采煤沉陷积水区漂浮水稻水上部生物量多光谱反演  

Multispectral Inversion of Floating Rice Biomass in Coal Mining Subsidence Water Area

作　　者：刘硕 陈晓辉 蔡跃章 胡睿鑫 曹正权 张世文[3] LIU Shuo;CHEN Xiaohui;CAI Yuezhang;HU Ruixin;CAO Zhengquan;ZHANG Shiwen(School of Spatial Informatics and Geomatics Engineering,Anhui University of Science and Technology,Huainan Anhui 232001,China;Resource and Environmental Protection Department,Huaihe Energy Group Corporation,Huainan Anhui 232001,China;School of Earth and Environment,Anhui University of Science and Technology,Huainan Anhui 232001,China)

机构地区：[1]安徽理工大学空间信息与测绘工程学院,安徽淮南232001 [2]淮河能源集团公司资源环保部,安徽淮南232001 [3]安徽理工大学地球与环境学院,安徽淮南232001

出　　处：《安徽理工大学学报(自然科学版)》2024年第6期64-70,共7页Journal of Anhui University of Science and Technology:Natural Science

基　　金：采煤地表沉陷区水面种植关键技术及模式研究与示范(HX2024012395);部省合作试点项目(2023ZRBSHZ002);安徽省自然资源科技项目(2023-K-6)。

摘　　要：目的生物量是判断作物生长发育的重要指标,为了对生长阶段作物生物量进行快速、准确、无损遥感监测,方法以采煤沉陷区种植的漂浮水稻为对象,使用无人机获取分蘖期多光谱影像,提取冠层波段反射率,构建反映水稻长势状况的14个植被指数,并分析其与水上部分生物量实测数据相关性,以显著性检验P<0.01的植被指数作为输入量,基于岭回归、随机森林(RF)和BP神经网络(BPNN),建立漂浮水稻生物量估算模型,并通过对比分析,确定最佳估算模型。结果试验表明,在水稻分蘖期阶段选取的植被指数中转换叶绿素吸收指数(TCARI)和土壤调整植被指数(SAVI)显著性最强且稳定,相关系数分别达到0.663(P<0.01)和0.627(P<0.01)。结论随机森林模型用于反演水面漂浮水稻分蘖期的生物量最佳,其在水稻分蘖期的验证集R2和RMSE分别为0.765%和4.275%。Objective In order to carry out rapid,accurate and non-destructive remote sensing monitoring of crop biomass at the growth stage,Methods The reflectance of the canopy band was extracted from the multispectral images of the tillering stage obtained by the UAV over the floating rice planted in the coal mining subsidence area to construct 14 vegetation indices reflecting the growth status of the rice,analyzing their correlation with the measured data of some biomass on the water.With the vegetation index of P<0.01 of the significance test as the input and based on the ridge regression,random forest(RF)and BP Neural Network(BPNN)floating rice biomass estimation models were established and the best one was determined through the comparative analysis.Results It was found that the conversion of chlorophyll uptake index(TCARI)and soil adjusted vegetation index(SAVI)were the most significant and stable among the vegetation indices selected at the tillering stage of rice,with the correlation coefficients of 0.663(P<0.01)and 0.627(P<0.01)respectively.Conclusion The random forest model is the best when used to invert the biomass at the tillering stage of floating rice on the water surface,with its validation set R20.765 and RMSE 4.275%.

关 键 词：生物量 光谱反射率 植被指数 多光谱遥感 漂浮水稻 

分 类 号：X832[环境科学与工程—环境工程]