无人机影像识别白喉乌头的相对高程阈值法适用性分析

本文以新疆伊犁州新源县那拉提草原的典型毒害草白喉乌头为研究对象，通过PhotoScan软件对无人机航拍影像进行数据处理并获取研究区DSM数据，在此基础上，运用相对高程阈值法识别提取白喉乌头的最佳阈值范围，并验证其有效性。试验结果表明：①在无人机影像中，相对高程阈值法能够清晰准确地体现地物分布特征，适用于提取与普通牧草高度误差明显的地物；②当10 cm≤T<20 cm时，白喉乌头能够被准确识别，且与实际白喉乌头分布点基本一致，分类精度达91%；③相对高程阈值法能够较好地分离出白喉乌头，既提升了分类依据的可靠性，又实现了白喉乌头高精度识别。该方法可应用于实际的草原监测工作。

Applicability analysis of relative elevation threshold method in recognizing diphtheria aconitum in UAV images

Taking the typical poisonous grass diphtheria aconitum in the Nalati grassland of Xinyuan county, Ili Kazak Autonomous Prefecture, Xinjiang as the research object, the PhotoScan software is used to process the drone aerial images and obtain the DSM data of the study area. On this basis, the relative elevation threshold is used. The method extracts the optimal threshold range for identifying diphtheria aconitum and verifies its effectiveness. The results show that: ① In UAV images, the relative elevation threshold method can clearly and accurately reflect the distribution characteristics of ground features, and is suitable for extracting ground features with obvious height errors from ordinary grass; ② When 10 cm≤T<20 cm, diphtheria aconitum can be accurately identified, and is basically consistent with the actual distribution of diphtheria aconitum, with a classification accuracy of 92%; ③ The relative elevation threshold method can better separate diphtheria aconitum, which improves the reliability of the classification basis. It also realizes the high-precision identification of diphtheria aconitum. This method can be applied to actual grassland monitoring.