首页 > 论文 > 中国激光 > 47卷 > 11期(pp:1110002--1)

基于迁移学习的小样本机载激光雷达点云分类

Small Sample Airborne LiDAR Point Cloud Classification Based on Transfer Learning

  • 摘要
  • 论文信息
  • 参考文献
  • 被引情况
  • PDF全文
分享:

摘要

已有基于深度学习的机载激光雷达点云分类方法存在训练时间长、对样本数据需求量较大等问题,无法广泛应用于不同情况下的复杂场景。针对该问题,提出了一种基于迁移学习的小样本机载激光雷达点云分类方法。该方法首先对机载激光雷达点云进行光谱信息的补充,通过提取点云数据的归一化高度、强度值和植被指数特征构建三通道点云特征图;通过设置不同的邻域大小和投影方向,生成多尺度和多投影特征图,并基于迁移学习方法进行多尺度、多投影的深层特征提取。针对上述提取的深层次特征,利用池化操作提取全局特征,并采用卷积神经网络进行初步分类,然后利用图割全局优化策略实现机载激光雷达点云的高精度分类。采用国际摄影测量与遥感协会提供的标准测试数据集对所提方法进行验证。与该协会网站上已公布的分类结果以及同样采用迁移学习方法的分类结果相比,所提方法在仅使用训练集中约0.6%的数据作为训练样本的情况下,总体分类精度可以达到94.9%,分类精度最高。

Abstract

Existing airborne LiDAR point cloud classification methods based on deep learning have problems such as long training time and large demand for sample data, which cannot be widely used in complex scenarios in different situations. To solve this problem, this paper proposes a small sample airborne LiDAR point cloud classification method based on transfer learning. First, the spectral information of the airborne LiDAR point cloud is supplemented, and then three-channel point cloud feature map is constructed by extracting normalized elevation, intensity value, and vegetation index features of point cloud data. Second, multi-scale and multi-projection feature maps are constructed by setting different neighborhood sizes and projection directions, and the multi-scale and multi-projection deep-level feature extraction is carried out based on the transfer learning method. Finally, for the above-mentioned deep-level features extracted, the global features are extracted by pooling operation, the convolutional neural network is used for preliminary classification, and then the graph cut global optimization strategy is used to achieve high-precision classification of airborne lidar point clouds. The proposed method is validated using the standard test dataset provided by the International Association of Photogrammetry and Remote Sensing. By comparing the classification results published on the association''s website and the classification results obtained by the transfer learning method, we find that the method in this paper can achieve an overall classification accuracy of 94.9% when only about 0.6% of the data in the training dataset is used as the training sample, and the classification accuracy is the highest.

广告组1 - 空间光调制器+DMD
补充资料

中图分类号:P237

DOI:10.3788/CJL202047.1110002

所属栏目:遥感与传感器

基金项目:河南省自然科学基金面上项目、河南省科技攻关、河南理工大学博士基金、河南理工大学基本科研业务费专项项目;

收稿日期:2020-05-08

修改稿日期:2020-07-09

网络出版日期:2020-11-01

作者单位    点击查看

雷相达:河南理工大学测绘与国土信息工程学院, 河南 焦作 454000
王宏涛:河南理工大学测绘与国土信息工程学院, 河南 焦作 454000
赵宗泽:河南理工大学测绘与国土信息工程学院, 河南 焦作 454000

联系人作者:王宏涛(wht_31@ hpu.edu.cn)

备注:河南省自然科学基金面上项目、河南省科技攻关、河南理工大学博士基金、河南理工大学基本科研业务费专项项目;

【1】Zhao Z F. Research on road extraction based on airborne LiDAR point cloud [D]. Kunming: Kunming University of Science and Technology. 2012.
赵振峰. 基于机载LiDAR点云的道路提取研究 [D]. 昆明: 昆明理工大学. 2012.

【2】Wang H T, Lei X D, Zhao Z Z. 3D deep learning classification method for airborne LiDAR point clouds fusing spectral information [J]. Laser & Optoelectronics Progress. 2020, 57(12): 122802.
王宏涛, 雷相达, 赵宗泽. 融合光谱信息的机载LiDAR点云三维深度学习分类方法 [J]. 激光与光电子学进展. 2020, 57(12): 122802.

【3】Zhao C, Zhang B M, Yu D H, et al. Airborne LiDAR point cloud classification using transfer learning [J]. Optics and Precision Engineering. 2019, 27(7): 1601-1612.
赵传, 张保明, 余东行, 等. 利用迁移学习的机载激光雷达点云分类 [J]. 光学精密工程. 2019, 27(7): 1601-1612.

【4】Zhang J X, Lin X G, Ning X G. SVM-based classification of segmented airborne LiDAR point clouds in urban areas [J]. Remote Sensing. 2013, 5(8): 3749-3775.

【5】Shi X S, Cheng Y L, Zhao Z Y, et al. Point cloud classification algorithm based on IPTD and SVM [J]. Laser & Optoelectronics Progress. 2019, 56(16): 161002.
释小松, 程英蕾, 赵中阳, 等. 基于三角网滤波和支持向量机的点云分类算法 [J]. 激光与光电子学进展. 2019, 56(16): 161002.

【6】He M Y, Cheng Y L, Liao X J, et al. Building extraction algorithm by fusing spectral and geometrical features [J]. Laser & Optoelectronics Progress. 2018, 55(4): 042803.
何曼芸, 程英蕾, 廖湘江, 等. 融合光谱特征和几何特征的建筑物提取算法 [J]. 激光与光电子学进展. 2018, 55(4): 042803.

【7】Hackel T, Wegner J D, Remote Sensing, Spatial Information Sciences. III- . 2016, 3: 177-184.

