Alignment Study of Oral Cavity Point Cloud based on Improved Deep-interaction Transformer

Zhiqiu Zhang; Ye Chen

doi:10.6919/ICJE.202604_12(4).0002

Authors

Zhiqiu Zhang
Ye Chen

DOI:

https://doi.org/10.6919/ICJE.202604_12(4).0002

Keywords:

Oral Medicine; Point Cloud Registration; Deep Learning; Attention Mechanism.

Abstract

In digital oral treatment, the accuracy of point cloud registration has a significant impact on treatment outcomes. To address the limitations of traditional point cloud registration methods when dealing with sparse and partially missing point clouds, an improvement based on the Deep Interaction Transformer (DIT) model is proposed. This study introduces new loss functions and an oral point cloud self-attention module to enhance the registration accuracy of single-frame oral point clouds.The proposed loss functions include symmetry loss of dental curves, smoothness loss of dental surfaces, and contour consistency loss, each optimized to cater to the geometric characteristics of oral point clouds. Additionally, the designed self-attention module, named Oral Point Cloud Self-Attention (OPCSA), enables more refined information interaction through point-wise calculation of queries, keys, and values, thereby enhancing feature extraction capabilities and robustness.Experimental results demonstrate that the improved DIT model significantly outperforms the baseline model in tasks of single-frame point cloud registration for oral and dental models. Not only does this research provide a new solution for high-precision registration of oral point clouds, but it also shows promising potential in terms of methodological effectiveness and application feasibility. The approach sets a foundation for future advancements in digital dental treatment and point cloud processing technologies.

Downloads

Download data is not yet available.

References

[1] Dhanda M, Kukreja A, Pande S S. Region-based efficient computer numerical control machining using point cloud data[J]. Journal of Computing and Information Science in Engineering, 2021, 21(4): 041005.

[2] P. J. Besl and N. D. McKay, “Method for registration of 3-D shapes,” Proc. SPIE, vol. 1611, pp. 586–606, Apr. 1992.

[3] J. Yang, H. Li, D. Campbell, and Y. Jia, “Go-ICP: A globally optimal solution to 3D ICP point-set registration,”IEEE Trans. Pattern Anal.Mach.Intell.,vol.38, no.11, pp. 2241–2254, Nov. 2016.

[4] Y. Aoki, H. Goforth, R. A. Srivatsan, and S. Lucey, “PointNetLK: Robust & efficient point cloud registration using PointNet,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Jun. 2019, pp. 7163–7172.

[5] X. Li, J. K. Pontes, and S. Lucey, “PointNetLK revisited,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Jun. 2021, pp. 12763–12772.

[6] Z. J. Yew and G. H. Lee, “RPM-Net: Robust point matching using learned features,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recog nit. (CVPR), Jun. 2020, pp. 11824–11833.

[7] Z. J. Yew and G. H. Lee, “REGTR: End-to-end point cloud corre spondences with transformers,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Jun. 2022, pp. 6677–6686.

[8] K. Fu, S. Liu, X. Luo, and M. Wang, “Robust point cloud registration framework based on deep graph matching,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Jun. 2021, pp. 8893–8902.

[9] G. Chen, M. Wang, Q. Zhang, L. Yuan and Y. Yue , “Full Transformer Framework for Robust Point Cloud Registration With Deep Information Interaction ” IEEE Trans Neural Netw Learn Syst 2024 Vol. 35 Issue 10 Pages 13368-13382, Oct. 2024

[10] Ligas M, Prochniewicz D. Procrustes based closed-form solution to the point-wise weighted rigid-body transformation in asymmetric and symmetric cases[J]. Journal of Spatial Science, 2021, 66(3): 445-457.

[11] A. Vaswani et al,“Attention is all you need,” in Proc. Adv. Neural Inf. Process. Syst., 2017, pp. 5998–6008.

[12] H. Zhao, L. Jiang, J. Jia, P. Torr, and V. Koltun, “Point transformer,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), Oct. 2021, pp. 16259–16268.

[13] Jin X, Xie Y, Wei X S, et al. Delving deep into spatial pooling for squeeze-and-excitation networks[J]. Pattern Recognition, 2022, 121: 108159.

[14] Deerwester S, Dumais S T, Fraser Marglin D, et al.Indexing by Latent Semantic Analysis[J]. Journal of the American Society for Information Science, 1990, 41(6): 391-407.

[15] Feng, Y., Xu, F., & Xu, R, "3D Shape Recognition via Deep Learning". IEEE Transactions on Image Processing, 2009,29, 5870-5880.

[16] Peng, H., & Zhang, X, "An Efficient Algorithm for 3D Surface Reconstruction with Smoothing Constraints". Computer Graphics Forum, 2009,35(7), 191-201.

[17] Aigerman, N., & Efrat, A,"Generalized-ICP: A Novel Framework for Point Set Registration". "Generalized-ICP". Robotics: Science and Systems (RSS), 2009.

[18] Dong, Q., & Wang, Z,"An Unsupervised Learning-based Gaussian Mixture Model for Point Cloud Registration". Computer Vision and Image Understanding,2022, 192, 102897.

[19] J. Li, C. Zhang, Z. Xu, H. Zhou, and C. Zhang, “Iterative distance-aware similarity matrix convolution with mutual-supervised point elimination for efficient point cloud registration,” in Proc. 16th Eur. Conf. Comput. Vis. Glasgow, U.K.: Springer, Aug. 2020, pp. 378–394.

[20] Zhou, Y., Wang, Y., Hu, Q., & Huang, J,"Ablation Studies in Machine Learning: The Importance of Feature Selection". Journal of Machine Learning Research, 2006,21(154), 1-26.