测绘通报 ›› 2017, Vol. 0 ›› Issue (11): 51-57,63.doi: 10.13474/j.cnki.11-2246.2017.0347

• 学术研究 • 上一篇    下一篇

一种基于主成分分析的协同克里金插值方法

卢月明1, 王亮1, 仇阿根1, 张用川1,2, 赵阳阳1   

  1. 1. 中国测绘科学研究院, 北京 100830;
    2. 武汉大学资源与环境科学学院, 湖北 武汉 430079
  • 收稿日期:2017-04-21 出版日期:2017-11-25 发布日期:2017-12-07
  • 作者简介:卢月明(1991-),男,硕士生,主要研究方向为空间数据挖掘和地理信息系统应用。E-mail:925651787@qq.com
  • 基金资助:
    测绘新技术系统开发与示范应用(2016KJ0104)

A CoKriging Interpolation Method Based on Principal Component Analysis

LU Yueming1, WANG Liang1, QIU Agen1, ZHANG Yongchuan1,2, ZHAO Yangyang1   

  1. 1. Chinese Academy of Surveying and Mapping, Beijing 100830, China;
    2. School of Resource and Environmental Science, Wuhan University, Wuhan 430079, China
  • Received:2017-04-21 Online:2017-11-25 Published:2017-12-07

摘要: 针对协同克里金插值方法在插值时,辅助变量较多造成计算复杂度增加,而辅助变量较少引起插值精度降低这一问题,提出了一种基于主成分分析的协同克里金插值方法(PCA-CoKriging)。该方法首先使用主成分分析对插值相关变量进行将维,得到较少几个综合指标,然后里利用这几个综合指标作为辅助变量进行协同克里金插值。为验证该方法的有效性和数据分布对该方法的影响,本文选取了2016年北京市范围内4个季节中PM2.5浓度满足正态分布效果不同的4组数据,分别使用PCA-CoKriging和普通克里金插值方法、常规协同克里金插值方法,进行了插值试验。结果表明,本文方法与普通克里金插值方法、常规协同克里金插值法在4组试验中的平均绝对误差分别为4.91、6.04、5.61,平均均方根误差分别为6.65、8.76、7.57。综合比较,本文方法比常规协同克里金插值的平均绝对误差与均方根误差分别提升了10.73%、12.56%,比普通克里金插值法的平均绝对误差与均方根误差分别提升了18.71%、24.09%。

关键词: 主成分分析, 协同克里金插值, Pearson相关系数, PM2.5

Abstract: Aiming at the problem that the cooperative Kriging interpolation method has higher computational complexity when the auxiliary variables are numerous, and the interpolation precision is lower when the interpolation variables are less, a cooperative Kriging interpolation method based on principal components analysis is proposed (PCA-CoKriging). This method first uses the principal components analysis to reduce the dimension of the related interpolation variable, obtains several comprehensive indexes, and then uses these comprehensive indexes as the auxiliary variables to conduct cooperative Kriging interpolation. In order to verify the effectiveness of the method and the influence of the data distribution on the method, four groups of data about PM2.5 concentration in the four seasons in Beijing in 2016 which meet different effects of normal distribution are selected, and PCA-CoKriging, ordinary Kriging interpolation method, and conventional Co-Kriging interpolation method are used to carry out interpolation experiments. The results show that the mean square error in the method of this article, ordinary Kriging interpolation method, and conventional Co-Kriging interpolation method are 4.91, 6.04 and 5.61 respectively, and the average root mean square errors are 6.65, 8.76 and 7.57 in the four groups. In comprehensive comparison, the mean absolute error and root mean square error in the proposed method have increased by 10.73% and 12.56% respectively compared with those of the conventional CoKriging interpolation, and 18.71% and 24.09% respectively compared with those of the ordinary Kriging interpolation method.

Key words: principal component analysis, CoKriging, Pearson correlation coefficient, PM2.5

中图分类号: