| Summary: | 碩士 === 國立中興大學 === 土木工程學系 === 93 === Abstract
Currently, the research area of most of the Geoid Model building in Taiwan area covers the entire Taiwan and usually the research method is the gravity measurement. However, because the gravity measurement data is difficult to obtain, and needs a great deal of manpower and working-hour, as well as the reason of the special terrain of Taiwan, the representation of the result is not good enough, and is not totally suitable for the precision requirement in small area. Therefore, this research utilizes GPS technique, the data from leveling, and different computation to calculate regional geodetic undulation. It is anticipated that the results must be under a specified precision. No matter on academic or practicality domain of engineering surveying, it should be a topic worthy of further study.
The research adopts the GPS points with known 1st order leveling data, acquiring the coordinate and the ellipsoid height by RTK, computing by 2nd curve surface fitting, and discussing on the integrality and the localization, with the changing of quantity of the referencing points, to achieve the best solution. At the same time, the BP Artificial Neural Network is adopted, implementing tests aiming to random point selecting, best fitting point and different training functions and concealed level nodes, to decide regional geoid precisely. According to this research, the root mean square could achieve and by adopting the curve surface fitting and BP Artificial Neural Network to build the geoid model. It fulfills the standard of engineering survey. Finally, compared with the program developed by Ministry of Interior and Taichung city government, then we found it could be applied for a better solution for the elevation measurement of GPS surveying, and could be the reference for future engineering survey.
|
|---|
