基于面向对象分类的南方水稻种植面积提取方法  

Object-oriented Classification for the Extraction of Rice Planting Area in South

作　　者：陈燕丽[1,2] 莫伟华[1,2] 莫建飞[1,2] 王君华[1,2] 钟仕全[1,2] 

机构地区：[1]广西气象减灾研究所,广西南宁530022 [2]国家卫星气象中心应用遥感试验基地,广西南宁530022

出　　处：《遥感技术与应用》2011年第2期163-168,共6页Remote Sensing Technology and Application

基　　金：广西科学研究与技术开发计划项目(桂科攻0816006-8);广西科学基金项目(桂科自0832204)

摘　　要：南方丘陵地区水稻种植具有分散、地块小、形状多样等特点,利用中低分辨率遥感数据提取水稻种植面积,难以满足精度要求。以SPOT5遥感影像为数据源,应用面向对象的分类方法提取了广西玉林市辖区晚稻种植面积。针对试验区不同稻作区的种植特点,选择其适合的尺度及参数进行多尺度影像分割,建立影像对象的层次结构,计算对象的光谱、几何及拓扑关系等特征,形成分类规则对不同稻作区进行信息提取。采用野外实地调查数据对分类结果进行类别和面积一致性检验,总体精度96.31%,Kappa系数0.9226,面积一致性精度99.92%。Rice planting in southern hills exhibits many characteristics, such as scattered, small land, diverse shape and so on. Using low-resolution remote sensing data for the extraction of rice planting area can not meet the required precision. Object oriented classification of SPOT5 image has been carried out to extract the rice planting area of Yulin city in Guangxi province. As different characteristics of the rice planting area,appropriate scale and parameters were selected to multi-scale image segmentation, a hierarchy of images objects is established. Then we calculated the object＇s spectral, geometric and topological characteristics and form the classification rules for different rice areas to extract rice information. Field survey data was used in area and category of classification consistency test, the overall accuracy of 96. 31%,Kappa coefficient of 0. 9226,an area of consistent accuracy of 99.92%.

关 键 词：面向对象分类 SPOT5 多尺度影像分割 信息提取 水稻 种植面积 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]