点云配准中多维度信息融合的特征挖掘方法

Feature Mining Method of Multi-Dimensional Information Fusion in Point Cloud Registration

数据挖掘是使用人工智能等方法在大型数据集中提取隐含潜在信息的过程,为从大量信息中获取有价值的知识提供了有效途径.在使用深度学习解决点云配准任务的过程中,数据挖掘也无处不在.全局特征提取和刚体变换估计是无对应点云配准的2个关键阶段,挖掘隐藏在2个阶段中的丰富信息是点云配准的重要任务之一.然而,最近提出的方法在提取全局特征时容易忽略低维局部特征,导致大量点云信息的丢失,使得后续刚体变换估计阶段求解变换参数时精度无法达到预期.首先,提出了一种基于多维度信息融合的特征挖掘网络,充分挖掘点云中的高维全局和低维局部信息,有效弥补了点云配准的全局特征提取阶段局部特征的缺失.其次,在刚体变换估计阶段使用了对偶四元数估计姿态,其可以在一个公共框架内同时表示旋转和平移,为姿态估计提供紧凑和精确的表示.最后,在ModelNet40数据集上进行的大量实验表明:与现有前沿的无对应点云配准方法相比,提出的方法可以获得更高的精度,同时对噪声具有较强的鲁棒性.

Data mining is the process of extracting hidden and potential information in large datasets by artificial intelligence and other methods,which provides an effective way to obtain valuable knowledge from a large amount of information.Data mining is omnipresent in the process of solving point cloud registration task by deep learning.Extracting global features and estimating rigid body transformation are two key stages of corresponding-free point cloud registration.Mining abundant information hidden in two stages is one of the important tasks of point cloud registration.However,recently proposed methods are easy to ignore low-dimensional local features when extracting global features,resulting in the loss of numerous point cloud information,which makes the accuracy of solving transformation parameters in the subsequent rigid body transformation estimation stage unable to reach the expectation.Firstly,a features mining network based on multi-dimensional information fusion is devised,which fully excavates the high-dimensional global information and low-dimensional local information in point cloud,and effectively offsets the lack of local features in the global feature extraction stage of point cloud registration.Secondly,dual quaternion is utilized to estimate pose in the rigid body transformation estimation stage,which can represent rotation and translation simultaneously within a common framework and provide a compact and precise representation for pose estimation.Finally,extensive experiments on ModelNet40 dataset are conducted.The results illustrate that,compared with the existing corresponding-free point cloud registration methods,the proposed method can obtain higher accuracy,while being highly robust with respect to noise.