准算法,能够满足复杂工况下目标精确定位要求.A point cloud registration algorithm based on Gaussian mixture model and point-to-plane metric was proposed to solve the problem of inaccuracy of object positioning under nonlinear illumination,clutter or occlusion environment.Firstly,assuming that elements of template cloud were generated by the Gaussian mixture model,the probability values of the components in Gaussian mixture model were allocated according to the point-to-plane distance.Further,a negative logarithmic likelihood function was constructed.Next,the EM algorithm was used to optimize the likelihood function,and the closed solution of Q function in maximum step was derived to improve the real-time performance of the algorithm.Using synthetic data and flange parts to test the proposed algorithm,the results showed that both the accuracy and robustness of the algorithm were superior to the traditional registration algorithm,and could meet the requirements of precise positioning in complex conditions.

关 键 词：点云配准 迭代最近点法 点到面距离 高斯混合模型 最大期望算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]