Abstract:Based on more than ten years of global GPS station data, we carry out preprocessing of the detection and reparation of jumps through the wavelet transform, and introduce the PCA method to measure the station time. The feasibility and results of sequence changes are analyzed and evaluated. The most important nonlinear periodic item is extracted from the series. It shows that the principal component analysis method uses the method of orthogonal decomposition and coordinate residual space-time matrix decomposition into a number of orthogonal components. The results reveal that the residual time series shows obvious cyclical terms, and the east direction has a linear drift trend. Through the Fourier transform, the periodic term in the coordinate time series is extracted, showing that most of the global GPS stations have nonlinear periodic laws, in which annual and semi-annual cycles dominate, and the information related to geophysical phenomena is extracted for feature recognition.