基于多季相光谱混合分解和决策树的干旱区土地利用分类

发布时间：2018-02-26 07:10

本文关键词： 土地利用决策树分类土地覆被光谱混合分解多季相遥感数据 Landsat OLI　出处：《农业工程学报》2016年19期 　论文类型：期刊论文

【摘要】：该文利用2015年Landsat 8多季相数据,以民勤县为例探讨了干旱区土地覆被/利用分类方法。首先进行了研究区多季相遥感数据的主成分变换,确定了Landsat OLI光谱空间的内在维数。其次通过对各季相主成分空间的分析,确定了光谱混合分解的端元类型及各自的代表性季相。获取端元光谱之后,采用全约束线性光谱混合模型分解各个季相的遥感影像,估计各端元组分占像元的面积百分比(端元丰度值)。最后利用端元丰度值的多季相估计结果,依据先验知识和训练样本建立决策树分类规则进行土地覆被/利用分类。分类方法的总体精度达到90.94%,Kappa系数为0.90。研究表明:将物理意义明确的端元丰度值作为分类变量,能够快速、有效地获取分类规则,生成树的结构简单、合理、清晰,对训练数据的依赖性较低,因此具有应用于整个旱地系统土地覆被/利用分类的潜力。
[Abstract]:In this paper, the classification method of land cover / utilization in arid area is discussed by using Landsat 8 multi-season data in 2015, and Minqin County as an example. Firstly, the principal component transformation of multi-seasonal remote sensing data in the study area is carried out. The intrinsic dimension of the Landsat OLI spectral space is determined. Secondly, through the analysis of the principal component space of each seasonal phase, the endmember types of the spectral mixing decomposition and their representative seasonal phases are determined. A fully constrained linear spectral mixing model is used to decompose the remote sensing images of each seasonal phase, and the area percentage of each end component in the pixel is estimated (end component abundance value). Finally, the multi-quarterly estimation results of the end component abundance value are used. According to the prior knowledge and training samples, the classification rules of decision tree are established for land cover / use classification. The overall accuracy of the classification method is 90.94 and Kappa coefficient is 0.90. The results show that the abundance value of the end element with clear physical meaning is taken as the classification variable. It can obtain classification rules quickly and effectively, and the structure of the spanning tree is simple, reasonable and clear, and the dependence on training data is low, so it has the potential to be applied to land cover / use classification in the whole dryland system.
【作者单位】：中国农业大学资源与环境学院;北京市农林科学院草业与环境研究发展中心;
【基金】：中国博士后科学基金资助项目(2016M591112)
【分类号】：F301.2

【相似文献】