当前位置: X-MOL 学术Precision Agric. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Canopy defoliation by leaf-cutting ants in eucalyptus plantations inferred by unsupervised machine learning applied to remote sensing
Precision Agriculture ( IF 5.4 ) Pub Date : 2022-06-15 , DOI: 10.1007/s11119-022-09919-x
Alexandre dos Santos , Isabel Carolina de Lima Santos , Jeffersoney Garcia Costa , Zakariyyaa Oumar , Mariane Camargo Bueno , Tarcísio Marcos Macedo Mota Filho , Ronald Zanetti , José Cola Zanuncio

Defoliation by leaf-cutting ants alters the physiological processes of plants, and this defoliation can be inferred from satellite imagery used to identify plant injuries. The aim of this study was to evaluate the spectral pattern of defoliation by leaf-cutting ants in eucalyptus plants on a pixel level using unsupervised machine learning techniques applied to remote sensing by satellites. The study was carried out in a eucalyptus plantation in the municipality of Telêmaco Borba, Paraná state, Brazil. The nests of leaf-cutting ants were located and georeferenced. Multispectral images were obtained from the Sentinel-2 (S-2) and planet scope (PS) satellites. The response variables were the RGB-NIR bands and four vegetation indices (VIs). The data obtained from these bands and vegetation indices was separated in an unsupervised method by the k-medoids clustering algorithm and input into a Random Forest (RF) model. The significance of the models was tested with permutational multivariate analysis of variance (PERMANOVA). The k-medoids algorithm classified the spectral response of the RGB-NIR and VIs bands into two main factors of variation in the tree canopy. The models selected were 1200 trees and 6 variables for the S2 satellite (accuracy = 97.74 ± 0.040%) and 900 trees and 5 variables for the PS (accuracy = 97.42 ± 0.026%). The unsupervised machine learning technique, applied to remote sensing, was effective to map defoliation caused by leaf-cutting ants, and this approach can be used in precision agriculture for pest management purposes.



中文翻译:

通过应用于遥感的无监督机器学习推断桉树人工林中切叶蚁的冠层落叶

切叶蚂蚁的落叶改变了植物的生理过程,这种落叶可以从用于识别植物损伤的卫星图像中推断出来。本研究的目的是使用应用于卫星遥感的无监督机器学习技术,在像素水平上评估桉树植物中切叶蚂蚁的落叶光谱模式。该研究是在巴西巴拉那州 Telêmaco Borba 市的一个桉树种植园进行的。切叶蚁的巢穴被定位和地理参考。多光谱图像是从 Sentinel-2 (S-2) 和行星范围 (PS) 卫星获得的。响应变量是 RGB-NIR 波段和四个植被指数 (VI)。从这些波段和植被指数获得的数据通过 k-medoids 聚类算法以无监督的方法分离,并输入到随机森林 (RF) 模型中。使用置换多变量方差分析 (PERMANOVA) 测试模型的显着性。k-medoids 算法将 RGB-NIR 和 VIs 波段的光谱响应分类为树冠变化的两个主要因素。选择的模型是 S2 卫星的 1200 棵树和 6 个变量(准确度 = 97.74 ± 0.040%)和 PS 的 900 棵树和 5 个变量(准确度 = 97.42 ± 0.026%)。应用于遥感的无监督机器学习技术有效地绘制了由切叶蚂蚁引起的落叶图,这种方法可用于精准农业中的害虫管理。

更新日期:2022-06-16
down
wechat
bug