基于总体最小二乘的点云三维配准及改进的ICP算法研究

发布时间：2018-01-13 06:20

本文关键词：基于总体最小二乘的点云三维配准及改进的ICP算法研究　出处：《中国矿业大学》2016年硕士论文　论文类型：学位论文

【摘要】：三维激光扫描仪作为一种新兴的测量手段,可以快速高效地获取实物或场景的三维点云信息。本文针对三维激光扫描点云拼接的核心问题,研究探讨了基于总体最小二乘和抗差估计的点云数据三维配准方法,并基于多元总体最小二乘理论研究改进了的迭代最近点算法(ICP),有效提高了配准的效率和稳健性。1.研究基于总体最小二乘附有限制条件的三维基准转换模型及其抗差算法。在论述杠杆观测值及其对抗差权函数和总体最小二乘模型的影响,并公式推导影响存在性的基础上,构建了面向三维坐标转换的三种抗差权函数:Huber、顾及杠杆观测值的IGG和IGGⅢ。实验研究表明基于总体最小二乘的模型精度优于最小二乘模型,IGG权函数和IGGⅢ权函数抗差效果优于Huber。2.引入多元总体最小二乘算法,研究了基于多元总体最小二乘的三维基准转换模型及其抗差算法。针对附有限制条件的三维基准转换模型中存在的模型误差和病态问题,结合重心化后三维基准转换模型函数前后对称的特点,建立了基于多元总体最小二乘三维基准转换模型,实现了模型抗差计算。验证表明,基于多元总体最小二乘的求解模型比传统的求解转换参数的模型精度更优,且计算过程效率较高;3.在三维点云拼接过程中,提出利用多元总体最小二乘改进ICP算法并推证了算法的可行性和有效性。研究通过主元分析法获取点云间的初始位置,利用KD-Tree进行最近点搜索,基于改进的ICP算法实现刚体转换参数的求解。实验研究表明,改进的ICP算法计算效率优于基于四元数法ICP算法,在保证精度的前提下,提高了计算效率。
[Abstract]:The 3D laser scanner as a new measurement method, 3D point cloud information can access the objects or scenes quickly and efficiently. The core problems in the registration of 3D laser scanning point cloud, is discussed based on total least squares and robust estimation of the point cloud data of three dimensional registration method, and based on the theory of multivariate iterative total least squares the improved nearest point algorithm (ICP), effectively improve the efficiency and robustness of.1. registration datum based on total least squares with constraints transformation model and robust algorithm. Influence on the function and the total least squares model on the lever right in the observation of the existence and influence of formula on the basis of constructed for three kinds of robust weighting function of 3D coordinate transformation: Huber, take into account the leverage observation values of IGG and IGG III. Experimental results show that based on the overall The accuracy of the model is better than the least squares model of least squares, IGG weight function and weight function of robust IGG III Huber.2. is better than using multivariate total least squares algorithm is studied, the total least squares multivariate datum transformation model and robust algorithm based on the model error and the ill posed problem existed in the model with a three-dimensional datum transformation, combined with 3D the characteristics of symmetric datum transformation model function before and after the center of gravity, established the multiple 3D datum transformation model based on total least squares, the model robust calculation. Results show that the accuracy of the model based on the model of multiple total least squares conversion parameters better than the traditional method, and the computation efficiency is higher; 3. in 3D point cloud in the process of using total least squares multivariate improved ICP algorithm and prove the feasibility and effectiveness of the algorithm. The initial position of access point clouds by principal component analysis method, by using KD-Tree to solve the closest point search, improved ICP algorithm based on rigid transformation parameters. Experimental results show that the improved ICP algorithm is computationally more efficient than four element method based on ICP algorithm, under the premise of ensuring the accuracy, the computational efficiency is improved.

【学位授予单位】：中国矿业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P225.2

【相似文献】