Regresar

Using an Artificial Neural Network to transformation of coordinates from PSAD56 to SIRGAS95

Abstract:

The technological advance allowed the improvement of existent Geodetic Reference Systems (GRS) in their definition (system) as well in their realization (frame). Usually, the coordinate transformation among the different GRS is done in an analytical way by formulae based on transformation parameters which are not able to model local deformations present in the classical networks. The objective of this work is to evaluate an Artificial Neural Network (ANN) as a basis of a transformation method. For this study coordinates in the system PSAD56 (Provisional South American Datum 1956) in Ecuador were chosen and the cartesian geodetic coordinates in the system SIRGAS95 (Geocentric Reference System for Americas 1995) were determined. Firstly, a set of point coordinates was transformed by a seven parameter transformation derived from the similarity mapping among known coordinates of homologous points in both systems. Secondly, an Artificial Neural Network with Radial Basis Functions (ANN-RBF) was trained to pbkp_redict coordinates from one system to the other. The results found that the ANN-RBF gave better results for the transformation. These better results would indicate that an ANN-RBF is superior to model the existent deformations in classical networks like PSAD56 when transferred to a new geocentric system like the SIRGAS95. © Springer-Verlag Berlin Heidelberg 2009.