融合几何注意力和多尺度特征点云配准网络  

Fusion of Geometric Attention and Multi-Scale Feature Network for Point Cloud Registration

作　　者：杜佳锦 柏正尧[1] 刘旭珩 李泽锴 肖霄 尤逸琳 DU Jiajin;BAI Zhengyao;LIU Xuheng;LI Zekai;XIAO Xiao;YOU Yilin(School of Information,Yunnan University,Kunming 650504,China)

机构地区：[1]云南大学信息学院,昆明650504

出　　处：《计算机工程与应用》2024年第12期234-244,共11页Computer Engineering and Applications

基　　金：云南省重大科技专项计划资助项目(202002AD080001)。

摘　　要：点云不同的几何特征影响着点云配准的难度。然而,大多数点云存在部分重叠,几何特征被噪声干扰,并且点云表面有部分不可区分的特征,这使得提取具有代表性的几何特征更加困难。结合以上原因提出一种融合几何注意力和多尺度特征的点云配准网络GMNet,利用几何Transformer提取几何特征并对点云点对距离和角度编码,使其在低重叠情况下更具鲁棒性;并使用多尺度特征架构聚合不同尺度上丰富的语义信息,提高点云配准的准确率;最后特征通过一致决策算法选择具有适当邻域大小的特征。GMNet分别在室内数据集3DMatch、3DLoMatch和室外数据集KITTI上进行实验,实验结果表明GMNet的整体配准精度较高,在3DMatch数据集和3DLoMatch数据集上配准召回率分别提升到93.4%和76.0%,在KITTI数据集上相对旋转误差和相对平移误差分别降低到6.2 cm和0.26°。该方法使用的几何Transformer提取有代表性几何特征,联合多尺度特征学习点云中的不同层次几何信息,有效提升点云配准精度。The different geometric features of point clouds affect the difficulty of point cloud registration.However,most point clouds have partial overlap,geometric features are disturbed by noise,and there are some indistinguishable features on the point cloud surface,which makes it more difficult to extract representative geometric features.Combining the above reasons,this paper proposes a point cloud registration network GMNet that fuses geometric attention and multiscale features,using geometric Transformer to extract geometric features and encode point cloud point pairs with distances and angles to make it more robust under low overlap.Using multi-scale feature architecture to aggregate abundant semantic information at different scales to improve the accuracy of point cloud registration.Finally the features are selected by consistent voting algorithm with the appropriate domain size.The experimental results show that the overall registration accuracy of GMNet is higher,and the registration recall is improved to 93.4%and 76.0%for 3DMatch and 3DLoMatch datasets respectively.The relative rotation error and relative translation error are reduced to 6.2 cm and 0.26°on the KITTI dataset,respectively.The geometric Transformer used in this method extracts representative geometric features and combines multi-scale features to learn different levels of geometric information in the point cloud,which effectively improves the point cloud registration accuracy.

关 键 词：点云配准 几何Transformer 多尺度特征 一致决策算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]