TY - GEN
T1 - An Improved ICP Point Cloud Registration Method Based on PCA and Point Normal
AU - Liu, Tong
AU - Hao, Can
AU - Gao, Chao
AU - Cui, Peiling
AU - Zhou, Weihu
N1 - Publisher Copyright:
© 2025 Copyright held by the owner/author(s).
PY - 2026/3/16
Y1 - 2026/3/16
N2 - Point cloud registration addresses the issue of incomplete point clouds caused by occlusion in industrial precision measurement. Improving the registration efficiency and accuracy is an important task. In this paper, we propose an improved ICP registration method based on PCA and point normal. First, the three principal directions of the point cloud are calculated by PCA method, to get the possible transformations. Then the similarity between the registration and the reference point cloud by each possible transformation is calculated, and the one with the highest similarity is selected as the initial transformation. At last, the fine registration is realized by using the nearest distance and point normal constraint ICP together. The registration experiments were carried out on both the public datasets and the real captured point cloud. Compared with another three methods, the results demonstrate our method’s efficiency and accuracy.
AB - Point cloud registration addresses the issue of incomplete point clouds caused by occlusion in industrial precision measurement. Improving the registration efficiency and accuracy is an important task. In this paper, we propose an improved ICP registration method based on PCA and point normal. First, the three principal directions of the point cloud are calculated by PCA method, to get the possible transformations. Then the similarity between the registration and the reference point cloud by each possible transformation is calculated, and the one with the highest similarity is selected as the initial transformation. At last, the fine registration is realized by using the nearest distance and point normal constraint ICP together. The registration experiments were carried out on both the public datasets and the real captured point cloud. Compared with another three methods, the results demonstrate our method’s efficiency and accuracy.
KW - Industrial precise measurement
KW - PCA based registration
KW - normal constraint ICP registration
KW - point cloud similarity
UR - https://www.scopus.com/pages/publications/105035396385
U2 - 10.1145/3772673.3772699
DO - 10.1145/3772673.3772699
M3 - 会议稿件
AN - SCOPUS:105035396385
T3 - ACMLC 2025 - Proceedings of 2025 7th Asia Conference on Machine Learning and Computing
SP - 105
EP - 110
BT - ACMLC 2025 - Proceedings of 2025 7th Asia Conference on Machine Learning and Computing
PB - Association for Computing Machinery, Inc
T2 - 2025 7th Asia Conference on Machine Learning and Computing, ACMLC 2025
Y2 - 25 July 2025 through 27 July 2025
ER -