Alternative Methods for Improving Transformation Consistency Between Geocentric and Non-Geocentric (Local) Coordinate Systems

Abstract

Today, GPS is the most popular surveying technique used in determination of geodetic networks. Using this technique, one can obtain point coordinates in a geocentric coordinate system. On the other hand, geodetic frameworks in many countries are referenced to local coordinate systems. Therefore, geodesists often come across transformation problems between geocentric and local coordinate systems. The 7-parameter similarity transformation method, which reveals the real geometrical relationship between terrestrial coordinate systems, is usually the basic solution to this problem. However, this method may not always provide a transformation with a good consistency when the existing coordinates in local systems are distorted. In this case, alternative methods to the basic solution should be introduced to acquire a better consistency. For this purpose, the polynomial method is the most common alternative in practice. In this study, two other methods, namely the least square filtering and the artificial neural network methods, are introduced into the transformation problem. These three methods are investigated using the data composed of the point coordinates in ITRF94 and ED50 frames which are located in the city of Istanbul in Turkey. This investigation, carried out with respect to various criteria, shows that the most consistent results are obtained by the artificial neural network method.