Application of Artificial Neural Network in GPS Height Transformation

Abstract

GPS technology has been widely used since it was put into use. At present, the plane locational accuracy of GPS can already achieve millimeter level. But in terms of height, in order to apply the GPS ellipsoidal height in engineering practice, the geoid seperation or height anomaly of the corresponding point must be achieved to transform GPS geodetic height to normal height. In this paper, by taking 12 points from the national GPS control network of Xuxiang Village of Haining City as sample data, a BP neural network using 3-4-2-1 model structure is adopted and a nonlinear coefficient 1.1 is added in the response function. The height anomalies of the 5 points of the test set are calculated and the residual errors are achieved by comparing with the measured values. The internal and external coincidence accuracies of the model are 0.824cm and 0.922cm separately. The result shows that the model can completely meet the precision requirement of the fourth-grade leveling survey and can be used to transform the heights of the study area.