Summary: | 博士 === 國立中興大學 === 土木工程學系所 === 107 === How to use GNSS and leveling data with several different geometric interpolation methods and indicators to obtain a better regional earth undulation model, in order to promote the reasonable point precision and its efficiency as the subject of the study. When the fluctuation of the earth''s undulation is considered to be gentle within a certain range, a mathematical function can be used to reflect the change of the spatial distribution, and then the earth fluctuation values of other undetermined unknown points are obtained through interpolation. In this study, if the (1) least squares support vector machine is used in the test area selected in this study, the resulting elevation accuracy can reach ±1.8 cm. (2) Using the particle swarm algorithm, the resulting elevation accuracy can reach ±1.9cm. (3) Using the gene representation planning method, the obtained elevation accuracy can reach ±2.0 cm. (4) Using the WTLS quadric surface method, the obtained elevation accuracy can reach ±1.9 cm. (5) Using the RWTLS quadric surface method, the resulting elevation accuracy can reach ±1.8 cm. Using the regional earth undulation mode established by these several methods, the root mean square error of the checkpoints is within ±5 cm of the engineering measurement standard.
Five kinds of indicators were used to evaluate the six preferred regional undulation models, which were judged by the sensitivity and stability of the index values. The comprehensive conclusion in this study is that the "confidence index (C)" evaluation index has the best stability. It is concluded how to choose the best regional earth undulation model for various mapping applications.
|