留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于高程感知多尺度图卷积网络的地物分类

文沛 程英蕾 王鹏 赵明钧 张碧秀

文沛,程英蕾,王鹏,等. 基于高程感知多尺度图卷积网络的地物分类[J]. 北京航空航天大学学报,2023,49(6):1471-1478 doi: 10.13700/j.bh.1001-5965.2021.0434
引用本文: 文沛,程英蕾,王鹏,等. 基于高程感知多尺度图卷积网络的地物分类[J]. 北京航空航天大学学报,2023,49(6):1471-1478 doi: 10.13700/j.bh.1001-5965.2021.0434
WEN P,CHENG Y L,WANG P,et al. Ground object classification based on height-aware multi-scale graph convolution network[J]. Journal of Beijing University of Aeronautics and Astronautics,2023,49(6):1471-1478 (in Chinese) doi: 10.13700/j.bh.1001-5965.2021.0434
Citation: WEN P,CHENG Y L,WANG P,et al. Ground object classification based on height-aware multi-scale graph convolution network[J]. Journal of Beijing University of Aeronautics and Astronautics,2023,49(6):1471-1478 (in Chinese) doi: 10.13700/j.bh.1001-5965.2021.0434

基于高程感知多尺度图卷积网络的地物分类

doi: 10.13700/j.bh.1001-5965.2021.0434
基金项目: 国家自然科学基金(61773396)
详细信息
    通讯作者:

    E-mail:Ylcheng@163.com

  • 中图分类号: V557+.3;TP391

Ground object classification based on height-aware multi-scale graph convolution network

Funds: National Natural Science Foundation of China (61773396)
More Information
  • 摘要:

    机载激光雷达获取的点云具有类别分布不均匀、样本高程差异大的复杂特点,现有方法难以充分识别其细粒度的局部结构。针对该问题,堆叠使用多层边缘卷积算子同时提取局部信息和全局信息,并引入高程注意力权重作为特征提取的补充,设计了一种用于机载激光雷达点云地物分类的端到端网络,提出基于高程感知多尺度图卷积网络的地物分类方法。对原始点云划分子块并采样到固定点数;采用多尺度边缘卷积算子提取多尺度局部-全局特征并进行融合,同时采用高程感知模块生成注意力权重并应用于特征提取网络;利用改进的焦点损失函数进一步解决类别分布不均问题,完成分类。采用国际摄影测量与遥感学会(ISPRS)提供的标准测试数据集对所提方法进行验证,所提方法的总体分类精度达到0.859,单类别分类精度特别是对建筑物的提取精度较ISPRS竞赛中公开的最好结果提高了4.6%。研究结果对实际应用和网络设计优化具有借鉴意义。

     

  • 图 1  边缘特征提取示意图

    Figure 1.  Schematic diagram of edge feature extraction

    图 2  多尺度特征提取示意图

    Figure 2.  Schematic diagram of multi-scale feature extraction

    图 3  高程感知模块原理图

    Figure 3.  Schematic diagram of Height-aware module

    图 4  基于HAMS-GCN的点云分类流程

    Figure 4.  Point cloud classification process based on HAMS-GCN

    图 5  实验数据集

    Figure 5.  The experimental data set

    图 6  递归深度测试结果

    Figure 6.  Recursive depth test results

    图 7  预处理分块示意图

    Figure 7.  Schematic diagram of preprocessed sub-blocks

    图 8  分类结果可视化

    Figure 8.  Visualization of classification results

    图 9  分类结果细节可视化

    Figure 9.  Detailed Visualization of classification results

    图 11  与其他方法结果对比

    Figure 11.  Comparison with results from other methods

    图 10  混淆矩阵

    Figure 10.  Confusion matrix

    表  1  实验数据集中不同类别的百分比

    Table  1.   The percent of different categories in the experimental dataset %

    类别训练集测试集
    电力线0.070.15
    低矮植被23.9923.97
    不透水表面25.7014.77
    车辆0.610.90
    篱笆/栅栏1.601.80
    屋顶20.1726.48
    外墙立面3.612.72
    灌木6.316.03
    树木17.9413.17
    下载: 导出CSV

    表  2  本文方法与其他模型在ISPRS数据集上的结果对比

    Table  2.   Comparisons between the proposed method and other models on the ISPRS dataset

    方法m1m2m3m4m5m6m7m8m9OAAvg F1
    PointNet0.5260.7000.8320.1120.0750.7480.0780.2460.4540.6570.419
    PointNet++0.5790.7960.9060.6610.3150.9160.5460.4160.7700.8120.656
    文献 [20]0.8280.6590.9420.6710.2520.9150.4900.6270.8260.8050.664
    文献 [21]0.6120.8770.9330.5560.6190.9160.3860.7270.7750.8520.693
    本文0.3030.8660.9030.3400.3700.9620.5550.4690.8830.8590.687
    下载: 导出CSV

