Comparison of three noise reduction methods for GNSS elevation time series
-
摘要: 为了探究经验模态分解(EMD)、整体经验模态分解(EEMD)和小波降噪三种方法的降噪性能,以中国区6个国际GNSS服务(IGS)站高程分量的5 a、10 a和20 a时序数据为例,对它们的降噪结果进行比较分析. 首先利用线性拟合分离趋势项,并采用3σ准则剔除异常值,得到满足符合降噪要求的样本序列;然后分别用这三种方法分离样本序列中的噪声,得到降噪后的序列;最后以信噪比(SNR)、相关系数、均方根误差(RMSE)为评价指标比较分析它们的降噪性能. 实验结果表明:1)当坐标时间序列质量较差时,EEMD和小波降噪可以很好的分离噪声;2)对于5 a和10 a时序数据,小波降噪的效果最好;对于20 a时序数据,EEMD和小波降噪效果接近,优于EMD;3)小波降噪抑制有色噪声的能力最佳.
-
关键词:
- 全球卫星导航系统(GNSS)坐标时间序列 /
- 经验模态分解(EMD) /
- 整体经验模态分解(EEMD) /
- 小波降噪
Abstract: In order to explore the noise reduction performance of the empirical mode decomposition (EMD), ensemble empirical mode decomposition (EEMD) and wavelet analysis, elevation time series data of different lengths from sin IGS stations are taken as examples. Firstly, the outliers and the trend items in the original data are removed to get the sample sequence meeting the experimental requirements. Then, the sample sequence is denoised by three methods and gets the real signal without noise. Finally, calculating the indexes of signal-noise ratio, correlation coefficient and root mean square error of data to compare the three noise reduction methods. The experimental results indicate that: 1) EEMD and wavelet analysis can well denoise when the quality of coordinate time series is poor. 2) Wavelet analysis has the best denoising performance on the Global Navigation Satellite System (GNSS) coordinate time series with time span of 5 a or 10 a; For 20 a time series samples, EEMD and wavelet analysis have similar denoising effects and are better than EMD. 3) The force of wavelet analysis to eliminate colored noise is better. -
表 1 各站点评价参数统计
m 站点 评价指标 EMD EEMD 小波 5 a 10 a 20 a 5 a 10 a 20 a 5 a 10 a 20 a BJFS Rsn 6.699 4 4.711 3 6.893 2 7.7963 5.9044 7.2519 6.6059 6.7721 5.7586 R 0.664 5 0.612 8 0.705 9 0.7036 0.6679 0.7183 0.7147 0.7254 0.3142 RMSE 0.004 4 0.005 0 0.004 8 0.0042 0.0047 0.0047 0.0044 0.0045 0.0071 KMIN Rsn 15.216 1 17.085 2 16.648 2 14.9817 14.4541 17.7433 12.6550 12.4947 15.9068 R 0.881 3 0.905 3 0.900 5 0.8790 0.8741 0.9113 0.8564 0.8547 0.8954 RMSE 0.004 9 0.004 5 0.005 5 0.0050 0.0052 0.0052 0.0056 0.0057 0.0057 LHAS Rsn 9.343 3 13.961 0 14.966 0 16.7810 17.3550 17.0938 15.9924 17.4586 18.8132 R 0.784 6 0.870 3 0.886 1 0.8977 0.9064 0.9050 0.9003 0.9132 0.9257 RMSE 0.005 3 0.004 4 0.004 5 0.0035 0.0037 0.0040 0.0036 0.0037 