摘要
本文首先分析了遥感影像尺度的三层次内涵。重点针对遥感像元尺度,分析了遥感像元尺度效应及其分形机理,由于现有分形方法没有考虑影像本身尺度(空间分辨率),造成尺度间分形维数的比较时像元尺度效应变化难以有效反映,本文针对此问题提出了基于表面积的加窗分形布朗运动和加窗双层地毯两种改进分形方法。为验证改进分形方法的可靠性,采用了不同像元尺度下系列监督分类进行验证。试验结果表明,每种地物的分维数都随着空间分辨率的降低或像元尺度的缩小,呈总体下降趋势,在某些特征尺度上会出现预示着某些地物结构的拐点,这些拐点对观测该区域地物具有一定指示意义。系列监督分类精度也一定程度上证明了以上两种改进分形方法在分析尺度效应中的可行性。因此本文的方法对于分析遥感像元尺度效应和探索地物尺度聚合规律具有一定的理论意义。
This article first analyzed three levels connotation of remote sensing image scale that is ＂pixel level-scale＂ ＂object level-scale＂ and ＂pattern-level scale＂, and then analyzed fractal mechanism of remote sensing image pixels scale effect focus on remote sensing image pixels scale emphatically, as the existing fractal methods do not consider the image itself scale（such as spatial resolution）, causing pixels scale effect change compared by fractal dimension between scale is difficult to effectively reflect, this paper proposes MFBM and MDBM two modified fractal approach aimed at this problem .To verify the reliability of modified fractal method, the analysis based on connotation mechanism of pixels scale effect uses series supervised classification in different observation scales. Experimental results show that the fractal dimensions of each surface feature are going down generally with spatial resolution lowered and observation scale narrowed, and appearing inflection points in some features scale, these inflection points has some indication significance to observation the surface features in the region. Series supervised classification accuracy have proved the feasibility of both modified fractal methods in analyzing scale effect to a certain extent. Therefore the method in this article has some theoretical significance on analyzing remote sensing image pixels scale effect and exploring the surface feature scale aggregation rules.
出处
《测绘学报》
EI
CSCD
北大核心
2015年第B12期89-95,101,共8页
Acta Geodaetica et Cartographica Sinica
基金
国家自然科学基金(41371347)
中央高校基本科研业务费专项资金(2652013084)~~
关键词
像元尺度效应
尺度内涵
改进分形方法
分维数
监督分类
pixel scale effect
scale connotations
modified fractal method
fractal dimension
supervisedclassification
