点击次数：

论文类型：期刊论文

发表时间：2022-06-28

发表刊物：Journal of Dalian University of Technology

卷号：61

期号：3

页面范围：316-323

ISSN号：1000-8608

关键字："color point cloud data; rigid registration; local descriptors"

CN号：21-1117/N

摘要：The continuous development of laser scanning technology makes it convenient to obtain the information of color point clouds in three-dimensional space.However,how to unify the color point cloud collected from multiple viewpoints in the same coordinate system is still a challenge.Therefore, an image descriptor based on deep learning is proposed,which is applied to 3Dcolor point cloud registration to obtain the initial pose of point cloud registration with high precision.Firstly,according to the one-to-one correspondence between point clouds and images,3Dcolor point clouds can be projected into images.Secondly,the convolutional neural network is used to extract the local features of point cloud key points,and the combined descriptor is formed by combining the directional gradient histogram.Thirdly,the matching point pairs of point clouds are calculated according to the calculated combined descriptor,and the transformation relationship between point clouds is obtained to realize the point cloud coarse registration.The validity of the proposed method is verified by comparing the actual 3Dcolor point cloud data with various registration algorithms.

备注：新增回溯数据