基于无人机多光谱影像的蔬菜种植监测技术研究  

Monitoring of vegetable planting information using UAV based multispectral imagery

作　　者：田明璐 班松涛[2] 袁涛[1] 王彦宇 马超[1] 李琳一[1] TIAN Minglu;BAN Songtao;YUAN Tao;WANG Yanyu;MA Chao;LI Linyi(Institute of Agricultural Science and Technology Information,Shanghai Academy of Agricultural Sciences,Shanghai Engineer Research Center for Digital Agriculture,Shanghai 201403,China;College of Natural Resources and Environment,Northwest A&F University,Xianyang 712100,China)

机构地区：[1]上海市农业科学院农业科技信息研究所,上海市数字农业工程技术研究中心,上海201403 [2]西北农林科技大学资源环境学院,咸阳712100

出　　处：《上海农业学报》2020年第5期115-121,共7页Acta Agriculturae Shanghai

基　　金：上海市农委科技兴农重点攻关项目[沪农科创字(2018)第2-1号];上海市科委“科技创新行动计划”农业领域项目(18391901100)。

摘　　要：利用无人机平台搭载多光谱成像仪获取研究区蔬菜地块的高分辨率多光谱遥感影像,对影像上各类蔬菜的光谱进行分析,发现在各个波段上,不同种类蔬菜之间光谱具有一定的差异。使用多种分类方法对影像上各类蔬菜进行分类,结果表明:基于像元的分类效果较差;面向对象分类法能够有效识别和区分不同种类的蔬菜,分类图中各类蔬菜的分布与实际情况一致,Kappa系数大于0.9,总精度达到90%以上,与基于目视解译的各类蔬菜种植面积结果相比较,误差在7.5%以内。综上,以无人机多光谱遥感影像为数据源,使用面向对象分类技术,可以实现对蔬菜种植信息的自动化监测。In order to acquire the high-resolution multispectral imagery of vegetable plots,a Micasense Rededge camera mounted on a six-axis multi-rotor UAV was used.Spectral analysis of different vegetable species showed differences with varying degrees at each band.Different methods were used to classify the vegetable species,the results showed that traditional pixel-based classification methods were not able to distinct different vegetable species effectively.The classification map was highly consistent with ground truth when using objectoriented classification method,and the accuracy test showed that the Kappa was higher than 0.9,and the overall accuracy was more than 90%.The areas of each vegetable species were calculated from the classification map and the errors were within 7.5%compared with the results of visual interpretation.In summary,the vegetable planting information could be precisely extracted from UAV based multispectral imagery using object-oriented classification method.

关 键 词：蔬菜 无人机 多光谱 面向对象分类 

分 类 号：S127[农业科学—农业基础科学]