A Study Of Validity Assessment Of Wavelet De-noising

Wavelet transform has played an important role in the field of geodetic data processing. Noise can be eliminated effectively based upon wavelet decomposition and reconstruction which is the fundamental process of geodetic data processing. Currently, wavelet de-noising has been widely used in deformation monitoring data process, GPS data process, geophysical inversion and so on. However, it is indeed hard to master true characteristics of the actual signals, thus the identification of the optimal de-noising results is difficult. On that account, this thesis mainly focus on the validity assessment of wavelet de-noising, and aims to develop a unified framework for validity assessment of wavelet de-noising.The primary contents of the thesis can be summarized as follows:(1) The framework of validity assessment of wavelet de-noising is specified, and it mainly consists of two parts:selection of wavelet analysis methods and identification of optimal decomposition and reconstruction scale.(2) A comparative study of validity assessment methods of wavelet de-noising is performed. The ability and applicability of existing validity assessment methods were obtained based on a series of special designed experiments.(3) Based on the characteristics of signals in the process of wavelet de-noising, the variation rule of the root mean square error, single to noise ratio and smoothness with regard to the decomposition scale is employed to develop a hybrid indicator for determine the best grading scale of wavelet de-noising. A series of experiments on different singles are utilized to test the effectiveness of the hybrid indictor. The comparisons with current methods are further used to demonstrate the advantages of the hybrid indictor.(4) The established framework for validity assessment of wavelet de-noising is further used in deformation monitoring data process in Chang ji Highway, Hunan Province. First, wavelet transform is used to eliminate noise from signals:root mean square error(RMSE) and signal to noise ratio(SNR) are utilized to select the wavelet analysis methods, and the hybrid indictor proposed above is used to identify the optimal decomposition and reconstruction scale. Then, Grey System Model (GM) is established for forecasting, and the accuracy of the prediction results is more than95percent.