【8】Ramiya A M, Nidamanuri R R, Krishnan R. Object-oriented semantic labelling of spectral-spatial LiDAR point cloud for urban land cover classification and buildings detection [J]. Geocarto International. 2016, 31(2): 121-139.

【9】Ni H, Lin X G, Zhang J X. Classification of ALS point cloud with improved point cloud segmentation and random forests [J]. Remote Sensing. 2017, 9(3): 288.

【10】Wei S F. Scene understanding based on point clouds and 2D images [D]. Beijing: University of Chinese Academy of Sciences. 2016.
魏书法. 基于三维点云与二维图像的场景理解 [D]. 北京: 中国科学院大学. 2016.

【11】Niemeyer J, Rottensteiner F, Soergel U. Contextual classification of lidar data and building object detection in urban areas [J]. ISPRS Journal of Photogrammetry and Remote Sensing. 2014, 87: 152-165.

【12】Yang S J, Zhang K S, Shao Y S. Classification of airborne LiDAR point cloud data based on multiscale adaptive features [J]. Acta Optica Sinica. 2019, 39(2): 0228001.
杨书娟, 张珂殊, 邵永社. 基于多尺度自适应特征的机载LiDAR点云分类 [J]. 光学学报. 2019, 39(2): 0228001.

【13】Zhang W M, Qi J B, Wan P, et al. An easy-to-use airborne LiDAR data filtering method based on cloth simulation [J]. Remote Sensing. 2016, 8(6): 501.

【14】Yousefhussien M, Kelbe D J, Ientilucci E J, et al. A multi-scale fully convolutional network for semantic labeling of 3D point clouds [J]. ISPRS Journal of Photogrammetry and Remote Sensing. 2018, 143: 191-204.

【15】Chen N C, Xing C J, Zhang X, et al. Spaceborne earth-observing optical sensor static capability index for clustering [J]. IEEE Transactions on Geoscience and Remote Sensing. 2015, 53(10): 5504-5518.

【16】Duan Y L, Ma X L, Jia D. A transfer learning method based on ResNet''''s authentication code obfuscation style [J]. Journal of Xi''''an Shiyou University (Natural Science). 2019, 34(6): 121-125.
段祎林, 马儇龙, 贾端. 基于ResNet验证码混淆风格的迁移学习方法 [J]. 西安石油大学学报(自然科学版). 2019, 34(6): 121-125.

【17】Kornblith S, Shlens J, Le Q V. Do better ImageNet models transfer better?[C]∥2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 15-20, 2019, Long Beach, CA, USA. New York: , 2019, 2656-2666.

【18】Zhang C L, Luo J H, Wei X S, et al. In defense of fully connected layers in visual representation transfer [M]. Advances in Multimedia Information Processing-PCM 2017, Cham: Springer. 2018, 10736: 807-817.

【19】Boykov Y, Kolmogorov V. An experimental comparison of min-cut/max- flow algorithms for energy minimization in vision [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2004, 26(9): 1124-1137.

【20】Nguyen A, Le B. 3D point cloud segmentation: , 2013, 225-230.

【21】Horvat D, ?alik B, Mongus D. Context-dependent detection of non-linearly distributed points for vegetation classification in airborne LiDAR [J]. ISPRS Journal of Photogrammetry and Remote Sensing. 2016, 116: 1-14.

【22】Yang Z S, Jiang W S, Xu B, et al. A convolutional neural network-based 3D semantic labeling method for ALS point clouds [J]. Remote Sensing. 2017, 9(9): 936.

【23】Niemeyer J, Rottensteiner F, Soergel U, Remote Sensing, Spatial Information Sciences, et al. XLI- . 2016, B3: 655-662.

【24】Wang S L, Suo S, Ma W C, et al. Deep parametric continuous convolutional neural networks[C]∥2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 18-23, 2018, Salt Lake City, UT, USA. New York: , 2018, 2589-2597.

【25】Zhao R B, Pang M Y, Wang J D. Classifying airborne LiDAR point clouds via deep features learned by a multi-scale convolutional neural network [J]. International Journal of Geographical Information Science. 2018, 32(5): 960-979.

【26】Yang Z S, Tan B, Pei H K, et al. Segmentation and multi-scale convolutional neural network-based classification of airborne laser scanner data [J]. Sensors. 2018, 18(10): 3347.

【27】Qi C R, Yi L, Su H, et al. -06-07)[2020-05-08] . https:∥arxiv. 2017, org/abs/1706: 02413.

【28】Arief H A, Indahl U G, Strand G H, et al. Addressing overfitting on point cloud classification using Atrous XCRF [J]. ISPRS Journal of Photogrammetry and Remote Sensing. 2019, 155: 90-101.

【29】Thomas H, Qi C R, Deschaud J E, et al. KPConv: , 2019, 6410-6419.

【30】Wen C C, Yang L N, Li X, et al. Directionally constrained fully convolutional neural network for airborne LiDAR point cloud classification [J]. ISPRS Journal of Photogrammetry and Remote Sensing. 2020, 162: 50-62.

【31】Zhao C, Guo H T, Lu J, et al. ALS point cloud classification with small training data set based on transfer learning [J]. IEEE Geoscience and Remote Sensing Letters. 2020, 17(8): 1406-1410.

引用该论文

Lei Xiangda,Wang Hongtao,Zhao Zongze. Small Sample Airborne LiDAR Point Cloud Classification Based on Transfer Learning[J]. Chinese Journal of Lasers, 2020, 47(11): 1110002

雷相达,王宏涛,赵宗泽. 基于迁移学习的小样本机载激光雷达点云分类[J]. 中国激光, 2020, 47(11): 1110002

您的浏览器不支持PDF插件,请使用最新的(Chrome/Fire Fox等)浏览器.或者您还可以点击此处下载该论文PDF