山西农业大学学报（自然科学版）2024，Vol.44Issue(1)：120-129,10.DOI:10.13842/j.cnki.issn1671-8151.202308034
基于无人机高光谱影像的水稻叶片SPAD值反演方法研究
Inversion method of rice leaf SPAD values based on UAV hyperspectral imagery
摘要
Abstract
[Objective]Achieving efficient and non-destructive monitoring of chlorophyll content in rice through UAV hyperspec-tral imagery is an important means for the development of modern precision agriculture.This study investigated different pre-processing methods and their combinations for raw spectral data of rice leaves,constructed models using different spectral pa-rameters,and obtained the optimal inversion model for relative chlorophyll content(SPAD value)of rice leaves in the study ar-ea.This research aimed to provide references for efficient and non-destructive monitoring of chlorophyll content in rice leaves.[MethodsThe rice fields in Yaodu Town,Qingbaijiang District,Chengdu City,Sichuan Province were selected as the research object.The SPAD value of rice leaves and hyperspectral reflectance in the range of 500-900 nm were measured.Different pre-processing methods,including first order differentiation(D1),Savitzky-Golay convolution smoothing(SG smoothing),stan-dard normal transformation(SNV),multiple scattering correction(MSC),and their combinations were applied to the original reflectance data.The characteristic bands with p<0.1 were selected as the first spectral parameter.Principle component analy-sis(PCA)was then conducted on the basis of the characteristic bands to obtain the second set of spectral parameters.These two parameters were used as input variables of the Extra Trees models to establish inversion model of rice chlorophyll content in the study area.[Results]Compared with models constructed using feature bands selected based on correlation coefficients,models constructed with PCA for dimensionality reduction of the feature bands achieved higher modeling accuracy.Specifical-ly,R2 values for ET_D1 and ET_SG_MSC increased from 0.769 and 0.782 to 0.793 and 0.825,representing improvements of 3% and 5.5%,respectively.The R2 value for ET_SG_SNV increased from 0.754 to 0.796,showing an improvement of 5.6% .Among the models,ET_PCA_features_SG_MSC demonstrated the highest accuracy,with R2 and RMSE values of 0.825 and 0.984,respectively,making it the optimal inversion model for SPAD values of rice leaves in the study area.[Con-clusion]The results of this study provided a reference and basis for achieving efficient and accurate monitoring of chlorophyll content in rice leaves.关键词
水稻/叶绿素/高光谱/相关系数/主成分分析/Extra TressKey words
Rice/Chlorophyll/Hyperspectral/Correlation coefficient/Principal component analysis/Extra Tress分类
农业科技引用本文复制引用
谢东,何敬,何嘉晨,王彬,林远杨,刘刚..基于无人机高光谱影像的水稻叶片SPAD值反演方法研究[J].山西农业大学学报（自然科学版）,2024,44(1):120-129,10.基金项目
自然资源部成都平原国土生态与土地利用野外科学观测研究站开放基金(CDORS-2023-04) （CDORS-2023-04）
成都市技术创新研发项目(2022-YF05-01090-SN) （2022-YF05-01090-SN）
成都理工大学研究生质量工程项目(2022YJG022) （2022YJG022）
