为了高效提取激光雷达点云数据的局部几何结构特征,实现三维(3D)目标的配准、检测和识别,提出了一种基于分层墨卡托投影(HMec)的局部点云特征描述子。首先,采用传统方法进行特征提取;然后,利用具有保角特性的墨卡托投影,将3D点云数据的局部邻域点分层投影到多个墨卡托平面上;最后,分别统计各墨卡托平面的分布直方图,得到特征点的局部特征描述子。HMec特征描述子能很好地保留点云的局部几何结构特征,从而提高特征描述子的辨别力。在Bologna和3DMatch数据集上的测试结果表明,相比其他9种局部特征描述子,HMec特征描述子的辨别力更强、噪声鲁棒性更好。

In order to efficiently extract the local geometric structure features of LiDAR point cloud data and realize the registration, detection and recognition of three-dimensional (3D) targets, a local point cloud feature descriptor based on hierarchical Mercator projection (HMec) is proposed in this paper. First, the traditional method is used for feature extraction. Then, the local neighborhood points of 3D point cloud data are projected onto multiple Mercator planes using the Mercator projection with conformal feature. Finally, the local feature descriptors of feature points are obtained by counting the histogram of each Mercator plane. HMec feature descriptor can retain the local geometric structure features of point cloud, so as to improve the discrimination of feature descriptor. The test results on Bologna and 3DMatch datasets show that HMec feature descriptors have stronger discrimination and better noise robustness than the other nine local feature descriptors