MULTISPECTRAL IMAGES CLASSIFICATION APPLIED TO THE IDENTIFICATION OF ARCHAEOLOGICAL REMAINS: A POST-DICTIVE PERSPECTIVE

Automated and semi-automated image classi!cations have made their way into archaeological applications, but early attempts have been strongly criticized. "is study examines semi-automated detection methods of archaeological evidence through a comparison of pixel-based and object-oriented data classi!- cation. "is research has been carried out on high-resolution imagery (WorldView-2) and the selected case study is located on the western slope of Etna (Sicily), the highest volcano in Europe, where a huge variety of settlements can be found from Prehistoric to Medieval times. "e methodology of both pixel-based and object-based data classi!cation is described and discussed over to speci!c case-study. "e di#erent nature of the two methods combined with the post-dictive approach adopted provides useful results in order to determine robustness and weakness of techniques presented here. In fact, our goal is to analyze advantages and disadvantages of the usage of pixel and object-based classi!cation techniques and shed light on the signi!cant change in pattern recognition. Finally, the obtained data are compared with manual visual interpretations and analyzed in terms of their accuracy.