    表  3  不同模块对实验结果的影响

    Table  3.   Effects of different modules for the proposed method on the Vaihingen dataset

    边缘卷积多尺度高程感知OAAvg F1
    0.8150.662
    0.8260.671
    0.8340.674
    0.8590.687
    注:“√”表示测试网络集成了该模块。
    下载: 导出CSV
  • [1] ZERMAS D, IZZAT I, PAPANIKOLOPOULOS N. Fast segmentation of 3D point clouds: A paradigm on LiDAR data for autonomous vehicle applications[C]//2017 IEEE International Conference on Robotics and Automation. Piscataway: IEEE Press, 2017: 5067-5073.
    [2] YANG B S, HUANG R G, LI J P, et al. Automated reconstruction of building LoDs from airborne LiDAR point clouds using an improved morphological scale space[J]. Remote Sensing, 2017, 9(1): 14-27. doi: 10.3390/rs9010014
    [3] POLEWSKI P, YAO W, HEURICH M, et al. Detection of fallen trees in ALS point clouds using a Normalized Cut approach trained by simulation[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2015, 105: 252-271. doi: 10.1016/j.isprsjprs.2015.01.010
    [4] ENE L T, NÆSSET E, GOBAKKEN T, et al. Large-scale estimation of change in aboveground biomass in miombo woodlands using airborne laser scanning and national forest inventory data[J]. Remote Sensing of Environment, 2017, 188: 106-117. doi: 10.1016/j.rse.2016.10.046
    [5] TCHAPMI L, CHOY C, ARMENI I, et al. SEGCloud: Semantic segmentation of 3D point clouds[C]//2017 International Conference on 3D Vision. Piscataway: IEEE Press, 2018: 537-547.
    [6] HU Q, YANG B, XIE L, et al. RandLA-Net: Efficient semantic segmentation of large-scale point clouds[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2020.
    [7] LI Y Y, BU R, SUN M, et al. PointCNN: Convolution on χ-transformed points[C]//Advances in Neural Information Processing Systems. San Francisco: Curran Associates Inc, 2018: 820-830.
    [8] COLGAN M, BALDECK C, FÉRET J B, et al. Mapping savanna tree species at ecosystem scales using support vector machine classification and BRDF correction on airborne hyperspectral and LiDAR data[J]. Remote Sensing, 2012, 4(11): 3462-3480. doi: 10.3390/rs4113462
    [9] WANG C S, SHU Q Q, WANG X Y, et al. A random forest classifier based on pixel comparison features for urban LiDAR data[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2019, 148: 75-86. doi: 10.1016/j.isprsjprs.2018.12.009
    [10] 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. doi: 10.1016/j.isprsjprs.2013.11.001
    [11] SU H, MAJI S, KALOGERAKIS E, et al. Multi-view convolutional neural networks for 3D shape recognition[C]//2015 IEEE International Conference on Computer Vision. Piscataway: IEEE Press, 2016: 945-953.
    [12] WANG L, HUANG Y C, SHAN J, et al. MSNet: Multi-scale convolutional network for point cloud classification[J]. Remote Sensing, 2018, 10(4): 612. doi: 10.3390/rs10040612
    [13] CHARLES R Q, HAO S, MO K C, et al. PointNet: Deep learning on point sets for 3D classification and segmentation[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2017: 77-85.
    [14] QI C R, YI L, SU 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. New York: ACM, 2017: 5105–5114.
    [15] WANG Y, SUN Y B, LIU Z W, et al. Dynamic graph CNN for learning on point clouds[J]. ACM Transactions on Graphics, 2019, 38(5): 1-12.
    [16] HSU P H, ZHUANG Z Y. Incorporating handcrafted features into deep learning for point cloud classification[J]. Remote Sensing, 2020, 12(22): 3713. doi: 10.3390/rs12223713
    [17] LI W Z, WANG F D, XIA G S. A geometry-attentional network for ALS point cloud classification[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2020, 164: 26-40. doi: 10.1016/j.isprsjprs.2020.03.016
    [18] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2): 318-327. doi: 10.1109/TPAMI.2018.2858826
    [19] CHEN Y, LIU G L, XU Y M, et al. PointNet++ network architecture with individual point level and global features on centroid for ALS point cloud classification[J]. Remote Sensing, 2021, 13(3): 472. doi: 10.3390/rs13030472
    [20] 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): 3347-3357. doi: 10.3390/s18103347
    [21] 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. doi: 10.1080/13658816.2018.1431840
  • 加载中
图(11) / 表(3)
计量
  • 文章访问数:  198
  • HTML全文浏览量:  59
  • PDF下载量:  26
  • 被引次数: 0
出版历程
  • 收稿日期:  2021-08-02
  • 录用日期:  2021-10-29
  • 网络出版日期:  2021-11-22
  • 整期出版日期:  2023-06-30

目录

    /

    返回文章
    返回
    常见问答