Fitting the GPS/Leveling Quasi-Geoid Using Bayesian-Regulation BP Neural Network
The 2.5′×2.5′resolution local quasi-geoid is calculated using the global gravity field model and GPS/leveling data of region which points spacing is about 10km with the Bayesian- regulation BP neural network in this paper. The inner and outer precision of quasi-geoid are both superior 0.05m.The result indicat that the Bayesian regulation BP neural network could improve the precision of fitting and restrain the over-fitting in fitting. The region quasi-geoid excelled than 0.05m can be computed using the global gravity field model and about 10km baseline GPS/leveling data in smoothness region.
L. Song and X. Q. Hu, "Fitting the GPS/Leveling Quasi-Geoid Using Bayesian-Regulation BP Neural Network", Applied Mechanics and Materials, Vols. 90-93, pp. 2903-2906, 2011