GPR Data Interpretation Using Self Organizing Maps

A systematic workflow is proposed which includes ground penetrating radar data pre-processing and attribute calculation. The attributes are classified using a competitive artificial neural network, namely Self Organizing Maps (SOM). This workflow is applied to 3-D ground penetrating radar datasets acquired with the PulseEkko1000 system from a tank model and the archaeological site of Aptera, Chania, Crete, Greece. This proposed methodology, proved to be useful for distinguishing different pipes from the surrounding medium with high resolution and buried relics.