摘要
针对现有的高分辨率遥感影像面向对象分类确定最优分割尺度研究中,大多仅考虑了对象光谱特征而忽略了对象空间特征的局限性,采用RMNE(the ratio of mean difference to neighbors(Abs)to entropy)方法,以高分二号(GF-2)影像为数据源,利用影像纹理信息熵作为对象内部同质性指标,对象光谱均值与邻域光谱均值差分绝对值作为对象之间异质性指标,并结合目视确定茶园最优分割尺度为170,进而利用面向对象分类方法实现了茶园提取。结果表明,基于RMNE方法确定最优分割尺度获取的分割结果,较为符合真实的茶园对象边界,并且该分割尺度下的茶园提取生产者精度达到96.76%,用户精度达到83.60%。
Aiming at the existing research on determining the optimal segmentation scale for object-oriented classification of high-resolution remote sensing images,most of which only considers the spectral characteristics of the object and ignores the limitations of the spatial characteristics of the object,this paper uses RMNE(the ratio of mean difference to neighbors(Abs)to entropy)method,using the Gaofen-2(GF-2)image as the data source,using the image texture information entropy as the internal homogeneity index of the object,and the absolute value of the object spectral mean and the absolute value of the neighborhood spectral mean difference between the objects.Qualitative indicators,combined with visual observation,determine the optimal segmentation scale of tea plantations as 170,and then use object-oriented classification method to achieve tea plantation extraction.The results show that the segmentation results obtained based on the RMNE method to determine the optimal segmentation scale are more in line with the real tea plantation object boundaries,and the tea plantation extraction accuracy under this segmentation scale reaches 96.76%and the user accuracy reaches 83.60%.
作者
陈慧
江洪
蒋世豪
CHEN Hui;JIANG Hong;JIANG Shihao(Key Laboratory of Spatial Data Mining&Information Sharing of Ministry of Education,Fuzhou University,Fuzhou,350108,China;National and Local Joint Engineering Research Center of Satellite Spatial Information Technology,Fuzhou 350108,China;Digital China Research Institute(Fujian),Fuzhou 350108,China)
出处
《测绘与空间地理信息》
2020年第12期17-20,共4页
Geomatics & Spatial Information Technology
基金
国家重点研发计划课题(2017YFB0504203)
福建省自然科学基金项目(2017J01658)资助。
关键词
面向对象
RMNE
最优分割尺度
茶园
object-oriented
RMNE
optimal segmentation scale
tea plantations
