| Summary: | 博士 === 國立中央大學 === 土木工程學系 === 102 === It is well-known that unmodeled biases, such as the multipath effect, are a major source of errors in GPS code and carrier phase measurements in the differential mode, which can hinder the achievement of the highest levels of accuracy. An alike multipath-characterized signal is spatially correlated within a small area that introduces slow varying errors in the measurements due to satellite dynamics, whose biases cannot be averaged out. These offset biases are unique, much like a portrait. According to the correlation between day-to-day time series residual estimates in the recent past, this relationship can be widespread and economically exploited to mitigate multipath errors.
In this study an innovative method, which involves empirical mode decomposition (EMD) in the Hilbert-Huang transform (HHT), is employed to analyze time-series phase residuals. After decomposition, statistical significance testing using a 95 percentile boundary line can identify a few short period components, whiles the white noise is determined using a threshold to eliminate the high frequency component. In this study show how to choose the best threshold. An extrapolation technique, which is rooted in grey relational analysis (GRA), is simultaneously utilized to predict the biases for the current positioning task and thus to correct for such systematic biases. When technically supported by the above-mentioned mode functions and grey modeling, classical least-squares adjustment with parametric weighting can yield more accurate three-dimensional coordinates.
The results also show that this mitigation technique is a necessary procedure, which allows the ambiguity solutions to become more reliable so that after correction there is a over 50% improvement in GPS kinematic OTF positioning and geodetic monitoring accuracy.
|
|---|
