一种基于K-means聚类的最小二乘配置法及在大尺度地壳速度场高精度拟合推估中的应用

A K-Means-Based Least Squares Collocation Method and Its Application in High-Precision Fitting and Estimation of Large-Scale Crustal Velocity Fields

ES评分 0

DOI 10.14075/j.jgg.2024.11.525
刊名
Journal of Geodesy and Geodynamics
年,卷(期) 2025, 45(12)
作者
作者单位

1 长安大学地质工程与测绘学院, 西安市雁塔路126号, 710054;
2 西部矿产资源与地质工程教育部重点实验室, 西安市雁塔路126号, 710054

摘要
针对传统最小二乘配置(traditional least squares collocation,TLSC)算法在大尺度区域地壳运动速度场高精度拟合中,在监测站点稀疏区域与块体边缘处速度场拟合结果会出现异常与不平滑的问题,结合K-means聚类算法与TLSC算法发展了一种基于K-means聚类的最小二乘配置法(KLSC),并在青藏高原GNSS地壳运动实测速度场中验证该方法的有效性。结果表明:1)相较于TLSC算法,KLSC算法利用K-means算法在无监督分类中的优势,基于GNSS速度场本身特征先将研究区域划分为多个速度相似的子区域,然后在每个子区域内分别利用TLSC进行速度场拟合,避免了局部复杂地质环境对区域速度场拟合精度的影响;2)KLSC算法以各网格点到各聚类中心的距离最近为依据选取拟合参数,解决了数据稀疏区域速度场拟合结果较差的问题;3)KLSC算法利用次近距离拟合并结合卷积滤波,有效解决了块体边缘处速度场拟合结果不平滑的问题;4)KLSC算法拟合的速度场的RMSE精度和相关性均优于TLSC算法,东、北向拟合速度场RMSE精度分别提高37%~48.2%和52.1%~67.2%,相关性分别提高24.1%~24.7%和4.7%~5.2%。
Abstract
To address the issues of abnormal and non-smooth fitting results in sparse monitoring station areas and block margins when using the traditional least squares collocation (TLSC) algorithm for high-precision fitting of crustal motion velocity fields in large-scale regions, a K-means clustering-based least squares collocation method (KLSC) is developed by integrating the K-means clustering algorithm with the TLSC algorithm. The effectiveness of this method is validated using the GNSS-measured crustal motion velocity field on the Qinghai-Xizang plateau. The results show that: 1) Compared to the TLSC algorithm, the KLSC algorithm leverages the advantages of the K-means algorithm in unsupervised classification to first divide the study area into multiple sub-regions with similar velocities based on the inherent characteristics of the GNSS velocity field, and then applies TLSC for velocity field fitting within each sub-region, thereby avoiding the impact of local complex geological environments on the accuracy of regional velocity field fitting; 2) The KLSC algorithm selects fitting parameters based on the proximity of each grid point to the cluster centers, resolving the issue of poor fitting results in data-sparse areas; 3) By utilizing secondary nearest-neighbor fitting combined with convolution filtering, the KLSC algorithm effectively addresses the non-smooth fitting results at block margins; 4) The RMSE accuracy and correlation of the velocity fields fitted by the KLSC algorithm are superior to those of the TLSC algorithm, with improvements in RMSE accuracy of 37% to 48.2% and 52.1% to 67.2%, and increases in correlation of 24.1% to 24.7% and 4.7% to 5.2% for the eastward and northward fitted velocity fields, respectively.
关键词
GNSS速度场; 最小二乘配置算法; K-means算法; 青藏高原; 地壳平滑速度场
KeyWord
GNSS velocity field; least squares collocation method; K-means algorithm; Qinghai-Xizang plateau; smoothed crustal velocity field
基金项目
页码 1294-1303
  • 参考文献
  • 相关文献
  • 引用本文

王程, 瞿伟, 李久元, 唐兴友, 王光耀, 杨凯. 一种基于K-means聚类的最小二乘配置法及在大尺度地壳速度场高精度拟合推估中的应用 [J]. 大地测量与地球动力学. 2025; 45; (12). 1294 - 1303.

  • 文献评论

相关学者

相关机构