面向线性光谱混合分解的邻域像元集螺线型构建方法

发布时间：2018-04-28 19:55

  本文选题：线性光谱混合分解 + 混合像元　； 参考：《测绘学报》2017年11期

【摘要】：高时间分辨率遥感影像在地表景观破碎区域易形成混合像元,难以发挥其高时间维度优势。现有方式多是基于线性光谱混合模型,借助邻域像元所构成的像元集合组成线性方程组,求出组分光谱值的最小二乘解,提高其空间分辨率。然而,现有方法依赖窗口形式来构建邻域像元集合,在某些区域易造成方程组无解的欠定问题。本文在分析其问题原因的基础上,引入阿基米德螺线代替传统的矩形窗口,对邻域各像元依次遍历,构建空间邻近、组分相近的邻域像元集合来解决该问题。在GlobeLand 30数据上的试验表明,螺线型构建方法对5种混合尺度上多种类型地物均具有稳定的精度,与传统窗口构建方法相比,可从构建邻域像元集合方面将总体理论精度提高2%,分解结果精度提高近1个数量级。
[Abstract]:High time resolution remote sensing images are easy to form mixed pixels in the broken area of surface landscape, so it is difficult to give full play to the advantage of high time dimension. Most of the existing methods are based on the linear spectral mixing model. By means of the pixel set formed by the neighborhood pixel, the least square solution of the spectral value of the component is obtained and its spatial resolution is improved. However, the existing methods rely on the window form to construct the set of neighborhood pixels, which can lead to the unsolvable problem of equations in some regions. In this paper, based on the analysis of the causes of the problem, Archimedes spiral is introduced to replace the traditional rectangular window, the neighborhood pixels are traversed in turn, and the adjacent pixel sets with adjacent space and similar components are constructed to solve the problem. Experiments on GlobeLand 30 data show that the spiral construction method has a stable accuracy for many types of ground objects on five mixed scales, compared with the traditional window construction method. From the aspect of constructing neighborhood pixel set, the total theoretical accuracy can be improved by 2 and the resolution result accuracy can be improved by nearly one order of magnitude.
【作者单位】： 吉林大学地球探测科学与技术学院;国家基础地理信息中心;中国矿业大学(北京)地球科学与测绘工程学院;
【基金】：国家自然科学基金重点项目(41231172)~~
【分类号】：P237
，

本文编号：1816643