0.0037 SHAO Rsn –13.326 9 −14.067 6 –18.729 2 6.9709 7.6652 7.4851 22.3927 14.3306 14.1328 R 0.233 0 0.249 5 0.180 5 0.6911 0.7315 0.7259 0.8731 0.8933 0.8874 RMSE 0.012 4 0.013 4 0.015 0 0.0039 0.0039 0.0041 0.0027 0.0028 0.0031 URUM Rsn 4.388 3 4.708 6 11.327 0 10.8715 11.9636 16.8493 11.5299 15.0094 16.8187 R 0.589 1 0.593 4 0.828 3 0.7771 0.8021 0.9026 0.8764 0.8946 0.9093 RMSE 0.005 5 0.005 7 0.004 8 0.0039 0.0040 0.0036 0.0034 0.0035 0.0036 WUHN Rsn −0.048 5 2.093 3 3.438 7 5.1568 6.8895 6.9276 7.3793 8.9143 9.0487 R 0.506 2 0.620 2 0.672 3 0.6052 0.7026 0.7065 0.7420 0.7721 0.7808 RMSE 0.007 7 0.007 5 0.007 1 0.0057 0.0058 0.0060 0.0052 0.0052 0.0054 -
[1] 黄立人. GPS基准站坐标分量时间序列的噪声特性分析[J]. 大地测量与地球动力学, 2006, 26(2): 31-33,38. [2] 马俊, 曹成度, 闵阳, 等. 基于主成分分析去除GNSS站坐标时间序列有色噪声[C]//第十一届中国卫星导航年会, 成都: 2020. [3] HUANG N E, SHEN Z, LONG S R, et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis[J]. Mathematical, physical and engineering sciences, 1998, 454(1971): 903-995. DOI: 10.1098/rspa.1998.0193 [4] 张双成, 何月帆, 李振宇, 等. EMD用于GPS时间序列降噪分析[J]. 大地测量与地球动力学, 2017, 37(12): 1248-1252. [5] 肖瑛, 殷福亮. 解相关EMD: 消除模态混叠的新方法[J]. 振动与冲击, 2015, 34(4): 25-29. [6] 胡爱军, 孙敬敬, 向玲. 经验模态分解中的模态混叠问题[J]. 振动、测试与诊断, 2011, 31(4): 429-434. DOI: 10.3969/j.issn.1004-6801.2011.04.006 [7] WU Z H, HUANG N E. Ensemble empirical mode decomposition: a noise-assisted data analysis method[J]. Advances in adaptive data analysis, 2009, 1(1): 1-41. DOI: 10.1142/S1793536909000047 [8] 张恒璟, 程鹏飞. 基于经验模式分解的CORS站高程时间序列分析[J]. 大地测量与地球动力学, 2012, 32(3): 129-134. [9] 马俊, 姜卫平, 周晓慧, 等. 联合小波和方差分量估计方法分析中国IGS测站时间序列变化特征[J]. 武汉大学学报(信息科学版), 2018, 43(4): 629-636. [10] 戴海亮, 孙付平, 姜卫平, 等. 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. [11] WU H, LI K, SHI W Z, et al. A wavelet-based hybrid approach to remove the flicker noise and the white noise from GPS coordinate time series[J]. GPS solutions, 2015, 19(4): 511-523. DOI: 10.1007/s10291-014-0412-6 [12] 魏冠军, 陈晨, 黄逸宇, 等. 基于小波变换的CORS高程坐标时序周期性分析与降噪方法研究[J]. 全球定位系统, 2019, 44(4): 61-67. [13] 马俊, 曹成度, 姜卫平, 等. 利用小波包系数信息熵去除GNSS站坐标时间序列有色噪声[J]. 武汉大学学报(信息科学版), 2021, 46(9): 1309-1317. [14] 张恒璟, 龙安森, 文汉江. EEMDAN的CORS站高程时间序列分析方法[J]. 测绘科学, 2020, 45(2): 29-34. [15] 姜卫平, 李昭, 刘万科, 等. 顾及非线性变化的地球参考框架建立与维持的思考[J]. 武汉大学学报(信息科学版), 2010, 35(6): 665-669. [16] 袁林果, 丁晓利, 陈武, 等. 香港GPS基准站坐标序列特征分析[J]. 地球物理学报, 2008, 51(5): 1372-1384. DOI: 10.3321/j.issn:0001-5733.2008.05.011