文献类型：期刊文献

中文题名：基于无人机多光谱影像的云南切梢小蠹危害监测反演研究

英文题名：Harm Monitoring and Inversion Study on Tomicus yunnanensis Based on Multi-spectral Image of Unmanned Aerial Vehicle

作者：马云强[1,2] 李宇宸[3] 刘梦盈[2] 石雷[2] 张军[3] 张忠和[2]

第一作者：马云强

机构：[1]西南林业大学生物多样性保护学院/云南省森林灾害预警与控制重点实验室,云南昆明650224;[2]中国林业科学研究院资源昆虫研究所,云南昆明650224;[3]云南大学资源环境与地球科学学院,云南昆明650504

年份：2021

卷号：34

期号：9

起止页码：1878-1884

中文期刊名：西南农业学报

外文期刊名：Southwest China Journal of Agricultural Sciences

收录：CSTPCD;;北大核心:【北大核心2020】；CSCD:【CSCD_E2021_2022】；

基金：国家重点研发计划课题“林业有害生物检测、监测与预警关键技术”子课题“云南松小蠹监测预警关键技术与GIS应用平台研究”(2018YFD0600201-4)。

语种：中文

中文关键词：无人机;多光谱;云南切梢小蠹;危害;反演

外文关键词：UAV(unmanned aerial vehicle);Multi-spectral image;Tomicus yunnanensis;Harm;Inversion

分类号：S771

摘要：【目的】探索如何利用多光谱遥感对云南切梢小蠹的危害状况进行快速准确的监测。【方法】基于多光谱无人机获取研究区影像结合实地采集样方数据,应用DNN深度学习模型定量反演研究区云南切梢小蠹危害信息,分析虫害等级光谱图像发现GNRE指数和NDVI指数对虫害造成的枯稍率的相关性。【结果】将两指数应用于模型拟合,对比结果发现植被指数NDVI与云南切梢小蠹危害的枯梢率相关性为R^(2)=0.67696,高于GNRE的R^(2)=0.45331,且危害等级分类总体分类精度为71.11%,Kappa系数为0.6751,危害等级由高到低的面积占比分别为1.18%、8.60%、73.73%、16.47%。【结论】利用无人机影像结合深度学习技术,可准确得到切梢小蠹分布信息,且危害等级分布呈现出由道路以及树林密度较小区域向密度较大区域其危害性逐渐减小,可为后续大范围切梢小蠹危害云南松的监测与防治提供参考。
【Objective】The present paper aimed to study how to monitor the damage of Pinus yunnanensis Tomicus yunnanensis quickly and accurately.【Method】The paper used the hyperspectral UAV to obtain the image of the research area combined with the field sampling data,and the DNN deep learning model to quantitatively retrieve the pest information in the research area.【Result】It was found that there was a significant correlation between GNRE index and NDVI index.The results showed that the correlation between NDVI and the damage rate of Tomicus yunnanensis was R^(2)=0.67696,which was higher than that of GNRE(R^(2)=0.45331),and the overall classification accuracy was 71.11%,kappa coefficient was 0.6751,the area ratio of hazard grade from high to low was 1.18%,8.60%,73.73%and 16.47%,respectively.【Conclusion】Using UAV image and deep learning technology,the distribution information of Tomicus yunnanensis could be accurately obtained,which showed a gradual decrease from the area with low density of road and forest to the area with high density.So it could be used for the subsequent large-scale Tomicus yunnanensis monitoring which could provide reference for measurement and protection.