基于体素格尺度不变特征变换的快速点云配准方法

Fast point cloud registration method based on voxel-SIFT feature

以三维激光雷达为主的多源融合同步定位与建图技术是无人系统室内外无缝定位的研究热点,点云配准是其中重要环节。为了提高点云配准效率,提出一种基于体素格尺度不变特征变换的快速点云配准方法。首先,基于点云体素格提取尺度不变特征变换的特征点,完成初始配准;然后利用随机抽样一致算法剔除边缘匹配点并优化初始变换参数;最后基于优化初始变换参数,采用K维树近邻搜索法,结合最近邻点迭代配准算法完成原始点云数据的快速精确配准。对开源数据集以及实际采集的6个典型场景数据的测试表明,该方法的平均配准时间较传统最近邻点迭代配准算法缩短了78%。实验结果表明了该方法的有效性。

Multi-source fusion simultaneous localization and mapping technology using 3 D Lidar has become a research hotspot of indoor and outdoor seamless positioning for unmanned systems, in which 3 D laser point cloud registration is an important part. In order to improve the efficiency of point cloud registration, a fast point cloud registration method using voxel-based Scale-Invariant Feature Transform is proposed. Firstly, the feature points of voxel-based Scale-Invariant Feature Transform are extracted, and the initial registration is completed. Then the Random Sample Consensus method is used to eliminate the mismatched edge points and optimize the initial transformation parameters. Finally, based on the optimized initial transformation parameters, the K-dimensional tree neighbor search method, combined with Iterative Closest Point registration algorithm, is used to complete the fast and accurate registration of the original point cloud data.The experimental results by testing six typical scenarios from open source data sets and field test data demonstrate that the average registration time of proposed method is 78% shorter than that of the traditional Iterative Closest Point registration algorithm, which show the effectiveness of the proposed method.