BP Neural Network Use In The GPS Height Fitting

The vertical component taken by GPS is based on the reference ellipsoid, and it has a Margin to the normal height that need for the engineer, that is height anomaly. Because of the geoid is irregular, the use of the vertical information taken by the GPS are restricted.GPS height transformation, through the height anomaly known point value calculation of height anomaly unknown point value then calculate normal high to the unknown point is the basic GPS Height conversion method. BP network because of its self-organizing, adaptive characteristics are widely used in GPS Height fitting.On the basis of introducing elevation and elevation systems and relationship about geoid height, normal height and orthometric height, we discussing the basic method and its limitations about application of BP neural networks in GPS Height fitting. Against some limitations about BP network ,put forward use genetic algorithm to optimize the network topology and BP weight threshold, in a bid to improve the accuracy of conversion. Based on the test analysis, summarized BP network topology, operating parameters selection rules, and discussed the Parameters and the effect of the evolution. Analysis the effect about different activation functions and data processing and the accessed some useful conclusions...