基于ICESat-2和Sentinel-2的河道管理范围内阻水植被提取研究  

Study on Water-blocking Vegetation Extraction within River Management Areas Based on ICESat-2 and Sentinel-2

作　　者：吴迪 袁晓宏 李冰 杨光 WU Di;YUAN Xiaohong;LI Bing;YANG Guang(Heilongjiang Geomatics Center of MNR,Harbin 150081,China;Heilongjiang Provincial Society of Surveying,Mapping and Geoinformation,Harbin 150081,China;Heilongjiang River and Lake Chief System Support Center,Harbin 150001,China)

机构地区：[1]自然资源部黑龙江基础地理信息中心,黑龙江哈尔滨150081 [2]黑龙江省测绘地理信息学会,黑龙江哈尔滨150081 [3]黑龙江省河湖长制保障中心,黑龙江哈尔滨150001

出　　处：《测绘与空间地理信息》2024年第12期14-18,共5页Geomatics & Spatial Information Technology

摘　　要：针对河道管理范围内阻水植被实地调查工作量大、耗时长、难度大等问题,本文利用星载雷达ICESat-2的ATL03和ATL08产品,通过关联分析的方法筛选和计算激光点的植被高度,结合Sentinel-2波谱变量,分别使用BP神经网络和随机森林回归算法,构建空间连续的植被高度信息反演模型,快速提取河道管理范围内的阻水植被。实验结果表明,随机森林回归模型在稳定性和准确性方面优于BP神经网络模型,能够更有效地提取河道管理范围内的阻水植被。本方法可提高河道管理中阻水植被调查的效率和准确性,为河道治理和管理工作提供科学的决策参考。This study addresses the issues of heavy workload,time consumption,and difficulty in field survey of water-blocking vegetation within river management areas.Utilizing the ATL03 and ATL08 products from the ICESat-2 satellite ATLAS,vegetation heights were screened and calculated through correlation analysis of laser points.Combined with spectral variables of Sentinel-2,spatially continuous vegetation height inversion models were constructed using BP neural network and random forest regression algorithm to rapidly extract water-blocking vegetation within river management areas.Experimental results indicate that the random forest regression model is superior to the BP neural network model in terms of stability and accuracy,effectively extracting water-blocking vegetation within river management areas.This method can improve the efficiency and accuracy of surveying water-blocking vegetation in river management,providing scientific decision-making support for river governance and management.

关 键 词：ICESat-2 BP神经网络 随机森林回归算法 阻水植被 

分 类 号：P237[天文地球—摄影测量与遥感]