应用无人机多光谱遥感对栎类食叶虫害危害程度的监测

Application of UAV Multispectral Remote Sensing to Monitor Damage Level of Leaf-feeding Insect Pests of Oak

为了探讨无人机多光谱遥感在森林病虫害危害程度的监测效果,以河南省鲁山县王沟林区的栎类树种为研究对象,将栎类虫害危害程度分为4个等级(正常、轻度、中度、重度),按照危害等级设置样本点。使用4旋翼无人机(搭载5波段镜头)获取研究样地的多光谱遥感影像,在R 4.1.3平台进行方差分析,从不同等级样点的13个植被指数中筛选出差异显著的植被指数作为特征参数,分别采用最大似然法(MLC)、面向对象多尺度分割法(eCognition)、随机森林(RF)、支持向量机(SVM)构建分类模型并检验精度。结果表明:最大似然法、多尺度分割法、支持向量机、随机森林等分类方法的总体精度分别为77.0%、82.8%、86.2%、90.8%,Kappa系数分别为为0.693、0.770、0.816、0.877,随机森林模型分类效果最佳,能够满足对栎类虫害危害程度进行准确监测。

In order to explore the effectiveness of UAV multispectral remote sensing in monitoring the damage degree of forest pests and diseases, oak species in Wanggou forest area of Lushan County, Henan Province were used as the research object, and the damage degree of oak pests was divided into four classes (normal, mild, moderate and severe), and sample points were set according to the damage class. A 4-rotor UAV (equipped with a 5-band lens) was used to acquire multispectral remote sensing images of the study sample sites, and ANOVA was conducted on the R 4.1.3 platform to select vegetation indices with significant differences from 13 vegetation indices of different classes of sample sites as the characteristic parameters, using the maximum likelihood method (MLC), object-oriented multiscale segmentation (eCognition), random forest (RF), and Support vector machine (SVM) was used to construct classification models and check the accuracy. The results showed that the overall accuracies of the maximum likelihood method, multiscale partitioning method, support vector machine and random forest were 77.0%, 82.8%, 86.2% and 90.8%, the kappa coefficients were 0.693, 0.770, 0.816 and 0.877, respectively, and the random forest model had the best classification effect and accurately monitored the damage degree of oak pests.