面向对象的高分辨率影像耕地信息提取  

Object-oriented Farmland Extraction from High Spatial Resolution Imagery

作　　者：王新军[1,2] 武红旗[1,2] 盛建东[1,2] 蒋平安 

机构地区：[1]新疆农业大学草业与环境科学学院,乌鲁木齐830052 [2]新疆草地资源与生态重点实验室,乌鲁木齐830052

出　　处：《新疆农业科学》2012年第2期371-378,共8页Xinjiang Agricultural Sciences

基　　金：国家公益性行业(农业)科研专项经费项目(200903001);新疆土壤学重点学科基金

摘　　要：【目的】高分辨率遥感影像是快速提取耕地信息的主要数据源。【方法】在对阿图什耕地特征分析基础上,运用面向对象的特征提取方法,利用高分辨影像QUICKBIRD,探讨光谱特征、空间特征、纹理特征、典型特征等在耕地信息提取中的最优参数选择及具体应用,分析、讨论该方法优势与缺陷。并与传统分类方法提取结果进行对比分析。【结果】总精度提高了27.06%,Kappa系数提高了0.413 6,避免了"椒盐"现象。【结论】面向对象提取信息的方法周期较短、精度较高。[ Objective ] High spatial resolution remote sensing imagery is primary data source of quick farmland extraction. [ Method] This study was based on the farmland feature of Artux City, by applying the object- oriented feature extraction method and High Spatial Resolution Imagery QUICKBIRD. The optimized parameter selection and particular application of spectral, spatial, textural and typical feature in extracted farmland information were discussed in this paper. Moreover, the advantage and disadvantage of this method were analyzed and discussed. [ Result] Total accuracy was improved by 27.06% , Kappa coefficient was in- creased by 0. 413 6, and ＂salt＂ phenomenon was avoided. [ Conclusion] The result of experiment proved that the method has superiority of less periods and better precision for information extraction by analysis and comparison with the result that was extracted by traditional classification method.

关 键 词：面向对象 高分辨率 遥感影像 耕地 

分 类 号：S517[农业科学—作物学]