Abstract: Taking the elevation coordinate time series data of six CORS stations in Hong Kong CORS network as the research object, wavelet multi-resolution analysis is used to detect the periodic variation and nonlinear motion trend in the coordinate time series from the time-frequency angle. The threshold method performs noise reduction processing on the coordinate timing, and compares the noise reduction effects according to different noise reduction quality evaluation indexes. The results show that there are very obvious characteristics of annual cycle and semi-annual periodic variation in the elevation coordinate timing. The wavelet noise reduction can effectively eliminate or reduce the noise information in the original time series, and obtain the coordinate time series closer to the theoretical analysis.