Study of whole fitting quasi-geoid using BP neural network method

Due to the model error and smooth joint problem of fitting quasi-geoid using polynomial fitting method,a new method of whole fitting quasi-geoid using BP neural network was proposed in this study.Using the gravimetric quasi-geoid and GPS/leveling data in an area,the BP neural network method was compared with the whole and divisional polynomial surface fitting method.The results show that the new method could reduce the model error in the case of a bigger area and the anomalous difference between two kinds of quasi-geoids.The new method can obviously improve the inner and outer precision of fitting two kinds of quasi-geoids than the whole and divisional polynomial surface fitting method.The BP neural network is a feasible fitting method for finishing local high precise and high resolution quasi-geoid.