中国森林病虫2025，Vol.44Issue(3)：9-15,7.DOI:10.19688/j.cnki.issn1671-0886.20250005
基于Sentinel-2影像的落叶松毛虫虫害遥感监测研究
Remote sensing monitoring of Dendrolimus superans based on Sentinel-2 image
摘要
Abstract
To quickly and accurately obtain the occurrence information of Dendrolimus superans,Saihanba Mechanized Forest Farm of Hebei was taken as research area,and Sentinel-2 images were used to extract spectral,vegetation and texture features,and the sensitive features related to occurrence area and degree of insect pests were optimized.Random forest(RF)and extreme gradient boosting(XGBoost)were used to identify occurrence area of insect pests.Pest occurrence grades were classified based on ensemble learning(XGBoost and CatBoost)and supervised learning(Mahalanobis distance and minimum distance)models.The results showed that:the red-edge band(B5,B6,B7),near-infrared band(B8,B8a)and short wave infrared band(B11)of Sentinel-2 remote sensing image had higher response to the occurrence of D.superans;The overall classification accuracy(OA)of the XGBoost model based on the first 12 features was 92.03%,and Kappa coefficient was 82.51%;The OA of Mahalanobis distance,minimum distance,XGBoost and CatBoost were 80.00%,94.30%,99.15%and 98.23%,respectively,Kappa coefficients were 72.97%,92.31%,98.84%and 97.60%,respectively;XGBoost algorithm had the best effect on grade classification for pest damage.XGBoost algorithm could be selected to construct the pest monitoring model,which could provide references for monitoring of D.superans.关键词
落叶松毛虫/Sentinel-2/虫害遥感监测/特征优选Key words
Dendrolimus superans/Sentinel-2/remote sensing monitoring of insect pests/feature selection分类
林学引用本文复制引用
张玉姣,卢伟,杨晋宇,黄塞松,刘璇,王子瑾..基于Sentinel-2影像的落叶松毛虫虫害遥感监测研究[J].中国森林病虫,2025,44(3):9-15,7.基金项目
国家自然科学基金项目"塞罕坝地区华北落叶松人工林土壤食物网结构与功能及其对经营措施的响应"(31971651) （31971651）
国家重点研发计划课题"华北落叶松优质大径材高效培育技术"(2023YFD2200803) （2023YFD2200803）
