基于改进ICP算法的室内环境三维地图创建研究

发布时间：2019-07-15 12:43

【摘要】：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15.989m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.
[Abstract]:An improved feature point ICP algorithm based on discrete selection mechanism is proposed, and a 3D map creation method based on the algorithm is designed. The method is divided into three stages. Firstly, the SURF feature points in RGB color image collected during camera motion are extracted and matched, then the initial registration is carried out with RANSAC algorithm, the initial pose of feature point set is optimized, the mismatching is removed, and the feature point ICP algorithm based on discrete selection mechanism is used for accurate registration. Finally, the g2o graph optimization algorithm and key frame are used to optimize the camera motion trajectory, reduce the cumulative error, and the point cloud data collected by the camera is constructed according to the camera's current position and pose to construct a three-dimensional point cloud map. The feasibility and effectiveness of the method are proved by experimental comparison in five open data set environments. The error is only 0.059 m when the camera motion length is 15.989 m, and the 3D map of the experimental environment can be accurately created.
【作者单位】：武汉工程大学计算机科学与工程学院;武汉工程大学智能机器人湖北省重点实验室;
【基金】：国家863计划项目(2013AA12A202) 国家自然科学基金项目(41501505)
【分类号】：TP391.41;TP242

【相似文献】