A machine learning algorithm to detect pine wilt disease using UAV-based hyperspectral imagery and LiDAR data at the tree level

基于无人机 (UAV) 的高光谱成像 (HI) 和光探测和测距（LiDAR）技术是一种有效的森林健康监测方法。然而，以往的研究很少使用机载 HI 和 LiDAR 检测松材线虫病（PWD）并比较预测 PWD 不同感染阶段单木受害阶段的能力。本文将 PWD 感染分为五个阶段（绿色、早期、中期、重度和灰色），以及结合HI 和 LiDAR 数据来检测 PWD。我们使用随机森林 (RF) 算法评估了高光谱方法的精度（仅HI 数据）、LiDAR（仅 LiDAR 数据）及其组合（HI 加 LiDAR 数据）来预测单木受PWD感染的不同感病阶段。

Unmanned Aerial Vehicle (UAV)-based hyperspectral imaging (HI) and Light Detection and Ranging (LiDAR) technologies are effective approaches for forest health monitoring. However, previous studies have rarely used airborne HI and LiDAR to detect Pine Wilt Disease (PWD) and compare their capabilities in predicting individual tree infection stages of PWD at different severity levels. This study divides PWD infection into five stages (green, early, middle, severe, and grey), and combines HI and LiDAR data for PWD detection. We employed the Random Forest (RF) algorithm to evaluate the performance of three approaches: the hyperspectral method using only HI data, the LiDAR method using only LiDAR data, and their combined approach using both HI and LiDAR data, for predicting individual tree infection stages of PWD at different severities.

搜集汇总

背景与挑战

背景概述

该数据集利用无人机高光谱成像和LiDAR技术，结合随机森林算法，检测松材线虫病（PWD）的五个感染阶段（绿色、早期、中期、重度和灰色）。数据属于人工林重大灾害防控关键技术研究项目，由北京林业大学团队负责，数据量为9.75 MB。