多光谱激光点云卷积神经网络的地物分类研究Study on object classification for multispectral LiDAR point clouds based on convolutional neural network
邹晓亮
摘要(Abstract)：
针对商用激光传感器Optech LiDAR Titan系统获取的多光谱激光点云数据进行地物分类试验的探索,该文提出一种基于卷积神经网络(CNN)模型的多光谱激光LiDAR点云数据地物分类方法。新数据源多光谱激光点云具有多通道和多次散射回波的典型特性,生成感兴趣的热力图,根据热力图特征值和nDSM辅助数据进行感兴趣地物分类。采用CNN模型学习结果与面向对象影像分析OBIA分类方法相结合对分类结果进行精化,并用随机采样参考点对地物分类结果进行精度评估,解决CNN模型分类的正确性和可靠性问题。实验表明,地物分类整体精度OA达到89.8%,Kappa值0.858,该方法在多光谱激光点云地物分类方面具有稳健性、有效性和通用性。
关键词(KeyWords)： 多光谱激光点云;卷积神经网络;热力图;样本标注;地物分类;精度评估
基金项目(Foundation): 地理信息工程国家重点实验室项目(SKLGIE2016-M-3-3)
作者(Author): 邹晓亮
DOI: 10.16251/j.cnki.1009-2307.2021.07.007
参考文献(References)：
- [1]Titan Brochure and Specifications.Optech titan multispectral LiDAR system-high precision environmental mapping[EB/OL].(2015-12-16)[2020-06-15].http：∥www.teledyneoptech.com/wp-content/uploads/TitanSpecsheet-150515-WEB.pdf．
- [2]BAKULA K.Multispectral airborne laser scanning-a new trend in the development of LiDAR technology[EB/OL].[2020-05-15].https：∥www.researchgate.net/publication/296486863_Multispectral_airborne_laser_scanning_-_a_new_trend_in_the_development_of_LiDAR_technology．
- [3]WICHMANN V,BREMER M,LINDENBERGER J,et al.Evaluating the potential of mutispectral airborne LiDAR for topographic mapping and land cover classification[EB/OL].[2020-05-15].https：∥www.isprs-ann-photogramm-remote-sens-spatial-inf-sci.net/II-3-W5/113/2015/．
- [4]SHAKER A,LAROCQUE P,SMITH B.The optech titan:a multi-spectral LiDAR for multiple applications[EB/OL].[2020-05-15].http：∥shoals.sam.usace.army.mil/Workshop Files/2015/Day_01_pdf/1130_LaRocque.pdf．
- [5]ZOU X L.ZHAO G H,LI J,et al.3Dland cover classification based on multispectral LiDAR point clouds[C]∥2017IEEE International Geoscience and Remote Sensing Symposium.Fort Worth,TX,USA:IEEE,2016:741-747．
- [6]EKHTARI N,CRAIG G,CARLOS F D G.Classification of airborne multispectral LiDAR point clouds for land cover mapping[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2018,11:2068-2078．
- [7]PAN S,GUAN H,YU Y,et al.A comparative landcover classification feature study of learning algorithms:DBM,PCA,and RF using multispectral LiDAR data[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2019,12 (4):1314-1326．
- [8]LECUN Y,BOTTOU L,HAFFNER P.Gradient-based learning applied to document recognition[J].Proceedings of the IEEE,1998,86(11):2278-2324．
- [9]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]∥Proceedings of the 25th International Conference on Neural Information Processing Systems.Lake Tahoe,Nevada:ACM,2012:1097-1105．
- [10]JIA Y P,SHELHAMER E.Caffe:convolutional architecture for fast feature embedding[EB/OL].[2020-06-15].https：∥arxiv.org/pdf/1408.5093.pdf．
- [11]ABADI M,AGARWAL M,BARHAM P,et al.Tensorflow:large-scale machine learning on heterogeneous distributed systems[EB/OL].[2020-05-15].https：∥Cornell University Library/arXiv.org/cs/arXiv:1603.04467v1/1603.04467v1.pdf．
- [12]龚健雅，季顺平．摄影测量与深度学习[J]．测绘学报，2018,47(6):693-704.(GONG Jianya,JI Shunping.Photogrammetry and deep learning[J].Acta Geodaetica et Cartographica Sinica,2018,47(6):693-704．）
- [13]曹爽，潘锁艳，管海燕．机载多光谱LiDAR的随机森林地物分类[J]．测绘通报，2019(11):79-84.(CAOShuang,PAN Suoyan,GUAN Haiyan.Random forestbased land-use classification using multispectral LiDARdata[J].Bulletin of Surveying and Mapping,2019(11):79-84．）
- [14]JENSEN J R.Introductory digital image processing:a remote sensing perspective[M].Second edition.Upper Saddle River,New Jersey,USA:Prentice Hall,1996:247-252．
- [15]RUSSELL G.CONGALTON.A review of assessing the accuracy of classifications of remotely sensed data[J].Remote Sensing of Environment,1991,37(1):35-46.