戴海亮, 孙付平, 姜卫平, 肖凯, 朱新慧, 刘婧. 小波多尺度分解和奇异谱分析在GNSS站坐标时间序列分析中的应用[J]. 武汉大学学报 ( 信息科学版), 2021, 46(3): 371-380. DOI: 10.13203/j.whugis20190107
引用本文: 戴海亮, 孙付平, 姜卫平, 肖凯, 朱新慧, 刘婧. 小波多尺度分解和奇异谱分析在GNSS站坐标时间序列分析中的应用[J]. 武汉大学学报 ( 信息科学版), 2021, 46(3): 371-380. DOI: 10.13203/j.whugis20190107
DAI Hailiang, SUN Fuping, JIANG Weiping, XIAO Kai, ZHU Xinhui, LIU Jing. Application of Wavelet Decomposition and Singular Spectrum Analysis to GNSS Station Coordinate Time Series[J]. Geomatics and Information Science of Wuhan University, 2021, 46(3): 371-380. DOI: 10.13203/j.whugis20190107
Citation: DAI Hailiang, SUN Fuping, JIANG Weiping, XIAO Kai, ZHU Xinhui, LIU Jing. Application of Wavelet Decomposition and Singular Spectrum Analysis to GNSS Station Coordinate Time Series[J]. Geomatics and Information Science of Wuhan University, 2021, 46(3): 371-380. DOI: 10.13203/j.whugis20190107

小波多尺度分解和奇异谱分析在GNSS站坐标时间序列分析中的应用

Application of Wavelet Decomposition and Singular Spectrum Analysis to GNSS Station Coordinate Time Series

  • 摘要: 为了有效地提取GNSS(global navigation satellite system)站坐标时间序列中的有用信息,提高坐标时间序列的建模精度,提出一种小波多尺度分解与奇异谱分析相结合的非线性运动建模方法,并利用全球11个测站20年(1999―2018年)的GPS(global positioning system)垂向坐标时间序列对所提方法进行了验证。首先,通过小波分解将坐标时间序列分解到不同尺度上; 然后,对分解后的各层高频部分和低频部分进行奇异谱分析; 最后,通过叠加合成得到原始坐标时间序列的拟合值,并对所提方法的拟合效果进行评估。结果表明,与单纯的奇异谱分析方法相比,所提方法能够更加准确地从含噪声的有限尺度时间序列中提取趋势和周期等有用信息,降低了部分周期项如季节周期项、月周期项被当作噪声剔除的概率,并且建模精度有26%的提高。

     

    Abstract:
      Objectives  In order to effectively extract useful information from time series of the global navigation satellite system(GNSS) sites, and improve the modeling accuracy of coordinate time series, this paper proposes a nonlinear motion modeling method combining wavelet decomposition and singular spectrum analysis. Experiments were carried out using global positioning system(GPS) vertical coordinate time series of 11 stations around the world from 1999 to 2018.
      Methods  Firstly, the coordinate time series is decomposed into different scales by wavelet decomposition.And then the singular spectrum analysis(SSA) is performed on the high-frequency part and the low-frequency part of each layer. Finally, combine the fitting value that is the fitting coordinate time series, and the fitting effect of the new method was evaluated.
      Results  The results show that, compared with the simple singular spectrum analysis method, the new method can extract useful information such as trend and period more accurately from the limited scale of the time series with noise. And reduce the partial period items in the singular spectrum analysis method to some extent, for example, the seasonal period item and the monthly period item are regarded as the probability of noise rejection, and the modeling accuracy is improved by 26%.
      Conclusions  The purpose is to propose a nonlinear motion modeling method based on wavelet decomposition and singular spectrum analysis, so as to improve the modeling accuracy of coordinate time series.

     

/

返回文章
返回