多光谱无人机不同飞行高度下苹果树高的提取  

Study on extraction of apple tree height at different flight altitudes using multispectral UAV

作　　者：张振飞 郭靖[2] 颜安 侯正清 袁以琳 肖淑婷 孙哲 ZHANG Zhenfei;GUO Jing;YAN An;HOU Zhengqing;YUAN Yilin;XIAO Shuting;SUN Zhe(College of Resources and Environment,Xinjiang Agricultural University,Urumqi 830052,China;Institute of Landscape Architecture,Xinjiang Academy of Forestry Sciences,Urumqi 830092,China)

机构地区：[1]新疆农业大学资源与环境学院,乌鲁木齐830052 [2]新疆林业科学院园林绿化研究所,乌鲁木齐830092

出　　处：《新疆农业科学》2024年第7期1710-1716,共7页Xinjiang Agricultural Sciences

基　　金：新疆维吾尔自治区重点研发计划项目“新疆杏李、杏等主要果树抗寒关键技术研究”(2023B02026)。

摘　　要：【目的】利用多光谱无人机影像快速、准确、无损的获取苹果树高信息,实现无人机遥感技术对苹果树生长状况的监测,并分析无人机飞行高度对树高提取结果的影响。【方法】利用大疆精灵4多光谱无人机分别获取30、60和90 m飞行高度的苹果树无人机影像,经大疆智图(DJI Terra)软件处理生成DOM和DSM影像数据,基于生成的DOM和DSM,利用克里金插值法生成研究区DEM,将DSM和DEM作差生成苹果树CHM提取树高,与实地测量的果树高值进行回归分析和精度验证。【结果】30 m飞行高度平均树高提取精度为88.49%,R^(2)为0.8378,RMSE为0.4031 m;60 m飞行高度平均树高提取精度为74.72%,R^(2)为0.6577,RMSE为0.8846 m;90 m飞行高度平均树高提取精度为56.20%,R^(2)为0.5273,RMSE为1.4767 m。【结论】利用多光谱无人机遥感技术可以实现对苹果树高的提取,提取精度随着无人机飞行高度的增加而降低,30 m飞行高度提取结果最佳,90 m飞行高度提取结果最差。在合适的飞行高度内,多光谱无人机遥感技术可以快速、准确、无损的实现对果园果树生长状况的监测,提高果园的管理效率。【Objective】The purpose of this study is to utilize multispectral unmanned aerial vehicle(UAV)imagery to rapidly,accurately,and non-destructively acquire height information of apple trees,aiming to achieve monitoring of apple tree growth conditions using UAV remote sensing technology and analyze the influence of UAV flight height on the extraction results of tree height.【Methods】The DJI Phantom 4 multispectral UAV was employed to acquire UAV imagery of apple trees at flight heights of 30,60,and 90 m,respectively.The acquired imagery was processed using DJI Terra software to generate digital orthophoto models(DOM)and digital surface models(DSM).Based on the generated DOM and DSM,a digital elevation model(DEM)of the study area was created using the Kriging interpolation method.The difference between the DSM and DEM was used to generate the canopy height model(CHM)for extracting tree height.Regression analysis and accuracy validation were conducted by comparing the extracted tree heights with field-measured values.【Results】The average accuracy of tree height extraction at a flight height of 30 m was 88.49%,with an R^(2) value of 0.8378 and an RMSE of 0.403,1 m.At a flight height of 60m,the average accuracy of tree height extraction was 74.72%,with an R^(2) value of 0.657,7 and an RMSE of 0.884,6 m.At a flight height of 90 m,the average accuracy of tree height extraction was 56.20%,with an R^(2) value of 0.527,3 and an RMSE of 1.476,7 m.【Conclusion】The use of multispectral UAV remote sensing technology enables the extraction of apple tree height possible.The extraction accuracy decreases with an increase in UAV flight height.The best results are obtained at a flight height of 30 m,while the poorest results are obtained at a flight height of 90 m.Within appropriate flight heights,multispectral UAV remote sensing technology can rapidly,accurately,and non-destructively monitor the growth conditions of orchard fruit trees,thereby improving the management efficiency for orchard operators.

关 键 词：多光谱无人机 飞行高度 苹果树 

分 类 号：S127[农业科学—农业基础科学]