The contribution of artificial intelligence to aerial photointerpretation of archaeological sites: a comparison between traditional and Machine Learning methods

On the basis of the research activity carried out as part of the Archeo 3.0 project 'Integration of key enabling technologies for the efficiency of preventive archaeological excavations', the authors explore the feasibility and limits of the automated approach for the recognition of archaeological marks. This approach is mainly motivated by the relevance that aerial photographs play in the reconstruction of ancient topography of human settlements. For this aim, a collection of historical aerial photographs of both the city and the necropolis of Vulci has been considered. These photographs, in colour and B/W, have been previously used in a PhD thesis in Ancient topography in which the traditional methodology (photointerpretation and cartographic restitution) has been fully exploited. In this work, a systematic study is presented in order to compare the results obtained with Machine Learning techniques and traditional ones. This comparison allows us to discuss the strengths and limits of both methodologies.

L'articolo riguarda l'utilizzo di tecniche di Intelligenza Artificiale applicate al riconoscimento delle tracce archeologiche da fotografia aerea, utilizzando alcune immagini della città di Vulci e delle sue necropoli. Il grado di precisione e la validità di questo metodo di tracciamento è stato comparato con una tesi di Dottorato condotta, sulle stesse fotografie, con tecniche di restituzione cartografica tradizionale.