基于无人机遥感时空谱分析的松材线虫病受害木早期定位与识别

Bursaphelenchus xylophilus is an invasive alien species, pine wilt disease (PWD) is caused by it and always be called the ‘cancer’ of pine trees because of its rapid spread and difficult to control. In recent years, there are more and more new PWD epidemic areas in China, and the scope of the epidemic areas is gradually expanding, resulting in great economic and ecological losses. Thus, timely and accurate disease identification and monitoring must be solved imperatively. Focus on the problems in forest diseases and pests monitoring using remote sensing technology: the disease information cannot be acquired timely, and it is difficult to realize the precise positioning and identification of damaged trees because satellite remote sensing is greatly affected by weather and has low spatial resolution. Based on the life cycle of Bursaphelenchus xylophilus and its vector insects, the time-series UAV-based hyperspectral and digital images are utilized as the main data sources in this project. Combined with field survey data and developed ‘temporal-spatial-spectral’ fusion and correlation algorithm, the main work of this project are as follows: (1) establishing the indicative spatial-spectral features for each stage of disease occurrence; (2) determining the time point of early disease identification; and (3) realizing the individual tree precise positioning and identification of early damaged trees and dead trees from two directions finally: spatial-spectral features extraction and classification framework construction. The successfully implementation of this project will provide the technology support for establishing an effective, timing and positioning UAV-based identification system of forest diseases and pests, and provide the basis for timely and effective prevention and control of PWD.

松材线虫为外来入侵物种，其引起的病害传播快速且防治难度大，被称为松树的“癌症”。近年来，在我国不断有新增疫区出现且疫区范围逐步扩大，造成了极大的经济和生态损失。及时准确的病害监测成为亟待解决的问题。针对当下存在的病害信息获取不及时，星载遥感数据空间分辨率低且受天气影响大，难以实现受害木的准确定位及识别等问题，本项目利用无人机搭载成像高光谱及高清数码相机，依据松材线虫以及媒介天牛的生命周期，获取病害主要疫区时间序列无人机遥感数据，研究“时-空-谱”数据融合及相关分析算法，实现病害发生各阶段指示性特征图谱的构建和病害早期识别时间点的确定，并在此基础上从空谱特征提取和分类框架构建两个方向进行早期受害木以及病害枯死木的识别及精确定位研究。为建立有效的定数据、定时间、定位置的森林病虫害无人机监测技术提供支撑，同时为松材线虫病的实际防控工作提供依据。

作为气候变化背景下具有快速传播趋势的检疫性疾病，近年来，松材线虫病害发生面积不断扩大，发生地区呈现出向北移动的态势，弄清松材线虫病害发生与无人机遥感数据的协作机理，发现并构建有效的分阶段识别特征，实现基于遥感影像的病害木定位，将更有利于减小病害在遥感识别中的不确定性，从而为保持森林健康和高生产力提供经验和方法，兼具理论意义和实际价值。本项目在辽宁、山东、安徽分别开展实地调查和试验，从松材线虫病害发生光谱响应机理出发，从时间、空间及光谱多个层面开展深入研究，回答松材线虫病害识别定数据、定时间和定位置的核心关键问题。项目主要研究内容及重要结果集中在以下几个方面：（1）分别在山东威海、辽宁抚顺和安徽巢湖开展“叶片—林分”尺度的松材线虫危害木调查及无人机数据获取试验，并在安徽省六安市开展区域尺度灾害发生调查试验，收集对应区域长时间序列Landsat数据，形成多源多模松材线虫病害识别数据资源，为松材线虫病害识别、监测及预警工作提供必要的数据支撑；（2）在“针叶—单木—区域”三个尺度上探明松材线虫侵染寄主植被的生理生化参数变化及确整个侵染过程中的光谱响应机理，通过光谱特征分析、“空-谱”融合算法及时间序列数据分析技术明确多尺度松材线虫病害等级划分的时空谱特征，实现松材线虫危害木识别数据推荐；（3）形成基于寄主植被生理生化参数融合空谱特征的病害发生响应机制，明确松材叶绿素b和含水量可以作为松材线虫病害早期识别的重点关注参数，同时通过空谱时特征在时间尺度上的变化规律研究，进一步明确接种后三周可较好的实现松材线虫危害木早期识别；（4）针对松材线虫危害木定位识别，分别从“时-空-谱”数据融合框架、深度学习算法优化及区域病害发生反演模型构建三个层面入手，实现了多尺度松材线虫危害木的定位识别。