Analysis of the applicability of PointNet++ deep learning model for semantic segmentation of typical elements of urban as-built mapping

doi:10.13474/j.cnki.11-2246.2024.0215

Abstract

Abstract: The traditional methods for processing urban completion mapping point cloud scene data obtained by 3D laser scanner have several limitations and cannot meet the demand for efficient processing of products in the information society. In this paper, we analyze the demand for classification of urban completion mapping point cloud scenes and study the automated processing of point cloud scenes using a deep learning network model. Firstly, we preprocess the input urban completion mapping data to achieve point cloud downsampling, denoising, and ground point and non-ground point segmentation. Secondly, manually labels five regional scenes with millimeter-level labels and performs data augmentation techniques. And finally tests the semantic segmentation performance and effect of the PointNet++ network in urban completion mapping point cloud scenes. The test results show that the PointNet++ network can achieve the semantic segmentation of laser point clouds in urban completion mapping point cloud scenes with a small number of samples, and the overall mIoU reaches 73.06%, meeting the demand for semantic automatic segmentation of urban completion mapping point clouds and offering a new approach to processing urban completion mapping point cloud data.

Key words: urban as-built mapping point cloud scenes, semantic segmentation, deep learning, model applicability

CLC Number:

HUANG Yinghua, DONG Zhenchuan, LI Hao, CHEN Zhuang, LIU Changrui, ZHANG Xianzhou. Analysis of the applicability of PointNet++ deep learning model for semantic segmentation of typical elements of urban as-built mapping[J]. Bulletin of Surveying and Mapping, 2024, 0(2): 85-89.

References

[1] 成都市工程建设项目“多测合一”技术细则(试行)[EB/OL].[2023-04-10].http://mpnr.chengdu.gov.cn/ghhzrzyj/sjwj/2019-12/20/fec7f6be51384dbab9d004533055be44/files/dcab4f283d3a41d189cfb36c82cdc81e.pdf.
[2] 赵若鹏.测绘新技术在建筑工程规划竣工测量中的应用研究[D].武汉: 湖北工业大学, 2018.
[3] 尚金光,张小波,陈超,等.三维激光扫描点云及其全景技术在“多测合一”中的应用[J].城市勘测,2020(2): 57-61.
[4] 倪福泽.加权总体最小二乘效率优化算法及其在测绘数据处理中的应用[D].太原: 太原理工大学,2020.
[5] 王峰,林鸿,李长辉.地面三维激光扫描技术在城市测绘中的应用[J].测绘通报,2012(5): 47-49.
[6] 王文曦,李乐林.深度学习在点云分类中的研究综述[J].计算机工程与应用,2022,58(1): 26-40.
[7] BOULCH A,GUERRY J,LE SAUX B,et al.SnapNet: 3D point cloud semantic labeling with 2D deep segmentation networks[J].Computers ＆ Graphics,2018,71: 189-198.
[8] MATURANA D,SCHERER S.VoxNet: a 3D convolutional neural network for real-time object recognition[C]//Proceedings of 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).Hamburg,Germany: IEEE,2015: 922-928.
[9] CHARLES R Q,HAO Su,MO Kaichun,et al.PointNet: deep learning on point sets for 3D classification and segmentation[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Honolulu: IEEE,2017: 77-85.
[10] QI C R,YI Li,SU Hao,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.New York: ACM Press,2017: 5105-5114.
[11] HU Qingyong,YANG Bo,XIE Linhai,et al.RandLA-net: efficient semantic segmentation of large-scale point clouds[C]//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Seattle: IEEE,2020: 11105-11114.
[12] ZHOU Yin,TUZEL O.VoxelNet: end-to-end learning for point cloud based 3D object detection[C]//Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Salt Lake City: IEEE,2018: 4490-4499.
[13] 郭昌龙,夏振平,李超超,等.结合改进半径滤波和局部平面拟合的点云去噪算法[J/OL].激光与光电子学进展: 1-11.[2023-12-20].http://kns.cnki.net/kcms/detail/31.1690.tn.20230920.1045.016.html.
[14] ZHANG Wuming,QI Jianbo,WAN Peng,et al.An easy-to-use airborne LiDAR data filtering method based on cloth simulation[J].Remote Sensing,2016,8(6): 501.
[15] 贾洋,李升甫,周城宇,等.高速公路典型要素语义分割深度学习模型适用性分析[J].北京测绘,2022,36(10): 1365-1369.