Improved ICP Algorithm Based on Normal Vector Weight
ZHU Yu-mei1,XING Ming-yi2,CAI Jing1
1. Beijing Changcheng Institute of Metrology and Measurement, Aviation Industry Corporation of China, Ltd., Beijing 100095, China
2. Beihang University,Beijing 100191, China
Abstract Aiming at the problem that the accuracy and speed of point cloud registration in the process of 3D reconstruction are not ideal, an iterative nearest point (ICP) algorithm based on normal vector weight improvement is proposed. By projecting the normal vector of the point cloud onto the Gaussian sphere, the distribution of normal vectors in different directions is counted, the corresponding weight is assigned by combining the geometric structure information of the object, and the normal vector weight combined with the error measurement method from point to plane is used to calculate the optimal rigid body transformation matrix. Experimental results show that taking spherical point cloud data as an example, compared with the iterative closest point (ICP) algorithm before improvement, the registration error is reduced to about 30% without reducing the registration speed, and the algorithm is suitable for various point cloud models with significant effects.
Jiang J, Cheng J, Chen X. Registration for 3-D point cloud using angular-invariant feature［J］. Neurocomputing, 2009, 72(16): 3839-3844.
Wu M Q, Li Z W, Zhong K. Adaptive splicing method of point cloud based on geometric features and image features［J］. Acta Optica Sinica, 2015, 35(2): 229-236.
［15］
Meinzer H, Fangerau M, Schmidt M. Convergent iterative closest-point algorithm to accomodate anisotropic and inhomogenous localization error［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(8) : 1520-1532.
Kong D M, Zhang N, Wang S T, et al. Complicated Point Cloud Model Segmentation Based on Multi-view Region Growing［J］. Acta Metrologica Sinica, 2021, 42(6): 704-709.
［1］
Besl P J, Mckay N D. A Method for Registration of 3-D Shapes［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1992, 14(2): 239-256.
Liu Z, Zhou T, Peng D D, et al. Research on an Improved PCA-based ICP Point Cloud Registration Algorithm［J］. Journal of Natural Sciences of Heilongjiang University, 2019, 36(4): 473-478, 505.
Zhu Y M, Jiang H Z. Point Cloud Registration Algorithm Based on DNSS Combined with Point-to-Plane ICP［J］. Measurement Technology, 2020, 40(6): 21-25.
Yang Q X, Wang C Y, Yang J,et al. Improved ICP algorithm based on angle of law arrow［J］. Computer Engineering and Design, 2016, 37(8): 2082-2086.
Zhang Y Y, Qü W H, Guo Q L, et al. Damage Parameter Calculation Based on 3D Point Cloud Mean Curvature Estimation［J］. Acta Metrologica Sinica. 2018, 39(5): 609-614.
［2］
Rusinkiewicz S, Levoy M. Effcient variants of the ICP algorithm［C］// International Conference on 3-D Digital Imaging and Modeling. Quebec City, Canada,2001.
［5］
Yang J, Li H, Jia Y. GO-ICP: Solving 3D Registration Effciently and Globally Optimally ［C］//Proceedings of the 2013 IEEE International Conference on Computer Vision. IEEE. 2013.
［10］
Makovetskii A , Voronin S , Kober V , et al. Affine Registration of Point Clouds Based on Point-to-plane Approach［J］. Procedia Engineering, 2017, 201:322-330.
Dai J L, Chen Z Y, Ye X Z. Application of ICP algorithm in point cloud registration［J］. Chinese Journal of Image and Graphics, 2007, 129(3): 517-521.
Zhang Y C, Li Y B, Fu X B. Point Cloud Segmentation De-noising Method Based on Curvature Constraint［J］. Acta Metrologica Sinica, 2020, 41(10): 1218-1225.
［3］
Yang J, Li H, Campbell D. Go-ICP: A globally optimal solution to 3D ICP point-set registration［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence, 38(11): 2241-2254, 2016.
［11］
Kwok T H. DNSS: Dual-Normal-Space Sampling for 3-D ICP Registration［J］. IEEE Transactions on Automation Science and Engineering, 2019, 16(1) : 241-252.
Yang X Q, Yang Q X, Yang J. ICP Algorithm Based on Normal Vector Improvement［J］. Computer Engineering and Design, 2016, 37(1): 169-173.
2023, Vol. 44 
