李彬彬, 谢欢, 童小华, 叶丹, 孙凯鹏, 李铭. 基于随机森林的ICESat-2卫星数据地表覆盖分类[J]. 红外与激光工程, 2020, 49(11): 20200292. DOI: 10.3788/IRLA20200292
引用本文: 李彬彬, 谢欢, 童小华, 叶丹, 孙凯鹏, 李铭. 基于随机森林的ICESat-2卫星数据地表覆盖分类[J]. 红外与激光工程, 2020, 49(11): 20200292. DOI: 10.3788/IRLA20200292
Li Binbin, Xie Huan, Tong Xiaohua, Ye Dan, Sun Kaipeng, Li Ming. Land cover classification using ICESat-2 data with random forest[J]. Infrared and Laser Engineering, 2020, 49(11): 20200292. DOI: 10.3788/IRLA20200292
Citation: Li Binbin, Xie Huan, Tong Xiaohua, Ye Dan, Sun Kaipeng, Li Ming. Land cover classification using ICESat-2 data with random forest[J]. Infrared and Laser Engineering, 2020, 49(11): 20200292. DOI: 10.3788/IRLA20200292

基于随机森林的ICESat-2卫星数据地表覆盖分类

Land cover classification using ICESat-2 data with random forest

  • 摘要: 该研究将ICESat-2卫星激光测高数据作为地表覆盖分类的新数据源,提出了一种基于随机森林的ICESat-2卫星地表覆盖分类方法,探索了光子计数卫星激光测高在地表覆盖分类中的应用潜力。该方法采用光子数目、不同类型光子水平和垂直分布比例、信噪比、太阳条件、大气条件作为分类的输入,并在中国长三角地区开展了多类地表覆盖类型分类实验进行了验证。实验结果表明,ICESat-2卫星的强波束和弱波束的激光数据在水体、森林、低植被以及城市/裸地四类地表的总体分类精度均能达到优于85%;在水体、森林以及低植被/城市/裸地三类地表的总体分类精度能达到优于90%的水平。

     

    Abstract: ICESat-2 data was considered as a new land cover classification data source, and a method was proposed to classify land cover using ICESat-2 data with random forest, to explore the application potential of the space-borne photon counting lidar in the land cover classification. The method used the photon number, the proportion of horizontal and vertical distribution of different types of photons, signal-to-noise ratio, solar conditions and atmospheric conditions as the input of classification, and was verified by the experiment of multi-category land cover in China's Yangtze River Delta. For four categories of water, forest, low vegetation and urban/barren, the classification results show that the overall accuracy of strong beam and weak beam is better than 85%. For three categories of water, forest, and low vegetation/urban/barren, the classification results show that the overall accuracy of strong beam and weak beam is better than 90%.

     

/

返回文章
返回