基于无人机多光谱的耐旱苎麻品种筛选方法  

Screening of Drought⁃tolerant Ramie Based on UAV Multispectral Imagery

作　　者：付虹雨 王薇 卢建宁 岳云开 崔国贤[1] 佘玮[1] FU Hongyu;WANG Wei;LU Jianning;YUE Yunkai;CUI Guoxian;SHE Wei(College of Agriculture,Hunan Agricultural University,Changsha 410128,China)

机构地区：[1]湖南农业大学农学院,长沙410128

出　　处：《农业机械学报》2023年第4期206-213,共8页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家重点研发计划项目(2018YFD0201106);财政部和农业农村部:国家现代农业产业技术体系项目(CARS-16-E11);国家自然科学基金项目(31471543)。

摘　　要：高温干旱是影响作物生长及最终生产力的主要胁迫源。当前,无人机遥感技术已在作物倒伏和病虫害的分级监测研究中取得重大进展,但有关利用无人机遥感进行作物抗旱等级监测的研究却鲜有报道。因此,以苎麻种质资源为研究对象,提出了苎麻抗旱性量化标准,并提供了一种利用无人机多光谱遥感鉴定苎麻种质资源抗旱性的方法。首先,由专家对36份苎麻种质资源进行抗旱性分级;然后,结合无人机多光谱遥感获取的植被指数,采用随机森林(Random forest,RF)、支持向量机(Support vector machine,SVM)、决策树(Decision tree,DT)3种机器学习方法分别构建苎麻抗旱性鉴定模型,并通过苎麻在高温干旱胁迫下的表型响应检验鉴定结果;最后,基于无人机获取的遥感表型,筛选高温干旱胁迫下优质苎麻种质资源。结果表明,利用SVM构建的苎麻抗旱性鉴定模型正确率达到0.74,不同抗旱级分类F1得分范围为0.69~0.79,说明该方法能用于苎麻种质资源抗旱性评估。利用无人机遥感数据反演得到的3项苎麻表型性状(叶绿素相对含量、叶面积指数、株高)均与人工测量值具有较强的相关性,在此基础上,研究从高温干旱胁迫中筛选出了3个优质苎麻种质资源PJ CD、WS XM、湘苎7号。High temperature and drought are the main stress sources affecting crop growth and final productivity.At present,UAV remote sensing technology has made great progress in the hierarchical monitoring of crop lodging and pests and diseases,but there are few reports on the use of UAV remote sensing for crop drought resistance grade monitoring.Therefore,taking ramie germplasm resources as the research object,quantitative criteria for ramie drought resistance was proposed,and a method to identify the drought resistance of ramie germplasm resources was providedby multi⁃spectral remote sensing of UAV.Firstly,totally 36 ramie germplasm resources were graded for drought resistance by experts.Then,combined with the vegetation index obtained by UAV multi⁃spectral remote sensing,and three machine learning methods,random forest(RF),support vector machine(SVM)and decision tree(DT)were used to construct ramie drought resistance identification models,and the results were evaluated by testing the phenotypic response of ramie under high temperature and drought stress.Finally,high⁃quality ramie germplasm resources under high temperature and drought stress were screened based on the remote sensing phenotypes obtained by UAV.The results showed that the accuracy of the ramie drought resistance identification model constructed by SVM reached 0.74,and the F1⁃score of different drought resistance classes was ranged from 0.69 to 0.79,indicating that the method could be used to evaluate the drought resistance of ramie germplasm resources.Three phenotypic characters of ramie(SPAD value,leaf area index and plant height)obtained from UAV remote sensing data were strongly correlated with the measured values.On this basis,three high⁃quality ramie germplasm resources PJ CD,WS XM and Xiangzhu 7 were selected from high temperature and drought stress.

关 键 词：高温干旱胁迫 苎麻 抗性筛选 无人机遥感 

分 类 号：S423.4[农业科学—植物保护] S252.9