基于旋翼无人机低空遥感的高原鼠兔危害等级划分技术研究

高原鼠兔是青藏高原特有的小型啮齿动物，是高寒草地生态系统的关键种。科学合理防控高原鼠兔危害，对其危害程度开展监测和评价十分必要。通过无人机空中遥感平台并结合地面调查，以甘肃省玛曲县高原鼠兔危害草地为研究对象，选择不同危害程度的典型样地，采用自主设计的软件提取无人机影像中高原鼠兔洞口数、地上生物量、草地盖度等指标；构建了由洞口数、地上生物量、盖度和可食牧草比例组成的高原鼠兔危害等级综合指标体系，并对研究区样地进行危害等级划分。最终形成一套适用于高原鼠兔危害的低空遥感监测方法，可为后期高效、准确监测高原鼠兔危害提供有力的技术支持。主要结果如下：1）经实地人工调查结果验证，无人机影像洞口数量和植被盖度解译精度分别达到95%和93%。2）筛选了可见光波段差值植被指数VDVI作为估测地上生物量的最佳植被指数，并构建了估测模型。经实地调查验证平均精度为86.54%。3）构建了综合危害指数DRI，并提出了划分危害级别的各指标量化值。通过聚类分析将研究区鼠害地划分为5个等级，即I-无危害草地，Ⅱ-轻度危害草地，Ⅲ-中度危害草地，Ⅳ-重度危害草地，V-极度危害草地。

A study of UAV remote sensing technology for classifying the level of plateau pika damage to alpine rangeland

The plateau pika is a unique small mammal and a key species among the fauna of the alpine rangeland ecosystem on the Qinghai Tibetan Plateau. Therefore， there is a need to monitor and evaluate the level of damage to alpine rangeland caused by the plateau pika. In this study， unmanned aerial vehicle （UAV） technology validated by field observation was used to monitor and evaluate the level of plateau pika damage to rangeland in Maqu county of Gansu Province. Several variables， including the number of pika holes， aboveground biomass， vegetation cover， and proportion of edible forage in the aboveground biomass， among others， were extracted from UAV images by self-designed software. Then a multi-factor index system， comprising the number of pika holes， aboveground biomass， vegetation cover and proportion of edible forage was developed to categorize the degree of damage by plateau pika. Finally， a low altitude remote sensing monitoring methodology suitable for collecting data on plateau pika damage was devised. The system is ready for implementation as a technical tool for efficient and accurate monitoring of plateau pika damage. The main results were： 1） The accuracy of photo interpretation for the pika hole number and vegetation cover was 95% and 93%， respectively， based on field verification. 2） The visible-band difference vegetation index （VDVI） was selected as the best vegetation index for estimating aboveground biomass； for the estimation model constructed， the accuracy was 86.54%. 3） A damage scale was formulated to classify plateau pika damage in the study area （damage to rangeland index， DRI） into five levels as follows： Level I， no damage； Level Ⅱ， light damage； Level Ⅲ， moderate damage； Level Ⅳ， severe damage and Level V， extreme damage.