联合FOD和CWT的水稻稻纵卷叶螟遥感监测  

Remote Sensing Monitoring of Cnaphalocrocis Medinalis(Guene)Based on FOD and CWT

作　　者：张运 茆红 方浩帆 孙李家旺 ZHANG Yun;MAO Hong;FANG Haofan;SUN Lijiawang(School of Geography and Tourism,Anhui Normal University,Wuhu,Anhui 241002,China;Resource Environment and Geography Information Engineering Anhui Engineering Technology Research Center,Wuhu,Anhui 241002,China)

机构地区：[1]安徽师范大学地理与旅游学院,安徽芜湖241002 [2]资源环境与地理学信息工程安徽省工程技术研究中心,安徽芜湖241002

出　　处：《遥感信息》2024年第5期20-28,共9页Remote Sensing Information

基　　金：安徽省重点研究与开发计划(202104g01020004);安徽省科技重大专项(202104a07020002、202003a06020002);安徽省2022年高校优秀青年人才支持计划重点项目(13);安徽省高等学校科学研究重点项目(2023AH050137)。

摘　　要：针对高光谱数据存在数据冗余、信息利用率低等缺点,采用分数阶微分和连续小波变换两种光谱细化方法,以提高光谱有效信息的利用率,并在现有光谱指数基础上构建分数阶微分光谱指数,结合连续小波变换选取的特征波段,基于PSO-SVR算法,实现水稻稻纵卷叶螟严重程度遥感监测。结果显示:水稻冠层光谱反射率和稻纵卷叶螟危害的水稻受害程度最大相关系数位于1.2阶处,与受害程度相关性最好的分数阶微分光谱指数是DI和GDI;在连续小波变换提取特征时,有效光谱信息主要集中在低分解尺度;分数阶微分光谱指数和小波特征的组合特征建模精度最高,R 2=0.861,RMSE=0.059,MAE=0.044。研究表明,FOD和CWT可以有效挖掘光谱中的细微信息,为实现水稻稻纵卷叶螟的精准监测提供依据。In response to the shortcomings of data redundancy and low information utilization in hyperspectral data,this article adopts two spectral refinement methods,which are fractional order differentiation and continuous wavelet transform,to improve the utilization of effective spectral information.It constructs a fractional order differential spectral index based on the existing spectral index,and combines it with the feature bands selected by continuous wavelet transform to achieve remote sensing monitoring of the severity of Cnaphalocrocis Medinalis(Guene)using the PSO-SVR algorithm.The results show that the maximum correlation coefficient between the spectral reflectance of rice canopy and the degree of rice damage caused by the Cnaphalocrocis Medinalis(Guene)is at the order of 1.2,and the best correlation between the fractional order differential spectral index and the degree of damage is found in DI and GDI.When extracting features using continuous wavelet transform,the effective spectral information is mainly concentrated at low decomposition scales.In addition,the combination of fractional differential spectral index and wavelet features has the highest modeling accuracy,with R 2=0.861,RMSE=0.059,and MAE=0.044.It indicates that FOD and CWT can effectively mine subtle information in the spectrum,providing a basis for precise monitoring of Cnaphalocrocis Medinalis(Guene).

关 键 词：机载高光谱 病虫害 微分光谱 光谱指数 粒子群优化算法 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]