无人机RGB影像植被分类方法研究——以福建省三明市尤溪县为例  

Study on Object-oriented UAV RGB Image Vegetation Classification Method--Case Study at Youxi County of Sanming City

作　　者：吴宏正 陈曦 陈文惠[1,2] Wu Hongzheng;Chen Xi;Chen Wenhui(College of Geographical Science,Fujian Normal University,Fuzhou Municipality,Fujian Province 350007;National Key Laboratory Training Base of Humid Subtropical Mountain Ecology,Fuzhou Municipality,Fujian Province 350007)

机构地区：[1]福建师范大学地理科学学院,福建福州350007 [2]湿润亚热带山地生态国家重点实验室培养基地,福建福州350007

出　　处：《亚热带水土保持》2020年第4期1-8,共8页Subtropical Soil and Water Conservation

基　　金：福建省科技厅公益类目(2017R1034-1)。

摘　　要：本研究以三明尤溪某地作为研究区,以DJI Phantom 4 Pro V2.0无人机航测影像为数据源,采用面向对象分类方法进行植被分类。经测试,取尺度参数100、形状参数0.3和紧致度参数0.5分割效果最好。利用光谱、纹理、形状等信息创建规则进行分类,将地物细分为芦苇、芒萁、杉木、马尾松、竹林、水田、百香果、柑橘、茶树、草地和3种农作物等13种植被类型,房屋、道路、水体、裸地等4种非植被类型,总体精度达到85.2%,Kappa系数为0.84。By using DJI Phantom 4 Pro V2.0 UAV aerial survey images as data sources,it was conducted the vegetation types classification with the object-oriented classification method at Youxi County of Sanming Municipality which was selected as a research area.After testing,the result shows that it has the best effect when the image segmented with the scale parameter of 100,shape parameter of 0.3 and compactness parameter of 0.5.Based on the creation rules of spectrum,texture,shape and other information,it was subdivided the features into 13 vegetation types including reeds,mandarin trees,fir trees,pine trees,bamboo forest,paddy fields,passion fruit orchard,citrus orchard,tea trees,grasslands and three crops,and 4 non-vegetation types including houses,roads,water bodies and bare land.The overall accuracy of study is 43.9%,and the Kappa coefficient is 0.4.

关 键 词：无人机RGB影像 多尺度分割 面向对象分类 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置] Q948.152[自动化与计算机技术—控制科学与工程]