Research On GPS Height Fitting Model And Application

The definition of GPS leveling is to translate geodetic height computed from GPS surveying into normal height by quasi-geoid; it is widely used in the application of survey engineering. However, the abnormal of GPS points can not be obtained directly, so at present, we cannot translate geodetic height into normal height straightly and accurately. Conventional geometry leveling is the main method to determine normal height, but this method has poor efficiency. Therefore, how to utilize GPS, which has high precision in position, rapid speed, simple manipulation and so on to deal with the height problem, it has become a hot spot in survey engineering.This paper studies the principle of some fitting models and how to optimizing and choosing model parameters, and gives the method for obtaining higher precision abnormal height using surface fitting with project examples, then contrasted the fitting results between these models, and summarizes suitable situation as well as the fitting precision to each kind of model.The paper sums up a number of issues related to GPS leveling firstly, includes the definition of level surface, geoid, and quasi-geoid; the definition of geodetic elevation, orthometric heigh, and normal height; the principle of GPS leveling; and the primary methods for studying geoid. Then this paper introduces several methods to ascertain abnormal height, especially to surface fitting method. There two kinds model used in surface fitting method, which is single model and comprehensive model. In the first instance we talk about single models, includes cubic spline interpolation, polynomial fitting model, multiple surface fitting model, BP neural network method, and so on.Then mainly introduces comprehensive model to everybody, such as weighted aggregative model, hybrid model, remove-restore method, and so on. Comprehensive model composed of two or more kinds of single models, its composing has two forms, the one is making models fitting the quasi-geoid through coordinating weight matric, the other is Unifies the long wave model and the short wave model, and the two single models can support each other. Afterward the paper introduces the ways of data processing and model optimization before taking GPS elevation conversion. Data processing is eliminating gross error from GPS data and leveling data; model optimization is take optimization and selection to model parameters with F-test. And the paper introduces the Akaike Information Criterion (AIC) proposed by Akaike to choose the better model for transforming GPS height.Finally, Applies several kind of different models in the new city construction project in Yuhuan County Zhejiang Province and the quasi-geoid construction project in Chongqing using leveling connection surveying, through computing and contrast, then obtains the general rule of using fitting method to get abnormal height. The research shows that using fitting method is feasible, the Comprehensive model is better than single model. In the situation of eliminating gross error and position distributing uniformly, the fitting precision may reach 2cm, and even better if we can enhances the accuracy of geometry leveling and GPS surveying.