基于改进ICP算法的室内环境三维地图创建研究

摘要

In this paper,an improved Iterative Closest Point (ICP) algorithm is proposed based on features with the discrete selection mechanism for motion estimation to reconstruct 3D map in indoor environment.Firstly,SURF features in consecutive RGB images are extracted and matched.Then,Random Sample Consensus (RANSAC) algorithm is used for initial registration to optimize the initial pose of features and remove the outliers.Furthermore,secondary registration is applied to calculate the refined transformation between point-clouds in the different coordinate systems combining with the Iterative Closest Point algorithm which based on features with the discrete selection mechanism.Finally,the trajectory of the moving camera is optimized using General Gragh Optimization (g2o) framework combining with key frames,and 3D map is reconstructed through projecting 3D points cloud observed by the camera into global map according to current camera poses.The performance of our proposed algorithm in five public datasets is tested.The results demonstrate that the algorithm is feasible and effectively with the translational error just of 0.059m and ability to generate the 3D map of environment accurately.%提出一种基于离散选取机制的改进特征点ICP算法,并设计了基于该算法的三维地图创建方法.该方法分为3个阶段,首先提取并匹配相机运动过程中采集的RGB彩色图像中的SURF特征点;然后结合RANSAC算法进行初始配准,优化特征点集初始位姿、去除误匹配,并结合基于离散选取机制的特征点ICP算法进行精确配准;最后利用g2o图优化算法结合关键帧实现对相机运动轨迹的优化,减少累计误差,并将相机采集到的点云数据根据相机当前位姿构建三维点云地图.经过在5个公开数据集环境下进行实验对比,证明本方法的可行性和有效性,在相机运动长度为15.989 m的情况下误差仅为0.059m,且能够准确地创建实验环境的三维地图.