摘要
【目的】探索基于GF-1遥感影像的干旱区绿洲地物的最优分割尺度,为民勤绿洲地物的高精度提取奠定基础.【方法】本研究采用了综合类内一致性与类间异质性两种指数构建的RMAS分割指数,以国产GF-1遥感影像为数据源,民勤绿洲为试验区,获取了建筑物、林地与耕地的最优分割尺度;通过参考对象与分割对象之间的光谱与位置关系,提出了使用光谱相似性指数SSI以及分割距离指数SDI对分割质量进行评价;使用最邻近分类法对试验区地物进行提取.【结果】1)试验区建筑物、林地以及耕地的最优分割尺度为50、30、80;2)最优分割尺度下,SSI指数接近0,且SDI指数最小;3)地物提取的总体精度为87.50%,Kappa系数为84.35%,3种地物在最优分割尺度下提取的精度分别为82.86%、89.29%、96.15%.【结论】通过RMAS指数能够准确获取最优分割尺度,以达到较高的地物提取精度.
【Objective】To explore the optimal segmentation scale of the arid region based on GF-1 remote sensing image,and lay the foundation for the high accuracy extraction of the ground objects of Minqin oasis.【Method】The RMAS segmentation index was constructed by combining interclass consistency and interclass heterogeneity,which was used to obtain the optimal segmentation scale of the building,woodland and cultivated land.GF-1 remote sensing image was used as the data source.Considering the relationship between spectral and position,the spectral similarity index(SSI)and the segmentation distance index(SDI)were used to evaluate the segmentation result.The nearest neighbor classification was used to extract the features of the research area.【Result】The optimal scale was 50,30 and 80 respectively.In the optimal segmentation scale,the SSI index was close to 0 and the SDI index was the smallest.The overall classification accuracy was 87.50%,and the kappa coefficient was 84.35%.The producer accuracy of the building,woodland and cultivated land were 82.86%,89.29%and 96.15%respectively.【Conclusion】The optimal segmentation scale could be accurately obtained by RMAS index for high extraction accuracy.
作者
张华
张改改
ZHANG Hua;ZHANG Gai-gai(College of Geography and Environment Science,Northwest Normal University,Lanzhou 730070,China;College of Life Sciences,Lanzhou University,Lanzhou 730000,China)
出处
《甘肃农业大学学报》
CAS
CSCD
北大核心
2018年第4期116-123,共8页
Journal of Gansu Agricultural University
基金
国家自然科学基金项目(41461011)
作者简介
张华(1978-),女,博士,副教授,主要从事生态水文与环境遥感方面的研究工作.E-mail:zhanghua2402@163.com。
