ABSTRACT
Many significant progresses have been made in the field of deep learning. This paper mainly discusses the 3D Match point cloud registration method based on deep learning and its improvement. The method introduced in this article is divided into four steps. The first is to obtain point cloud data. This step uses a bilateral filtering algorithm, which plays a good role in removing noise points from point clouds. The second step is to use 3d match to register key points. This step uses the truncated distance function (TDF) to perform preliminary processing on the point cloud data, and input the Siamese network matching with metric learning to learn the features of the point cloud. The third step eliminates the wrong point pair, this step uses the classic RANSAC algorithm. The fourth step is similarity measurement. The 3D Match network will output a set of 512-dimensional feature vectors, and the spatial dimension is relatively high. Therefore, a cosine similarity that is more suitable for multi-dimensional feature similarity measurement is used to replace the commonly used Euclidean distance.
- Liwei Huang Review of recommendation system based on deep learning [J]. Chinese Journal of Computers, 2018, 41(7):29. 1619-1647Google Scholar
- Jiang Y, huang h g, Shu Q, Scale point cloud registration algorithm in high-dimensional orthogonal subspace mappinglj. Acta Optica Sinica, 2019, 39(3): 0315007.Google ScholarCross Ref
- Hu Q, Yang B, Xie L, Randla-net: Efficient semantic segmentation of large-scale point clouds [C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 11108-11117.Google Scholar
- Lowe D G. Distinctive Image Features from Scale-Invariant Keypoints [J]. International Journal of Computer Vision, 2004, 60(2): 91-110.Google ScholarDigital Library
- Sipiran I, Bustos B. Harris 3D: a robust extension of the Harris operator for interest point detection on 3D meshes [J]. Visual Computer, 2011, 27(11): 963.Google ScholarDigital Library
- Yu Z. Intrinsic shape signatures: A shape descriptor for 3D object recognition [C]// IEEE International Conference on Computer Vision Workshops. IEEE, 2010. 159-166Google Scholar
- Han J, Peng Y, He Y, Enhanced ICP for the Registration of LargeScale 3D Environment Models: An Experimental Study [J]. Sensors, 2016, 16(2): 228.Google ScholarCross Ref
- Zeng A, Song S, Nießner M, 3dmatch: Learning local geometric descriptors from rgb-d reconstructions [C]// Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1802-1811.Google Scholar
- Qi C R, Su H, Mo K, Pointnet: Deep learning on point sets for 3d classification and segmentation [C]// Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 652-660.Google Scholar
- J. Li, B.M. Chen and G. H. Lee, "SO-Net: Self-Organisation Network for Point Cloud Analysis", 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018, pp. 9397-9406, doi: 10.1109/CVPR. 2018.00979.Google ScholarCross Ref
Recommendations
Deep learning architectures in emerging cloud computing architectures: Recent development, challenges and next research trend
AbstractThe challenges of the conventional cloud computing paradigms motivated the emergence of the next generation cloud computing architectures. The emerging cloud computing architectures generate voluminous amount of data that are beyond ...
Graphical abstractDisplay Omitted
Highlights- The article presents literature review on the application of deep learning in emerging cloud computing architectures.
Research on Point-wise Gated Deep Networks
Display Omitted We introduce pgRBMs into DBNs and present Point-wise Gated Deep Belief Networks.Similar to pgDBNs, Point-wise Gated Deep Boltzmann Machines are presented.We introduce dropout and weight uncertainty methods into pgRBMs.We discuss the ...
Comments