基于随机森林和决策树的马尾松松材线虫病监测方法

Monitoring method of pine wood nematode disease based on random forest and decision tree

为了快速掌握森林病虫害的感病区域,及时有效地采取防治措施。文章以湖北省宜昌市夷陵区马尾松松材线虫病为背景,将地面调查数据和Landsat-8遥感影像作为基础数据源,结合随机森林和决策树cart算法,将七个波段原始光谱值作为输入,进行五次重复实验并计算平均检测精度和AUC值,提出了适用于湖北省宜昌市松材线虫病的遥感监测方法;同时,相对于传统的光谱指数分析法获取数据耗时耗力,文章提出的基于原始光谱建立病虫害监测模型的方法,平均精度在76%以上,能较好地实现森林病虫害的快速检测,对松材线虫病的防治具有一定的指导意义。

In order to quickly master the susceptible area of forest diseases and insect pests, timely and effective control measures are taken. In this paper, based on the background of Pinus massoniana wood nematode disease in Yiling District, Yichang City,Hubei Province, the ground survey data and Landsat-8 remote sensing image are used as the basic data source, combined with random forest and decision tree cart algorithm, the original spectral values of seven bands are used as inputs, five repeated experiments are carried out and the average detection accuracy and AUC value are calculated, and a remote sensing monitoring method suitable for pine wood nematode disease in Yichang City, Hubei Province is proposed. At the same time, compared with the traditional spectral index analysis method, the method of establishing pest surveillance model based on original spectrum is more than 76%. It can realize the rapid detection of forest diseases and insect pests, and has certain guiding significance for the control of pine wood nematode disease.