Semantic segmentation of indoor 3D point cloud by joint optimization of geometric features and neural networks

doi:10.13474/j.cnki.11-2246.2023.0136

Abstract

Abstract: A precise semantic segmentation of indoor 3D point cloud is the basis for realizing deep applications of interior space. To address the problem of incomplete and inconsistent segmentation objectives of existing semantic segmentation methods for 3D point clouds. In this paper, an semantic segmentation method for point clouds is proposed, it uses geometric features of point clouds and deep neural networks. First of all, it uses deep learning to achieve the initial extraction of semantic labels of indoor structural information. Secondly, it uses the segmentation method of point cloud with geometric features and color features to accurately segment the original data.Finally, a probabilistic model has proposed to cross-validate the initial segmentation results with the segmentation results of geometric features to achieve joint optimization of the results for semantic segmentation. The accuracy and validity of the segmentation method proposed in this paper are verified based on open-source datasets, and three sets of indoor point cloud data from simple to complex indoor scenes are tested respectively, and the experimental results show that the method proposed in this paper can effectively improve the semantic segmentation accuracy of the indoor 3D point cloud.

Key words: neural network, point cloud, semantic segmentation, multi-level plane extraction, color region segmentation

CLC Number:

P237

YAO Mengmeng, LI Xiaoming, WANG Weixi, XIE Linfu, HUANG Junjie, HUANG Hongsheng, TANG Shengjun. Semantic segmentation of indoor 3D point cloud by joint optimization of geometric features and neural networks[J]. Bulletin of Surveying and Mapping, 2023, 0(5): 56-61.

References

[1] KANG Z, YANG J, YANG Z, et al.A review of techniques for 3D reconstruction of indoor environments[J].International Journal of Geo-Information, 2020, 9(5):1-31.
[2] CHOI J, CHOI J, KIM I.Development of BIM-based evacuation regulation checking system for high-rise and complex buildings[J].Automation in Construction, 2014, 46:38-49.
[3] TAIRA H, OKUTOMI M, SATTLER T, et al.InLoc:indoor visual localization with dense matching and view synthesis[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(4):1293-1307.
[4] 杨必胜, 董震.点云智能研究进展与趋势[J].测绘学报, 2019, 48(12):1575-1585.
[5] BELLO S A, YU SHANGSHU, WANG CHENG.Review:deep learning on 3D point clouds[J].Remote Sensing, 2020, 12(11):1729.
[6] LI B S, CEN H Y, BAO Y P, et al.Edge detection algorithm for point cloud based on rolling-circle[J].Computer Engineering and Design, 2013, 34(5):1836-1840.
[7] XU W, KANG Z, JIANG T.Segmentation approach for terrestrial point clouds based on the integration of graph theory and region growing[C]//Proceedings of 2009 Joint Urban Remote Sensing Event.[S.l.]:IEEE, 2009.
[8] ZHAN Q, YU L, LIANG Y.A point cloud segmentation method based on vector estimation and color clustering[C]//Proceedings of the 2nd International Conference on Information Science and Engineering.Hangzhou:IEEE, 2010.
[9] 李宝顺, 岑红燕, 包亚萍, 等.基于平面提取的点云数据分割算法[J].计算机应用与软件, 2014, 31(7):145-148.
[10] WU BICHEN, WAN A, YUE XIANGYU, et al.SqueezeSeg:convolutional neural nets with recurrent CRF for real-time road-object segmentation from 3D LiDAR point cloud[EB/OL].[2022-03-02].https://arxiv.org/abs/1710.07368.
[11] WANG F, ZHUANG Y, GU H, et al.OctreeNet:A novel sparse 3D convolutional neural network for real-time 3D outdoor scene analysis[J].IEEE Transactions on Automation Science and Engineering, 2020, 17(2):735-747.
[12] FENG Y, ZHANG Z, ZHAO X, et al.GVCNN:group-view convolutional neural networks for 3d shape recognition[C]//Proceedings of 2018 IEEE Conference on Computer Vision and Pattern Recognition.Lake City:IEEE, 2018.
[13] QI C R, SU H, MO K, et al.PointNet:deep learning on point sets for 3D segmentation and segmentation[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Honolulu:IEEE, 2017.
[14] QI C R, YI L, SI H, et al.PointNet++:deep hierarchical feature learning on point sets in a metric space[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems.[S.l.]:NIPS, 2017.
[15] HU Q, YANG B, XIE L, et al.RandLA-Net:efficient semantic segmentation of large-scale point clouds[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.[S.l.]:IEEE, 2020.
[16] TARSHA-KURDI F, LANDES T, GRUSSENMEYER P.Hough-transform and extended ransac algorithms for automatic detection of 3D building roof planes from lidar data[C]//Proceedings of ISPRS Workshop on Laser Scanning 2007 and SilviLaser 2007.Espoo:[s.n.], 2007.
[17] WU Y X, LI F, LIU F F, et al.Point cloud segmentation using Euclidean cluster extraction algorithm with the smoothness[J].Meas Control Technol, 2016, 35(3):36-38.
[18] 王涛, 王文举, 蔡宇.基于深度学习的三维点云语义分割方法研究[J].计算机工程与应用, 2021, 57(23):18-26.