Conversion Of Ann-based Gps Elevation

GPS (Navigation by Satellite Timing And Ranging Global Positioning System) has the merits of not only high precision, multiple function, concise manipulation and short observation time, but also fitting to all weather conditions, So it is widely used in the area of surveying and mapping.We can obtain the baseline vectors and high-accuracy discrepancy of geodetic height between GPS points after observation datum was dealt with. If the geodetic height of one point has already been known, we ean get the geodetic height of each point in the whole net. Geodetic height is based on ellipsoidal height coordinate system, however, our country adopted the similituded geoiding as the normal high system, then the GPS geodetic height need to be converted into normal height. At present, GPS and level fitting method is adopted extensively as GPS height conversion method. Certain model error always exists when the mathematics model fits as the similituded geoid.Artificial neural network (ANN) is a new developing interdisciplinary disciplines, therefore, it has some advantages as follows: firstly, it is simplified approximate to biological neural network. Neural network have non-linear ability of mapping and have characteristic of non-model estimation; secondly, it is an effective tool of non-linear problems. ANN based GPS height conversion is a kind of self-adaptive method of mapping. ANN has not supposed terms but can acquire more rational results, through the limited times of samples' training, better mapping relation of function can be acquired from trained samples. So unknown influence of factor is avoided, error of models is reduced, thus can improve the precision of conversion of GPS height.Based on the analysis of conversion of GPS height and BP network algorithm, this paper designed three layer feed-forward BP network to convert GPS height, combined with an example, computation and analyses are made. And some questions about GPS height conversion by ANN are referred.