Approaches to Improving the Pre-Excavation Detection of Inhumations.

Abstract

As large scale landscape surveys continue to increase in commercial and research archaeogeophysics, there is still a markedly low ability to geophysically detect and interpret archaeological and forensic inhumations in some instances. The aim of this ongoing research project is to improve data acquisition by implementing an interactive ad hoc workflow model for determining appropriate methodologies for ground-penetrating radar (GPR) surveys, to improve data processing speed, and reduce observer error. Can the confidence of manual interpretations of GPR data be improved by adapting machine learning libraries for automatic object extraction and classification to GPR data based on a training dataset comprised of ground-truthed real GPR data and simulated GPR data?